Cancer stem cells are malignant cells with the capacity of self-renewal, the potential to develop into any cell in the overall tumor population, and the proliferative ability to drive continued expansion of the population of malignant cells (Jordan et al., 2006). It is widely accepted that CSCs play a critical role in cancer initiation, progression, recurrence, and metastasis (Nguyen et al., 2012; Shi et al., 2013). CSCs, possessing elevated tumor-initiating ability (Pattabiraman and Weinberg, 2014), can be identified by biomarkers (regarded as stem-like markers) including epithelial cell adhesion molecule (EpCAM), CD44, CD24, CD133, and aldehyde dehydrogenase (ALDH), etc. (Ajani et al., 2015; Lu et al., 2016). Moreover, CSCs display the ability to self-renew, which is often implicated in several signaling pathways, such as Wnt, Notch, and Hedgehog (Hh) signaling pathways (Takebe et al., 2011).

Evidence for stem-like CTCs existence was convincingly documented based on the expression of stem cell markers in CTCs (Table 1). ALDH1 is a classical stem cell marker that is frequently expressed in CTCs from patients with breast cancer (Reuben et al., 2011; Kasimir-Bauer et al., 2012), pancreatic cancer (Ting et al., 2014), and lung cancer (Hanssen et al., 2016). Theodoropoulos et al. (2010) showed that 35.2% of CTCs in metastatic breast cancer patients had the CD44⁺/CD24^-/low phenotype, which is highly associated with stemness and tumorigenic potential. CD133, an identified CSCs indicator, was overexpressed in CTCs and contributed to resistance to cancer therapies in breast cancer (Nadal et al., 2013) and colorectal cancer (Iinuma et al., 2011). Schölch et al. (2016) reported that the stem cell marker Bmi1 was significantly upregulated in CTCs, suggesting that CTCs possess a stem cell-like phenotype and an increased capacity of self-renewal and tumor formation. Moreover, EpCAM, which is considered a CSC marker, is often expressed in CTCs and commonly used for their enrichment and detection. Furthermore, the tumor-initiating capacity of CTCs was proven through xenografts from freshly isolated CTCs. Hodgkinson et al. (2014) enriched EpCAM⁺cytokeratin⁺ CTCs from the blood of small-cell lung cancer (SCLC) patients and subcutaneously injected them into immunocompromised mice to form palpable tumors within 4 months. Blood samples from metastatic breast cancer patients were depleted of hematopoietic cells and transplanted into the femoral medullar cavity of immunocompromised mice to test whether CTCs are tumorigenic. Within 6–12 months after transplantation, EpCAM⁺CD44⁺CD47⁺MET⁺ CTCs developed multiple bone, lung, and liver metastases in mice (Baccelli et al., 2013). In addition, a xenograft model was successfully generated with CTCs from isolated prostate cancer patients (Williams et al., 2015).

Due to shed off from primary tumors followed by intravasation into the circulation system, cancer cells with stem-like properties performed as circulating tumor stem cells (CTSC) (Clevers, 2011) (Figure 1). In the early steps of the metastatic cascade, CTCs have been thought to be involved the epithelial-mesenchymal transition (EMT) process (Aceto et al., 2014), which leads epithelial cells to weaken their cell–cell adhesion and gain migratory and invasive properties, thus becoming mesenchymal-like cells (Ksiażkiewicz et al., 2012). Once induced in cancer cells, EMT may allow them to escape from the primary tumor, invade into the circulation, and reach the site of future metastasis (Krawczyk et al., 2014). Upon arriving at their destinations, the interactions between such CTCs and the local microenvironment or “niche” modulate tumor metastatic colonization (Li et al., 2007). The survival and stemness of CTCs can similarly benefit from interactions with supportive niches, which are rich in developmental and self-renewal signals, such as Hh, Wnt, and Notch (Massagué and Obenauf, 2016). Factors secreted by stromal cell can also influence the stem cell niche, providing a suitable microenvironment for cancer initiation and development. For example, breast carcinoma clones are primed for metastasis in the CXCL12-rich microenvironment of the bone marrow (Zhang et al., 2013), while bone morphogenic proteins (BMPs) can induce CD133⁺ CSCs to differentiate and markedly attenuate the tumor-forming ability of CD133⁺ cells (Visvader and Lindeman, 2008). Due to protection by the niche, cancer cells are able to adapt to the microenvironment, initiate proliferation and eventually seed a successful metastasis. Finally, CTCs usually undergo a process of mesenchymal epithelial transition (MET) and recover their epithelial character to proliferate and differentiate into different secondary or metastatic tumors (Nieto, 2013). EMT is now known to not only facilitate the metastatic spread and the progression of cancer cells (Mitra et al., 2015), but also to support the induction of a stem cell phenotype with properties such as self-renewal associated with high invasiveness and resistance to apoptosis and therapy (Mani et al., 2008). Stem-like CTCs may proliferate as tumor-initiating cells to form aggressive tumors during relapse/metastasis.

Circulating tumor cells often occur at very low concentrations in the blood, with estimates of just one CTC per ∼10⁷ white blood cells (WBCs) per milliliter of blood, suggesting that any assay for CTC enumeration must be able to handle many surrounding/normal cells and the error rates will be high. Therefore, the different properties of CTCs that distinguish them from the surrounding normal hematopoietic cells can be utilized for their enrichment and detection. Generally, CTCs have large size distributions of 20∼30 μm in diameter, which is much larger than that of capillary bores (∼8 μm) (Chaffer and Weinberg, 2011). However, their physically plastic properties facilitate the pass of CTCs into circulation, and indeed CTC clusters can rapidly and reversibly reorganize into single-file chain-like geometries and transit through capillary-sized vessels (Au et al., 2016). In addition, CTCs can be cloaked by platelets or by coagulation factors to protect themselves from the immune system and oxidative stress (Labelle et al., 2011; Le et al., 2015). Thus, it is extremely difficult to accurately detect CTCs. However, CTC-based biomarkers provide useful approaches for capturing and detecting CTCs.

EpCAM frequently overexpressed on the surface of CTCs can be used as a cell biomarker (Maetzel et al., 2009; Huch and Dollé, 2016). It has demonstrated that CTCs show a high level of expression of EpCAM in numerous human malignancies (Went et al., 2006; Barsan et al., 2013). Yamashita et al. proved that EpCAM⁺ CTCs are an independent risk factor for HCC recurrence (Yamashita et al., 2013; Zhou et al., 2016). Consequently, anti-EpCAM is usually used for CTC screening. However, EpCAM is expressed in not all tumors, and absent EpCAM expression has been cited as a reason for the lack of CTC capture (Khan et al., 2011; Mikolajczyk et al., 2011). In addition, cytokeratins (CKs), including CK8, CK18 and CK19, are largely associated with CTCs, and fluorescent anti-CKs antibodies are often used to recognize and mark CK-positive cells.

Certainly, CTCs can provide an ideal approach to strongly affect cancer diagnosis and treatment strategies. Therefore, the detection of CTCs in blood samples is becoming one of the most active areas of translational cancer research. To improve the detection efficiency and speed, various sophisticated systems have been developed, which generally rely on the different properties and characteristics between WBCs and CTCs within the blood (Ee et al., 2016). Currently, methods based on physical properties, antibodies and nucleic acid are used to isolate and detect CTCs from whole blood samples. For instance, the large size, mechanical plasticity and dielectric properties of CTCs may endow their significant distinction compared with normal cells (Gascoyne and Shim, 2014; Sollier et al., 2014; Park et al., 2016). These physical properties are logically utilized for the isolation and enrichment of CTCs which are fast, simple and label-free; however, these methods are restricted by their non-specificity, which can be overcome by an antibody-combined method (Arya et al., 2013). Antibody-based functional assays include immunocytochemistry, immunomagnetic, and adhesion-based methods. Presently, the CellSearch assay, involving an immunomagnetic technique and image cytometry, is the only FDA-approved CTC diagnostic technology for metastatic breast, prostate, and colorectal cancer (de Wit et al., 2014). The sensitivity and specificity of markers are critical for the efficiency of such techniques, although none of the markers are expressed exclusively. Thus, two or more main approaches can be generally envisioned for this purpose. In nucleic acid-based approaches, CTCs can be detected using DNA- or RNA-based technologies with impressive sensitivity. For example, AdnaTest kits (AdnaGen, Germany), which utilize the technique of multiplex reverse transcription PCR (RT-PCR), allow simultaneous amplification and detection of multiple transcripts of circulating DNA or RNA (Markou et al., 2011). Additionally, several other methods, such as microscopy imaging, CTC filters, acoustic-based separation, etc., are available to detect CTCs (Zheng et al., 2011; Lopez-Riquelme et al., 2013; Li et al., 2015). Even though there are various approaches that have been developed for the detection of CTCs, none are ideal to meet the application needs, as issues of CTC loss, low purity, and narrow detection spectrum still need to be addressed (Viswanath and Kim, 2016). Most of these methods contain a series of complicated processes, such as erythrocyte lysis, cell centrifugation and washing, which may lead to insufficient capture and even a waste of detection time and money. Overall, because of the extreme rarity of CTCs and inadequate sensitivity and specificity, CTC detection and enumeration is still not a part of routine tumor staging in clinical practice.

The challenge of CTC detection is related to the requirement of high sensitivity combined with high specificity (Paterlini-Brechot and Benali, 2007), however, several factors still hinder the analysis of CTCs which can be summarized as: (i) CTCs are very rare in the circulation, (ii) there is no marker that can reliably and efficiently distinguish these CTCs from other blood-borne cells (Esmaeilsabzali et al., 2013) and (iii) downstream molecular and genomic characterization are challenging due to the low number of cells that can be isolated (Friedlander et al., 2014). To date, the CTC analysis techniques should be urgently developed with high sensitivity and selectivity, which can also be used for the rapid purification of CTCs, paving the way for downstream CTC characterization, and obtaining viable CTCs for ex vivo expansion (Lin et al., 2014). Currently, nanotechnology provides extensive applications in biology and medicine, and many researchers take advantage of nanotechnology to improve the efficiency and sensibility of CTC capture and to accelerate detection speed. Nanomaterials possess many unique and excellent physical properties that can be used to overcome the limitations of traditional CTC detection methods and make viable CTCs more accessible.

Immunomagnetic technology is used extensively in CTC enrichment and detection, because it is easy to manipulate and exhibits high capture efficiency and specificity. Based on antibody-antigen binding, immunomagnetic technologies have good sensitivity that makes it especially suitable for rare CTC separation. Additionally, in immunomagnetic assays, a magnetic field can be introduced without direct contact with cells and attract cells over a broader spatial domain (Chen et al., 2013). Thus far, various types of immunomagnetic technologies for CTC separation have been invented. In the earlier stage, magnetic particles (microbead) were in range more than 0.5 μm, while MNPs emerged with a smaller diameter in 5–200 nm (Bhana et al., 2015). MNPs commonly composed of magnetic elements, such as cobalt (Co) and iron (Fe), show alignment of their magnetic moment in the presence of magnetic field. MNPs reveal higher cellular binding capability and excellent stability in whole blood. Their smaller size makes the attachment to CTCs with many MNPs easy and leads to a higher magnetic susceptibility. Furthermore, MNPs with various biomarkers can be exploited to characterize CTCs.

CellSearch system, a commercially available device for CTC detection, uses Fe₃O₄ MNPs coated with anti-EpCAM antibody to confer magnetic properties to the EpCAM positive cells, resulting in magnetic separation of CTCs from the bulk of other cells in the blood (Truini et al., 2014). The CellSearch system can enumerate CTCs as low as two CTCs in 7.5 mL of peripheral blood (Allard et al., 2004). MNPs functionalized with anti-EpCAM antibodies were used to bind selected cells in the presence of a magnetic field in a reversibly bonded nanotextured polydimethylsiloxane (PDMS) chamber using NdFeB block magnets (Hoshino et al., 2011). Kim et al. (2013) presented a microseparator for the isolation of CTCs using immunomagnetic nanobeads with bound EpCAM antibodies, revealing that the CTC microseparator isolated approximately 90% of spiked CTCs in human peripheral blood and purified to approximately 97%. In another micromagnet array, which was fabricated integrating inkjet-printing technology combined with immunomagnetic microchip, the whole blood sample is labeled with MNPs and separated due to the magnetic momentum when the mixture flows through the microchannel (Figure 2A) (Chen et al., 2015). Immunomagnetic nanospheres (IMNs) fabricated by a convenient and highly controllable layer-by-layer assembly method successfully captured more than 94% of tumor cells in whole blood via only a 5 min incubation (Wen et al., 2014). These efforts of using immunomagnetic nanotechnology as a CTC capture platform promoted their potential use in clinic. Meanwhile, criticism is often expressed regarding the fact that detecting only EpCAM⁺ cells via this platform would miss some EpCAM^- CTCs and yield a false negative detection (Andree et al., 2015).

Nanoscale components present in the tissue microenvironment, including the extracellular matrix (ECM) and cell-surface structures, have been documents. Due to their mutual scale as a fundamental asset, nanostructured substrates with nanoscale topography that mimic the natural ECM or basement membrane can effectively interact with cell surface components. The major feature of nanostructured substrates is their larger biomimetic surface area. This feature not only enhances the interaction between substrates and cell surface targeting but also benefits the isolation and detection of CTCs (Wang et al., 2013). Furthermore, nanostructured substrates can be coated with higher densities of ligand and then increase the binding affinity to cells (Bhana et al., 2015). Therefore, nanostructured substrates with improved enrichment and detection accuracy may be developed as promising bio-platforms for CTC detection.

Various nanostructure substrates have been employed for CTC detection, including nanopillars (Wang et al., 2011), nanowires (Shen et al., 2013), nanofilms (Wang et al., 2016), nanocoatings (Figure 2B) (Eduardo et al., 2015), and PDMS (Islam et al., 2015), etc. Silicon nanopillars (SiNP) with diameters in the range of 100–200 nm on a silicon wafer have been used for cancer cell isolation, and the capture yield of MCF-7 cells in culture medium was 45–65% on SiNP; 10 times more than what was achieved on flat silicon (Wang et al., 2009). Tseng and colleagues pioneered a concept of “NanoVelcro” cell-affinity substrates, in which CTC capture agent-coated nanostructured substrates were utilized to immobilize CTCs with a high efficiency. Generations of NanoVelcro CTC chips based on silicon nanowire substrate (SiNS) were developed with a high CTC capture efficiency (Wang et al., 2011; Liu and Wang, 2014; He W. et al., 2016). The thermoresponsive NanoVelcro chip was created by grafting thermo-responsive polymer brushes [poly (N-isopropylacrylamide), PIPAAm] onto SiNS and incorporating with an overlaid PDMS chaotic mixer. This chip can increase the contact frequency between CTCs and NanoVelcro substrates, as well as the capture and release of CTCs with fast temperature responsiveness at 37 and 4°C, respectively (Ke et al., 2015). Furthermore, the CTCs can be purified and subjected to amplifications of their genomic DNA. Recently, a degradable zinc-phosphate-based hierarchical nanosubstrate (HZnPNS) (400–800 nm) emerged for the capture and release of CTCs (Shan et al., 2016). HZnPNS was functionalized with anti-EpCAM antibody and detected 6–75 CTCs/mL from metastatic cancer patient blood samples, which outperformed CellSearch system (1–125 CTCs/7.5 mL). In addition, HZnPNS allowed 88 ± 4% of captured cells to be gently released with a high viability of 92 ± 1%. Although nanostructured substrates can be used as an ultrasensitive tool for enriching rare CTCs from blood samples and keeping most captured cells viable for subsequent molecular analysis, further testing on clinical samples is necessary before reaching conclusions. There is still a challenge for purification platforms to pave the way for subsequent molecular and functional analysis.

Microfluidics, another recent technological development, can also improve the capabilities of CTC capture and detection. In microfluidics systems, low volumes of fluids are processed to achieve multiplexing capabilities and high-throughput screening (Volpatti and Yetisen, 2014). Taking advantage of small sample-volume requirements, fast processing times and large surface areas, various microfluidic platforms have been developed for rare CTC capture and enrichment (Hajba and Guttman, 2014). Various cell separation mechanisms have been devised in microfluidic platforms, including mechanisms that rely on magnetic forces, affinity chromatography, size and/or deformability-based isolation, and dielectrophoresis (Peng et al., 2013). Microfluidic technology is of great promise for detecting viable CTCs with satisfactory efficiency and purifying CTCs for down-stream biology analyses (Trifanny et al., 2016).

Recent progresses in nanotechnology have aided the development of advanced CTC-detecting microfluidic devices for the recovery and purification of CTCs. Immunomagnetic assays are often combined with microfluidic technology to enrich and detect CTCs. MNPs are commonly used to label target tumor cells. When blood sample flows through the microchannel on top of an array of permanent magnets, nanoparticle-labeled cells can be separated and captured on the substrate of the microchannel. The average capture rates for SkBr3, PC3, and Colo205 cells were 97, 107, and 94%, respectively (Huang et al., 2013). A nanoelectronic microfluidic chip fabricated by silicon nanograss (SiNG) electrodes was developed for the label-free distinguishing of both epithelial and mesenchymal CTCs (Hosseini et al., 2016). This system showed a great preference for MCF-7 and MDA-MB231 cancer cells with high capture yields between 92 and 97%. In another recent report, SiNS was embedded in the microfluidic chip (Figure 2C) in the combined use of rationally designed aptamer cocktails to achieve enhanced and differential capture of CTCs from non-small cell lung cancer (NSCLC) patients through a synergistic effect (Zhao et al., 2016). Kwak et al. (2017) utilized MNPs functionalized with anti-EpCAM antibodies to integrate the surface of the Mag-Gradient Chip based on microfluidics. The simple and multi-functional Mag-Gradient Chip can isolate 3 mL of heterogeneous CTCs sample in 1 h and realize 95.7% EpCAM-positive and 79.3% EpCAM-negative CTC isolation (Kwak et al., 2017). Although advances in microfluidic technologies are likely to accelerate CTC detection, many challenges hinder the clinical applications of CTC technologies.

CD44, which generally acts as a cell surface hyaluronic acid (HA)-binding glycoprotein, is frequently exploited to target CSCs. Prospectively, CD44 has also been investigated to identify stem-like CTCs. Galanzha et al. (2009) developed a photoacoustic (PA) and photothermal (PT) flow cytometry platform for in vivo detection and killing of CTCs with a stem-like phenotype (CD44⁺ CTCs). MNPs functionalized with anti-CD44 were used to bind and capture the stem-like CTCs under a magnet, and golden CNTs (GNTs) conjugated with anti-CD44 were exploited to kill the stem-like CTCs at a low confluence. Magnetic-induced clustering of MNPs within single cell, yielded 6.6-fold enhanced PT signals compared to the intact area before magnetic action. Then, the PT technique was applied to demonstrate ablation of the CD44⁺ CTCs labeled by GNT-CD44, confirmed by cell membrane damage and changed in the optical and fluorescent images. Recently, Hong et al. (2016) reported a multifunctional magnetic nanowire (NW) offering a significant improvement in CTC isolation efficiency. On these magnetic NWs, five different types of antibodies, including DAPI, anti-EpCAM, anti-CD44, anti-vimentin, and anti-CD45, were employed to differentiate CTCs from the surrounding leukocytes. In particular, anti-CD44 was employed to capture CTCs with stem-like properties. Furthermore, anti-CD44, combining anti-EpCAM, anti-Keratin18, and anti-insulin-like growth factor antigen (anti-IGF-I), was also used to functionalize silver-coated gold nanorods (AuNR/Ag) as surface-enhanced Raman scattering (SERS) multispectral contrast agents for CTC detection (Nima et al., 2014).

Although nanomethods are certainly capable of identifying stem-like CTCs through surface stem cell markers, there is a lack of nanomethods capable of capturing individual stem-like CTCs with a high sensitivity. Because none of these stem cell markers can be used exclusively to isolate CTCs in every type of tumor, a combination of multitype antibodies might help to specifically target CTCs. In addition, newer isolation and detection nanotechnologies such as nanostructured substrates and/or the incorporation of microfluidics devices might achieve that level of sensitivity. Another critical need is the development of nanotherapeutics for the stem-like CTCs. Beyond their highly sensitive nano–bio interactions, drug carriers with nanotechnologies may be potentially developed for the delivery of various anticancer therapeutic agents to kill stem-like CTCs.

A new SERS nanoparticle system was proposed for the direct detection of CTCs in the blood (Wu et al., 2015). The hydrodynamic diameter was determined to be 21 nm for SERS nanoparticles. Under the optimized experimental conditions, these SERS nanoparticles exhibited excellent specificity and high sensitivity for the direct detection of cancer cells in rabbit blood. The limit of detection (LOD) was 5 cells/mL, and a good linear relationship was obtained between the SERS intensity and the concentration of cancer cells in the range of 5–500 cells/mL, which demonstrates that the SERS nanoparticles can be used for the quantitative analysis of cancer cells in the blood. Overall, although further testing of clinical blood samples is necessary, SERS nanoparticles are promising for use in the direct detection of CTCs in blood with excellent specificity and high sensitivity.

Carbon nanotubes have remarkable electronic properties and have been applied for the electric detection of cancer cells in the blood. Yang et al. (2014) demonstrated a quantitative CNT-based sensor for direct detection of cancer cells in whole blood using real-time electrical impedance sensing. Multi-walled CNTs (MWCNTs) are used to increase the surface area and electrical conductivity of sensors. The binding of tumor cells to EpCAM antibodies causes increased electron-transfer resistance. The detection limit of the proposed system for the target cancer cells is at a minimum concentration of 10 cells/mL in whole blood samples. This method, via the simple process of fabricating the sensors, accurate structural control over the electrode surfaces, and high sensitivity and specificity for the target cells, provides a promising platform for the early diagnosis of cancer. In addition, Banerjee et al. (2015) reported a non-invasive strategy for isolating cancer cells by autonomously propelled CNT microrockets. This designed microrocket had the ability to rapidly target (∼5 min) and efficiently capture (∼85%) TfR⁺ cancer cells coming from an artificial CTC-like suspension, then it can also function as a magnetic isolation platform to capture cells from peripheral blood cells following high resolution imaging. As a result, it may provide an approach for the rapid and efficient extraction of CTCs.