Analytical Technologies for Liquid Biopsy of Subcellular Materials

Abstract

Liquid biopsy markers, which can be secured from a simple blood draw or other biological samples, are used to manage a variety of diseases and even monitor for bacterial or viral infections. Although there are several different types of liquid biopsy markers, the subcellular ones, including cell-free DNA, microRNA, extracellular vesicles, and viral particles, are evolving in terms of their utility. A challenge with liquid biopsy markers is that they must be enriched from the biological sample prior to analysis because they are a vast minority in a mixed population, and potential interferences may be present in the sample matrix that can inhibit profiling the molecular cargo from the subcellular marker. In this article, we discuss existing and developing analytical enrichment platforms used to isolate subcellular liquid biopsy markers, and discuss their figures of merit such as recovery, throughput, and purity.

1. INTRODUCTION

Biomarkers secured from a liquid biopsy are generating significant interest in the research and medical communities due to the minimally invasive nature of acquiring them and because they can enable precision medicine, which seeks to manage a variety of diseases, including oncology- and non-oncology-related diseases, from the molecular content secured from the markers (1, 2). These biomarkers include but are not limited to rare cells such as circulating tumor cells (CTCs), cell-free molecules such as cell-free DNA (cfDNA) and microRNA (miRNA), and extracellular vesicles (EVs).

Each of these markers has unique characteristics that can provide complementary information to manage a variety of diseases. For example, in the case of oncology, DNA mutations can be detected with the use of either CTCs or cfDNA. The RNA content of EVs can be used for expression profiling either from messenger RNA (mRNA) or miRNA. However, before any of the aforementioned biomarkers can be analyzed, they must be enriched from the biological sample because they are typically a vast minority in a mixed population.

There is now a pressing need to not only enumerate the liquid biopsy markers but also analyze their molecular cargo to provide important information such as diagnosis, prognosis, and treatment options and to determine whether a patient is responding to treatment or whether the disease is recurring. The challenge with liquid biopsy markers in terms of their molecular analysis is the mass limits they impose on the molecular assay. In addition, components present in a sample may interfere with the molecular processing of the cargo; enrichment can obviate these issues.

Enrichment techniques can take advantage of either the physical properties of the liquid biopsy marker (size, density, and electrical properties) or their biological properties (antigen expression). There have been several reports on the enrichment of CTCs from an assortment of clinical samples as well as reviews written on these enrichment techniques (3–7). As such, we do not cover CTC enrichment technologies in this article. Instead, we focus on the subcellular liquid biopsy markers, including the cell-free molecules (cfDNA and miRNA) and EVs.

cfDNA as a subcellular marker was first reported by Mandel and Métais in 1948 (8). In 1977 (9), the correlation between the level of cfDNA in the plasma of a patient with cancer and that in patients without cancer was demonstrated. In terms of cancer, circulating tumor DNA (ctDNA) (cfDNA describes the entire population of DNA coming from tumor cells and normal cells) was shorter than cfDNA from nondiseased cells (10). cfDNA from healthy individuals is 200–10,000 base pairs (bp) long, whereas the majority of ctDNA is <150 bp (10, 11).

Interest in cfDNA has evolved with the ability to detect mutations by next-generation sequencing (NGS) (12, 13). However, the quality of the sequencing data depends intimately on the quality of the input (14). As such, demands on the efficient extraction/enrichment of cfDNA from a sample are important. Besides NGS, the US Food and Drug Administration (FDA) has approved the Cobas^™ EGFR (epidermal growth factor receptor) mutation test for non-small-cell lung cancer, which is a cfDNA polymerase chain reaction (PCR)-based mutation detection assay (15). Even PCR tests require high-quality cfDNA in order to provide reliable results (9, 16, 17).

The cellular release of various types of membrane vesicles has been studied, and according to reported results, biological cells release vesicles of varying sizes both through the endosomal pathway and by budding from the plasma membrane. These vesicles are known by different names, including exosomes, microvesicles (MVs), and apoptotic bodies, but collectively are termed EVs. The particular subtype classification of EVs is based on their biogenesis (18). MVs are heterogeneous, membrane-bound vesicles shed from the surface of myriad cell types (19) and can range from 100 nm to 1 μm in size. Exosomes are the smallest of the EV family, with sizes ranging from 30 to 150 nm, and are released to the extracellular environment after the fusion of late endosomes/multivesicular bodies with the plasma membrane. Finally, apoptotic bodies are released as a product of apoptotic cell disassembly.

Finally, we include in this article enrichment techniques for viral particles, which are similar in size to exosomes, but their molecular composition is different. The inclusion was predicated on the recent pandemic outbreak of coronavirus disease 2019 (COVID-19), which is precipitated by the human-to-human transference of the virus severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) (20).

2. CELL-FREE NUCLEIC ACIDS

The relatively high abundance of cfDNA in blood (~80 ng/mL for patients with cancer) makes it a suitable marker for assessing the presence of mutations associated with several diseases (21). cfDNA are relatively short fragments, ~160 bp (10), but this can vary. miRNA, on the other hand, consist of fragments of small noncoding RNA that are 19–24 nucleotides in length (22). The ~80 ng/mL of cfDNA in patients with cancer corresponds to ~20,000 genome equivalents/mL (weight of a haploid genome is 3.5 × 10⁻¹² g per molecule). The fraction of ctDNA in the cfDNA population varies tremendously with type and stage of disease, but it can be as low as 0.1% (23). Additionally, the small length and dilution of cfDNA fragments when launched into circulation pose challenges for downstream molecular analysis. For example, the use of NGS to detect mutations in DNA requires concentrations on the order of 15 to 50 nM (~2 ng/μL), which is much higher than the average concentration typically seen in patient samples.

Therefore, extraction/enrichment of cfDNA from samples is critical in order to allow molecular processing to search with high reliability for mutations of interest. The small size of those fragments also poses a limitation to the extraction of cfDNA by methods that were originally developed to extract genomic DNA from cell lysates (24–27).

Extraction of cfDNA and miRNA from complex biological samples is challenged by the high abundance of contaminating molecules, for example, proteins; the protein concentration in plasma can range from 60 to 80 mg/mL. Unfortunately, many of these materials can interfere with downstream nucleic acid molecular processing, such as sequestering the activity of the polymerase used for PCR. Several methods have been developed to extract cfDNA from body fluids, such as plasma, serum, urine, and cerebrospinal fluid. In the next sections, we present several technologies currently available and those at a research stage of development for the extraction/enrichment of cfDNA and miRNA from samples.

2.1. Existing Approaches for Circulating Nucleic Acid Enrichment

Most of the currently available technologies for the extraction of cfDNA are based on spin columns or magnetic beads. In both cases, a solid surface with the appropriate chemistry typically extracts all nucleic acids from the sample irrespective of source (normal cells or diseased cells). This is important because many molecular analysis platforms require a certain number of diseased fragments to secure reliable results. For example, the percentage of patients with detectable ctDNA fragments is highly dependent on the stage of the disease for many cancers (28); only 42% of patients with stage I disease had detectable ctDNA and this number increased to 83% for those with stage IV disease when analyzed via NGS.

2.1.1. Spin columns and magnetic beads.

Spin columns (Figure 1a) bind nucleic acid fragments to silica in the presence of chaotropic agents such as guanidinium salts. Magnetic beads bind nucleic acid fragments to functionalized magnetic bead surfaces. Following nucleic acid binding, the magnetic beads with bound nucleic acids are guided by magnetic forces to extract them from the sample. The use of magnetic forces rather than centrifugation or vacuum manifolds required for spin columns makes it possible for magnetic bead–based methods to be automated. Evaluation of the performance of extraction kits in terms of recovery, size bias, and reproducibility gives an indication of their utility for specific studies.

Figure 1.

The bind-wash-elute procedure of magnetic bead and spin column extraction methods and the performance of selected kits in terms of recovery of cfDNA. (a) The general workflow for the operation of spin column and magnetic bead kits, which involve several washing and centrifuging steps. (b) cfDNA extraction yield of six kits; kits A, B, and C used magnetic beads and kits D, E, and F used spin columns. Panel b adapted with permission from Reference 33; copyright 2020 SAGE Publications. (c) Performance of six different kits in terms of total recovery (%) for fragment sizes ranging from 75 to 808 bp. Panel c reproduced with permission from Reference 29; copyright 2018 Elsevier. (d) Extraction of cfDNA fragments in the size range of 50 to 250 bp and 250 to 10,000 bp by magnetic bead– versus spin column–based methods. Panel d adapted with permission from Reference 33; copyright 2020 SAGE Publications. Abbreviations: bp, base pair; cfDNA, cell-free DNA.

A comparison of six kits (two magnetic bead–based kits, one automated magnetic bead–based kit, and three manual spin column–based kits) for the extraction of cfDNA from cultured human bone cancer cells showed that QIAamp (spin column) demonstrated higher yields (Figure 1b), a conclusion reached by others (29–31). In a similar study, Diefenbach et al. (29) observed that the processing time and cost per sample are higher for spin columns than for magnetic beads. However, spin columns offered higher recovery of fragments in the range of 75 to 808 bp (Figure 1c). Magnetic beads provided higher recovery of fragments in the range of 50 to 250 bp when compared with spin columns, which showed higher recovery for longer DNA fragments (250–10,000 bp) (31–33) (Figure 1d).

The type of liquid biopsy sample also has a strong influence on analytical performance. Lee et al. (31) studied urinary cfDNA extractions using two different spin columns, one bead based and the other bead and column based, and showed that genomic DNA contamination can lead to differences in results. However, genomic DNA contamination was not an issue when cerebrospinal fluid was used (34, 35).

2.1.2. microRNA.

Being a circulating nucleic acid, miRNA can be extracted using methods similar to those used for cfDNA. However, studies comparing the analytical performance of kits for miRNA and cfDNA provided different results. For example, Wright et al. (36) compared six kits for the extraction of miRNA from plasma. Overall, miRNeasy kits outperformed the other kits, yielding the highest amounts of miRNA. Fromm et al. (37) compared six total RNA (TRNA) extraction kits for miRNA preparations from monogenean Gyrodactylus salaris samples. The authors concluded that the phenol-free kit from Amresco showed the highest TRNA yield with high purity, whereas the MicroPrep kit from Zymo Research showed the highest miRNA-to-TRNA ratio.

2.2. Novel Analytical Approaches for Cell-Free DNA and microRNA Extraction

Despite the plethora of existing techniques available for the extraction of circulating nucleic acids, efforts to develop new technologies for cfDNA and miRNA extraction from samples have been reported. Three main methods have been reported: (a) liquid-phase methods, (b) solid-phase or chromatographic methods, and (c) electrokinetic methods (Figure 2).

Figure 2.

Different methods for cfDNA extraction and enrichment incorporated into microfluidic devices. (a, i) Microfluidic device for liquid-liquid extraction using a CD-type disc. (ii) Results obtained with this device for the recovery of spiked cfDNA fragments in plasma with comparison between the manual and disc methods as well as using blood versus plasma. Panel a adapted with permission from Reference 38; copyright 2019 Royal Society of Chemistry. (b) Results obtained when the same method was carried out, replacing centrifugation by vacuum. Panel b reproduced with permission from Reference 40 (CC BY). (c, i) Principle of SPRI using a plastic that was UV/O₃ activated (i.e., photoactivated). (c, ii) Recovery results for 122- and 290-bp DNA models obtained when varying the immobilization buffer composition, which consisted of PEG, MgCl₂, and ethanol. (c, iii) Plastic microfluidic device using SPRI for the solid-phase extraction of cfDNA. Panel c reproduced with permission from Reference 41; copyright 2018 Royal Society of Chemistry. (d) Capture of cfDNA using the application of an electric field. Panel d reproduced with permission from Reference 45; copyright 2015 Royal Society of Chemistry. (e) Device based on the use of DEP to capture and concentrate cfDNA. Panel e reproduced with permission from Reference 48; copyright 2012 John Wiley & Sons. Abbreviations: C-IFAST, centrifugal immiscible filtration assisted by surface tension; CD, compact disc; cfDNA, cell-free DNA; DEP, dielectrophoresis; EtOH, ethanol + water; GADPH, glyceraldehyde-3-phosphate dehydrogenase; IB, immobilization buffer; NAGK, N-acetyl-d-glucosamine kinase; PDMS, polydimethoxysilane; PEG, polyethylene glycol; RPPH1, ribonuclease P RNA component H1; SPRI, solid-phase reversible immobilization; TERT, telomerase reverse transcriptase.

Liquid-phase extractions are the basis of many traditional methods for DNA extraction, which commonly uses phenol and chloroform. Traditional phenol or chloroform extraction can take up to 48 h to complete and performs poorly when targeting small fragments such as cfDNA. A liquid extraction method using a lab-on-a-disk microfluidic with immiscible filtration assisted by surface tension (IFAST) was reported (38, 39) (see Figure 2a). A recovery of ~70% for cfDNA spiked into plasma was reported, and this value is similar to that obtained with a commercial kit (Figure 2a). Centrifugation was replaced by a vacuum (40), which was more convenient for microfluidics and resulted in recoveries near 100% (Figure 2b).

Another microfluidics approach used the concept of solid-phase reversible immobilization (SPRI) (41). SPRI consists of engineering a surface such that under optimum immobilization buffer conditions DNA specifically condenses onto a solid surface with the appropriate chemistry while other components do not (Figure 2c, subpanel i). After rinsing, the DNA is released in a small volume of aqueous liquid, resulting in enriched DNA. The physics of this phenomenon is well described by Rittich & Spanova (42). The authors of this article (41) demonstrated that the use of 16% polyethylene glycol (PEG), 20% ethanol, and 10 mM MgCl₂ recovered ~90% of 120-bp cfDNA fragments. The authors also demonstrated how the composition of the immobilization buffer could help tune the size of extracted cfDNA fragments (Figure 2c, subpanel ii). SPRI was transitioned to a polymer-based microfluidic device containing a high-density array of pillars (Figure 2c, subpanel iii).

A variety of materials and surface functionalization techniques to create cfDNA extraction solid phases have been described. A simple approach is the use of microfluidic devices in which the walls are modified to extract the cfDNA. Many of these devices use thermoplastics (Figure 2c) or silicon (41). Alternatively, beads packed inside a microchannel (43) or magnetic beads with chemistries appropriate for cfDNA extraction (44) can be used.

DNA can also be enriched with electrokinetic forces by way of a microfluidic device. The movement of charged molecules within an electric field has been utilized for the extraction of DNA at the interface between an aqueous phase and a gel (45) (Figure 2d), for example, a hydrogel matrix (46) or Nafion membrane (47).

Others have explored applying alternating electric fields using a phenomenon termed dielectrophoresis (DEP), which relies on the polarization of molecules in a nonuniform electric field. Using this strategy, Sonnenberg et al. (48) purified cfDNA using a microfluidic device (Figure 2e). A minimal concentration of 260 ng/mL was necessary to accumulate detectable amounts of cfDNA. In comparison to other solid-phase extraction methods, the use of DEP requires high voltages and the correct positioning of electrodes, which can add additional cost to the enrichment assay.

3. EXTRACELLULAR VESICICLES

EVs can be found in several types of biological fluids, including saliva, urine, blood, ascites, and cerebrospinal fluid. Blood plasma is an excellent source of EVs and is estimated to contain approximately 3 × 10⁶ particles per microliter (49). As with cfDNA and miRNA, EVs are generated from both diseased and nondiseased cells, and as such, enrichment is critical to improving the quality of the analysis of the molecular cargo contained within diseased EVs. As noted above, the term EV denotes a variety of particles produced from cells, for example, MVs and exosomes. MV biogenesis takes place through direct outward blebbing and pinching of the cell’s plasma membrane, while exosomes are released into the extracellular environment after the fusion of late endosomes and multivesicular bodies with the plasma membrane (49). The composition of EVs is determined mostly by the cell type from which they originated (50).

3.1. The Molecular Composition of Extracellular Vesicles

The cargo of EVs contains various types of proteins, RNA, and, in some cases, DNA. Cytoskeletal, cytosolic, and plasma membrane proteins are commonly found in EVs, as are proteins that show posttranslational modifications (51). EVs typically contain the tetraspanins such as CD9, CD63, CD81, and CD82 as well (51).

Although cellular mRNA has a size of 400 to 12,000 nucleotides, EV mRNA typically has <700 nucleotides (52, 53), but this can vary depending on the type of EV. EVs can contain mRNA fragments, long noncoding RNA, miRNA, and fragments of transfer RNA (51). Analysis of the contents of EVs has shown that specific mRNA fragments are enriched, especially the 3′ untranslated region fragment of mRNAs (52). EVs also contain miRNA. Deep RNA-sequencing has revealed that EVs contain 76% miRNAs and ~2% mRNAs (54). Because EVs are released from diseased and nondiseased cells, the diseased EVs should be enriched in high purity and high yields from body fluids so that the EVs’ molecular cargo can be analyzed free from interferences (55).

3.2. Existing Extracellular Vesicle Enrichment Techniques

For researchers to efficiently analyze EVs’ molecular cargo, EVs must be enriched from body fluids, in particular enrichment of the diseased EVs originating from nondiseased EVs. With the increasing number of research studies of EVs and their molecular content, a plethora of techniques have been developed to isolate EVs. These techniques typically fall under two broad domains: (a) those that isolate the entire EV population irrespective of the cell of origin and (b) those that isolate specifically EVs that are disease associated. In the sections below, existing techniques of both types are discussed.

3.2.1. Ultracentrifugation.

Ultracentrifugation (UC) is based on separation of particles according to their buoyant density. To affect the enrichment of EVs, researchers undertake several UC steps. First, particles with high buoyant density, such as cells (centrifugation at 300–400 × g), cell debris (centrifugation at 2,000 × g), aggregates of biopolymers, and other structures with a density higher than that of EVs, are sedimented (56) (Figure 3a). The resulting supernatant with EVs is then ultracentrifuged at >100,000 × g for 2 h, which yields an EV pellet (57, 58).

Figure 3.

(a) Differential UC for exosome isolation (sEV). Panel a reproduced with permission from Reference 56; copyright 2018 MDPI. (b) Summary of yield and purity of sEVs isolated by SEC or UC. Normalization of APOB signal to CD81 content as an estimate of sEV purity. Also, the lipoproteins demonstrated an APOB-to-CD81 ratio in the peak sEV fraction of SEC (5.5 mL) that was 60 times higher than that of UC. Panel b reproduced with permission from Reference 63; copyright 2018 Taylor and Francis. (c) Filtration and ultrafiltration for EV isolation. Normal prefiltration can collect sEVs into the bottom layer of the culture dish. The bottom solution layer needed to be processed through tangential ultrafiltration and the retentate collected. Panel c reproduced with permission from Reference 65; copyright 2014 Elsevier B.V. (d) Detection of EVs by lateral flow immunoassay using anti-CD9 and anti-CD81 capture antibodies and reflectance measurements of AuNPs on each test line (estimated as the peak area of the signal in mV × mm). EV-depleted plasma was used as a negative control (C⁻). Unbound anti-CD63 AuNPs captured with anti-IgG were used as system functional verification (C_t). Panel d reproduced with permission from Reference 69; copyright 2019 MDPI. Abbreviations: APOB, Apolipoprotein B; AuNP, gold nanoparticle; SEC, size-exclusion chromatography; sEV, small extracellular vesicle; UC, ultracentrifugation.

Density gradient UC uses two methods for the formation of a gradient: a continuous density gradient or a stepwise gradient based on sucrose (56) (Figure 3a). High spin speeds for long periods result in a concentration of exosome-like EVs in a band with similar densities (1.1–1.9 g/mL). As different EV types can have similar densities, the isolation of EVs by density gradient does not provide a pure fraction of exosomes depleted of other EV types (59). In some cases, UC-enriched EVs can be further purified by filtration or subsequent washing steps, but this does decrease the yield (59, 60).

Although UC can isolate EVs from large volumes of sample, some drawbacks include long isolation times (140–600 min), nonexosomal impurities, low reproducibility, and poor recovery (57, 60, 61). Although UC methods typically yield low-purity EVs compared with many other enrichment methods, Alvarez et al. (62) reported that UC with a sucrose density gradient can improve the purity.

UC and size-exclusion chromatography (SEC) have been systematically compared for enriching small EVs (sEVs) from rat plasma (Figure 3b), and results revealed that SEC-enriched sEVs had higher numbers, higher protein content, and higher particle-to-protein ratios compared with UC-enriched sEVs. However, SEC-enriched sEVs also contained large amounts of APOB⁺ lipoproteins and high quantities of non-sEV proteins (63).

3.2.2. Filtration.

Filtration has been used to isolate small particles on the basis of size (64, 65). Sequential filtration or a combination of filtration and UC is commonly used to provide high-grade exosomes (Figure 3c). A double-layer cell culture chamber with prefiltration using a modified polyethersulfone membrane was designed. Cells grew on the top layer and were blocked by the modified polyethersulfone, whereas EVs and proteins could pass through the membrane to reach the bottom layer. Tangential flow filtration with a 500-kDa hollow fiber was used to filter out proteins. A final step with low-pressure filtration selected the desired size of particles. Sequential filtration in which the size distribution of isolated EVs was controlled generated a throughput of 0.96 mL/h. However, the study showed that the high fluidic internal pressure and shear stress damaged the surface structure of the EVs, and as a result, the characterization and function of EVs were affected (66, 67).

3.2.3. Precipitation.

Hydrophilic polymers, such as PEG, reduce the solubility of EVs by lowering their hydration, leading to precipitation (68, 69) (Figure 3d). PEG 6000 or kits such as ExoQuick^™ and miRCURY are generally used for EV isolation. These reagents can be used to separate EVs at lower spin speeds with higher EV yields compared with UC. Upon addition of precipitation reagents, solubility of proteins is also decreased (70). Some of the advantages of precipitation reagents include preservation of EV integrity, no need for additional equipment, pH close to the physiological range, and the possibility to process many samples simultaneously (71). However, poor reproducibility, impurities, and retention of polymer are drawbacks (72–74). Analysis of four precipitation kits showed differences in their performances. The size distribution of the isolated particles using ExoQuick^™ was 40–150 nm and generated a high yield of exosomes. However, albumin impurity was abundant (75).

3.2.4. Affinity selection.

EVs contain important proteins that represent the cells from which the EVs were sourced. Tumor-derived EVs can express essential tumor-related proteins that can be used to diagnose cancer and its progression (51, 76). When specific proteins on the surface of EVs are targeted with affinity agents, a specific subtype of EV can be enriched. A variety of proteins can be targeted for EV affinity enrichment; these include disease-specific agents such as epithelial cell adhesion molecule (EpCAM), CD24, and CA125, which are cancer-related markers.

Affinity enrichment with high specificity and purity can be achieved for certain EV subtypes (77). However, the binding sites of affinity selection agents can get saturated by modest amounts of EVs. As a result, protein or RNA can be extracted from only a limited quantity of EVs, which can bring challenges. Because affinity enrichment can target only diseased EVs, the quality of the molecular analysis of the EV cargo can help improve the diagnosis of patients with disease (78).

3.3. Novel Approaches for Extracellular Vesicle Enrichment

Many of the recently reported platforms for the enrichment of EVs have been based on the use of microfluidics for several reasons, including their ability to be integrated into postenrichment processing steps such as enumeration and molecular profiling of the EV cargo. The enriched EVs can be enumerated (79–84), surface and cargo protein markers analyzed (79, 80, 85–88), and RNA profiled (81, 84, 89). By including the appropriate microscale or nanoscale structures within the chip, approaches such as affinity selection, filtration, centrifugation, viscoelasticity, and acoustic waves can be used for EV enrichment using a microfluidic device.

3.3.1. Affinity enrichment.

Affinity techniques can enrich primarily disease-associated EVs, which can improve the quality of the molecular data secured from the isolate (90). The ExoChip is an early example of a microfluidic chip used for affinity-enriching EVs (80). The ExoChip was fabricated using soft lithography and polydimethylsiloxane (PDMS) with surface-attached antibodies targeting CD63. Clinical serum samples were analyzed with immune electron microscopy and Western blot analysis was used to confirm isolation.

Many microfluidic devices used EV-specific markers such as the tetraspanins for affinity enrichment because in some cases, disease-specific markers may not provide viable molecular processing information. For example, in the later stages of cancer progression, cancer-related markers can be downregulated. A new version of the ExoChip used a phosphatidylserine (PS) targeting strategy (88) (Figure 4a). PS is expressed on the outer surface of cancer-related EVs. The ^newExoChip achieved 90% capture efficiency of cancer-related EVs, with the affinity-captured EVs released by Ca²⁺ chelation.

Figure 4.

(a) ^newExoChip design, which featured 30 × 60 circular patterns with a diameter of 500 μm configured with a footprint the size of a standard microscope glass slide. The mechanism of capture and release of cancer-associated exosomes used Ca²⁺-dependent binding between PS, annexin V, and EDTA-based Ca²⁺ chelation. The micrograph shows capture and release of exosomes. Panel a reproduced with permission from Reference 88; copyright 2019 John Wiley & Sons. (b) The nanointerfaced microfluidic exosome platform. (i) Close-up view of a single-channel PDMS/glass device highlighting the coated PDMS chip containing an array of Y-shaped microposts. (ii) Surfaces of the channel and microposts coated with graphene oxide and polydopamine as a nanostructured interface for the sandwich enzyme-linked immunosorbent assay with fluorescence signal amplification. Panel b reproduced with permission from Reference 79; copyright 2016 Royal Society of Chemistry. (c) Three-dimensional HB nanopatterns designed on a microfluidic device. (i) The device with solid HB mixers compared with nano-HB mixers that possessed a flow-through architecture. (ii) The fabrication method for the device. This device demonstrated the ability to detect tumor-associated EVs in plasma, with a minimum of 200 vesicles per 20 μL. The nanostructures were used to increase the surface area, content mass transfer, and EV capture speed and to reduce the hydrodynamic resistance. Panel c reproduced with permission from Reference 85; copyright 2019 Springer Nature. (d) Exodisc chip, which used a combination of centrifugal force and nanofiltration for the size selection of EVs. The chip was configured to fit the form of a CD (i) so that it could fit into a conventional CD player. (ii) A photograph of the Exodisc chip. The Exodisc was equipped with two different filters (iii), one for removing large entities, such as cells (Filter I), and the other for trapping the EVs (Filter II) but allowing free proteins to pass through the filter. (iv) Scanning electron micrographs of the filters. Panel d reproduced with permission from Reference 89; copyright 2017 American Chemical Society. (e) Size sorting of particles with an acoustic force generated by ultrasound standing waves in a microfluidic device. The particles are hydrodynamically focused with the use of a sheath flow and directed into the appropriate receiving reservoir on the basis of their size (see fluorescence micrograph at the bottom for sorting 190- and 1,000-nm beads). Panel e reproduced with permission from Reference 95; copyright 2018 ACS Publications. Abbreviations: CD, compact disc; EDTA, ethylenediaminetetraacetic acid; EV, extracellular vesicle; HB, herringbone; PDMS, polydimethylsiloxane; PS, phosphatidylserine.

A graphene oxide/polydopamine (GO/PDA) nanointerface was used to increase the EV-capturing surface area (79) (Figure 4b, subpanel i). Similar to sandwich enzyme-linked immunosorbent assays, the capture antibody targeting CD81 and detection antibodies targeting CD81, CD63, and EpCAM were used to characterize the EVs and remove interferences in the plasma sample (Figure 4b, subpanel ii). The assay provided a detection limit of 10⁶ particles/mL. Compared with the direct surface modification of GO or PDA only, the GO/PDA nanomatrix increased antibody capture efficiency of EVs by approximately twofold.

An approach using multiscale integration by designed self-assembly (MINDS) 3D nanostructures for the capture of EVs (Figure 4c, subpanel i) was fabricated using the steps shown in Figure 4c, subpanel ii (85). With MINDS, flow streams can pass through a bumper structure and a nanostructured herringbone (nano-HB) results in less hydrodynamic resistance and enhanced contact time of the EVs with the capture surface (Figure 4c, subpanel i). This approach offered a limit-of-detection of 10 EVs/μL. The microfluidic device also used affinity selection with EpCAM, CD24, and folate receptor α (FRα) as capture targets. For verification of the platform, 20 plasma samples from patients with ovarian cancer and 10 noncancer control plasma samples were processed and differences were achieved between the two groups in terms of the number of enriched EVs. The assay also enriched early disease stages by targeting the subpopulation of FRα-expressing EVs.

It is difficult to mass-produce PDMS-based microfluidic devices (91). As an alternative, thermoplastics are attractive because of their ability to be mass-produced and the simple modification protocols that can be employed to change their surface chemistry (92–94). A cyclic olefin copolymer ^EVHB-chip was manufactured with microinjection molding and designed to isolate tumor-specific EV-RNA within 3 h (81). The HB structure was compared with a flat channel surface and the results indicated that the HB device captured ~60% more EVs. The device could process a wide range of sample volumes (100 μL–5 mL) with a limit-of-detection of 100 EVs/μL.

Another group developed microfluidic devices using thermoplastics made via microinjection molding (78). A 7-bed EV microfluidic affinity purification (EV-MAP) device contained diamond-shaped pillars with a 10-μm diameter and 10-μm spacing to allow for high-throughput processing for enriching EVs via affinity selection. The device used mRNA to diagnose patients with ischemic stroke. mRNA expression of CD8⁺ EVs indicated that for genes upregulated during an ischemic stroke event, the EV-MAP device enriched EVs from clinical samples and profiled the genes of the EVs via droplet digital PCR, used for identifying patients with stroke, with a total processing assay time of 220 min. When the EVs were isolated by PEG precipitation, which isolates the entire EV population, mRNA expression differences for patients with stroke were not observed.

3.3.2. Filtration and centrifugal enrichment.

With the design flexibility of microfluidic devices, filtration can also be used. An Exodisc using a combination of centrifugal force and nanofiltration was reported (89) (Figure 4d, subpanels i,ii). With a centrifugal force limit of 500 × g, EV sizes of 20 to 600 nm could be collected between two nanofilters. Filter I (600-nm pore) was used to remove large particles (Figure 4d, subpanel iii), and Filter II (20-nm pore) was used to enrich the EVs and exclude free proteins (Figure 4d, subpanel iii). Scanning electron micrographs of Filters I and II are shown in Figure 4d, subpanel iv. The entire EV population was collected in 30 min with a recovery of 95%. Another platform with a combination of centrifugal force and filters for inline EV detection by flow cytometry was reported (82). The EVs were isolated by anti-CD81 antibodies, and with affinity microbead incubation, the enriched EVs could be concentrated and reconstituted and stained with a fluorescent dye.

3.3.3. Contactless and label-free extracellular vesicle enrichment methods.

These methods consist of using either fluid dynamics in a microchannel or microstructures in the channel to affect EV enrichment. A microfluidic viscoelastic flow device was developed for size-dependent and label-free isolation of EVs (83). Poly(oxyethylene) was added into a sheath fluid at a concentration of 0.1%, which maintained the feed solution at a particular viscosity. The particles were driven by an elastic force that situated particles in certain flow lines on the basis of particle size; larger particles traveled more toward the center of the channel. With the application of viscoelastic isolation, the platform demonstrated a sEV recovery of ~80% and a purity of >90%.

Microfluidic viscoelastic flow that used an acousto-fluidic device for label-free and contactless EV isolation was also demonstrated (84, 95). In one example, the platform included two surface acoustic wave modules that were operated at 19.6 MHz for cell isolation and 39.4 MHz for sEV isolation (84). The acoustic isolation was based on size because the deflection caused by the acoustic pressure and the drag force was proportional to the size of the particles. The cell removal rate by the first module was >99.999%, which resulted in a 75–90% reduction of red blood cells. With the series of modules, the isolation of 110-nm particles from whole blood yielded >99% recovery, and the purity of the sEVs reached ~98.4%. Ultrasound standing waves were used in a microfluidic device with sheath flow to create a tunable acoustic force that could sort particles on the basis of size as well (95) (Figure 4e). The acoustic force could be controlled electronically in situ to adjust the size of particles shuttled into the appropriate receiving reservoirs.

4. ENRICHMENT OF VIRAL PARTICLES

Viruses are subcellular particles similar in size to EVs (96) (Figures 5a,b). In the case of enveloped viruses (see SARS-CoV-2 in Figure 5c), they possess transmembrane proteins such as the S (spike) protein, which is unique to a certain class of viruses and allows for specific binding to host cells. Typically contained within the envelope is an RNA or DNA molecule coated with a capsid protein shell. Viruses are grouped into families on the basis of their shape, size, chemical composition, mode of replication, and genome structure. Each family of viruses gives rise to infection by the fusion of the viral particle (i.e., virion) to a host cell. The viral particle can be considered a subcellular material secured from an infected individual via a liquid biopsy.

Figure 5.

(a) Size comparison of various subcellular particles with respect to the size of a biological cell. As can be seen, viral particles are similar in size to sEVs (i.e., exosomes). Panel a reproduced with permission from Reference 96; copyright 2011 Springer Nature. (b) Transmission electron micrograph of SARS-CoV-2 viral particles. Panel b reproduced with permission from the National Pathogen Resource Collection Center. (c) Structure of a SARS-CoV-2 virus particle containing a single RNA molecule that is approximately 30,000 base pairs in length. (d) Viral load of SARS-CoV-2 as a function of days after hospitalization for 12 patients from Hong Kong. Viral particles were found in 11 of 12 patients (91.7%). Each line represents a different patient. Panel d reproduced with permission from Reference 101; copyright 2020 Oxford University Press. (e) Selection of viral particles with the use of affinity-coated magnetic beads situated within a microfluidic chip. The bottom panels show histograms for the assay’s specificity between the two influenza subtypes, H7N9 HA (left) and H9N2 HA (right), using HBV and Ebola virus as controls. Panel e reproduced with permission from Reference 104; copyright 2020 Elsevier. (f, top) Sized-based selection of viral particles with the use of a microchip containing pSiNW forests with a specific interwire distance. (Bottom) A control containing no virus particles (left) and a sample containing H5N2 virus particles stained with a fluorescent dye attached to an antibody that targeted the H5 antigen (right). Panel f reproduced with permission from Reference 109; copyright 2016 John Wiley & Sons. Abbreviations: HA, hemagglutinin; HBV, hepatitis B virus; pSiNW, porous silicon nanowire; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; sEV, small extracellular vesicles.

Coronaviruses (CoVs) are a family of enveloped single-stranded RNA viruses (97). CoVs are believed to be responsible for 5–10% of acute respiratory infections globally, with an estimated 2% of the population classified as healthy carriers of these viruses (98). Globally, current diagnosis of COVID-19 commonly uses real-time reverse transcription quantitative PCR (RT-qPCR) (99, 100). Several commercial RT-qPCR tests have been FDA-approved for COVID-19 testing, including the Centers for Disease Control and Prevention test, which has a limit-of-detection of ~10,000 copies/mL and thus may not be sufficient for detecting viral infections. For example, viral load from saliva indicated a range of 9.9 × 10² to 1.8 × 10⁸ particles/mL (101) (Figure 5d). Therefore, enrichment may be necessary to increase test positivity rates.

Most research in the field of viral enrichment has focused on utilizing affinity selection to isolate viral particles. In the case of affinity enrichment, these approaches take advantage of the specific interactions between viral particles mediated through their S protein and the host cell. For example, SARS-CoV-2 has a transmembrane S protein that has a high binding affinity toward the angiotensin converting enzyme-2 (ACE2) (102, 103). Proteins, antibodies, enzymes, and aptamers have been explored to enrich a wide variety of viruses. Other techniques, such as those utilizing size and dielectrophoretic properties, have been employed to enrich viral particles as well.

4.1. Affinity Enrichment

An example of a viral affinity target is hemagglutinin or neuraminidase for influenza A and B. Both hemagglutinin and neuraminidase are transmembrane glycoproteins of influenza virus membranes that promote viral invasion through binding to the sialic acid of host-cell-surface glycoproteins. Wang et al. (104) developed a microfluidic chip that separated two influenza A subtypes, H7N9 and H9N2, on the basis of antigenic differences between hemagglutinin glycoproteins found on the surfaces of viral particles. The authors immobilized anti-H7N9 and anti-H9N2 hemagglutinin antibodies to differently sized magnetic beads. As seen in Figure 5e, their chip consisted of three subunits: computer-operated multiplex microvalves, a magnetism zone, and a size-mediated zone that separates magnetic beads on the basis of size. Their results displayed high specificity between the two influenza subtypes (Figure 5e). The intra-assay variability was 4.5% for H7N9 hemagglutinin and 3.7% for H9N2 hemagglutinin.

Microfluidics can be combined with magnetic beads to form a single enrichment and detection platform (104). As an example, Lien et al. (105) reported a microfluidic device that integrated a pneumatic peristaltic rotary mixer and planar microcoils to incubate, purify, and enrich viral particles with magnetic beads conjugated to specific viral antibodies. Using the Dengue virus strain, they showed bead trapping efficiency of 87% at a mean flow rate of 0.83 mm/s and a starting concentration of 10⁷ beads/mL. The group demonstrated amplification of complementary DNA taken from the purified sample as a result of their enrichment process with a limit-of-detection of 10² pfu/mL.

Nanotrap^® particles (Ceres Nanosciences) are porous hydrogel beads with variable and tunable properties based on their polymer makeup, degree of cross-linking, and environmental factors such as temperature and pH. The Nanotrap particles enrich target viruses on the basis of the particle’s pore size and keep the target within the particle’s core through affinity binding. Shafagati et al. (106) used Nanotrap particles to capture whole virus particles, using Rift Valley fever virus as a model. They demonstrated that the Nanotrap particle NT53 allowed for 78% detection and was able to bind 99.35% of the virus particles at an input concentration of 10⁶ pfu/mL. NT53 allowed for a limit-of-detection of 2.5 pfu/mL.

In 2020, two RNA aptamers directed against SARS-CoV-2 RGD were reported (107). The aptamers were selected by an ACE2 competition strategy. The aptamers have K_d values of 5.8 and 19.9 nM. These could be attractive affinity agents for the selective enrichment of SARS-CoV-2 viral particles from a variety of samples, especially when placed within a microfluidic device.

4.2. Size-Based Enrichment

Physical barriers can be used to enrich viral particles on the basis of their size. Jeon et al. (108) fabricated a self-organized monolithic nanoporous membrane through stepwise anodization of aluminum. They demonstrated that their membrane had an efficiency of 93.3% for a 100-fold enrichment of hepatitis C virus (HCV) from a stock solution of 4.95 × 10⁴ pfu/mL, compared with centrifugation, which had an efficiency of only 22%.

Xia et al. (109) developed a microfluidic chip with porous silicon nanowire (pSiNW) forests extending from the bottom and sidewalls of microfluidic channels (Figure 5f). These pSiNW forests were fabricated to have controlled interwire spacing to allow viruses of a certain size to enter and remain trapped within the forest. Using the H5N2 avian influenza virus (AIV), the authors demonstrated a capture efficiency of 48%, a release efficiency from the forest of 60%, and a subsequent recovery efficiency of 29%. The capture efficiency was estimated by fluorescence staining of the captured H5N2 virus particles (Figure 5f).

Carbon nanotubes (CNTs) have been integrated into a microfluidic chip to enrich viral particles on the basis of size (110). Yeh et al. (110) vertically aligned nitrogen-doped multiwalled arrays of CNTs within a microfluidic chip. The capture efficiency for the CNTs varied depending on the intertubular spacing (25, 95, and 325 nm), with smaller intertubular distances resulting in higher capture efficiency. These results also showed improved lower limits-of-detection when the CNT enrichment was used prior to PCR detection. The microchip also effectively isolated H5N2 avian influenza virus particles.

4.3. Dielectrophoresis Enrichment

DEP relies on the viral particles’ selective response to applied electric fields. Grom et al. (111) presented a technique using a combination of electrohydrodynamic flow and DEP forces to trap hepatitis A virus particles. Through their microfluidic DEP system, they achieved an enrichment factor of 10⁷ to 10⁸.

5. CONCLUSIONS

Subcellular liquid biopsy markers are providing unique opportunities to manage a variety of diseases in order to enable precision medicine. One type of subcellular marker is the cell-free nucleic acids such as cfDNA and miRNA. What is attractive about these markers is that many established extraction/enrichment techniques can be used to provide high-quality inputs into molecular analysis platforms such as NGS. cfDNA has already been used in prenatal diagnostics (112), cystic fibrosis (113), and many cancer-related diseases (28, 114–116). Although several commercial kits can be used to extract cell-free nucleic acids from samples, there is a need for newer platforms that can provide higher levels of process automation and that are better tuned to enrich short DNA fragments (<100 bp) prevalent in cancer diseases (10).

An advantage of EVs as liquid biopsy markers over cell-free markers is the inherent protection of the EV cargo from degradation (117, 118). EVs also have a longer half-life in circulation than do cell-free molecules (118). Analysis of the molecular cargo of disease-specific EVs can provide a fingerprint for disease diagnosis and pathogenesis. For instance, tumor-derived EVs are rich in miRNAs that may serve as tumor markers (119). As an example, the RNA content in plasma EVs from patients with glioblastoma multiforme differs from that of healthy subjects (120). Also, patients with hepatocellular carcinoma after liver transplantation showed increased levels of certain miRNAs in their EVs (121). Proteoglycan glypican-1 (GP1)-positive exosomes were found in the plasma of patients with pancreatic cancer with absolute specificity and sensitivity (122). EVs have also been studied as biomarkers for many noncancer diseases in the central nervous system (123), liver (liver damage in viral hepatitis and hepatocyte injury in alcoholic, drug-induced, and inflammatory liver diseases) (124), kidney (intrinsic kidney disease) (125), brain (stroke) (126), lung (asthma) (127), and arteries (atherosclerosis) (128), and for radiation injury (129). In any case, enrichment of EVs is required despite their high number density found in plasma samples due to the small amounts of cargo they carry and because diseased and nondiseased EVs can be found in samples such as plasma.

The challenge with EVs is their enumeration, which requires special requirements because they are smaller than biological cells. EVs can be enumerated by the use of such techniques as flow cytometry and nonlabeling approaches such as resistive pulse sensing (130–132).

In the case of viral particles, few reports on enrichment platforms have appeared, and with the evolution of the new coronavirus pandemic, this demands attention. Enrichment of viral particles is important on two fronts: (a) removal of the target viral particles from sample-dependent inhibitors of enzymatic reactions, such as PCR polymerases; and (b) improvement of test positivity rates under low viral loads. Because the physical and chemical properties of viral particles and sEVs are similar, many of the platforms discussed for EV enrichment can be repurposed for viral particle enrichment.

6. PERSPECTIVE

Liquid biopsy markers can be considered to be in low abundance from a mixed population when present in biological samples such as blood. Therefore, some form of enrichment before analyzing their molecular content must be undertaken; this is true for all three major types of liquid biopsy markers: CTCs, EVs, and cell-free nucleic acids. It is insightful to compare and contrast in general the enrichment strategies used for the highly reported CTCs versus the subcellular markers (cell-free nucleic acids and EVs). Similar to CTCs, EVs can use both physical and biological techniques for enrichment—but with some precautions. For size-based (physical) properties, filtration can be used to enrich CTCs from blood because they are generally larger (~16 μm in diameter) than blood cells (white blood cells, ~7 μm) and UC or filtration can be used for EVs. In the case of CTC filtering, impurities typically arise from white blood cells, and for EVs, even in the most efficient size-based enrichment method, filtration isolates both diseased and nondiseased EVs owing to size similarities. For affinity (biological) selection of EVs, contamination from nondiseased EVs can be prevalent. For example, affinity enrichment of cancer EVs with EpCAM can result in the selection of EVs from normal epithelia due to their ability to secrete EVs continuously into circulation. In the case of CTCs, typically no epithelial cells are found in blood. However, for both EVs and CTCs, cell-free EpCAM protein can be present that can compete for affinity binding sites in the platform used for enrichment. For cell-free nucleic acids, enrichment involves using a solid-phase extraction approach with properly engineered surfaces to surface-bind nucleic acids. When using solid-phase extraction techniques, in the case of the oncology diseases, discriminating between ctDNA and non-ctDNA in a cfDNA population cannot be done. As such, for sequencing or PCR to find disease-associated mutations, the fraction of ctDNA must be sufficiently high to detect the target mutations.