In vitro diagnostic technologies for the detection of extracellular vesicles: current status and future directions

Extracellular vesicles (EVs) are natural vehicles carrying and transferring bioactive molecules for intercellular communication and play key roles in physiological and pathological processes. EVs are promising biomarkers for disease diagnosis due to their rich biological information and minimally invasive sampling. In this review, we summarize the current methods for isolating EVs from the body fluids and the techniques for analyzing EVs’ biomolecules including the proteins, nucleic acids, and lipids. The advantages, feasibility, and challenges of EV‐based in vitro diagnosis toward clinical application are discussed. Finally, an outlook on the marketing and routine clinical application of EV‐based in vitro diagnosis is provided.

such as plasma, saliva, urine, amniotic fluid, pleural and pericardial effusion, bile, and cerebrospinal fluid, and are significantly aberrant in pathological situations, making EVs' potential circulating biomarkers. 5,6 Compared with the traditional histopathological examination, EVs' detection is noninvasive, which facilitate it convenient for early screening and dynamic monitoring. 7 EVs' detection provides various of information on genomics, transcriptomics, proteomics, and metabolomics of tumorigenesis, then offers in vitro diagnosis (IVD) values in early diagnosis, curative effect monitoring, medication guidance, and prognosis of diseases. Despite the rapid development of EV-based IVD technologies, while the progress of EV detection toward the practical clinical applications is hindered following points. EVs' purification from the complicated body fluids is difficult. Isolation of disease-related EVs from the normal cell-derived EVs is essential for improving the detection sensitivity and specificity but is challenging. Moreover, most detection methods are based on bulk analysis without considering the heterogeneity of EVs, although analysis of the nanoscaled single EV helps to precise medicine.
Herein, we summarize the methods for EVs' isolation and the IVD technologies for the detection and analysis of EVs' biomolecules from the aspects of proteins, nucleic acids, and lipids, and highlight the pivotal studies that are prospectively feasible for clinical applications. We also provide insights on the challenges and future development of these technologies for the real-world clinical practice.

ISOLATION, CHARACTERIZATION, AND PRESERVATION OF EXTRACELLULAR VESICLES
Isolating EVs from body fluids with complex components has been challenging and is one of the main bottlenecks for the clinical application of the EV-based IVD technology. Currently, common methods for EVs' purification include ultracentrifugation, polymer precipitation, immunoaffinity capture, immunomagnetic isolation, and size-based chromatography or ultrafiltration. 8 So far, there is no universally recognized "gold standard" for the separation of EVs. Although there are concerns about the time and cost consumption and the relatively low yield and activity of EVs, ultracentrifugation is still the most applicable method for EVs' separation. 8 By using a combination of separation methods such as size ultrafiltration combined with exclusion chromatography, high-purity EVs can be efficiently extracted from complex body fluids like plasma with just one large volume of elution in 22 min. 9 Moreover, on-chip isolation methods have been designed in a variety of ways, ExoTIC is a highly modular chip that allows size-based EV classification from different kinds of EV populations, and the results of this convenient device are consistent with those obtained by high labor-intensive ultracentrifugation by comparing the microRNA and proteomic profiles of EVs isolated by these two methods. 10 The polymer co-precipitation technique involves incubating pretreated samples with a polymer-containing precipitation solution overnight at 4 • C and then collecting the precipitated EVs by low-speed centrifugation, and a number of commercial kits based on this (e.g., ExoQuick, Mountain Vie, Exo-spin) have been developed for EVs' enrichment. 1 Immunoaffinity-based methods use common EVs markers including four transmembrane proteins (e.g., CD9, CD63, and CD81) and tumor-related biomarkers (e.g., EGFR and EpCAM) to capture EVs, and tumor-derived EVs can be distinguished from normal EVs through immune-based separation. 11 Despite these efforts, purity and low yield are still the main concerns of EVs' isolation.
The International Society for Extracellular Vesicles (ISEV) recommends the necessary identification and characterization assays for EVs. It is first necessary to identify whether the membrane proteins (e.g., CD63, CD81, CD82, HLA, and integrin) and membrane-binding proteins (e.g., ALIX and TSG101) or exosome-free proteins (e.g., HSP70 and ACT) are present in the sample through Western blotting (WB). Morphological characterization by electron microscope, and particle concentration and diameter distribution analysis of EVs through nanoparticle tracking analysis (NTA) or other means are also required to confirm the purity and integrity of the isolated EVs. 1 Concerning the preservation of EVs' samples, in 2013, ISEV recommended preserving samples in phosphate buffer (PBS) in silica containers at −80 • C. 12 However, EVs may fuse during storage at −80 • C, which can reduce EVs' concentration and sample purity in a time-dependent manner, so ready-to-use is recommended in the majority of cases. 13

DETECTION TECHNOLOGIES OF EVS FEASIBLE FOR CLINICAL APPLICATIONS
Detection of EVs' biomarkers can be generally classified into three subtypes including the analysis of proteins, nucleic acids, and lipids as the representative biomolecules ( Figure 1). To address the problems of sensitivity and specificity of conventional detection techniques, advanced signal amplification technologies have been successively developed based on different detection principles, such as fluorescence, thermophoresis, surface-enhanced Raman scattering (SERS), electrochemistry, surface plasmon resonance (SPR), nuclear magnetic resonance (NMR), or F I G U R E 1 Technologies for extracellular vesicle (EV) bioactive molecule detection mechanics. 14,15 Recently, a batch of highly sensitive, specific and efficient EVs detection techniques have emerged, and the following methods are some of the relatively feasible for clinical applications. Many of the below-mentioned studies are summarized in Table 1.

Extracellular vesicle protein detection
EVs are rich in protein biomarkers which are the moststudied contents over the past 10 years. 3 EV proteins derived from parental cells are mainly referred to the membrane and intravesicular proteins indicative of the bioregulatory process. 1 Conventional methods used to analyze EV proteins include WB, enzyme-linked immunosorbent assays (ELISA), flow cytometry (FC), and mass spectrometry (MS) for EV proteomic profiling. 16 Using MS and protein profiling, thousands of proteins from microliterlevel samples could be analyzed to identify the candidate proteins as biomarkers with diagnostic value. 17 Although MS can provide high-throughput and quantitative EVs' proteomics analysis, it requires lengthy workflows (days) and also high cost. So, MS-based EV proteomics is feasible for EV biomarker discovery in scientific research, while not feasible at present as diagnostic techniques for practical clinical utility.
Microfluidic technologies have been widely employed in the separation and detection of EVs since the emergence of micro-nano manufacturing processes. 18 To date, microfluidic-based EVs' detection methods include fluorescence imaging, colorimetric detection, electrochemical detection, magnetic detection, and methods combined with other techniques such as SPR, SERS, and so on. 19 Bring EVs detection to clinical point-of-care testing (POCT) settings, an integrated magnetic-electrochemical exosome (iMEX) sensor was developed for portable and streamlined EVs-proteins analysis. 20 Employing this platform, EVs were captured and identified by magnetic beads and then detected by an electrochemical sensor, thus directly achieving the effect of "sample in, result out"; moreover, only 10 μl plasma sample was needed to complete the detection in 1 h, which offers rapid, highthroughput analysis for multiplex screening and enables affordable and on-spot analysis. 20 The combination of microfluidics and SPR technology offers the advantages of label-free, fast and high-throughput and the number of EVs detected was greater than 2000 particles/μl. The primary premise is the functionalization of the metal film surface with nanohole structures by antibodies targeting at EV surface proteins. Then, the surface binding of the target EV could induce the alteration of the local refractive index, which influences the plasmon resonance of sensor chip. The quantity of EV protein attachment may be calculated using the standard calibration curve by measuring the wavelength and intensity of transmitted light through the metal film. 16 Likewise, sensors based on SERS have shown much promise in EV protein detection. To construct an immunoassay using Raman scattering intensity as the quantitative signal for the detection of EV protein, self-polymerization of dopamine was used to capture antibodies on a substrate and then incubated with polydopamine encapsulated antibody-reporter-Ag (shell) -Au (core) multilayer (PEARL) SERS tags to form a "chip-exosome-PEARL tag" sandwich structure. 21 Using this platform, only 2 μl of clinical serum sample is required to achieve early diagnosis, classification, and metastasis monitoring of pancreatic cancer patients. Zheng's group constructed a droplet microfluidic chipbased EVs digital detection technique to achieve absolute quantification of EV subpopulations and demonstrated that EV subpopulations with positive GPC-1 expression can distinguish breast cancer patients from benign breast patients and healthy individuals. 22 The field of microfluidic chip-based EVs' isolation and detection is still at early stages, with a wide number of concepts and technical issues to be addressed and long way to go before clinical implementation. For example, given the heterogeneity of EVs, achieving high purity and selective separation of EVs and improving the accuracy and sensitivity are ongoing hurdles.
Nanoscale flow cytometry (nFCM) is a powerful highsensitivity technique that can detect scattered light signals from EVs as small as 30 nm in diameter, allowing protein detection at the level of single EV, which can overcome the heterogeneity of EVs. 23 nFCM may be used to assess protein profiles and quantify protein expression at a single EV level, and the high-sensitivity, high-throughput, and high-speed features of this technique facilitate it a feasible technique for clinical translation. 8 Apogee A50 nanoscale ultra-high-resolution nFCM, developed by Premedical Lab, is specifically designed for micro-particle detection and multi-purpose experiments. EVs can be labeled with fluorescent proteins or lipid-specific dyes, eliminating the need to remove uncoupled fluorescent dyes by ultracentrifugation, and no correction is required before data acquisition, further accelerating the analysis of EVs. It can analyze exosomes and microvesicles isolated from cell culture media and human samples quantitatively and specifically due to its unique 80 nm scattered light resolution and 10 nm sensitivity. 24 Yan's group pioneered the combination of Rayleigh scattering and sheath flow single molecule fluorescence detection technology, developed the first nFCM with independent intellectual property rights in China, which achieved high-throughput multiparameter simultaneous characterization of the particle size (7-500 nm), distribution, concentration, and biochemical properties of single nanoparticles and biomolecules. 25 Recently, Yan's group has attempted to combine nFCM with enzymatic digestion to detect single EV as small as 40 nm in diameter and SYTO 16-stained 200 bp DNA fragments, elucidating the large heterogeneity in the amount of EVs-DNA in a single EV. 26 Most of the EV protein analysis techniques mentioned above are mainly based on antibodies as protein capture probes. However, the disadvantages of tedious antibody preparation and functionalization modification steps, poor reproducibility, high price and instability of antibodies have limited their development and translation to a certain extent. Aptamer is a short single-stranded DNA or RNA sequence that is combined with a target molecule through spatial configuration complementation. It has high stability and is easy to store for an extended period, and is convenient for artificial chemical synthesis and modification. 27 Furthermore, the binding force between the aptamer and the target molecule is high, possibly even higher than that between an antibody and an antigen. Various nucleic acid complexes and DNA self-assembled structures have recently been used to quantify the surface and interior molecules of EVs. 28,29 Sun et al. recently developed a thermophoretic aptamer sensor to address the problem of soluble proteins in peripheral blood interfering with EVs' protein markers and analyzed the cancer-related protein profile of EVs from the plasma sample. 30 Utilizing this system, DNA tetrahedral probes were combined with the thermophoretic convergence method to amplify the fluorescence signal with the intensity indicating the expression level of the EV surface proteins. The accuracy for discrimination of metastatic breast cancer patients, non-metastatic breast cancer patients, and healthy control subjects is as high as 91.1%. 30 Pu et al. synthesized a near-infrared (NIR) afterglow semiconducting polyelectrolyte to form a complex with quencher-tagged aptamers through electrostatic interactions. When EVs were injected, quencher-aptamers were discharged and fluorescence was activated simultaneously, allowing multiplex protein profiling of EV to be performed. 31 Zheng's group also developed a fluorescent aptamer sensor based on the combination of aptamers, aggregation-induced emission luminogens, and graphene oxide to assess tumor-related surface proteins in EVs, and reached a detection limit of 3.43 × 10 5 particles/L. 32 Nevertheless, selecting the right protein-specific aptamer is critical for specificity and sensitivity of this nanosensor, which is a time-consuming process, so there is still a long way ahead to achieve true clinical translation of aptamer-based multiplex protein profilings.

Extracellular vesicle nucleic acid detection
The nucleic acids of EVs include different forms of RNA (mRNA, miRNA, and noncoding RNA) and DNA. The majority of the studies currently focus on RNA. Studies have demonstrated that RNAs are transported to recipient cells and sustain functions. 1,33,34 They reflect the transcriptional and regulatory information of specific functions performed by EVs, and show diagnostic value for diseases. 35 Researchers commonly utilize sequencing or hybridization chip technique to screen the nucleic acid differently expressed in EVs, and then employ real-time polymerase chain reaction (PCR) for verification in EVs' nucleic acid test. Due to the low intrinsic abundance of RNA in EVs, the multi-step sample extraction procedure is inefficient and also prone to introduce contamination, which may result in false-positive with biased amplification during PCR. The existing EVs' nucleic acids' analysis technology has introduced quantitative real-time PCR (qPCR), digital PCR (dPCR), fluorescent probes, next-generation sequencing (NGS), localized SPR (LSPR), and so on. 16 In the conventional qPCR procedure, since the proportion of the mutant transcript in the EVs-inherent nucleic acid is low compared to the wild-type transcript, this nonspecific amplification increases false-positive results and background signal, making it impossible to obtain an absolutely quantitative result. 36 The advent of dPCR has made the absolute quantification of nucleic acids possible. The main principle is to first execute a restricted dilution of PCR reagents, followed by PCR amplification in different reaction chambers, and finally, the initial copy number or concentration of the target nucleic acid molecule is obtained according to the Poisson distribution and the number and proportion of positive droplets. 37 The sample fluid can be separated into numerous independent units and perform multi-step parallel reactions by means of the microfluidic chip technique. Droplet dPCR (ddPCR) and microarray dPCR are two types of dPCR, both of which are high-throughput, low-sample consumption, and low-cost. Employing ddPCR platform, Lee et al. provided the basis for the diagnosis of tuberculosis by detecting the content of mycobacterium tuberculosis-specific DNA in EVs isolated from bronchoalveolar lavage fluid. 38 The bulk of the EV nucleic acid analysis listed above need prior EV isolation and nucleic acid extraction techniques, both of which are time-consuming processes that do not fit the clinical objectives of POCT. Integrated microfluidic chips provide a one-stop platform for EVs nucleic acid extraction and analysis. Integrated EV lysis and RNA detection microfluidic chips have been developed for the diagnosis of a variety of diseases. [39][40][41] Specially, Shao et al. designed an integrated platform for tumorspecific EVs enrichment based on immunomagnetic, RNA capture, and qPCR detection. The chip cartridge contains all of the essential components and is inserted into a customized PCR system with a thermocycler and a fluorescence monitor, with fluid flow modulated by a torque actuated valve. The device can directly detect the significant changes of the EVs-mRNA transcription level of key enzymes involved in repairing DNA damage in human serum samples from patients with glioblastoma (GBM), offering the possibility to monitor chemotherapy response of GBM patients. 39 However, in this immunomagnetic isolation-RNA extraction-qRT-PCR workflow, the agglomeration of magnetic microspheres easily produces mutual interference of fluorescence signals on the surfaces of different microspheres, lowering the detection sensitivity. Yang et al. recently developed an immunobiochip for in situ quantification of EVs RNA by electrostatic action employing cationic lipid complexes containing molecular beacons. In comparison to conventional immunomagnetic separation chips, this chip enables detection of miRNA-21 and TTF-1 mRNA of EVs in human serum samples, achieving absolute sensitivity and specificity in the diagnosis of patients with non-small cell lung cancer (NSCLC), and consuming less sample volume and detection time (4 h). 41

Extracellular vesicle lipid detection
In 2016, the potential application of EV lipids as disease biomarkers has been revealed for the first time in prostate cancer from urine samples. 42 Previous reviews have summarized the role of lipids in the biogenesis of EVs and in the performance of biological functions, and the associated lipid molecules involved in diverse EV pathophysiologies. [43][44][45] However, EV lipids contain selectively enriched classes and species in comparison to the parent cell membrane. 46 Moreover, the accurate validation of candidate lipid markers cannot be achieved due to the existence of many isomers and structural similarities of lipid molecules and the lack of available stable standards. 47 The diversity and complexity of EVs lipids, and the heterogeneity of the EVs increase the difficulty for the identification and quantification of EV lipids. EV lipids include cholesterol, ceramide, phosphatidylserine (PS), and so on. The detection approaches for them are currently limited, the main methods are lipidomics and metabolomics, with most commonly used highperformance liquid chromatography (HPLC) and high-throughput MS (LC-MS). Hough et al. employed LC-MS lipidomics to reveal that ceramides and glycerophospholipids in EV lipids in bronchoalveolar lavage fluid samples of asthmatic patients were significantly lower than in a control group not exposed to second-hand smoking. 48 In another study, researchers comprehensively analyzed the differences among different EVs' lipid groups using the non-targeted lipidomics method in the discovery status. In the verification status, due to the lack of standards and corresponding internal standards for lipid biomarkers in EVs, the candidate lipid biomarkers found in the previous phase were verified using the quasi-targeted lipidomics method, with the sensitivity and specificity reaching 100% and 71.2%, respectively. 49 Besides, the chemical composition of a single EV is studied through Raman spectroscopy based on laser tweezers. Smith et al. used Raman spectroscopy to measure the spectra of single EV from 8 cell lines and discovered that single EV separated from the same cell type also showed high spectral variability. They revealed that EVs derived from cancer and non-cancer cell lines differed from each other in the relative expression of cholesterol and phospholipid, suggesting that EV lipids have the potential to be tumor biomarkers. 50

PRESENT AND FUTURE OF EXTRACELLULAR VESICLE-BASED IN VITRO DIAGNOSIS FOR CLINICAL APPLICATION
In 2021, the State Drug Administration released the "Management of In Vitro Diagnostic Reagent Registration" stipulated CLASS II and CLASS III IVD reagent products that satisfy the following conditions can be exempted from clinical trials: first, the reaction principle is well-defined, the design is standardized, and the manufacturing process is well-established; second, the IVD reagents applied for exemption from clinical trials should prove that they are safe and effective; last, there is no record of serious adverse clinical events for that have been on the market for several years. 51 The biological features of EVs make it an ideal biomarker with great potential for early diagnosis, dynamic disease monitor, and prognostic assessment. The development of various nano-biosensors, the introduction of novel nanomaterials, the design of sensitive probes and signal amplification strategies have significantly improved the specificity and sensitivity of the analysis of EV biomolecules. Furthermore, single EV detection methods overcome the heterogeneity of EV for even deeper and more accurate functional investigations. In 2015, the M.D. Anderson Cancer Center reported that glypican-1 (GPC1) is more enriched in early pancreatic tumor cell-derived exosomes than that of the normal people, in which the specificity and sensitivity reached 100%, suggesting that GPC1 is an ideal early biomarker of pancreatic cancer. 52 Subsequently, the concept of "precise exosome" was proposed, and the basic research and clinical transformation in the field of EVs diagnosis and treatment exploded, resulting in the establishment of a large number of emerging biotechnology companies. The world's first EV-based cancer diagnosis kit-ExoDx Lung (ALK) was released by Exosome Diagnostics company in September 2016, EML-4-ALK mutation can be accurately and real-time detected in NSCLC patients, and the sensitivity and specificity can reach 88% and 100%, respectively, which marked a step forward from basic research to clinical application of EV-based IVD. 53 It was not long before this company launched its second exosomal RNAbased liquid biopsy kit, ExoDx R Prostate (IntelliScore), the first to use specific genetic information captured from a urine sample to detect high-grade prostate cancer. 54 Notably, this is the first Food and Drug Administration (FDA) approval of an EV-based liquid biopsy product, which is a milestone event for EVs moving toward clinical application. Currently, a multitude of companies have committed to applying EV detection in clinical diagnosis, including Exosome Diagnostics, Exosome Science, Codiak BioScience, 3D Medicine, Theragnostex, Biocept, Exovita Biosciences, and so on. China's first exosome diagnostic product-exosome ovarian cancer adjuvant diagnostic kit was launched by 3D Medicine and initiated clinical trials for diagnostic product registration in March 2019. The kit combines a sensitive chemiluminescent immunoassay with three biomarkers, CA125, HE4, and C5a, to reach an 87.5% sensitivity and an 81.0% specificity, and was given priority approval by the National Medical Products Administration (NMPA) this year due to an urgent clinical need. 55 Despite these great progresses, transforming EVs detection from differentially expressed biomarkers in the laboratory to full-fledged clinical diagnosis is still challenging. First, there is a lack of standardized extraction and analysis techniques for the EV testing technology. Second, the reproducibility of EVs' test results is still debatable, and it is difficult to compare test results across platforms horizontally, and the traceability chain of results is lacking. Third, there is a gap in the unified clinical reference interval, all of which limit the clinical translation of EVs. Overall, with multi-disciplinary cooperation in molecular biology, physical chemistry, engineering, and medicine, EV-based IVD technologies hold great potential for the routine clinical application in the future.

C O N F L I C T O F I N T E R E S T S TAT E M E N T
The authors declare no conflict of interest.