Optical nanomaterial-based detection of biomarkers in liquid biopsy

As a traditional biomarker, protein in the biofluid has long been utilized to monitor individuals’ health status. For now, this procedure, which is often described as a “blood test,” is included in part of routine medical check-ups [44]. The representative examples are alpha-fetoprotein (AFP) for liver cancer, carcinoembryonic antigen (CEA) for colorectal cancer, carbohydrate antigen 19-9 (CA 19-9) for pancreatic cancer, cancer antigen 125 (CA 125) for ovarian cancer, and prostate-specific antigen (PSA) for prostate cancer. Because these protein markers usually have an established reference range to discriminate the normal and abnormal concentration, the required sensing performance is relatively obvious. For example, the clinical cut-off range of the CEA marker is around 5 ng/mL for both lung and colorectal cancers [45]. Commercial ELISA kits can detect down to 0.2 ng/mL, and most biosensing studies report more sensitive LODs than ELISA [46‐48]. Likewise, PSA, the most common screening target in prostate cancer tests, is usually under 4 ng/mL in the blood obtained from healthy individuals (Table 1). If it is elevated to above borderline (> 10 ng/ml), the possibility of having prostate cancer reaches around 50% [26]. However, the background level and borderline can be shifted by age and other health conditions. The sensitivity of the commercial ELISA kits is around 8 pg/mL, and the recently suggested biosensing studies claimed an impressive performance over ELISA with sub-picomolar detection [49, 50]. However, these traditional biomarkers have limitations in the criteria of liquid biopsy. First, these markers are naturally present in the blood at a certain level, regardless of disease or other health issues. Second, protein LB makers do not provide a holistic view of the disease because they usually designate one or two specific cancers. For instance, a PSA level in the normal range tells us the patients might not have prostate cancer, but it does not mean they do not have any type of cancer. Furthermore, some cancers do not have analogous protein biomarkers. Thus, there is a limitation that the detection of protein level cannot offer comprehensive information on the health status of the patients, and separate tests may be required for each marker. Third, the background level of these protein markers in the blood is usually not low and highly varied according to individual differences, such as age, sex, race, and other factors. Fourth, these biomarkers are not able to contribute to the significant promises of liquid biopsy, including prediction of prognosis and therapeutic responses. The protein level itself is not relevant to the understanding of tumor heterogeneity. For this reason, there has been a question over its classification: do we need to include these traditional protein markers as a part of the novel concept of liquid biopsy? Some researchers agree, but some disagree [51, 52]. Nevertheless, protein LB biomarkers also face a turning point in the liquid biopsy era thanks to the advances in technologies. In spite of the fact that protein detection is far from an alternative concept to tissue biopsy, proteins also carry out information derived from disease-related cells and are able to support clinal implementation [52].

Traditional biomarkers like protein have been a target analyte since the first page of the biosensor history. To evaluate the risk of diseases at the early stage, the primary goal of biosensors is ultrasensitive and quantitative detection. The representative optical nanomaterial-based detections of protein LB markers are described in Table 3. The studies conducted multiplexed detection are provided in Table 4.

Xu et al. presented a novel immunoassay for the detection of AFP using a plasmon-induced silver photoreduction system [59]. The silver crystals were generated on the surface of AuNPs by only the visible light illumination without using reducing agents. Thanks to this enzyme-free amplification, the sensitivity of the sensor was largely enhanced in a simple manner. The LOD was 3.3 fg/mL, which is more than 3 orders of magnitude lower compared to the LOD of commercial ELISA (around 6 pg/mL).

Recently, more studies have adopted two or more nanomaterials to achieve synergetic effects. Wang et al. reported an aptasensor for the detection of CEA based on fluorescence resonance energy transfer (FRET) between UCNPs and GO [48]. When CEA was added, the structure of aptamer was changed, and UCNPs were separated from the GO, resulting in fluorescence recovery. The LOD was 7.9 pg/mL in aqueous solution and 10.7 pg/mL in serum, and it is almost 2 orders of magnitudes lower than the LOD of commercial ELISA mentioned above (around 0.2 ng/mL). Li et al. developed a hybrid SERS immunosubstrate consisting of Au nanoflowers and red phosphorus (RP) nanoplates [73]. The anisotropic growth of 3D NPs having sharp edges on the 2D RP substrate, which is advantageous in electron conductivity and visible-light-responded bandgaps, provides a sensitive and robust platform. The LOD of the sensor was 7.41 × 10^–5 U/mL and it is much lower than the cut-off value (37 U/mL) and the LOD of the commercial ELISA kits (around 0.3 U/mL). More importantly, the presented immunosubstrates were recyclable through the photocatalytic degradation of antigens and antibodies.

Medetalibeyoglu et al. utilized three different types of nanomaterials to develop sensitive and selective SERS-based sandwich immunoassays [62]. In this design, 2-dimensional transition metal dichalcogenides (TMDCs) and AuNPs are hybridized to prepare SERS probes. In the meantime, Ti₃C₂T_x MXenes and Fe₃O₄ NPs@Au NPs are incorporated to fabricate SERS substrates. It is one example of the rational design of nano material-based immunoassay because metal NPs cover the limited efficiency and low functionality of TMDCs. At the same time, 2-dimensional materials like TMDCs can provide chemical enhancement to AuNP-based systems. MXenes, another 2-dimensional nanomaterial, also provide similar advantages, and the incorporation of Fe₃O₄ NPs@Au NPs makes the resulting sheets into magnetic substrates for the enhancement of sensitivity and specificity via magnetic separation. As a result, the system showed 0.033 pg/mL of LOD and a wide dynamic range covering 6 orders of magnitudes.

Another important direction of protein LB biomarkers is multiplexed detection. The simultaneous measuring of two or more biomarkers from identical samples can clarify the complex relationship between biomarkers and disease, so eventually, it may provide the opportunity for early detection. The simultaneous detection of multiple biomarkers is demonstrated by two different combinations: (a) representative biomarkers but not limited to specific cancer; (b) clinically related biomarkers of s single cancer subtypes.

Lee et al. proposed an example of the former concept. Their nanoplasmonic biosensor based on AuNPs targeted AFP, CEA, and PSA, and those protein biomarkers are renowned indicators of liver, lung, and prostate cancers. The LOD of the sensor was 91 fM, 94 fM, and 10 fM for AFP, CEA, and PSA, respectively, from the serum samples. These are much lower levels than both cut-off values (picomolar level) and even the background of healthy individuals. In addition, a wide dynamic range, from the femtomolar level to the nanomolar level, traces the changes over the biological range; therefore, this kind of approach may contribute to the early screening of the potential disease. On the other hand, Wu et al. designed an example of the latter concept, particularly focusing on lung cancer. Their sandwich immunoassay using multicolor QDs and magnetic microbeads targeted three protein biomarkers (CYFRA 21-1, CEA, and NSE) for lung cancer. Thanks to the different colors of three types of QDs, which are designated three different biomarkers, the concentration of each or the ratio between them was evaluated. The LOD of the sensor was 38 pg/mL, 364 pg/mL, and 370 pg/mL for CYFRA 21-1, CEA, and NSE, respectively. This kind of approach may contribute to the accurate detection of lung cancer regardless of the concentration of a specific biomarker. In addition, the diagnosis based on multiple biomarkers provides valuable information for future treatment decision-making.

Peptides are short chains of amino acids that are linked via peptide bonds [119]. Like proteins, peptides also are amino acid-based building blocks in living organisms. The difference between them is size and structure, thus rendering distinct biological functions [120]. When the liquid biopsy was first introduced, it mainly focused on oncology because its concept was a counterpart of tissue biopsy. Later, its range has been expanded to other diseases that can find biomarkers from the biofluids. The most famous peptide biomarkers are amyloid-beta (1–40) and amyloid-beta (1–42), which have long been considered as biomarkers of Alzheimer’s disease (AD) [121]. Historically, much effort has been made to detect these peptides, as well as tau protein, from cerebrospinal fluid (CSF) and even plasma in advance of the diagnosis by the medical imaging system. Detecting AD biomarkers is helpful in early diagnosis, and early diagnosis is beneficial in disease management and treatment. In the past, this idea had never been described as a “liquid biopsy.” For now, more and more literature set the expanded range of liquid biopsy, including biomarkers of other diseases [122, 123]. Although the CSF, a special kind of biofluid that cannot be accessible without an invasive procedure, is not perfectly fit for the philosophy of liquid biopsy, amyloid-beta in plasma or other fluid is more matched to the concept of liquid biopsy. The difficulty in detecting amyloid-beta is a relatively high background level that is not differentiated between the patient group and the control group.

Another peptide biomarker can be found in the field of cardiovascular diseases (CVD). Like AD biomarkers, CVD biomarkers had not been considered the LB biomarkers in the past; but more recent articles have started to discuss CVD detection as a part of liquid biopsy [124]. The most widely used CVD markers are brain natriuretic peptide (BNP) and N-terminal proBNP (NT-proBNP). They are significant indicators in heart failure and cardiac dysfunction. These peptides are secreted from the walls of the heart chamber directly into the bloodstream. The clinical cut-off of BNP and NT-proBNP is 100 pg/mL and 300 pg/mL, respectively [30], but the background level usually increases in the older age groups [125]. Various commercial test kits with analyzers for detecting peptide biomarkers have recently been on the market. For example, Roche Elecys® (Roche diagnostics) is one of the widely used methods in clinics to test BNP and NT-proBNP. This electrochemiluminescence immunoassay (ECLIA)-based system displays a high degree of diagnostic accuracy. With an 18-min testing time, the system is capable of detecting NT-proBNP as low as 5 pg/mL. Roche Elecys® systems are also developed for amyloid beta peptide detection. It shows 90% of concordance with amyloid PET imaging [126].

The representative cases of peptide LB marker detection using optical NPs are described in Table 5. The NPs discussed in these studies are metallic NPs, silica-coated metallic NPs, and MWCNTs. Among peptide LB biomarkers, the detection of BNP and pro-BNP are relatively similar to that of protein LB biomarkers. Because there already is a confirmed reference level in specific biofluids, setting a guideline for detecting them is relatively clear. The performance of the developed methods needs to be sensitive and accurate around the cut-off levels. Most studies report a better LOD than the general cut-off level of BNP (100 pg/mL) and proBNP (300 pg/mL) and even find a way to reach a sub-picogram level for early detection.

On the other hand, another type of peptide LB biomarker, like beta-amyloid, is far more complicated. The absence of enough clinical evidence and somewhat contracted reports among the studies are problematic when setting a guideline for the detection of beta-amyloid. The concentration range of these peptides is broadly distributed with individual differences, and even background level keeps increasing along with normal aging. According to the previous studies, the lower limit of plasma concentration of Aβ_(1–40) and Aβ_(1–42) is 10^–11 and 10^–12 g/mL, respectively [34]. In the case of beta-amyloid, multiplexed detection of Aβ_(1–40) and Aβ_(1–42) is an essential requirement because the ratio between them is more prominent than each concentration [145]. Kim et al. [133] suggested a shape-code plasmonic biosensor for the detection of three kinds of AD biomarkers, Aβ_(1–40), Aβ_(1–42), and tau proteins. Each biomarker was coded using 50 nm AuNPs, AuNRs (aspect ratio = 1.6), and AuNRs (aspect ratio = 3.6), respectively. The LOD of the sensor for each biomarker was 34.9 fM, 26.0 fM, and 23.6 fM, respectively. These are much lower levels compared to the background concentration of these biomarkers. Therefore, the presented one-step multiple detection offers an opportunity for sensitive and accurate detection of AD biomarkers. In the meantime, Yang et al. presented SERS-based multiplexed detection of Aβ_(1–40) and Aβ_(1–42) using silver nanogap shells on Si NPs and magnetic beads [134]. In the format of sandwich immunoassay, LOD was 0.25 pg/mL and 0.33 pg/mL for Aβ_(1–40) and Aβ_(1–42), respectively. This performance based on the intense and stable SERS signals also indicates the detection of a very low amount of biomarkers from the complex matrix like serum. Further, Wang et al. fabricated Pt@Au plasmonic chiral triangular nanorings to detect both Aβ_(1–42) monomers and fibrils [131]. Based on the intense chiral response of triangular nanorings modified with L- and D-glutathione, the proposed methods took advantage of the SERS-chiral anisotropy effect. The LOD of the system was 0.045 × 10^–12 M and 4 × 10^–15 M for monomer and fibrils, respectively. This study provides the opportunity to investigate the process of amyloid peptide misfolding and aggregation.

Aβ oligomers (AβO) are one of the important themes in AD research. Fang et al. reported a detection method for AβO based on fluorescence ratio using ZIF-8-doped UCNPs-SiO₂@metal–organic framework/black hole quencher [138]. The authors utilized optical tweezer microscopic imaging. It is an interesting approach because optical trapping prevents interference with fluid viscosity. The microsphere embedding nanomaterials are advantageous in both marker enrichment and laser focusing. The LOD of the sensors was 28.4 pM, and quantitative detection was demonstrated between 100 pM and 10 μM. Yin et al. designed a 3-dimensional fluorophore-labeled DNA walker nanoprobe immobilized on the AuNPs [137]. These nanoprobes can detect AβO and provide real-time imaging in living cells and in vivo. When the AβO was present in the samples, the fluorophores were cleaved and released, thus enabling a signal amplification effect without enzyme. Under in vitro demonstration, LOD was 22.3 pM, and the dynamic range was confirmed in the concentration range of 0.1 to 1.0 nM.

Circulating tumor DNAs are tumor-derived fractions of cell-free DNAs (cfDNAs) [146]. Although the amount of cfDNAs fluctuated in healthy individuals, an elevated level of cfDNA in cancer patients was found in the early studies. Leon et al. reported that the plasma concentration of cfDNA in healthy control was in the range between 0 and 100 ng/mL (mean = 13 ± 3 ng/mL) [38]. On the other hand, the concentration of cfDNA in cancer patients was highly varied from 0 to 5000 ng/mL (mean = 180 ± 38 ng/mL). Interestingly, there was a huge disparity between the upper 50 percent and lower 50 percent, and this result indicates that the cfDNA level is usually high in cancer patients.

Currently, there are two approaches to detecting mutations in ctDNA [147]. The first one is a targeted detention using complementary oligonucleotides. Because this approach mainly focused on the known mutations in specific genes, the patients who do not have these mutations cannot be distinguished. Conversely, the second one is untargeted detection based on next-generation sequencing (NGS). This approach sequences millions of DNA fragments at the same time via the “sequencing by synthesis” method, so a large amount of information can be obtained, including unknown mutations. Therefore, it is a time-consuming procedure conducted by highly trained experts, and it also generates an extensive volume of data requiring elevated costs [148]. For these reasons, many efforts have been made to develop a sensitive detection system comparable to an NGS-based assay. For example, Nesvet et al. developed magnetic NP-based giant magnetoresistive sensors that detect 0.01% mutant allelic fraction in ctDNA. It achieved both high analytical sensitivity and rapid testing time [149].

Especially, ctDNAs recently gained more attention than CTCs due to the advances in sequencing technology and relatively simple preprocessing procedures. For example, Grail Inc. developed NGS-based ctDNA detection tests (“Galleri”) for multi-cancer early screening [150]. They have constructed the mutation library from large-scale discovery to distinguish the usual mutations and tumor-related mutations.

However, there are difficulties in ctDNA detection. First, the actual fraction of ctDNA is extremely low, like other LB biomarkers. There is a report that around 1% to 2% of overall cfDNA are accounted for ctDNA in cancer patients [151]. Second, its status is highly varied due to the short half-life and thus dependent on the sampling moment [152]. It implies the possibility of real-time monitoring of the tumor; but. It also is a technical huddle in developing sensing methods. Third, a large-volume sample is usually required to reach a satisfactory sensitivity [153]. These issues are getting even worse in the case of circulating free RNAs or circulating tumor RNAs, more rare and more unstable targets. In order to overcome this limitation, the detection of nucleic acid LB biomarkers requires both rapid and ultrasensitive sensing mechanisms.

In the biosensing field, ctDNA detection methods are basically based on the historical achievement in aptasensors [154, 155]. Therefore, the specific direction of the research has been tuned to find a disease-related sequence in cfDNAs. Key mutations like rat viral sarcoma (RAS), EGFR, PIK356, BRAF, and TP53 were targeted to estimate the actual fraction of ctDNA from total cfDNA [156‐158]. The representative studies using optical nanomaterials are described in Table 6. The lowest LOD is down to the attomolar range, and the widest linear range was 5 orders of magnitudes. Although these results cover the concentration range of cfDNA in plasma (Table 1), it is still hard to estimate the actual concentration of the mutated cfDNAs.

There are several criteria for the evaluation of ctDNA detection techniques. First of all, ultrasensitive detection and accurate quantification are essential, considering the minuscule amount of ctDNAs in the blood. Unlike traditional LB biomarkers, the sensitivity of the system is required to be down to a single-nucleotide level, distinguishing point mutation precisely and accurately. Eventually, simultaneous detection of multiple genetic mutations is a significant criterion for maximizing clinical feasibility.

Several studies reported multiplexed ctDNA detection technologies to conform to two kinds of point mutations simultaneously. Nguyen et al. present a strategy for the dual detection of ctDNAs via targeting two bio-signatures, E542K and E545K, tumor-specific genetic and epigenetic markers of ctDNA of PIK3CA gene [159]. The probe was designed using AuNPs functionalized with peptide nucleic acids (PNA). The capture and the enrichment of ctDNA induced the change of reflective index and can be detected as the peak change of LSPR. Moreover, the authors utilized the coupling plasmon mode to detect both epigenetics changes and enhanced the signal of specific genetic mutations. SERS-based ctDNA detection was also demonstrated by the advances in SERS immunoprobes and/or SERS immune-substrates. Lin et al. developed a SERS-active substrate for the detection of tumor-related DNAs. With a dual signal amplification method, using metal carbonyls (metal-COs) onto SiO₂@Au as interference-free SERS labels, the LOD of the system was 57.74 M, and the linear range is between 100 and 1000 nM. Bellassai et al. investigated ctDNA detection using SPR imaging systems [167]. The sensor interface, poly-L-lysine (PLL)-based dual functional layer, was designed to achieve two purposes: anti-fouling surface and immobilization of PNA probes. The sensor detects wild-type and Kirsten rat viral sarcoma (KRAS) p.G12D- and p.G13D-mutated genomic DNAs in plasma. The LOD of the sensor was 5 pg/μL level and it is equivalent to approximately 2.5 aM. It does not require preprocessing for DNA isolation and PCR amplification.

Cao et al. developed pump-free SERS microfluidic chips to detect both BRAF V600E-mutated and KRAS G12V-mutated ctDNAs [171]. The identification of BRAF V600E mutation, which is discovered in 3% of non-small cell lung cancer, is important in the decision of therapy. Likewise, the identification of KRAS mutation is related to poor survival rate. Therefore, simultaneous quantification of both mutations from ctDNA provides detailed information about the characteristics of the primary tumor. The authors especially combined SERS nanoprobe (Pd-Au nanorod@magnetic bead), catalytic hairpin assembly, and microfluidics. The high sensitivity of this study is derived from a dual-signal amplification strategy, CHA-based amplification, and magnetic beads-based aggregation. The LOD was 3.116 aM and 3.921 aM for BRAF V600E and KRAS G12V, respectively. Further, the authors confirm that attomolar level sensitivity and accurate quantification are present in mouse serum. Later, the authors proposed another micro fluid-based platform for evaluating expression levels of TP53 or PIK3CA-Q546K in ctDNAs [170]. TP53 is usually considered to be related to a worse prognosis and resistance to chemotherapy, and PIK3CA-Q546K plays an important role in the pathogenesis of NSCLC. The design of this study is basically similar to the previous one, but Au–Ag nano shuttles were utilized as SERS nanoprobes, and the Au–Ag nano bowl array was prepared as SERS substrates. The LOD was 2.26 aM and 2.34 aM for TP53 and PIK3CA-Q546K, respectively. Finally, the clinical feasibility was verified by comparing with qRT-PCR tests, using patients’ samples and healthy donors’ samples.

In the same context, extracellular RNAs in the biofluid can also be a potential biomarker. Almost all kinds of them are released through the death of the cells or the active release mechanism of the cells [173]. However, they are extremely unstable, and their half-life is estimated to be just a few seconds, so most RNA-related liquid biopsy studies tend to focus on the complexed form with the proteins or encapsulated form in the exosomes [174]. Among them, miRNAs are the most notable biomarkers in the RNA family. These nonprotein-cording RNAs, having a length of 19 to 25 nucleotides, are relatively stable compared to other nucleic acids [175]. In addition, the expression of these post-transcriptional regulators for gene expression is presumed to be dysregulated in various cancers. Because miRNA expression levels in blood have correlated with miRNA levels in tumor tissue, monitoring its level in blood can be a feasible approach to liquid biopsy [176]. Unfortunately, the evaluation of miRNA expression level has similar issues to other LB biomarkers. Current methods, including Northern blotting and RT-PCR, require complex and time-consuming procedures. There also is a risk of contamination. More importantly, high sensitivity is required to monitor the changes derived from diseases.

The representative studies using optical nanomaterials are described in Table 7. The lowest LOD is down to the femtomolar range, and the widest linear range was 10 orders of magnitude. There are several criteria for the evaluation of miRNA detection techniques. First, ultrasensitive detection and wide dynamic range are key criteria. Zhu et al. developed ECL biosensors for the detection of miR-182 [177]. The miRNAs were successfully separated and detected via enzymatical enhancement with the combination of the AuNP-decorated magnetic particles and QD-embedded mesoporous silica nanoparticles. The LOD of the sensor was 33 fM, and the linear range was in the range between 100 fM and 100 pM.

One of the approaches to overcome the limitations of the singular sensing mechanism is dual-mode sensing. Huang et al. developed SERS/Fluorescence biosensors consisting of well-arranged Au nanoarray substrates [184]. With the CHA-based amplification using fluorophore-labels hairpin DNAs, a stable, reliable, and reproducible signal was obtained from the miRNA assay. Based on the system integrating SERS and MEF effects, the authors detect HCC-related miR-224. The LOD of the system was 0.34 fM and 0.39 fM for SERS mode and fluorescence mode, respectively. The linear range was ranged from 1 fM to 10 nM via a triple enhancement system. In the validation using clinical samples obtained from the HCC patients, the level of miR-224 was largely reduced after hepatectomy.

Second, multiplexed detection is also essential in miRNA sensing technologies. Jiang et al. established the nanoparticle-based sandwich assay for the simultaneous detection of multiple miRNAs [183]. The authors separately encoded miR-21, miR-155, and miR-16 with AuNPs, PtNPs, and AgNPs, respectively, and collected via magnetic separation. The first two miRNAs are breast cancer-associated oncogenic miRNAs, whereas the last one is endogenous control. The results were analyzed by single-particle ICPMS, which can distinguish the signal differences derived from individual nanoparticles. The LOD of the sensor is 1.1 pM, 1.1 pM, and 1.2 pM for miR-21, miR-155, and miR-16, respectively, without requiring amplification steps.

Extracellular vehicles (EVs) are lipid-bounded particles that are involved in intercellular communications [186]. Because they mirror the mother cells and thus carry proteins and nucleic acids originating from the mother cells, there has been a hypothesis of their physiological and pathological roles. EVs are usually classified by their mechanism in biogenesis, concept, and characteristics (e.g., size). Among them, exosomes are endosome-originated nanosized vesicles that are secreted from the cells and circulate until reaching recipient cells [187]. They are considered signaling molecules involved in cell-to-cell communications. Unlike the other three members, exosomes are abundant in concentration, from 10⁷ to 10⁹ particles per milliliter of plasma [188]. Because the cellular origin defines their composition, the ultimate objective of CTC or ctDNA research is also achievable by the strategy of detecting exosomes [189]. In addition, exosomes have advantages over CTCs or ctDNAs. First of all, they are plentiful in amount and the only LB biomarker free from rarity issues. Second, they are covered by a double-layered membrane and thus are considered a sort of cargo containing a package of nucleic acids (DNA, mRNA, and miRNA) and proteins. For these reasons, there have been efforts to isolate exosomes with other rare markers (CTCs and ctDNAs) simultaneously from identical samples to gather more information and to enhance the feasibility of the test [10]. Unfortunately, little is known about their characteristics and mechanisms. In the aspect of engineering, exosomes are hard-to-collect targets due to their broad size range and various surface markers. Furthermore, there is a purity issue because almost all cells generate exosomes. Therefore, the enrichment methods to separate tumor-derived fractions from the normal cell-derived vesicles are essentially required. The concentration of exosomal protein and exosome itself is relatively high in patients’ plasma compared to the blood obtained from healthy plasma [190].

Exosomes are the strangest targets in liquid biopsy research. Unlike other LB biomarkers, extracellular vesicles have never been an interest of the biosensing field before the liquid biopsy era. Therefore, there were no precedent schemes like aptasensors and cytosensors. Until now, various enrichment and isolation methods have been introduced, including ultracentrifugation, polymer-based precipitation, immunoaffinity-based separation, and acoustic-based purification [10, 191‐193]. There also have been demonstrations using various optical detection methods. The representative studies are described in Table 8. Currently, several studies have reported approximately 10³ particles per mL, even down to around 10² exosomes per mL. However, the results of these previously reported studies are hard to analyze systemically due to the ambiguousness of exosomes. All procedures, from sample preparation to final identification, are not established yet. For example, most exosome separation methods cannot guarantee that the impurities are negligible, so there is a possibility that the particles having similar characteristics to exosomes can be counted. For these reasons, it is difficult to set a minimally required sensing performance for exosome detection.

The early studies focused on the accurate quantification with high sensitivity of cancer-derived exosomes. Xia et al. demonstrated the colorimetric exosome detection method using CD63-specific aptamer-capped SWCNTs [197]. Since SWCNTs have peroxide-like activity, they can catalyze H₂O₂-mediated oxidation of TMB. This reaction was reduced by the addition of exosomes, which are expressed CD63 on the surface; thus, the amount of TMB oxidation is reduced and can be confirmed by the naked eye. The LOD of the sensor was 5.2 × 10⁵ particles/μL with a linear range between 1.84 × 10⁶ and 2.21 × 10⁷ particles/μL. The authors also found that approximately 1.5-fold more exosomes were found in the patients’ samples. Thakur et al. reported the LSPR biosensing method based on Au nanoislands (AuNIs) [196]. In their sensor design, randomly distributed nanostructures like AuNIs provide a convenient way to fabricate mass-producible and low-cost substrates for biosensors. By using an LSPR interferometer, the authors distinguished exosomes from other background vesicles. The LoD of the sensor was 0.194 μg/mL, and the linear range was in the range from 0.194 to 100 µg/mL. In the meantime, Zong et al. presented a SERS-based sandwich immunoassay method using a combination of magnetic nanobead (MB@SiO₂) and silica-coated Au@Ag nanorod (Au@Ag NR@SiO₂). Because the resulting signal is dependent on the amount of immunocomplex, the amount of the exosome in the sample can be measured qualitatively and quantitatively. The LOD of the system was 1200 exosomes with the detection ability of up to 10⁵ exosomes.

Second, the evaluation of multiple surface markers, which might be shared from their parental cells, is one of the important approaches. In this context, SERS-based detection also offers new perspectives in exosome profiling considering the complex and ambiguous nature of exosomes. Wang et al. proposed SERS-based detection methods for the screening of multiple exosomes simultaneously. The magnetic beads with gold shells (160 nm) contribute as a SERS substrate, while gold nanoparticles (17 nm) are utilized as SERS nanoprobes. The sandwich assay was conducted using three different aptamers and Raman reporters. The authors demonstrated the system with the exosomes that derived from three different cancer types (SKBR3, T84, and LNCaP cancer cells for breast, colorectal, and prostate cancer), and the LOD of the system was 32, 73, and 203 particles per microliter, respectively.

Zhang et al. Developed a simultaneous detection method for exosomal proteins using AuNPs and UCNPs. In this core-satellite design of probes, AuNPs served as a core, and three different types of UCNPs (yttrium, europium, and terbium) were arranged as satellites through three types of different aptamers (CD63, HER2, and EpCAM). Because the UCNPs were released when the aptamer recognized the specific marker on the exosomes, the authors collected and analyzed the detached UCNPs using ICP-MS and profiled the marker expression level. Zhang et al. utilized bimetallic nanoparticles and graphene oxide to construct both SERS nanoprobe and SERS substrates [210]. In this design, GO on the SERS substrate contributes to the enhanced surface area and the improved functionality of the receptor (V-shaped double-stranded DNA). The exosomes recreated from MCF-7 cells were analyzed with LOD down to 1.5 × 10² particles/mL without any amplification strategy. Finally, the proposed system was validated using clinical samples and proved the ability to distinguish breast cancer patients, pancreatic cancer patients, and healthy individuals.

Circulating tumor cells (CTCs) are rare cells that have been shed from the primary tumor to the bloodstream. Its frequency is usually in the range between 0 and 10 cells per millimeter of blood obtained from cancer patients [212]. It is extremely low level compared to red blood cells (RBCs, ~ 1 × 10⁹ per milliliter) and white blood cells (WBCs, ~ 5 × 10⁶ per milliliter). Although the presence of CTCs was first documented more than 150 years ago, their clinical utility was not validated until the late 1990s [213]. In 2004, the first CTC isolation method, called the CellSearch® system, had cleared by the US Food and Drug Administration (FDA). Allard et al. conducted large-scale clinical tests using this FDA-cleared system based on magnetic bead separation. With 2183 blood samples from 964 metastatic cancer patients having eight types of cancer, they found 0 to 23,618 CTCs per 7.5 mL of the blood. In contrast, the healthy individuals and patients having non-malignant disease did not have more than 2 CTCs, except for one case among 344 cases [43]. These early reports stimulated the CTC research to isolate the cancer cells from the whole blood. However, the advances had slowed at a certain point and caused troubles in verifying the clinical utility. There are several reasons that make CTC research challenging. Not to mention that the emergence of these cells is a very rare event, there has been speculation that the marker expression of CTCs usually changes during the detachment process [214, 215]. To address these issues, numerous studies have suggested various isolation and enrichment methods, including immunoaffinity-based methods and size-based methods [214, 216‐218].

It is worth noting that CTCs are the most significant LB biomarkers because they are detached parts of the tumor, thus representing its origin. For these reasons, the detection of CTCs is somewhat different from the above-mentioned LB biomarkers. A comprehensive analysis should be accompanied by sensitive detection to detect down to a few cells. Because CTCs are whole packages containing proteins and nucleic acids, avoiding cell rupture and gently retrieving of viable CTCs for downstream analysis are also important. In addition, the heterogeneity of CTCs and a lack of their specific surface marker should be considered in the development of sensing technologies [219]. Notably, CTCs are the largest biomarker in liquid biopsy. The size of the cancer cells is usually above 10 μm in diameter, so arithmetically, it is approximately 10³ to 10⁴ times bigger than usual NPs. It means that multiple NPs can encode a single target cell, enabling signal accumulation [220]. Therefore, the distribution of multiple NPs on a cell may be equivalent to the mapping of cell surface marker expressions [2].

For the optical detection of CTCs, various detection methods using optical nanomaterials have been introduced. The representative studies are described in Table 9. Despite technical challenges, setting a guideline for CTC detection is relatively simple compared to ctDNAs and exosomes. The eventual performance needs to be reached for single-cell detection. Although the performance tends to vary by the setting of experimental conditions, several studies reported that the lowest LOD is close to a single-cell level in whole blood samples. Ruan et al. developed a SERS-based CTC detection system using nanoprobes consisting of triangular Ag nanoprisms and magnetic NPs. By a combination of FA-based isolation, magnetic enrichment, and SERS-based detection, the LOD of the system reached one CTC per mL. Afterward, the authors designed SERS-active magnetic NPs consisting of superparamagnetic iron oxide NPs with outer-arranged AuNPs. The LOD of CTC detection also reached 1 cell for mL. Wu et al. proposed a SERS-based method for the detection of CTCs in the blood. The authors prepare the Raman probes by encoding 4-mercaptobenzoic acid (4-MBA), followed by the functionalization with reductive bovine serum albumin (BSA) and folic acid (FA). They reported the LOD of 5 cells/mL with a linear range of 5 to 500 cells/mL.

The evaluation of surface markers is a key to addressing heterogeneity issues of CTCs. Lin et al. designed 3-dimensional amorphous nitrogen-doped carbon nanocages as an SERS nanoprobe to image triple-negative breast cancer (TNBC) cells [220]. The identification of TNBC cells is important because they do not express the representative surface expression of breast cancer cells, such as estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER-2). It means that TNBC subtypes are not effectively treated by HER-2-targeted therapy. The authors quantitatively identified two types of TNBC cells, HCC 1806 and MDA-MB-231.

The attractiveness of nanomaterials originates from the unique features that are different from their bulk corresponding materials. In the nanometric dimension, the property of the materials drastically changes in every aspect, such as optical, electrical, and mechanical characteristics [13]. In addition, nanomaterials have significant surface-to-area ratios, thus enhancing the efficiency of the reaction. Various nanomaterials discussed above can be utilized as a sensitive optical nanoprobe solely or cooperatively via versatile detection strategies. The factors that can be considered in designing optical probes are size, shape, morphology, arrangement, structure, composition, physical/chemical characteristics, and compatibility with incorporated materials. As mentioned above, more studies have exploited two or more kinds of nanomaterials to induce a synergetic effect. In addition, the selection of nanoparticles is closely linked to the selection of appropriate detection systems.

Each nanomaterial has its own advantages and disadvantages, and these differences need to be considered during the selection of nanomaterials and/or the design of optical probes. Inorganic nanomaterials usually play a key role in optical probes due to their excellent optical properties. Metallic NPs like Au and Ag display excellent optical properties strongly dominated by the collective oscillation of free electrons on the metal surface; thus, their localized surface plasmon resonance (LSPR) can be tuned by size, shape, morphology, and interparticle distance [244]. However, individual metallic NPs are often not satisfactory for detecting ultra-low amounts of target analytes. Meanwhile, semiconductor nanomaterials like QDs show discrete electronic states through the “quantum confinement effect”, having broad absorption and narrow and symmetric emission bands with high photostability [245]. However, their applications are often restricted by probe size, toxicity, blinking effect, and difficulties in bioconjugation. Similar to QDs, carbon-based fluorescence nanodots can display size-tunable photoluminescence behavior through surface passivation with organic molecules. These biocompatible and chemically inert carbon NPs have recently gained much attention as a sensing probe for optical detection owing to favorable properties, such as low toxicity and non-blinking effect [246]. Unlike other carbon-based materials, however, these carbon NPs still have not established systematic and scalable protocols. UCNPs are an emerging class of optical nanomaterials based on upconversion luminescence, an interesting phenomenon defined as the conversion of long-wavelength radiation to short-wavelength radiation (e.g., from infrared or near-infrared (NIR) to the visible range) [247, 248]. So, they are useful in biosensing and bioimaging due to the low autofluorescence background, low photobleaching, as well as narrow emission bandwidth. However, several challenges remain due to their complicated synthesis methods, causing trade-offs between toxicity and efficiency [249].

Sometimes, these nanomaterials themselves are not suitable for liquid biopsy applications. First of all, most nanomaterials are hard to control in complex media because of their inherently sensitive nature to ionic substances. It implies the difficulties in exploiting them under physiological conditions or real clinical samples, decolorizing the distinct merit of optical detection. Furthermore, some nanomaterials are not fully evaluated in terms of their biocompatibility. For these reasons, the hybridization of one or more nanomaterials by constructing nanoarchitectures can also be a rational solution. Nanoarchitectures can be designed in the aspect of composition (silica and polymeric materials), structure (core–shell and yolk-shell), or function (magnetic and catalytic reaction). Magnetic NPs or microbeads not only have a long history in biomedical applications as preferred solid support but are also one of the first successful strategies in liquid biopsy because of signal enrichment and effortless purification. Silica NPs have been considered an ideal matrix for phosphors or small metal crystals because they are transparent to light [250]. In addition, silicon NPs have several advantages, including chemical/physical stabilities and hydrophilic surfaces, allowing easy modification [251]. The silica encapsulation offers increased detectability compared to the individual phosphors or small metal crystals and expands the usability with size controllability and multifunctional properties based on the water solubility [252‐254]. Polymeric NPs are also an ideal candidate due to their soft and biocompatible nature. More importantly, polymeric material has the ability to react against the environment (e.g., stimuli-responsive behavior); so, their adoption can provide additional dynamic functionality to the probes [255].

The combination of the optical nanomaterials to detect LB biomarkers is a fascinating point in this subject. The LB biomarkers are very different from each other in characteristics and exist in different metric regimes from nanoscale to microscale. In Fig. 2, we displayed their degree of relativity with the illustration. The largest LB biomarkers are larger than 10 µm (e.g., CTCs), whereas the smallest LB biomarkers are smaller than 10 nm (e.g., microRNA). It offers interesting points in the selection of proper optical NPs and their size. Most protein and LB biomarkers are comparable to QDs and slightly smaller than metallic NPs (less than one order). Exosomes (30 to 150 nm) have a wide size distribution range comparable to various nanomaterials. (Fig. 2b). On the other hand, CTCs are much bigger than the largest nanoparticles (~ 150 nm), with almost two orders of differences (Fig. 2c). Therefore, smaller optical NPs detect CTCs by one-to-many correspondence. It is sort of analogous to the relation between protein and cell via encoding the surface of the cells with NPs; thus, it can provide information regarding the heterogeneity of the CTCs [256].

Liquid biopsy is now entering the plateau phase. Despite much anticipation, there still are many hurdles to overcome until it is established as a part of standard procedure for clinical decisions. The challenges stem from both the conceptual difficulties in liquid biopsy and the exceptional characteristics of each LB marker. Although we cover both traditional biomarkers and newly emerging LB biomarkers together under the broad concept of liquid biopsy, there will be slight differences in the future approaches. In the case of the traditional biomarkers (e.g., protein and peptide), enhanced sensitivity and accurate quantification are the primary objectives. On the other hand, newborn LB biomarkers (i.e., CTC, exosome, ctDNA) inherently possess more complicated problems, such as rarity, heterogeneity, and short half-life, compared to amino acid-based short chains or macromolecules. Aside from the sensing performance, an effective enrichment process should also be considered for these special classes of LB biomarkers. For example, the adoption of magnetic nanobeads or microbeads is a representative example. Since the first FDA-cleared CTC detection assay was introduced in 2007, magnetic separation occupies a large portion of liquid biopsy technology, and it is very suitable for optical detection using various NPs. Numerous studies and product prototypes utilize magnetic beads as a sort of substrate of the immunoreaction to be designated by optical NPs, thus achieving highly sensitive detection from the enriched samples.

In terms of clinical utility, traditional biomarkers have two different sides of points. They have been utilized in the diagnosis for a long time, whereas their already-proven limitation is obvious. The enhancement of analytical sensitivity and specificity might be helpful in the early screening of the disease. However, these biomarkers cannot provide in-depth information for treatment as well as prognosis. On the other hand, newborn LB biomarkers have potentials that have never been unveiled, even though their feasibility has not been fully proven due to the investigation with small cohorts and/or results in disagreement. As we mentioned above, the major reason for these conflicted results is a lack of a standard sample preparation method [255]. Currently, there is no consentaneous sampling, handling, or storage method in liquid biopsy. Likewise, one critical problem in biosensing development is a lack of reproducibility [257]. Although the studies discussed in this review report excellent sensing performance with additional functionalities to solve the current limitations in liquid biopsy, there will be a gap between lab-scale testing and practical applications. In conclusion, the actual clinical utility of LB biomarkers could be evaluated when accurate and reproducible methods are established with an efficient enrichment.

In the meantime, combined analysis would be a reasonable approach to overcoming the current obstacles in liquid biopsy. Two or more LB biomarkers provide complementary information about the disease. In addition, the drawback of each LB biomarker can sometimes be covered by other LB biomarkers. Traditional biomarkers have always been a starting point to validate the clinical meaning of ctDNA. Cohen et al. conducted a test to compare KRAS mutation in both ctDNA and protein LB biomarkers (AFP, CA15-3, CEA, CA-125, etc.) from 221 pancreatic cancer patients and 182 control patients [258]. Rossi et al. conducted a combinational analysis of CTCs and cfDNAs of metastatic breast cancer patients using CellSearch® systems and Guardant360, respectively [259]. Similarly, Ye et al. analyzed the samples obtained from metastatic breast cancer patients using CellSearch® systems and RT-PCR, respectively [260]. The role of exosomes as the only abundant LB biomarkers has also been investigated. Kim et al. designed a hydrogel-based immunoassay to isolate both CTCs and exosomes from colorectal cancer patients’ blood samples [10]. Although there was no clear evidence of the correlation between CTCs and exosomes in terms of concentration, the degradable hydrogel-based effortless collection of two different LB biomarkers from the identical sample provides an opportunity to contribute to further analysis.

New diagnostic technologies should satisfy diagnostic accuracy requirements for utilization in hospital routines. There are many criteria for adopting new diagnostic technologies in clinical practice. Guatt et al. categorized these requirements, including technological capability, range of possible use, diagnostic accuracy, impact of healthcare providers, therapeutic impact, and patient outcome [261]. In addition, they need to be technologically and psychologically accepted by physicists, biochemists, physiologists, and other healthcare providers.