Exosomes as a new frontier of cancer liquid biopsy

As the gold standard for exosome separation, ultracentrifugation is the most commonly used method [43], including differential ultracentrifugation and gradient density ultracentrifugation. Conventional differential ultracentrifugation was first proposed by Johnstone et al. to isolate exosomes from culture medium of reticulocytes [44]. Typically, low-speed centrifugation (300 g) is first employed to remove cellular debris, while 20,000 g speed of centrifugation is utilized to eliminate other large vesicles. A high force (100,000 g) is finally utilized to sediment exosomes. However, this method needs costly instrument and has contamination with aggregated proteins. More centrifugation cycles may obtain a purer outcome but lead to lower recovery. Gradient density ultracentrifugation is a better alternative to obtain exosomes with higher purity [45]. During centrifugation, different sizes of particles from two or more solutions are separated into different layers, whose density increases from top to bottom. Based on this, the method has been applied to separate exosomes, which have been found to float with densities of 1.15 to 1.19 g/mL [45, 46]. Although with higher purity than ultracentrifugation, limitations such as time-consuming process and the requirement of large biofluid volume have largely restrained its use in clinical application.

The feature of fixed-range diameters allows for the possibility to separate exosomes by size-based methods. Filtration is one of the size-based approaches for exosome separation by using membrane filters with specific pore sizes, which has the advantages of simple operation and effective purification but disadvantage of low yield [47]. Ultrafiltration is usually applied to concentrate exosomes from large amounts of original materials such as cell culture medium [48]. Currently, the combination of ultrafiltration with ultracentrifugation has been widely employed, in which filtration is used to remove cells and large vesicles while the purification of exosomes is achieved by ultracentrifugation [49]. Moreover, size-exclusion chromatography (SEC) could separate biomolecular components according to the size of sample and pore size of gel. During the separation, large molecules are eluted early, while small molecules or particles directly diffuse into the pores [50]. Anita N Böing et al. developed an SEC-based protocol by cross-linked sepharose CL-2B column, which could efficiently isolate exosomes with a diameter larger than 70 nm from platelet-free supernatant [50]. Guo et al. showed that CL-6B column had better performance than CL-2B column in particle yields and purity of exosomes [51]. Compared with centrifugation and filtration methods, SEC has the advantages of gentle processing and nondestructive outcomes [52]. Moreover, the combination of SEC with ultracentrifugation may have an improved recovery and purity [53].

As for precipitation techniques, highly hydrophilic polymer is used to competitively bind to water molecules around the exosomal membrane, thereby reducing the solubility and finally achieving low input volume of exosome separation. To date, polyethylene glycol (PEG) is the most commonly used polymer for exosome separation [54]. An exosome purification method called ExtraPEG was proposed by Rider et al., which could rapidly enrich exosomes and obtain sufficient contents harvested from vesicles for downstream analysis [55]. Recently, many commercial kits that rely on precipitation techniques have been developed for exosome isolation and enrichment, such as ExoQuick™ and Total Exosome Isolation. Study from Ding et al. indicated that ExoQuick™ could generate a relatively high yield of exosomes compared with other kits [56]. However, this method has been often criticized for the high cost and contamination of coprecipitated protein aggregates.

Owing to the priority of simplicity and specificity, immunoaffinity isolation strategy has been used for exosome enrichment in many studies (Fig. 3A). For example, anti-CD81 functionalized microfluidic chip was fabricated by Zhang et al. to isolate exosomes from plasma samples [70]. Yang et al. reported an integrated microfluidic device for exosome separation through forming a sandwich structure of AuNC-exosome-AuR complexes [58] (Fig. 3A_a). This device achieved a yield of 5 × 10⁹ particles from 5 mL urine sample in 30 min. Nevertheless, the dissociation of captured exosomes remains a big challenge, which arises from the strong affinity between antigen and antibody. To address this challenge, Kang et al. developed an exosome-specific dual-patterned immunofiltration (ExoDIF) device [59]. The biotinylated anti-CD63 antibody was immobilized on the surface of inner channel that sequentially pretreated with 3,3′-dithiobissulfosuccinimidylpropionate (DTSSP), biotinylated BSA, and avidin. As a result, nearly 87.1% of exosomes were captured from high dilution of cell culture media. Notably, the captured exosomes could be dissociated by DTT (dithiothreitol) through breaking the embedded disulfide bond of DTSSP (Fig. 3A_b).

Recently, EVs on demand chip (EVOD) was proposed by Kang et al., in which the capture of cancer-related exosome subpopulations was achieved by the reaction between tetrazine-conjugated anti-EpCAM/anti-EGFR antibody (TzAb) and TCO (trans-cyclooctene) functionalized microfluidic surface [82]. This chip was able to selectively isolate 76% more EGFR⁺ exosomes from cancer patients than that from healthy donors, which exhibited great potential for the early detection of NSCLC. Sun et al. designed an interesting covalent chemistry-mediated EV click chip to recognize, enrich, and recover hepatocellular carcinoma (HCC)-specific exosomes from plasma samples by multi-marker cocktail (anti-EpCAM, anti-ASGPR1 (asialoglycoprotein receptor 1), anti-CD147) [61] (Fig. 3Ac). The proposed exosome purification system achieved more than 81% of recovery yield and more than 85% of purity, providing a novel liquid biopsy tool to detect hepatocellular carcinoma. Despite great significance, these methods are high-cost and marker-dependent. In addition, more attention should be paid to the non-destructive release of captured exosomes.

Magnetic bead-based immunoaffinity enrichment has attracted much attention in recent years due to the advantages of convenience and high efficiency (Fig. 3A). Generally, exosomes are captured by antibody-modified magnetic beads, which are then separated by magnetic force. For example, Fang et al. conducted CD63 antibody-conjugated magnetic nanoparticles to isolate exosomes [71]. Moreover, novel immuno-affinitive superparamagnetic nanoparticles (IS-NPs) were proved to possess high efficiency, which combined anti-CD63 antibodies with superparamagnetic nanoparticles through the interactions between β-cyclodextrin (β-CD) and 4-aminoazobenzene (AAB) [60] (Fig. 3A_d). α-CD, a competitive agent extracting AAB from the β-CD-AAB inclusion compound, was adapted for the elution of exosomes. As a result, the capture and release efficiency of exosomes from artificial model samples was as high as 80% and 86.5%, respectively. IS-NPs method exhibited higher yield, increased purity and well-retained structural and functional integrity of exosomes than conventional separation methods.

The isolation of exosome subsets, especially cancer-derived exosomes, could be achieved by coupling magnetic beads with antibodies targeting tumor-specific biomarkers. Luo et al. performed a rapid separation and capture of exosomes by immunoaffinity magnetic beads and DNA origami-based aptamer. The quantitative detection of exosomes could be achieved through the combination with DNA fluorescence probe [83]. Li et al. developed a homogenous magneto-fluorescent exosome (hMFEX) nanosensor to separate GPC-1 positive exosomes in 80 μL of plasma from breast cancer patients [84]. He et al. conducted a microfluidic platform to capture tumor-derived exosomes by mixing samples with anti-EpCAM or anti-CA125 antibodies-labeled magnetic beads [36]. In general, the immunomagnetic separation methods hold the potential to facilitate rapid separation and clinical implication of circulating exosomes in desired areas.

Size-based microfluidic chip was used to separate exosomes from large cell debris or other membranous vesicles [76, 85] (Fig. 3B). Liu et al. designed a size-based exosome total isolation chip (ExoTIC) in which exosomes ranging in 30-200 nm were enriched and purified by multiple nanoporous membranes [62] (Fig. 3B_a). Compared with ultracentrifugation and commercial PEG precipitation kits, ExoTIC obtained much higher yields of exosomes from small volumes of human plasma. However, the blockage of membrane pores greatly limited the continuous separation of exosomes. To overcome this problem, Chen et al. introduced an ultrafast-isolation system, EXODUS, which integrated double coupled harmonic oscillations into a dual membrane filter configuration. By periodic negative pressure and air pressure switching, periodic negative pressure oscillations were generated on the nanoporous anodic aluminum oxide membrane, allowing small particles (i.e., proteins and nucleic acids) and fluids to pass through, while larger exosomes remained in the central chamber. Moreover, two pairs of oscillators enabled the resuspension of particles into the liquid via transverse waves and acoustofluidic streaming, which effectively limited fouling and particle aggregation [86]. In addition, tangential flow filtration (TFF) is a technology that effectively reduces the potential of pore clogging due to the perpendicular state of flow direction and filtration direction. Sunkara et al. developed a microfluidic tangential flow filtration device, Exodisc, to separate exosomes from human plasma and urine, which showed better exosome yield compared to conventional methods [63] (Fig. 3B_b). Deterministic lateral displacement (DLD) has been used for exosome separation since particles with different sizes perform distinct trajectories in a platform with certain angle-displayed micropillars. Wunsch et al. proposed a nanoscale lateral displacement (nano-DLD) array, in which larger vesicles laterally displaced across the array and were collected at a side channel while smaller vesicles flew out of the array in a zigzag mode, finally achieving the collection of urine-derived exosomes [75]. However, DLD was limited by low throughout and the requirement of high voltage and the density and stiffness of vesicles may interfere with DLD-based exosome isolation.

Lipid molecules on the surface of vesicles enable affinition-mediated exosome capture (Fig. 3C). Wan et al. reported a lipid nanoprobe for the rapid separation of exosomes from plasma [80]. The lipid bilayer of exosome was labeled with biotin-tagged 1,2-distearoyl-sn-glycero-3-phosphethanolamine-poly (ethylene glycol) (DSPE-PEG) probes in which DSPE was inserted into the exosome membrane by hydrophobic effect and PEG provided solubility in the aqueous phase. NeutrAvidin-coated magnetic sub-micrometer particles were used for the collection of exosomes through avidin-biotin affinity. As a result, it took only 15 min to isolate exosomes, which greatly shortened the isolation procedure [87]. Notably, cholesterol-PEG₁₀₀₀ of ~ 6.4 nm was considered to minimize steric hindrance in surface immobilization of lipid nanoprobes and thus achieved a much higher capture efficiency. Additionally, molecules absorbed to lipids on exosome membrane have also been applied to capture exosomes [88]. T-cell membrane protein 4 (Tim-4), which has high affinity to phosphatidylserine (PS), has been verified to facilitate simple exosome separation from serum samples, making it easier for rapid downstream analysis [65, 89] (Fig. 3C_a). In addition, the affinity interaction of TiO₂ shell and phosphate groups of exosomes is a novel strategy for the enrichment of phosphorylated peptide. Pang et al. used Fe₃O₄@TiO₂ nanoparticles to enrich and separate exosomes within 5 min with a capture efficiency of 96.5% [90]. The combination of Fe₃O₄@TiO₂ nanoparticles and surface-enhanced Raman scattering (SERS) tags could enrich exosomes and achieve in-situ qualification of target miRNAs simultaneously [66] (Fig. 3C_b).

Acoustic-based microfluidics is a simple and efficient method for exosome separation. Typically, ultrasonic waves are applied to samples and the particles undergo different forces and are separated depending on their physical properties such as size and density [81] (Fig. 3D). Anson et al. utilized the ultrasonic waves scattering between micrometer-sized seeding particles and nanoparticles in a resonant cavity to enrich exosomes. The integrated acoustic device enabled fast operation, non-contact and continuous separation of exosomes from urine and plasma samples [91]. Wu et al. developed an acoustofluidic platform for direct, label-free, and contact-free enrichment of exosomes from whole blood. The device integrated two separation modules, one of which was able to remove particles larger than 1 μm in diameter and the other one could isolate exosomes from larger microvesicles and other particles [92]. Recently, Gu et al. reported an acoustically driven spinning droplets device, in which slanted interdigitated transducers (IDTs) were used to allow nanoparticles to move according to the sound waves of varying frequencies. Particles of fixed size could be specifically concentrated by placing two droplets of different sizes next to each other. As a result, the method could isolate exosomes from 5 μL samples in less than 1 min [93]. In addition, Tayebi et al. combined acoustic radiation force (ARF) with dielectrophoretic (DEP), in which high frequency (> 10 MHz) interdigital transducer was placed in the flow path to simultaneously generate ARF and dielectrophoretic force field. Particles in the medium presented lateral translation by the competition between fluid drag forces, ARF, and DEP force fields, resulting in an active separation of extracellular vesicles [67] (Fig. 3D_a).

Thermophoresis refers to a phenomenon that particles migrate from space with high- temperature to low-temperature areas via a temperature gradient induced by localized laser heating. To address the challenge of time-consuming isolation and purification procedures, Liu et al. developed a sensitive thermophoretic method to enrich tumor-derived exosomes. By using aptamers that targeted tumor-specific markers, thermophoresis could achieve rapid isolation and enrichment of exosomes from other components without exosome pre-isolation (Fig. 3D). The accumulation of exosomes produced an amplified fluorescence signal of aptamers, which enabled profiling of surface biomarkers of exosomes as well as detection of miRNAs [68] (Fig. 3D_b). Tian et al. performed a similar assay to analyze exosomes in 1 μL plasma, which offered a low-cost, sensitive method for liquid biopsy in metastatic breast cancer [69] (Fig. 3D_c).

Transmission electron microscopy (TEM) is considered as the common method to identify and characterize a single exosome with typically cup-shaped structure (Fig. 4A) [62]. For scanning electron microscopy (SEM), images are presented by collecting the electrons ejected from the samples (Fig. 4C) [96]. Notably, the morphology of exosomes may be affected by dehydration during sample handling procedures. On the contrary, cryo-electron microscopy (cryo-EM) is a better choice because it avoids sample fixation and dehydration. Exosomes are analyzed at a very low temperature and exhibit a round structure that is different from TEM images [97] (Fig. 4B). In addition, atomic force microscopy (AFM) could provide information on both surface morphology and material properties (stiffness, adhesion) by amplitude modulation and phase modulation [96] (Fig. 4D). Moreover, specific exosome membrane markers functionalized AFM tips permit the identification and detection of proteins in single exosome [96, 98].

Dynamic light scattering (DLS) is a technique for measuring the distribution of exosome size and zeta potential. The changes of scattered light interfere and intensity can be identified by a sensor, allowing for estimating the size distribution of particles [100‐102] (Fig. 4E). However, the contamination of protein aggregates and large vesicles makes it challenging for DLS to distinguish them from exosomes [103]. Nanoparticle tracking analysis (NTA) is a widely used method for determining the concentration and size distribution of particles (Fig. 4F). Particles are illuminated by a beam of light and the scattering light signals are collected by optical microscope [100, 104]. Tunable resistive pulse sensing (TRPS) detects the electrical signals generated by changes in ion conductivity when exosomes pass through pores filled with conductive media [105] (Fig. 4G).

Both membrane and cytoplasmic proteins could be detected in exosomes [4]. Some membrane proteins on exosomes are involved in cancer development and progression and thus have been used as targets for exosome isolation and purification [28, 106]. Western blot and enzyme-linked immunosorbent assay (ELISA) are two regular approaches widely used to detect exosomal proteins [7, 107]. However, these methods are faced with complex procedure and low sensitivity problems [108].

Up to now, multiply novel techniques have been developed to detect exosomal proteins (Table 2). Colorimetric detection is a method to determine the content of a component by comparing or measuring the color intensity of chromogenic substances (Fig. 5A). Similar to other nanomaterials like Fe₃O₄ NPs and AuNPs, DNA has the ability of significantly increasing the peroxidase activity of single-walled carbon nanotubes (s-SWCNTs). Xia et al. developed a visible and simple method for the detection of exosomes [109]. Briefly, CD63 aptamer improved the minic peroxidase activity of s-SWCNTs and effectively catalyzed oxidation of 3,3′,5,5′-tetramethylbenzidine (TMB), resulting in a colorless solution turning blue. On the contrary, the addition of exosomes induced the release of aptamer from the surface of nanotubes and the color of solution to turn from dark to light, which could be observed by naked eye or UV-visible spectrometry with a detection limit (LOD) of 5.2 × 10⁵ particles/μL. As previously discussed, ZnO-chip was designed to effectively isolate exosomes. After incubation with primary antibody mixture (anti-CD9/CD63 antibody), exosomes were recognized by HRP (horseradish peroxidase)-labeled secondary antibody (Fig. 5A_a). Finally, a minimal detectable concentration of 2.2 × 10⁴ particles/μL was obtained by UV-visible spectrometry or microplate analyzer [110]. Liang et al. designed a microfluidic system integrated with double membrane filter and ELISA to detect the content of exosomes in urine samples of bladder cancer patients [111]. In addition to common exosomal proteins, colorimetry could be applied to detect cancer-specific proteins as well. For instance, a PSA (prostate-specific antigen) aptamer-based sensor was used for the visual detection of prostate cancer-specific exosomes in 500 μL human plasma [72]. Moreover, Di et al. reported a nanozyme-assisted immunosorbent assay (NAISA), which enabled sensitive and rapid multiplex profiling of exosomal proteins [112] (Fig. 5A_b). The surface proteins of exosomes could be specifically captured by antibodies immobilized on the surface of a microplate and catalyzed a colorimetric reaction. Signal intensity obtained from microplate reader is proportional to the number of target proteins. As a result, NAISA allowed the rapid profiling of multiple exosomal proteins such as CD63, CEA (carcinoembryonic antigen), GPC-3 (Glypican-3), PD-L1 (programmed death-ligand 1), and HER2 (human epidermal growth factor receptor 2) from clinical samples.

Fluorescence spectrophotometry is a method for substance identification and content determination according to the positivity and intensity of fluorescence spectral line (Fig. 5B). He et al. conducted a microfluidic chip to detect surface and intravesicular biomarkers from 30 μL plasma samples. This chip obtained a markedly improved detection sensitivity of IGF-1R (type 1 insulin growth factor receptor), 0.281 pg/mL of IGF-1R and 0.383 pg/mL of p-IGF-1R, which was 100-fold higher than that achieved by ELISA [36] (Fig. 5B_a). Liu et al. developed an immunosorbent assay, in which immunomagnetic beads were utilized to capture exosomes, followed by the conjugation of an enzymatic reporter, which could produce a fluorescent signal for quantitation of GPC-1⁺ exosomes in the droplet microfluidic system [124]. Wei et al. proposed similar single molecule array (SiMoa) platform, by which universal exosomes and tumor-derived exosomes could be ultrasensitively detected with an LOD of 34 particles/μL and 25 particles/μL, respectively [125]. Yu et al. conducted a CD63 aptamer-based detection method. CD63 on exosomes could bind to aptamers modified on the surface of magnetic beads, resulting in the shedding of a Cy3-labeled short sequence into the supernatant. The fluorescence intensity in the supernatant was used to quantify exosomes in complex clinical samples [126].

Graphene oxide has the ability of quenching fluorescence via fluorescence resonance energy transfer (FRET) when conjugated with fluorescent dyes. For instance, the fluorescence of FAM-labeled aptamers was quenched when absorbed onto graphene oxide membranes, while target exosomes competitively bound to aptamers and re-exhibited fluorescent signals with an LOD of 1.6 × 10⁵ particles/mL [113] (Fig. 5B_b). Li et al. developed a facile fluorescent aptasensor based on aggregation-induced emission luminogens (AIEgens). Graphene oxide absorbed (tetrafrylene-containing tertiary amine) TPE-TAS/aptamer complex allowed fluorescence quenching in the absence of exosomes. When target exosomes were introduced, the aptamer preferentially bound to its target, resulting in the separation of the TPE-TAS/aptamer complex from the graphene oxide surface, followed by a “turn-on” fluorescence signal. The linear range of tumor-derived exosomes was around 4.0 × 10⁵-1.8 × 10⁷ particles/μL under optimized conditions [99].

Electrochemical detection is used to detect the analytes by measuring the electrochemical potential or current of the sample, which has the advantages of high sensitivity and wide measurement range (Fig. 5C). In recent years, electrochemical biosensors have been developed by researchers since the altered electrochemical signals could quantify exosomes when recognition elements such as antibody and aptamer specifically bind to exosomes [127, 128]. Cao et al. proposed an electrochemical biosensor to accurately detect PD-L1⁺ exosomes [114] (Fig. 5C_a). Exosomes were firstly captured by anti-CD63 functionalized magnetic beads and then bound with anti-PD-L1-linked DNA strand, introducing a hyperbranched rolling circle amplification (HRCA). The HRCA could decrease the environmental pH, leading to the decomposition of PVP@HRP@ZIF-8 and release of HRP, which resulted in amplified electrochemical responses and thus achieved the identification and detection of cancer-derived PD-L1⁺ exosomes. This biosensor displayed a wide dynamic range for PD-L1⁺ exosomes (1 × 10³ to 1 × 10¹⁰ particles/mL) and the detection limit was 334 particles/mL. Jeongmin et al. reported a HiMEX approach which integrated magnetic exosome separation and electrochemical detection of exosome-bound proteins after enzymatic amplification. The combined detection of tumor biomarkers (EGFR, EpCAM, CD24 and GPA33) in exosomes from 20 μL plasma samples were helpful for the diagnosis and monitoring of colorectal cancer [129]. Electrochemical methods, especially aptamer-based biosensors have shown great potential in the detection and profiling of exosomal proteins [65]. Kashefi-Kheyrabadi et al. introduced a detachable microfluidic device implemented with an aptamer-based electrochemical biosensing method (DeMEA). In this system, aptamer targeting EpCAM was immobilized on the electrode surface that was pre-electroplated with gold nanostructures and microfluidic vortexes could increase the collision between exosomes and sensing surface (Fig. 5C_b). Consequently, DeMEA was able to quantify exosomes from plasma samples of breast cancer patients at different stages, which provides a highly sensitive and early detection of cancer-specific exosomes [115].

Surface plasmon resonance (SPR) is a physically optical phenomenon caused by total reflection of light at the metal film/liquid level interface to analyze molecular interactions (Fig. 5D) [130, 131]. Im et al. developed a nano-plasmonic exosome (nPLEX) assay based on transmission SPR through periodic nanohole arrays (Fig. 5D_a). With functionalized antibodies in each array, the nPLEX sensor displayed spectral shifts or intensity changes that were proportional to the levels of target exosomal proteins [116]. In another study, the markedly improved scattering wavelength shift and scattering intensity were observed on AuNC-Exosome-AuR due to the plasmon effect [58]. The differential expression of LRG1 (leucine rich alpha-2-glycoprotein 1) in urinary exosomes between lung cancer patients and healthy individuals was evaluated by anti-LRG1 antibody-conjugated AuR probes. Another surface plasmon resonance imaging (SPRi)-based biosensing assay was developed by Fan et al. The bioaffinity interactions between antibodies (anti-CD63/anti-EGFR/anti-EpCAM) modified-gold chip and different recognition sites permitted the multiplex characterization of NSCLC-derived exosomes. The LOD of this biosensor was estimated to be 10⁴ particles/μL. Despite multiple merits, the broad applications of nanoplasmonic biosensors are restrained by the difficult fabrication of nanostructures. To address this challenge, Liu et al. developed an intensity-modulated SPR biosensor free from nanostructure. In this sensor, the reflection intensity and reference intensity of lasers were recorded by two photodetectors and were used to quantify the expression levels of exosomal proteins, which exhibited a higher detection sensitivity than ELISA [132].

Surface enhanced Raman scattering (SERS) is capable of enhancing the Raman signal of small molecules attached to the rough metal surface through electromagnetic and chemical mechanisms [133] (Fig. 5D). A new assay was reported for the real-time detection and protein profiling of exosomes. Generally, gold-plated slides combined with 3D-printed antibody arrays were fabricated to capture exosomes, and QSY21-coated gold nanorods were used as the label agent to quantitatively detect target proteins. The levels of plasma-derived exosomes of breast cancer patients were quantitatively determined by using this assay targeting HER2 and EpCAM. The proposed 3D-printed array template enabled cheap, portable, and easily available establishment of detection platform, providing a new strategy for the development of novel cancer liquid biopsy [134]. Dong et al. showed that the analysis of protein phosphorylation status may provide new possibilities for cancer diagnostics [117] (Fig. 5D_b). A beehives-like Au-coated TiO₂ macroporous inverse opal structure was developed to capture and analyze exosomes without any labeling process. The intensity of 1087 cm^− 1 SERS peak referred to the P-O bond within the phosphoproteins of exosomes and the intensity of peak was at least two times from plasma of cancer patients than that from healthy donors. However, the above-mentioned two assays both require the pre-separation of exosomes, which greatly hinders the rapid analysis. Pang et al. presented a simple immunoassay to capture and analyze exosomal PD-L1 directly from serum samples [90]. Fe₃O₄@TiO₂ nanoparticles were designed to isolate exosomes and anti-PD-L1 antibody modified Au@Ag@MBA SERS tags were applied for exosomal PD-L1 labeling and SERS detection. This assay was confirmed to quantify exosomal PD-L1 in 4 μL serum sample within 40 min.

In CRISPR/Cas9 system, the Cas9 nuclease could efficiently shear the double-stranded DNA (dsDNA) sequence by recognizing specific complementary dsDNA containing protospacer adjacent motif (PAM) sequences with the help of guide RNA (gRNA) [135]. Through aptamers that specifically target exosomal proteins, the detection of proteins could be transformed into the quantification of nucleic acids. Recently, Zhao et al. reported a detection method which combined aptamer-based exosomal membrane protein recognition with CRISPR/Cas12-assisted fluorescence signal amplification [136]. In this system, CD63 aptamer specifically targeted exosome membrane proteins, triggering the conformational change of aptamer and release of blocker strands (with complementary sequences to aptamer). The released blocker was then recognized by CRISPR/Cas12a, resulting in the trans-cleavage toward TaqMan probe and the separation of fluorescence reporter group and quenching group, finally leading to the generation of amplified fluorescence signal. As a result, this method achieved a linear detection range of 10³-10⁷ particles/μL and was successfully applied in the direct detection of plasma exosomes without ultracentrifugation. Xing et al. developed an apta-HCR-CRISPR assay to detect circulating nucleolin⁺ or PD-L1⁺ exosomes from 50 μL serum of nasopharyngeal carcinoma cancer (NPC) patients [137]. Nucleolin or PD-L1-targeted aptamer was first amplified by HCR (hybridization chain reaction) to produce a long-repeated CRISPR-targetable DNA unit. Through collecting the fluorescence signal induced by collateral cleavage activities of CRISPR-Cas12a, the assay enabled a detection limit of 10² particles/μL. A similar assay was conducted by Li et al. for the ultrasensitive detection of CD109⁺ and EGFR⁺ exosomes [138]. CRISPR/Cas system is expected to be a sensitive tool for the identification and quantification of exosomal proteins and may contribute to the diagnosis and therapeutic monitoring of cancer.

Intrinsic heterogeneity is one of the main factors hindering exosome analysis in body fluids. Single exosome detection may provide more accurate information of tumor progression (Fig. 5E). Guo et al. presented an ssDNA-assisted single exosome detection platform. Rolling circle amplification (RCA) contributed to an amplified fluorescence signal from the surface protein, enabling easy visualization of individual exosomes with an LOD of 82 vesicles/μL [139]. In addition, Liu et al. developed a λ-DNA and aptamer-mediated approach, allowing for two-dimensional analysis of single exosome by size and tumor-associated marker expression [140].

Since nanosized exosomes cannot be sensitively identified by conventional flow cytometry, aldehyde/sulfate latex beads have been used to bind to vesicles, which are then stained with fluorescent antibodies and characterized for their protein markers. For example, anti-GPC-1 antibody and Alexa-488-tagged secondary antibody were introduced to the exosome-attached beads and the percentage of positive beads was therein referred as the percent of GPC-1⁺ exosomes [38]. In addition, the detection of expressed EpCAM, HER2 and EGFR in exosomes indicated their potential role in cancer diagnostics [118, 141] (Fig. 5E_a). Moreover, the development of nano-flow cytometry has provided a new option for multi-parameter analysis of single particle. The expression of CD9, CD63, CD81, CD235a, CD45, CD41a and CD144 of single exosome was measured via immunofluorescent labelling using nano-flow cytometry to evaluate the quality of exosome preparations isolated by six different methods. Liu et al. used nano-flow cytometry to analyze the expression of CD9, CD63, CD81, CD47, CD45, CD24, and EpCAM in tear-derived exosomes and found that the exosome concentration in tear fluid was approximately 100-fold higher than that of plasma exosomes [142].

Droplet digital PCR (ddPCR) is a technique in which the PCR reaction mixture is divided into tens of thousands of aqueous droplets in an oil emulsion (Fig. 5F). Each single droplet contains no more than one copy of target gene and is labeled as positive or negative according to the fluorescence amplitude. The concentration of target nucleic acids is then estimated by the Poisson distribution and ratio of the positive droplets [177]. The comparison between ddPCR and qPCR showed that ddPCR had higher accuracy and sensitivity in analyzing urinary exosomal miRNAs [163]. Chen et al. used ddPCR to identify IDH1 (isocitrate dehydrogenase 1) transcripts in exosomes derived from serum or cerebrospinal fluid (CSF) of glioblastoma (GBM) patients. Mutant IDH1 mRNA was identified in CSF-derived exosomes of patients bearing mutant IDH1 glioblastoma and higher level of IDH1 mRNA was found in exosomes from patients with tumors than healthy controls [164]. Sun et al. recently used ddPCR to quantify 10 HCC-specific mRNA from plasma samples of HCC patients (Fig. 5F_a). The diagnostic value of HCC exosome-derived mRNA signatures was evaluated by computing the digital scoring [61]. Moreover, exosomal miR-15a-5p was detected by ddPCR to distinguish endometrial cancer (EC) patients from healthy subjects [40]. Shen et al. reported that more copy numbers of lncRNAs RP11-77G23.5 and PHEX-AS1 were quantified by ddPCR in EpCAM-specific exosomes from malignant lung cancer patients compared to benign lung tumors [178].

Molecular beacon (MB) is a hairpin-like oligonucleotide labeled with fluorescent dye and quencher at two ends of the probe (Fig. 5G). MB is designed to spontaneously hybridize with the targeted sequence, thereby destroying the hairpin ring structure and inducing the appearance of fluorescence [179]. Lee et al. observed that high fluorescent signals were obtained by the hybridization of molecular beacons and miRNA-21 in exosomes of cancer cells and human serum [122] (Fig. 5G_b). Moreover, miR-375 and miR-574-3p were detected from exosomes of human urine by MB-based biosensors [168]. The expression levels of miR-21, miR-375, and miR-27a were detected in exosomes of human serum [121] (Fig. 5G_a). Oliveira et al. utilized CPP (cell-penetrating peptides) to deliver MB across the plasma membrane and then to detect miRNA-451a in red blood cell (RBC)-derived exosomes [170]. Chen et al. designed a 2′-O-methyl and phosphorothioate modified molecular beacon to quantitatively analyze exosomal miRNA-1246 from human plasma. After rupturing the exosome membrane with Triton X-100, the probe could specifically target miRNA-1246 inside and display quantitative fluorescence signals [169]. Zhang et al. developed an integrated exosome isolation and detection system, in which exosomes could be separated by microfluidic technology via using a little volume of samples. Meanwhile, nanopore detection technology effectively improved the detection efficiency of tumor-related miRNA without the need for amplification and fluorescence labeling of detected objects [180]. The molecular beacon-based biosensor has the priority of simple procedure, free from exosome pre-isolation and nucleic acid extraction, indicating their great potential in liquid biopsy for cancer diagnosis and prognosis.

DNA tetrahedron, a DNA nanoarchitecture with high controllability, can provide diverse amplified signal tags through chemical modification and DNA self-assembly [181]. Gao et al. conducted a DNA tetrahedron nanoprobe-based FRET sensing platform to sensitively detect miR-146b-5p in different cell lines [182]. Chen et al. reported a hairpin-tetrahedron nanoprobe for quantitative measurement of exosomal miR-21 in human serum [171]. In the absence of miR-21, the construction of fluorophore donor-quencher pairs resulted in a low FRET effect, while the presence of target miRNA induced the damage of stem-loop structure and occurrence of strong FRET. Consequently, the assay obtained a good linearity in the range from 1 × 10^− 12 to 10 × 10^− 9 M (mol/L) and a detection limit of 45.4 × 10^− 15 M. Zhang et al. proposed a similarly electrochemical biosensor based on two multifunctional DNA tetrahedrons assisted catalytic hairpin assembly. By leverage localized reaction and cascade amplification, the sensor enabled sensitive detection of tumor-associated exo-miRNAs down to 7.2 aM in 30 min [172].

Localized surface plasmon resonance (LSPR) occurs when the incident photon frequency matches the overall vibration frequency of precious metal nanoparticles or metal conducting electrons (Fig. 5H). Joshi et al. developed a biosensor based on LSPR for the label-free and nondestructive measurement of exosomal miR-10b [173]. Gold nanoprisms were chemically synthesized onto a silanized glass substrate and then functionalized with HS-C6-ssDNA and PEG6-SH. The direct hybridization of target miR-10b and HS-C6-ssDNA induced the formation of double-strand DNA which could increase the index of refraction of nanoprisms and change the wavelength of LSPR dipole peak (λLSPR). The concentration of miR-10b could be evaluated by ΔλLSPR. This platform was sensitive enough to distinguish between miR-10b and miR-10a (only one nucleotide difference) even in the subattomolar concentration range. Wu et al. developed an SPRi-based biosensor to detect multiple exosomal miRNAs for accurate diagnosis of NSCLC, in which Au-on-Ag heterostructure and DNA tetrahedral framework were utilized to enhance SPR signal and each exosomal miRNA could be ultrasensitively identified with different SPR signals [123] (Fig. 5H_a).

As a single-exosome analytic method, total internal reflection fluorescence (TIRF) has emerged for providing additional information about the heterogeneity of exosomes (Fig. 5E). He et al. proposed a TIRF-based single-vesicle imaging assay which delivered molecular beacon probes into exosomes and thus induced an amplified fluorescence of target miRNA (Fig. 5E_b). They performed a direct visualization of single vesicles and in-situ quantitative analysis of miR-21 in human serum samples and found that this assay showed better performance than conventional PCR assay in distinguishing cancer patients from healthy subjects [119].

Zhao et al. developed a thermophoretic sensor for in situ detection of exosomal miRNAs without the need of RNA extraction or target amplification (Fig. 5F). Aptamer modified in nanoflares could bind to target exosomal miRNA, inducing the appearance of fluorescence. The fluorescent signal became amplified after thermal electrophoretic accumulation, allowing the sensitive detection of 0.36 fM exosomal miRNA in a small volume of serum sample [175]. In addition, Han et al. proposed a DNA tetrahedron-based thermophoretic assay (DTTA) which achieved a detection limit of 14 aM mRNA in serum exosomes. After internalized by exosomes, the two fluorophore-labeled recognition sequences of DNA tetrahedron could bind to target mRNA, leading to an increase in FRET signal. Thermophoretic effect was applied to further amplify the FRET signal, which achieved a highly sensitive detection of PSA mRNA in exosomes. The DTTA assay showed that exosomal PSA mRNA performed better than serum PSA protein in discriminating prostate cancer from benign prostatic hyperplasia (AUC: 0.93 versus 0.74), providing a new approach for precise detection of prostate cancer [183].

Recently, CRISPR/Cas system has offered new opportunities to develop analytical methods for the detection and quantification of nucleic acids in exosomes. The strategy of integrating nucleic acid amplification with CRISPR/Cas has been proposed to improve analytical specificity and sensitivity [184]. For instance, the platform of rolling circular amplification-assisted CRISPR/Cas9 cleavage (RACE) was conducted by Wang et al. to detect multiple exosomal miRNAs [176]. During RCA process, padlock probe recognized single base differences in the presence of HiFi Taq DNA ligase. The amplification product of long ssDNA, which consisted of a large number of repeated target sequences and PAM structures, was specifically recognized by Cas9 nuclease. Consequently, the TaqMan probe hybridized with ssDNA was completely cleaved by Cas9 protein, allowing “turn on” fluorescence change that could be conveniently measured by a spectroscopy. As a result, this RACE platform could be used for highly specific detection of single or multiple exosomal miRNAs from human plasma. However, the large number of naturally PAM structures in the genome may increase the risk of off-target effects, greatly impairing the specificity of analysis [185].

Hepatocellular cancer is the fourth cause of cancer-related death. Molecular signatures loaded into HCC-derived exosomes may be used to for diagnosis. Exosomal miR-21 was found to suppress the apoptosis of HCC cells and be upregulated in HCC patients [112]. Exosomal proteins such as CEA and GPC-3, were able to distinguish HCC patients from healthy subjects [193], which may serve as promising biomarkers for noninvasive cancer diagnosis. In a study containing 158 samples, plasma was collected for the detection of 10 HCC-specific genes, including alpha-fetoprotein (AFP), GPC3, albumin, apolipoprotein H, etc. [61]. It was observed that higher fluorescent signals were obtained in HCC cohort via ddPCR in comparison with noncancer cohorts. The 10 gene signatures were further computed by machine learning, which showed great potential to distinguish early-stage HCC from at-risk liver cirrhosis with an AUC of 0.93. Nakano et al. compared the expression levels of exosomal miR-92b and circulating AFP among HCC patients who received liver transplantation [194]. It was revealed that exosomal miR-92b could predict early recurrence of HCC with an AUC of 0.925, while the AUC of AFP was 0.651. Zhu et al. integrated plasma exosomal RNA-sequence, cell-free RNA-sequence and TCGA tissue RNA-sequence datasets to identify 5 noncoding RNAs (circ-0073052, circ-0080695, SNORD3B-1, LINC01226 and HULC) as potential biomarkers of liver cancer. In addition, a panel of SNORD3B-1, circ-0080695 and miR-122 showed the highest AUC (89.4%) to classify liver cancer patients from healthy donors in comparison with other marker panels. By the selected panel, 79.2% AFP-negative samples and 77.1% early-stage liver cancer samples were successfully detected in the testing and validation sets, which indicated the potential of exRNAs panel in the early diagnosis of liver cancer [210].

As one of the most recognized indicators associated with pancreatic cancer progression, miRNA-10b is now being widely studied for the early diagnosis of pancreatic ductal adenocarcinoma (PDAC) [211]. Joshi et al. indicated that the levels of exosomal miR-10b were remarkably different among pancreatic cancer patients, at-risk patients with chronic pancreatitis (CP) and healthy individuals, which suggested the potential of miR-10b in diagnosing pancreatic cancer and predicting CP patients who may develop to PDAC [173]. The same conclusion was obtained by Pang’s group by using a dual-SERS biosensor for one-step detection of mRNAs in exosomes [195]. Melo et al. found that GPC-1⁺ circulating exosomes (GPC-1⁺ crExos) exhibited high specificity and sensitivity to identify PDAC patients from healthy individuals and chronic pancreatitis (AUC = 1.0), which was superior to CA199 (AUC = 0.739). Moreover, the levels of GPC-1⁺ crExos were correlated with tumor burden and survival in patients before and after surgery, suggesting the potential of GPC-1 serving as a reliable biomarker for treatment effect and prognostic monitoring [38]. In addition, the combination of GPC-1 and CD63 was demonstrated to show 99% sensitivity and 82% specificity in distinguishing PDAC patients from healthy subjects [197]. Furthermore, migration inhibitory factor (MIF) was reported to have better performance than GPC-1 and EGFR in the discrimination of different stages of pancreatic cancer [199].

Colorectal cancer (CRC) is one of the most common malignancies with high morbidity and mortality [1]. CircRNA in exosomes is related to the occurrence and development of cancer. Due to the circular structure, it provides a stable biomarker for cancer diagnosis. CircRNA-0004771 was found to be upregulated in serum exosomes of CRC patients [200]. The AUC of exosomal circ-0004771 to identify CRC from healthy controls was 0.88, and that to distinguish patients with stage I and II from other benign intestinal diseases was 0.816, which indicated that circ0004771 could be used as a potential diagnostic marker for colorectal cancer. Wei et al. explored novel exosome biomarkers for the early diagnosis and prognosis of colorectal cancer [125]. The detection of plasma samples revealed that the expression of CD9, CD63 and EpCAM were significantly higher in CRC patients compared with healthy and benign controls (with an AUC of 0.90 and 0.96, respectively). In addition, miR-15b, miR-21, and miR-31 were reported to be highly expressed in serum exosomes of CRC patients. ROC curves showed that AUC of miR-15b was 0.86 and the combined miR-15b, miR-21, and miR-31 panel exhibited 81.21% sensitivity and 81.03% specificity [212].

Non-small cell lung cancer accounts for over 80% of lung cancer-induced deaths [1]. High level of PD-L1 was detected in serum exosomes of NSCLC patients compared to normal individuals [132]. The diagnostic value of exosomal PD-L1 was further confirmed by Pang et al. in a larger cohort of patients with an AUC of 0.97 [90]. Ma et al. found that the expression of exosomal miR-21 was higher in patients with recurrent NSCLC compared with healthy individuals [201]. However, high levels of exosomal miR-21 were also identified in other cancers such as colorectal cancer, breast cancer, and liver cancer [123, 189, 204]. Therefore, the detection of multiple miRNAs may be more valuable for NSCLC diagnosis. Wu et al. demonstrated that miR-21, miR-378, miR-139, and miR-200 were differentially expressed between NSCLC patients and healthy donors, providing more optional biomarkers for the early diagnosis of NSCLC [123]. Notably, the experimental results showed that more amounts of exosomal RNAs were obtained in urine-derived exosomes of NSCLC patients than that from plasma and bronchoalveolar lavage fluid, which may provide a new direction of sample selection to analyze exosomes [62].

Breast cancer (BC) is the most common cause of cancer-related death among females. Early diagnosis can effectively suppress the mortality and achieve better treatment outcomes. Despite as a widely used biomarker, CA15-3 (carbohydrate antigen 15-3) is not sensitive enough to diagnose primary and metastatic breast cancer (MBC) [203]. Lee et al. demonstrated that cancer cell-derived exosomal miR-21, miR-222, and miR-200c could be quantified in body fluids and used for breast cancer diagnosis [204]. Lu et al. reported that the combination of RDW (red blood cell distribution width), MPV (mean platelet volume), and CA15-3 showed better specificity and sensitivity to identify breast cancer than single biomarker [203]. In addition, an exosome signature was identified by computing weighted sum of eight biomarkers (CA15-3, CA125, CEA, HER2, EGFR, PSMA, EpCAM, and VEGF) through machine learning. Consequently, the signature showed a high discriminative accuracy to differentiate MBC from non-metastatic ones and age-matched healthy donors (91.1%). Moreover, the exosome signature was reported to accurately monitor MBC treatment response and serve as an independent prognostic factor for progression-free survival in MBC patients.

PSA is a widely used biomarker for the detection of prostate cancer. However, increased PSA was also identified in inflammatory diseases such as benign prostatic hyperplasia (BPH) [213]. Therefore, the development of sensitive and specific biomarker is urgently needed to diagnose patients at early stage and monitor the progression of cancer. Li et al. reported that exosomal ephrinA2 had superior capability to blood circulating PSA to differentiate prostate cancer patients from BPH patients with an AUC of 0.906 [207]. Moreover, increased EpCAM and PSMA were found in the serum exosomes of prostate cancer [127]. In addition, the expression of exosomal TUBB3 mRNA was reported to be associated with poor progression-free survival of abiraterone in metastatic castration-resistant prostate cancer patients [208].

Due to the priority of living-cell secretion, large amounts and stable circulation compared to CTC and ctDNA, exosome-based liquid biopsy has been tested in clinical trials and several of them have been approved and reached the market. In 2016, Exosome Diagnostics proposed the first exosome-based liquid biopsy in the world, ExoDx™ Lung (ALK), for the isolation and analysis of exosomal RNA from blood samples. At CLIA-certified laboratory, ExoDx™ Lung (ALK) was proved to be an accurate, real-time tool to detect EML4-ALK mutations in NSCLC patients with 88% sensitivity and 100% specificity, which provides a more direct and sensitive method to detect gene fusions than cfDNA. In addition, the ExoDx Prostate IntelliScore (EPI) has been certified by FDA. Based on the detection of ERG, PCA3, and SPDEF RNA in exosomes, EPI provides a risk score to predict whether a patient with PSA from 2 to 10 ng/mL is likely to develop higher-grade prostate cancer [214]. According to ExoDx, 93% of sensitivity was achieved in prospective studies, and 26% of unnecessary needle biopsies were avoided when the EPI threshold was set at 15.6 [215]. Three independent, prospective, and multicenter clinical trials declared that EPI outperformed standard of care and could be used to assist in the early diagnosis of prostate cancer and eliminate unnecessary prostate biopsy [64]. Moreover, MedOncAlyzer 170 is a newly developed liquid biopsy system capable of detecting both exosomal RNA and ctDNA in a single trial. It can identify significant and functional mutations in multiple cancer types from small volumes (0.5 ml) of patient blood or plasma. Due to the unique formation manner of exosome and ctDNA, MedOncAlyzer 170 is accurate and highly sensitive to detect mutation at all stages of cancer progression and treatment. Although the clinical application value has been verified, larger clinical samples, populations and trials are still needed to confirm the role of exosome-based liquid biopsy in cancer diagnosis and treatment.