Ultrafast Detection of Exosomal RNAs via Cationic Lipoplex Nanoparticles in a Micromixer Biochip for Cancer Diagnosis

Abstract

Exosomes are cell-derived, nanosized extracellular vesicles for intercellular communication. Exosomal RNAs have been shown as one type of promising cancer liquid biopsy biomarkers. Conventional methods to characterize exosomal RNAs such as quantitative reverse transcription polymerase chain reaction (qRT-PCR) are limited by low sensitivity, large sample consumption, time-consuming process, and high cost. Many technologies have been developed to overcome these challenges; however, many hours are still required to complete the assays, especially when exosome lysis and RNA extraction are required. We have developed a microfluidic cationic lipoplex nanoparticles (mCLN) assay that utilizes a micromixer biochip to allow for the effective capture of exosomes by cationic lipoplex nanoparticles and thus enables ultrafast and sensitive exosomal RNA detection for cancer diagnosis. The sensing performance and diagnostic performance of the mCLN assay were investigated using non-small cell lung cancer (NSCLC) as the disease model and exosomal microRNA-21 and TTF-1 mRNA as the biomarkers. The limits of detection of the mCLN assay were 2.06 × 10⁹ and 3.71 × 10⁹ exosomes/mL for microRNA-21 and TTF-1 mRNA, respectively, indicating that the mCLN assay may require as low as 1 μL of serum for exosomal RNA detection. The mCLN assay successfully distinguished NSCLC from normal controls by detecting significantly higher microRNA-21 and TTF-1 mRNA levels in exosomes from both NSCLC patient serum samples and A549 NSCLC cells than those from normal controls and BEAS-2B normal bronchial epithelial cells. Compared with conventional qRT-PCR assay, the mCLN assay showed a higher diagnostic accuracy in lung cancer, required less sample volume (30 vs 100 μL), and consumed much less time (10 min vs 4 h).

Graphical Abstract

INTRODUCTION

Exosomes are cell-secreted extracellular vesicles with diameters ranging from 50 to 150 nm. They present stably and abundantly in many bodily fluids, such as blood, urine, breast milk, and saliva. With the protection of a membrane structure, exosomes effectively transfer various biomolecules such as nucleic acids, proteins, and lipids between cells for intercellular communication.^1,2 Recent evidence has indicated that exosomes participate in not only physiological regulations but also pathological processes of cancer, including tumorigenesis, angiogenesis, metastasis, and drug resistance; therefore, they have become promising biomarkers for cancer liquid biopsy.^3–6 Among various cargos, exosomal RNAs, including both microRNAs (miRs) and mRNAs, have shown great diagnostic values in cancer.^6–10 For example, a panel of exosomal miRNAs (let-7b, let-7e, miR-24, miR-21) from plasma detected non-small cell lung cancer (NSCLC) with an area under the curve (AUC) value as high as 0.899.¹¹ Exosomal GATA2 mRNA was found to be a predictive, urine-based biomarker for the diagnosis of clinically significant prostate cancer.¹²

The expression of exosomal RNAs is typically measured by conventional approaches such as quantitative reverse transcription polymerase chain reaction (qRT-PCR), next-generation sequencing (NGS), and microarray. However, these methods are limited by low sensitivity, large sample consumption, tedious and time-consuming process, and high cost. In order to overcome these limitations, many biosensing technologies have been developed recently to offer sensitive, simple, and cost-effective exosomal RNA detection. For example, a nanoplasmonic biosensor utilizing localized surface plasmon resonance (LSPR) was developed to quantify exosomal miR-10b for pancreatic cancer diagnosis.¹³ Electrochemical biosensors were developed for exosomal RNA quantification and showed promising diagnostic values in breast cancer, colon cancer, and gastric cancer.^14–19 Fluorescent sensing probes, such as gold nanoflare probe,²⁰ DNA-labeled carbon dots (DNA-CDs), and 5,7-dinitro-2-sulfo-acridone (DSA)-based fluorescence resonance energy transfer (FRET) bioprobe,²¹ molecular beacons,^22,23 and split DNAzyme probes²⁴ were used to detect exosomal microRNAs for the diagnosis of breast cancer,^20–24 melanoma,²⁴ and cervical cancer.²⁴ Cationic nanoparticles-based biochips were developed to detect exosomal microRNAs and mRNAs for the detection of lung cancer,^25–30 liver cancer,³¹ and pancreatic cancer.^29,32,33 Microfluidics-based devices that utilized ion-exchange nanomembrane RNA sensing^34,35 and exponential rolling circle amplification (eRCA)³⁶ were developed to detect exosomal miR-550, miR-21, and let-7a for the diagnosis of pancreatic cancer,³⁴ live cancer,³⁵ lung cancer, and glioblastoma.³⁶ An immunomagnetic exosomal RNA (iMER) microfluidic biochip was developed to realize on-chip exosome capture, RNA extraction, and qRT-PCR to detect exosomal MGMT and APNG mRNAs to predict treatment responses in patients with glioblastoma multiforme.³⁷ A microwell-patterned microfluidic digital analysis biochip was developed to detect exosomal EWS-FLI1 mRNA for Ewing sarcoma (EWS) diagnosis.³⁸ Surface-enhanced Raman scattering (SERS)-based biosensors were developed to offer the sensitive detection of exosomal miR-21, miR-222, and miR-200c for breast cancer diagnosis³⁹ and exosomal miR-10b for pancreatic cancer early detection.⁴⁰ Although these new technologies have shown strong advantages over conventional methods, most of them still need hours to complete the measurements, especially when exosome lysis and RNA extraction are required.

Herein, we have developed a microfluidic cationic lipoplex nanoparticles (mCLN) assay to enable ultrafast and sensitive exosomal RNA detection (Figure 1a). In the mCLN assay, a micromixer biochip was used to overcome the challenges of mass transfer and effectively mix exosomes and cationic lipoplexes containing molecular beacons (CLP-MBs), enabling the fusion between negatively charged exosomes and positively charged CLP-MBs through electrostatic interaction.^25,30 After the formation of exosomes/CLP-MBs complexes, molecular beacons bind to exosomal RNA targets, fluorescence signals from molecular beacons are thus restored and used to characterize the expression of exosomal RNAs. Compared with conventional qRT-PCR assay (Figure 1b), which requires a large sample volume (100 μL serum), a tedious procedure, and hours to complete, the mCLN assay is a simple, ultrafast (~10 min), and low sample consumption (30 μL serum) test that sensitively measures the levels of exosomal RNAs in blood for cancer diagnosis.

Figure 1.

Schematics of the detection procedures of the mCLN assay and conventional RNA isolation–qRT-PCR workflow. (a) In the mCLN assay, the exosomes and CLP-MBs are effectively mixed to form exosomes/CLP-MBs complexes through electrostatic interactions, allowing for the hybridization of exosomal RNAs with MBs and the restoration of fluorescence signals from MBs. The restored fluorescence signals are then measured by the microplate reader and converted to exosomal RNA expression. (b) For a conventional workflow, total RNAs are extracted from exosomes and then quantified by qRT-PCR. The mCLN assay is a simpler, faster (10 min vs 4 h), and lower sample consumption (30 vs 100 μL) assay compared with the RNA isolation–qRT-PCR workflow.

In this study, the clinical utility of the mCLN assay as a cancer liquid biopsy test was demonstrated using lung cancer as the disease model. Lung cancer is one of the leading causes of cancer deaths. About 235 760 new cases are expected in 2021 in the United States. The 5 year survival rate of lung cancer decreases from 55% for early stage lung cancer to 6% for late stage lung cancer.⁴¹ The high incidence and high mortality underscore an urgent need for sensitive, simple, and fast tests that not only detect lung cancer at an early stage to improve the overall survival but also can be easily adapted to clinical settings to meet the needs of such a large patient population. Currently, low-dose computed tomography (LDCT) is recommended as a screening test for lung cancer. However, the high false-positive rate (95%) and radiation risk largely compromise the applications of LDCT in lung cancer screening and early detection.^42–44 Liquid biopsy tests, such as the mCLN assay developed in this work, detect tumor-derived biomarkers non-invasively, complement LDCT, provide additional information on cancer, and thus represent potent strategies to improve diagnostic accuracy. The diagnostic performance of the mCLN assay was evaluated using exosomal miR-21 and thyroid transcription factor-1 (TTF-1) mRNA as the biomarkers because these two RNAs have been shown as promising biomarkers for lung cancer.^{25,27,28,30,45–51} Since TTF-1 expression is more frequently observed in cancers of lung origin, it can be used as a biomarker to assist in distinguishing lung cancer from other types of cancer. The mCLN assay distinguished NSCLC patients from normal controls with a higher sensitivity and specificity than the gold standard qRT-PCR assay. Our results have demonstrated that the mCLN assay offers an ultrafast and sensitive detection of exosomal RNAs and, thus, is a promising liquid biopsy assay for lung cancer diagnosis.

RESULTS AND DISCUSSION

Sensing Mechanism of the mCLN Assay.

The mCLN assay utilizes a micromixer biochip to overcome the limits of mass transfer in biosensing and effectively mix exosomes with CLP-MBs and thus achieves the rapid detection of exosomal RNAs. As shown in Figure 1a, in the micromixer biochip, a stream that carries exosomes and a stream that carries CLP-MBs are fed into the microfluidic channel. The micromixer units in the microfluidic channel split the two streams into multilayered substreams through repeated split-and-recombination (SAR) structures,^52–54 which significantly increases the interface of the exosomes stream and the CLP-MBs stream and enhances the fusion between negatively charged exosomes and positively charged CLP-MBs via electrostatic interaction. The fusion between exosomes and CLP-MBs was demonstrated by cryogenic transmission electron microscopy (cryo-TEM), biological atomic force microscopy (bioAFM), and scanning electron microscopy (SEM) in our previous studies.^25,30 Following the fusion, MBs in the CLP hybridize with the target RNAs in exosomes. MBs are hairpin-structured oligonucleotides with fluorophores and quenchers attached at each end. The fluorophores are originally quenched by the quenchers. The hybridization between MBs and target RNAs separates the fluorophores from quenchers and thus restores the fluorescent signals of the fluorophores. The intensities of restored fluorescence signals are measured by a microplate reader and converted to the expression of exosomal RNAs.

Fabrication and Flow Pattern of the Micromixer Biochip.

Static micromixers are commonly used to effectively mix two or multiple liquids at the microscale by creating up to thousands of interfaced streams between these liquids. Because the thickness of the streams is greatly reduced (up to thousands of times), the diffusion distance is significantly shortened and the mixing is substantially enhanced. In this work, a stacking manufacturing method was used to fabricate the micromixer biochip in a simple and cost-effective manner. This method stacks and bonds laser-cut multiple layers into one device. As shown in Figure 2a, two static micromixer layers (left) and one splitter layer (right) were first prepared by laser cut through an acrylic board. The width of the inlet/outlet channels was 2 mm. The width of the mixer unit was 4 mm. Then, the biochip was assembled by sandwiching the splitter layer between two micromixer layers (Figure 2b). The two micromixer layers were mirror image to each other along the center line.

Figure 2.

Structure and working mechanism of the micromixer biochip. (a) Schematic diagrams of the micromixer layer (left) and the splitter layer (right). (b) Schematic diagram of the assembled micromixer biochip. (c) Split-and-recombination process was used to split the streams and generate layer-by-layer laminated structure to improve the mixing efficiency. (d) Experimental flow pattern of micromixer biochip. Sucrose solutions (50%, w/v) with and without 0.05% (w/v) Rhodamine B were mixed in the biochip with 10 micromixer units at a flow rate of 2 mL/h. The flow pattern was visualized under a fluorescence microscope.

The effective mixing of liquids in the micromixer biochip was realized through a SAR process (Figure 2c). Each micromixer unit holds a SAR structure, which first splits the stream perpendicularly to the orientation of lamella flow and then recombines the substreams.^52–54 Therefore, once a stream passes through one micromixer unit, it is split into two substreams and the width of each substream is halved. This process repeats when the substreams continue passing through the following identical micromixer units, and a layer-by-layer laminated structure is thus formed. Theoretically, the two streams at the inlets will be divided into 2^m+1 substreams at the outlet and the width of each substream will be reduced by 2^m folds, where m represents the number of micromixer units. As a result, the interface of substreams is increased by 2^m folds, which significantly enhances the interaction between negatively charged exosomes and positively charged CLP-MBs and achieves a much more effective mixing of these two types of nanoparticles compared with bulk mixing where the mixing simply relies on diffusion.

The laminated flow pattern was demonstrated in a micromixer biochip with 10 micromixer units. The width of each stream was reduced from 2 mm at the inlet to 2 μm at the outlet after the stream passed through 10 micromixer units. To visualize the flow pattern, one stream was labeled with 0.05% (w/v) Rhodamine B and the other stream was unlabeled. Sucrose (50% w/v) was added to increase the viscosity of the liquid and decrease the diffusion of Rhodamine B. As shown in Figure 2d, at a flow rate of 2 mL/h, a layer-by-layer laminated flow pattern was clearly observed, especially when the streams passed through the first four micromixer units. Close to the outlet of the biochip, streams with uniform fluorescence intensity were observed. The fluorescence intensity of Rhodamine B in the solution collected at the outlet of the biochip was 50% of that of the original, unmixed solution, indicating an effective mixing performance of the micromixer biochip (Figure S1).

Characterization of Exosomes and CLP-MBs.

Exosomes were isolated from the cell culture media of A549 NSCLC cells and BEAS-2B normal bronchial epithelial cells. Exosomes were also isolated from serum samples of NSCLC patients (n = 10) and normal controls (n = 5). The size, size distribution, and number concentration of exosomes were determined by nanoparticle tracking analysis (NTA). The morphology of the exosomes was observed with cryo-TEM. The representative size distributions of exosomes isolated from A549 cell culture conditioned medium and the serum of a NSCLC patient are shown in Figure 3a,b. The representative cryo-TEM images of exosomes from a NSCLC patient serum are shown in Figure 3d. The mean diameters of A549- and BEAS-2B-cell-derived exosomes were 141 ± 31 and 150 ± 41 nm, respectively. For all human-serum-derived exosomes, the mean diameter ranged between 65 and 135 nm and the exosome number concentration was around 4.18 × 10¹² exosomes/mL (Table S1). The sizes of human-serum-derived exosomes were slightly smaller than the sizes of cell-derived exosomes. This may be because exosomes in sera are released by all cells in the body including red blood cells and platelets, and smaller cells (such as red blood cells and platelets) may secret smaller exosomes than larger cells (such as cancer cells and cell lines). The sizes and number concentrations of human-serum-derived exosomes did not show any significant difference between NSCLC patients and healthy donors (Figure S2). NTA was also used to characterize CLP-MBs used in this study, i.e., CLPs encapsulating both FAM-miR-21 and Cy5-TTF-1 MBs for the detection of miR-21 and TTF-1 mRNA, respectively. A representative size distribution of CLP-MBs is shown in Figure 3c. The CLP-MBs had a mean diameter of 150 ± 25 nm and a similar size distribution as the exosomes. The CLP-MBs were typically prepared at a number concentration of 1.6 × 10¹¹ particles/mL.

Figure 3.

Characterization of exosomes and CLP-MBs by nanoparticle tracking analysis (NTA) and cryo-TEM. Representative size distributions and fields of view of (a) exosomes isolated from A549 cell culture conditioned medium (100-fold dilution), (b) exosomes isolated from 30 μL of serum from a NSCLC patient (10 000-fold dilution), and (c) CLP-MBs (2000-fold dilution). Inserted field of view images are screenshots from recorded videos obtained during NTA analysis. (d) Representative cryo-TEM images of exosomes from the serum of a NSCLC patient.

The ζ potentials of CLP-MBs and exosomes isolated from cell culture medium and human serum samples were measured. The ζ potential of CLP-MBs was 33.92 ± 3.07 mV. The ζ potentials of A549-cell-derived exosomes, BEAS-2B-cell-derived exosomes, and human-serum-derived exosomes were −13.85 ± 1.96, −12.845 ± 2.08, and −15.33 ± 1.65 mV, respectively. The opposite surface charges of CLP-MBs and exosomes allow these two types of nanoparticles to fuse with each other through electrostatic interaction as demonstrated by cryo-TEM, bioAFM, and SEM in our previous studies.^25,30

The quality of exosomes was examined using the Exo-Check exosome antibody arrays. The representative results of exosomes isolated from serum samples of a normal control and a NSCLC patient are shown in Figure S2. Exosomes showed a strong expression of exosome markers such as CD63, CD81, ICAM1, and TSG101 and a weak expression of GM130, a cis-Golgi marker as the negative control. The strong expression of multiple exosome markers and the lack of GM130 demonstrated the presence and good quality of exosomes.

Optimization of the Operating Parameters of the mCLN Assay.

The sensing performance of the mCLN assay mainly depends on the interaction efficiency between CLP-MBs and exosomes in the micromixer biochip. In order to achieve the best sensing performance, two parameters, the CLP-MBs to exosomes (CLP-MBs/exosomes) number ratio and the number of micromixer units, were optimized.

Exosomes isolated from A549 cell culture medium and two exosomal RNAs (miR-21 and TTF-1 mRNA) were used in the optimization study. To measure the exosomal RNA expression using the mCLN assay, 100 μL of exosomes and 100 μL of CLP-MBs were flowed through the micromixer biochip. The hybridization of MBs to exosomal RNAs restored the fluorescence signals from MBs, which were measured by the microplate reader as I_{exosome/CLP-MB}. PBS solution containing no exosomes was also flowed through the micromixer biochip with CLP-MBs, and the fluorescence intensity was collected as the background signal (I_PBS/CLP-MB). The expression of exosomal RNA (E_{exosomal RNA}) was calculated using the following equation:

$$E_{\text{exosomal RNA}}=\frac{\left(I_{\text{exosome/CLP-MB}}-I_{\text{PBS/CLP-MB}}\right)}{I_{\text{PBS/CLP-MB}}}\times 100\%$$ (1)

To optimize CLP-MBs/exosomes number ratio, A549 exosomes were prepared at the concentration of 8 × 10¹⁰ exosomes/mL. CLP encapsulating both FAM-miR-21 MBs and Cy5-TTF-1 MBs were prepared at concentrations of 5.3 × 10⁹, 8 × 10⁹, 1.6 × 10¹⁰, 8 × 10¹⁰, 1.2 × 10¹¹, and 1.6 × 10¹¹ particles/mL to achieve CLP-MBs/exosomes number ratios of 1:15, 1:10, 1:5, 1:1, 1.5:1, and 2:1. Then, CLP-MBs were mixed with A549 exosomes using the micromixer biochip with 10 mixer units at a flow rate of 6 mL/h in order to enable ultrafast detection within 10 min. The levels of exosomal miR-21 and TTF-1 mRNA were measured on the basis of the restored fluorescence intensities of FAM-miR-21 MBs and Cy5-TTF-1 MBs. As expected, the expression of exosomal miR-21 and TTF-1 mRNA was first increased with the increase of the amount of CLP-MBs, i.e., the increase of the CLP-MBs/exosomes number ratio (Figure 4a). When the CLP-MBs/exosomes number ratio was further increased from 1:5 to 2:1, the expression of exosomal miR-21 and TTF-1 was decreased significantly. This may be because the use of extra CLP-MBs did not further improve the exosome capture efficiency and increase the signals from exosomal RNAs. Instead, the extra CLP-MBs led to an increase of background signals. According to eq 1, when the signals from exosomal RNAs did not increase significantly while the background signals increased, the calculated expression of the exosomal RNAs was reduced. Therefore, the CLP-MBs/exosome number ratio was set at 1:5 for the following studies. At this number ratio, one CLP-MB was able to capture five exosomes.

Figure 4.

Optimization of the operating parameters of the mCLN assay. (a) Effects of CLP-MBs/exosomes number ratio on detection sensitivity. CLP-MBs at the concentrations of 5.3 × 10⁹, 8 × 10⁹, 1.6 × 10¹⁰, 8 × 10¹⁰, 1.2 × 10¹¹, and 1.6 × 10¹¹ particles/mL were mixed with A549-cell-derived exosomes at a concentration of 8 × 10¹⁰ exosomes/mL through the biochip with 10 micromixer units to achieve CLP-MBs/exosomes number ratios of 1:15, 1:10, 1:5, 1:1, 1.5:1, and 2:1. Highest levels of exosomal miR-21 and TTF-1 mRNA were observed at a CLP-MBs/exosomes number ratio of 1:5. (b) Effects of micromixer unit number on detection sensitivity. The CLP-MBs (1.6 × 10¹⁰ particles/mL) and A549-cell-derived exosomes (8 × 10¹⁰ exosomes/mL) were mixed through the biochips with 10, 12, and 14 micromixer units. Highest levels of exosomal miR-21 and TTF-1 mRNA were observed at the micromixer number of 10.

The number of micromixer units was also optimized. The micromixer biochips with 10, 12, and 14 micromixer units were used, which narrowed down the width of the streams from 2 mm at the inlet to 2000, 500 and 125 nm at the outlet, respectively. CLP-MBs were prepared at a concentration of 1.6 × 10¹⁰ particles/mL, and A549 exosomes were prepared at a concentration of 8 × 10¹⁰ exosomes/mL to achieve a CLP-MBs/exosomes number ratio of 1:5. The flow rate was set at 6 mL/h for each liquid. As shown in Figure 4b, the biochip with 10 micromixer units detected the highest restored fluorescence intensities for both miR-21 and TTF-1 mRNA. However, when the number of micromixer units was increased, the expression of exosomal miR-21 and TTF-1 mRNA was decreased. Although the biochips with 12 and 14 micromixers provided thinner streams than the 10 micromixer biochip and seemed to enhance the interaction between CLP-MBs and exosomes, the split of the streams through additional two or four micromixer units could break up the already formed CLP-MB–exosome complexes and, thus, negatively affected the detection sensitivity. Therefore, the biochip with 10 micro-mixer units was used in the following studies.

Characterization of the Sensing Performance of the mCLN Assay.

With the optimized operating parameters, the sensing performance of the mCLN assay was evaluated to determine the dynamic range, the limit of detection (LOD), and the limit of quantification (LOQ). A serial dilution of A549 exosomes was prepared at the concentrations of 8 × 10⁷, 8 × 10⁸, 8 × 10⁹, 2.5 × 10¹⁰, and 8 × 10¹⁰ exosomes/mL. To maintain the CLP-MBs/exosomes number ratio of 1:5, the number concentrations of CLP-MBs were set at 1.6 × 10⁷, 1.6 × 10⁸, 1.6 × 10⁹, 5 × 10⁹, and 1.6 × 10¹⁰ particles/mL, respectively. The nanovesicles isolated from a cell culture medium that was not cultured with A549 cells were used as the blank control. Exosomes and CLP-MBs were flowed through the biochips with 10 micromixers at a flow rate of 6 mL/h. The expression of exosomal miR-21 and TTF-1 mRNA was measured. As shown in Figure 5, with the increase of exosome concentrations, the levels of exosomal miR-21 and TTF-1 mRNA were increased. When a semilog regression model was applied to the results, we found that the dynamic range of detection was from 8 × 10⁷ to 8 × 10¹⁰ exosomes/mL with R² values greater than 0.96 for both exosomal miR-21 and exosomal TTF-1 mRNA. The LOD and LOQ of the mCLN assay were calculated as LOD = 3σ/k and LOQ = 10 σ/k, where σ was the standard deviation of data collected from blank control samples and k was the slope of the calibration curve.⁵⁵ The LODs of the mCLN assay were 2.06 × 10⁹ exosomes/mL for exosomal miR-21 and 3.71 × 10⁹ exosomes/mL for exosomal TTF-1 mRNA. The LOQs were 6.86 × 10⁹ and 1.24 × 10¹⁰ exosomes/mL for exosomal miR-21 and exosomal TTF-1 mRNA, respectively. In this work (Table S1 and Figure S2) and our previous studies,^27,30,56 we observed that a typical exosome concentration in human serum samples was about 5 × 10¹² exosomes/mL. Therefore, the mCLN assay may only require a very small sample volume, as low as 1 μL of serum, for exosomal RNA detection. These results demonstrated that the mCLN assay is a sensitive, robust, and low sample consumption liquid biopsy assay for cancer diagnosis.

Figure 5.

Characterization of LOD, LOQ, and dynamic range of the mCLN assay. (a) Serial dilution of A549-cell-derived exosomes with concentrations ranging from 0 to 8 × 10¹⁰ exosomes/mL were prepared. The levels of exosomal miR-21 and TTF-1 mRNA were measured with a 10 micromixer biochip at a CLP-MBs/exosomes number ratio of 1:5. (b) Semilog regression models were applied to determine the LOD and LOQ of exosomal miR-21 and TTF-1 mRNA.

mCLN Assay Distinguished A549 Cells from BEAS-2B Cells.

Exosomes derived from A549 NSCLC cells and BEAS-2B normal bronchial epithelial cells were tested on the biochips with 10 micromixer units. The concentrations of CLP-MBs and exosomes were 1.6 × 10¹⁰ particles/mL and 8 × 10¹⁰ exosomes/mL, respectively, to make a CLP-MBs/exosomes number ratio of 1:5. With the flow rate of 6 mL/h, the assay was completed within 10 min. As shown in Figure 6a, the expressions of A549-cell-derived exosomal miR-21 and exosomal TTF-1 mRNA were 152.9-fold and 22.6-fold higher than those of BEAS-2B-cell-derived exosomes, respectively, demonstrating that the mCLN assay successfully distinguished NSCLC from normal controls. In addition, the levels of miR-21 and TTF-1 mRNA of A549- and BEAS-2B-cell-derived exosomes were measured using the mCLN assay at a higher CLP-MBs/exosomes number ratio of 2:1, where the exosome concentration was maintained at 8 × 10¹⁰ exosomes/mL but the concentration of CLP-MBs was increased to 1.6 × 10¹¹ particles/mL. We observed that the levels of miR-21 and TTF-1 mRNA of A549-cell-derived exosomes were only 3.3-fold and 3.1-fold higher than those of BEAS-2B-cell-derived exosomes, respectively (Figure S3). The differences of exosomal RNA expression between A549-cell-derived exosomes and BEAS-2B-cell-derived exosomes were much higher when the CLP-MBs/exosomes number ratio was at 1:5 rather than 2:1, which further confirmed the superior sensing performance of mCLN assay with optimized parameters. To investigate the detection specificity of the mCLN assay, MBs detecting cel-miR-39 were used as the negative controls. Compared with Cy5-TTF-1 MBs, little fluorescence was observed with Cy5-cel-miR-39 MBs, indicating great sensing specificity of the mCLN assay (Figure S4).

Figure 6.

MCLN assay showed high sensitivity in distinguishing A549 NSCLC cells from BEAS-2B normal cells. The expressions of miR-21 and TTF-1 mRNA of exosomes from A549 cells and BEAS-2B cells were measured via the (a) mCLN assay, (b) bulk mixing, and (c) qRT-PCR. The CLP-MBs concentration was 1.6 × 10¹⁰ particles/mL and the exosome concentration was 8 × 10¹⁰ exosomes/mL to achieve a CLP-MBs/exosomes number ratio of 1:5. With an assay time of 10 min, the mCLN assay successfully detected exosomal miR-21 and TTF-1 mRNA, distinguished A549 cells from BEAS-2B cells, and showed a higher sensitivity than the bulk mixing method. The qRT-PCR assay detected a significantly higher expression of exosomal miR-21 from A549 cells than those from BEAS-2B cells, but it showed a lower sensitivity than the mCLN assay (n = 3; * p < 0.05; *** p < 0.0005).

To further demonstrate the superior sensing performance of the mCLN assay, the detection sensitivity of the mCLN assay was compared with conventional bulk mixing method and qRT-PCR assay. For bulk mixing, the exosome concentration was 8 × 10¹⁰ exosomes/mL and the concentration of CLP-MBs was 1.6 × 10¹⁰ particles/mL, which were kept same as those used in the mCLN assay. At the same assay time, i.e. ten minutes after bulk mixing of exosomes and CLP-MBs, little fluorescence was observed for miR-21 and very low fluorescence signals were observed for TTF-1 mRNA (Figure 6b). There were no significant differences between A549-cell-derived exosomes and BEAS-2B-cell-derived exosomes. These results demonstrated that the mCLN assay enabled a much more effective mixing of exosomes and CLP-MBs than bulk mixing, and thus realized an ultrafast detection of exosomal RNAs.

The expression of exosomal miR-21 and TTF-1 mRNA was also measured by qRT-PCR. The quantification cycle (Cq) values of the endogenous control, miR-191, were used as the normalization factor.^57–59 At the concentration of 8 × 10¹⁰ exosomes/mL, qRT-PCR successfully detected miR-21 in both A549-cell-derived exosomes and BEAS-2B-cell-derived exosomes (Figure 6c). However, at this low concentration, qRT-PCR was not able to measure the expression of TTF-1 mRNA in exosomes from both cell lines. Because the mCLN assay and qRT-PCR assay produced data with different units, to allow for the comparison between these two methods, the difference of miR-21 levels between A549-cell-derived exosomes and BEAS-2B-cell-derived exosomes was used. The mCLN assay detected an 152.9-fold higher level of miR-21 in A549-cell-derived exosomes than that of BEAS-2B-cell-derived exosomes; however, the qRT-PCR assay only detected an 1.9-fold difference in miR-21 expression between A549-cell-derived exosomes and BEAS-2B-cell-derived exosomes. These results suggested that the mCLN assay had a much higher detection sensitivity than conventional qRT-PCR assay.

mCLN Assay Distinguished NSCLC Patients from Normal Controls.

To demonstrate the clinical utility of the mCLN assay in lung cancer diagnosis, the expressions of exosomal miR-21 and TTF-1 mRNA in human serum samples from 5 normal controls and 10 NSCLC patients were measured. Table S1 summarized the characteristics of all patients enrolled in this study. Exosomes were isolated from total 30 μL serum samples, resuspended in PBS, and characterized by nanoparticle tracking analysis. Exosomes from normal controls and NSCLC patients showed a similar size and exosome number concentration, and no significant difference was found between these two groups (Table S1 and Figure S2). Exosomes were diluted in PBS to a concentration of 8 × 10¹¹ exosomes/mL. The number concentration of CLP-MBs was set at 1.6 × 10¹¹ particles/mL to maintain the CLP-MBs/exosomes number ratio at 1:5. The CLP-MBs and exosomes were mixed in the 10 micromixer biochips at a flow rate of 6 mL/h. The restored fluorescence signals from MBs were used to calculate the expression of exosomal miR-21 and TTF-1 mRNA. As shown in Figure 7a,b, the mCLN assay successfully distinguished NSCLC patients from normal controls. The averaged levels of exosomal miR-21 and TTF-1 mRNA of NSCLC patients were 1530.5-fold and 6.1-fold higher than those of the normal controls, respectively. Conventional qRT-PCR assay was used to measure the levels of miR-21 and TTF-1 mRNA in exosomes isolated from total 100 μL serum samples. Exosomal miR-191 was used as the endogenous control.^57–59 As shown in Figure 7c, the qRT-PCR assay detected an 1.4-fold higher exosomal miR-21 level in serum samples from NSCLC patients than that of normal controls; however, the difference between these two groups was found to be not significant. Due to the low sample volume (100 μL of serum), the qRT-PCR assay was not able to quantify the expression of exosomal TTF-1 mRNA in these samples. Compared with the qRT-PCR assay, the mCLN assay successfully detected exosomal miR-21 and TTF-1 mRNA in serum samples and distinguished NSCLC patients from normal controls with statistical significance, demonstrating a superior detection sensitivity and specificity than qRT-PCR.

Figure 7.

MCLN assay showed high sensitivity and specificity in distinguishing NSCLC patients from normal controls. The expression of miR-21 (a) and TTF-1 mRNA (b) of human-serum-derived exosomes was measured using the 10 micromixer biochips. The concentration of CLP-MBs was 1.6 × 10¹¹ particles/mL and the concentration of exosomes was 8 × 10¹¹ exosomes/mL to achieve 1:5 CLP-MBs/exosomes number ratio. Significantly elevated expression of exosomal miR-21 and TTF-1 mRNA were observed in serum samples from the NSCLC patients than normal controls. (c) Conventional qRT-PCR assay detected higher levels of exosomal miR-21 in serum samples from NSCLC patients than normal controls, but no statistical significance was observed between these two groups. (d) The mCLN assay was more sensitive than qRT-PCR because it effectively distinguished TEXs from non-TEXs. The strong fluorescence signals from CLP-MBs/TEXs were easily detected by the microplate reader, while the weak fluorescence signals from CLP-MBs/non-TEXs complexes may not be effectively detected. (* p < 0.05).

The superior diagnostic performance of the mCLN assay over the qRT-PCR assay may be due to the effective differentiation of TEXs from non-TEXs. In mCLN assay, CLP-MBs interacted with exosomes directly and formed CLP-MBs/exosomes complexes. All fluorescence signals were localized within the CLP-MBs/exosomes complexes. Our previous studies^25,30 and this study (Figure 6) demonstrated that TEXs contained more tumor-associated RNA biomarkers than non-TEXs. Therefore, the fluorescence signals from CLP-MBs/TEXs complexes would be much stronger than CLP-MBs/non-TEXs complexes. The stronger fluorescence signals from CLP-MBs/TEX complexes could be easily detected by the microplate reader, while the much weaker fluorescence signals from CLP-MBs/non-TEXs complexes may not be effectively measured due to the detection limit (Figure 7d), enabling the differentiation between TEX RNAs and non-TEX RNAs and realizing a much higher sensitivity and specificity in distinguishing NSCLC from normal controls. However, for qRT-PCR assay, RNA molecules from TEXs and non-TEXs were mixed together during the RNA extraction step (Figure 1b); therefore, it was less capable in distinguishing TEX RNAs from non-TEX RNAs and showed poor performance in discriminating NSCLC from normal controls.

Exosomes are cell-secreted nanosized vesicles, which present abundantly in various types of bodily fluids. Exosomal RNAs are actively involved in tumorigenesis, angiogenesis, metastasis, and drug resistance and, thus, have become promising biomarkers for cancer liquid biopsy. Conventional methods such as qRT-PCR have been used to characterize exosomal RNAs; however, they are challenged by poor sensitivity, large sample consumption, tedious processes, and high cost. To overcome these challenges, new technologies such as a LSPR biosensor,¹³ electrochemical biosensors,^14–19 fluorescent probes-based assays,^20–24 cationic nanoparticles-based biochips,^25–33 microfluidic devices,^34–38 and a SERS-based biosensor^39,40 have been developed to offer more sensitive, simple, and cost-effective assays. However, due to the limits of mass transfer in biosensing, a majority of current technologies need hours to complete the measurements of exosomal RNAs and an even longer time if the assay requires exosome isolation and RNA extraction (Table 1). In this study, we developed a mCLN assay that utilized a micromixer biochip (Figure 2) to overcome the challenges of mass transfer, efficiently mix negatively charged exosomes with positively charged CLP-MBs, and enable an ultrafast and sensitive quantitation of exosomal RNAs. Compared with the existing technologies, the mCLN assay directly detected exosomal RNAs without requiring RNA isolation procedures and significantly reduced the assay time from hours to 10 min. We recognize that exosome isolation and concentration measurement are still required for the mCLN assay; therefore, the total time-to-result is approximately 1 h. However, when the time used for exosome isolation, RNA extraction, or exosome characterization is also considered for existing technologies, the mCLN assay is still a much faster assay (Table 1).

Table 1.

Comparison of mCLN Assay with Existing Sensing Technologies.

detection mechanism	disease model	biomarker	assay timea	LOD	sample conditions	exosome isolation	RNA isolation	reference
localized surface plasmon resonance	pancreatic cancer	miR-10b	>12 h (overnight incubation is required)	83.2 aM miRNAs	RNA extracted from 500 μL of plasma-derived exosomes	yes	yes	Joshi et al.13
electrochemical	breast cancer	miR-21	1 h 40 min	67 aM miRNAs	RNA extracted from serum-derived exosomes	yes	yes	Zhang et al.14
	NA	miR-143 and miR-146a	20 min + RT-PCR assay time	20 pM miRNAs	2 nM heat-denatured PCR amplicon	yes	yes	Goda et al.15
	colorectal cancer	miR-21	>1 h	1 pM miRNAs	10 μL of RNA extracted from serum-derived exosomes	yes	yes	Boriachek et al.16
	breast cancer	miR-21	7 h	2.75 fM miRNAs	RNA extracted from plasma-derived exosomes	yes	yes	Tang et al.17
	breast cancer	miR-21	1.5 h	2.3 fM miRNAs	RNA extracted from cell-derived exosomes	yes	yes	Luo et al.18
quantum dots-based photoelectro-chemical	gastric cancer	Homo sapiens HOXA distal transcript antisense RNA (HOTTIP)	2 h	5 fg/mL RNAs	RNA extracted from 250 μL of serum-derived exosomes	yes	yes	Pang et al.19
Au nanoflare probes	breast cancer	miR-1246	4 h	0.68 nM miRNAs, 2 × 1010 exosomes/mL	40 μL of plasma	no	no	Zhai et al.20
DNA-labeled carbon dots (DNA-CDs) and 5,7-dinitro-2-sulfo-acri-done (DSA)-based FRET bioprobe	breast cancer	miR-21	>2 h	3 fM miRNAs	RNA extracted from cell-derived exosomes	yes	yes	Xia et al.21
molecular beacons	breast cancer	miR-21	1–2 h	1 × 1013 exosomes/mL	50 μL of exosomes or 15 μL 3 × 1013 exosomes/mL spiked in 35 μL of human serum	yes	no	Lee et al.22
	breast cancer	miR-21, miR-27a, and miR-375	1 h	NA	6 × 1013 exosomes/mL spiked in 70% human serum	yes	no	Lee et al.23
split DNAzyme probes	melanoma, breast cancer, cervical cancer	miR-21	2 h	378 copies/μL	10 μL mixture of streptolysin O-treated exosomes and split DNAzyme probes	yes	no	He et al.24
tethered cationic lipoplex nanoparticles	lung cancer	miR-21 and thyroid transcription factor 1 (TTF1) mRNA	2 h	exosomes from 2 × 104 A549 cells spiked in serum	70 μL of exosomes derived from 70 μL of serum	yes	no	Wu et al.25
	lung cancer	TTF1 mRNA and transketolase 1 (TKTL1) mRNA	2 h	NA	20—80 μL of plasma	yes	no	Lee et al.26
	lung cancer	miR-21, miR-25, miR-155, miR-210, and miR-486	2 h	NA	30 μL of exosomes derived from 60 μL of serum	yes	no	Liu et al.27
	lung cancer	PD-L1 mRNA and miR-21	6 h	exosomes derived from 1 μL of cell culture media	exosomes derived from 90 μL of plasma	yes	no	Zhou et al.28
	pancreatic cancer, lung cancer	KRAS mRNA and EGFR mRNA	2 h	40 amol miR-21 ssDNA oligo	20 μL of serum	no	no	Hu et al.29
	hepato-Cellular carcinoma	α-fetoprotein (AFP) mRNA and Glypican-3 (GPC-3) mRNA	2 h	NA	20 μL of exosomes derived from 20 μL of plasma	yes	no	Wang et al.31
	pancreatic cancer	miR-21, miR-lOb, and miR-212	2 h	NA	plasma	yes	no	Pu et al.32
immunoselection and cationic lipoplex nanoparticles	lung cancer	miR-21 and TTF-1	4 h	1 × 106 exosomes/mL	30 μL of exosomes derived from 30 μL of serum	yes	no	Yang et al.30
tethered cationic lipid-polymer hybrid nanoparticles	pancreatic cancer	Glypican-1 (GPC1) mRNA	2 h	57 550 exosomes/mL	10 μL of serum	no	no	Hu et al.33
surface acoustic wave lysis and ion-exchange nanomembrane sensing microfluidic biochip	pancreatic cancer	miR-550	1.5 h	2 pM miRNAs	100 μL of cell culture medium	no	no	Taller et al.34
	liver cancer	miR-21	30 min	1 pM miRNAs	20 μL of plasma	no	no	Ramshani et al.35
microfluidic exponential rolling circle amplification (MERCA)	glioblastoma and lung cancer	miR-21 and let-7a	>4 h	5–8 zmol miRNAs, 2 × 106 exosomes	10 μL of cell-derived exosome lysate	yes	no	Cao et al.36
immunomagnetic exosomal RNA (iMER) microfluidic biochip	glioblastoma	O6-methyl-guanine DNA methyltransferase (MGMT) mRNA and alkylpurine-DNA-N-glycosylase (APNG) mRNA	2 h	108 exosome/mL spiked in human serum	100 μL of serum	no	no	Shao et al.37
microwell-patterned microfluidic digital analysis biochip	Ewing sarcoma (EWS)	GAPDH mRNA and EWS-FLI1 mRNA	>4 h	20 aM RNAs	RNA extracted from cell-derived exosomes	yes	yes	Zhang et al.38
surface-enhanced Raman scattering	breast cancer	miR-21, miR-222, and miR-200c	>20 h	1 aM miRNAs	RNA extracted from cell-derived exosomes	yes	yes	Lee et al.39
	pancreatic cancer	miR-10b	>110 min	1 aM miRNAs	RNA extracted from plasma-derived exosomes	yes	yes	Pang et al.40
mCLN assay	lung cancer	miR-21 and TTF-1 mRNA	10 min	miR-21, 2.06 × 109 exosomes/mL; TTF-1 mRNA, 3.71 × 109 exosomes/mL.	100 μL of exosomes derived from 30 μL of serum	yes	no	this work

^aAssay time does not include the time used for exosome isolation and RNA extraction.

We optimized the CLP-MBs/exosomes number ratio and the number of micromixer units to achieve the best sensing performance (Figure 4). With exosomes derived from A549 cells, the CLP-MBs/exosomes number ratio of 1:5 and the micromixer unit number of 10 provided the highest sensitivity in detecting both exosomal miR-21 and exosomal TTF-1 mRNA; therefore, these two parameters were used in our following studies. A serial dilution of A549-cell-derived exosomes was used to determine the LOD and LOQ of the mCLN assay (Figure 5). For exosomal miR-21, the LOD and LOQ were 2.06 × 10⁹ and 6.86 × 10⁹ exosomes/mL respectively. For exosomal TTF-1 mRNA, the LOD and LOQ were 3.71 × 10⁹ and 1.24 × 10¹⁰ exosomes/mL, respectively. Given that the typical exosome concentration in human serum samples is about 5 × 10¹² exosomes/mL (Table S1),^27,30,56 the mCLN assay may require as low as 1 μL of serum for exosomal RNA detection.

The diagnostic performance of the mCLN assay was first evaluated using cell-derived exosomes. The mCLN assay detected 152.9-fold and 22.6-fold higher expressions of miR-21 and TTF-1 mRNA in A549-cell-derived exosomes than those from BEAS-2B cells, respectively, successfully distinguishing NSCLC from the normal control (Figure 6a). On the contrary, the bulk mixing method did not generate any detectable signals within the same 10 min assay time (Figure 6b). The conventional qRT-PCR assay detected less difference in the expression of exosomal miR-21 between A549 cells and BEAS-2B cells. These results demonstrated that the mCLN assay is a more sensitive assay than the conventional bulk mixing method and the qRT-PCR assay.

The diagnostic performance of the mCLN assay was then investigated using serum samples from NSCLC patients and normal controls. The mCLN assay detected 1530.5-fold and 6.1-fold higher expressions of exosomal miR-21 and TTF-1 mRNA in exosomes from NSCLC patients than those from normal controls, successfully distinguishing the NSCLC group from the normal control group (Figure 7a,b). The levels of exosomal miR-21 were also measured via qRT-PCR assay. The exosomal miR-21 expression of NSCLC patients was 1.4-fold higher than that of normal controls; however, no significant difference was observed between these two groups (Figure 7c). The mCLN assay showed a higher diagnostic accuracy than the qRT-PCR assay. This may be because the mCLN assay more effectively distinguished TEXs from non-TEXs (Figure 7d), while the RNA extraction step in the qRT-PCR assay mixed RNAs from TEXs with those from non-TEXs (Figure 1b), making it less sensitive in detecting TEX-associated RNA biomarkers and in distinguishing NSCLC from normal controls.

CONCLUSIONS

A mCLN assay was developed to allow for the effective capture of exosomes by CLP-MBs in a micromixer biochip and to enable the ultrafast and sensitive detection of exosomal RNAs within 10 min. The mCLN assay was much faster than conventional methods, such as qRT-PCR and newly developed technologies. The potential clinical application of the mCLN assay was demonstrated in lung cancer diagnosis. The mCLN assay successfully distinguished A549 cells from BEAS-2B cells and NSCLC patients from normal controls using exosomal miR-21 and TTF-1 mRNA as the biomarkers. The mCLN assay showed a much higher detection sensitivity than the bulk mixing method and the gold standard qRT-PCR assay. In the future, we will further optimize the design and performance of the mCLN assay. We will integrate the exosome isolation function in the micromixer biochip to enable one stop exosome isolation and exosomal RNA detection. We will optimize the structure of the micromixer units to improve the mixing efficiency and thus the sensing sensitivity. Multichannel design will be used to achieve multiplex sensing. We will evaluate the clinical utility of the mCLN assay in lung cancer diagnosis using a larger cohort of human subjects, including normal controls, smokers at high risk of lung cancer, and early stage and late stage NSCLC patients. As different MBs can be easily encapsulated in the CLPs, we will also explore the applications of the mCLN assay in detecting different exosomal RNA biomarkers for other types of cancer, such as breast cancer and colorectal cancer. Since viruses, such as coronavirus COVID-19, share some similarities with exosomes, we may also explore the use of the mCLN assay to detect viral RNAs and develop it into an ultrafast diagnostic assay for infectious diseases. We hope to develop the mCLN assay into a sensitive, ultrafast, low sample consumption, and cost-effective in vitro diagnostic test that broadly assists with the clinical diagnosis of various diseases such as cancer and infectious diseases.

MATERIALS AND METHODS

Materials.

1,2-Di-O-octadecenyl-3-trimethylammonium propane (DOTMA, 890898) was purchased from Avanti Polar Lipids (Alabaster, AL). Cholesterol (C3045–5G) and d-α-tocopherol polyethylene glycol 1000 succinate (TPGS, 57668–5G) were purchased from Sigma-Aldrich (St. Louis, MO). Sucrose (97061–426) was purchased from VWR (Philadelphia, PA). Rhodamine B (AC29657–0250) was purchased from Fisher Scientific (Waltham, MA). Molecular beacons (MBs) detecting miR-21 and TTF-1 mRNA were synthesized by Sigma-Aldrich. The sequence of FAM-miR-21 MB was 5′-[6FAM]-CGCGA TC-[+T]CA[+A]CA[+T]CA[+G]TC[+T]GA[+T]AA[+G]CTA-GATCGCG-[BHQ1]-3′.^25,27,30 The sequence of Cy5-TTF-1 MB was 5′-[cyanine5]-CGCGATC-[+C]TA[+G]GC[+A]TT[+T]AG[+T]CC[+A]AC[+T]TT-GATCGCG[BHQ3]-3′.^25,30 [+N] represented locked nucleic acid (LNA) bases, which have a methylene bridge bond that links the 2′-O to the 4′-C of the ribose ring and “locks” the ribose ring in the ideal conformation for Watson–Crick binding. Compared with unmodified bases, LNA bases provide an increased binding affinity for their complementary strands, especially for short RNA targets such as microRNAs.

Fabrication of the Micromixer Biochip.

The micromixer biochip was fabricated by assembling five layers of structures: top and bottom cover layers, top and bottom micromixer layers, and a middle splitter layer. The top and bottom cover layers were fabricated using a 1/4” transparent acrylic board (Plaskolite, Columbus, OH). The top cover layer had two holes as the inlets and one hole as the outlet. The holes were cut by using a laser engrave machine (Rayjet 300, rayjetlaser). The top and bottom micromixer layers were mirror image to each other and were fabricated by using the same laser engrave machine on a 1/8” transparent acrylic board (Plaskolite). The splitter layer played a key role in separating and recombining substreams. The splitter layer needed to be thin enough to avoid generating too much resistance to the flow and undermining the mixing performance. However, if the layer is too thin, it is difficult to engrave the structure by a laser cut (PMMA film will be melted and lose geometry accuracy). A 1/32” transparent acrylic film was used for the splitter layer. The SAR structures in the splitter layer were cut by using the same laser engrave machine. For better assembly, alignment pin holes were also cut on each layer. In the final assembly process, the five layers were stacked and aligned by using metal pins. Between each layer, a thin layer of isopropyl alcohol (IPA) was preapplied. The assembled device was then clamped between two glass plates and submerged into IPA at 70 °C for 30 s. After being cooled down to room temperature in air, the device was boned and ready to use.

Cell Culture.

Human A549 NSCLC cells and BEAS-2B normal bronchial epithelial cells were obtained from American Type Culture Collection (ATCC, Manassas, VA). Cells were cultured in 10 mL of RPMI 1640 media (11875093, Thermo Fisher Scientific, Grand Island, NY) containing 10% exosome-depleted fetal bovine serum (FBS, 26140079, Thermo Fisher Scientific) and 1× penicillin–streptomycin (PS, 15140122, Thermo Fisher Scientific) in P100 Petri-dishes. The exosome-depleted FBS was prepared by centrifuging FBS at 100 000g for 3 h at 4 °C to remove the exosomes. A549 and BEAS-2B cells were seeded at densities of 1.5 × 10⁶ and 1.0 × 10⁶ cells per dish, respectively. The cells were subcultured every 2 days.

Isolation of Exosomes from Cell Culture Medium.

A549 and BEAS-2B cells were seeded in 10 mL of RPMI 1640 media containing 10% exosome-depleted FBS and 1× PS at densities of 1.5 × 10⁶ and 1.0 × 10⁶ cells per dish, respectively. At 2 days post seeding, the cell culture conditioned media were centrifuged at 4000g at 4 °C for 30 min and then at 10 000g at 4 °C for 1 h to remove cells and cell debris, respectively. The supernatant was collected, aliquoted, and stored at −80 °C. To prepare the blank control, RPMI 1640 media containing 10% exosome-depleted FBS and 1× PS were added to empty P100 Petri dishes with no cells, incubated in the incubator for 2 days, and processed following the same centrifugation process as the cell culture conditioned media.

To isolate exosomes from cell culture media, a total exosome isolation (from cell culture medium) kit (4478359, Thermo Fisher Scientific) was added into the cell culture media at a volume ratio of 1:2. The mixture was incubated at 4 °C overnight and then centrifuged at 10 000g for 1 h at 4 °C to precipitate exosomes. Finally, exosomes were resuspended in 1× PBS for further analysis. Exosomes from every 1 mL of cell culture medium were resuspended in 20 μL of 1× PBS.

Preparation of CLP-MBs.

The preparation of CLP-MBs followed the method used in our previous studies to ensure a high encapsulation efficiency.^25,27,30 DOTMA, cholesterol, and TPGS were mixed at a DOTMA/cholesterol/TPGS molar ratio of 49:49:2 in ethanol to get a lipid mixture at a concentration of 10 mg/mL. FAM-miR-21 MBs and Cy5-TTF-1 MBs in 1× PBS were mixed at a mass ratio of 1:1. Then, the MB mixture was added to the lipid mixture at a lipids to MBs mass ratio of 12.5:1 and a volume ratio of 2:3. Finally the lipids/MBs mixture was injected into 1× PBS at a volume ratio of 1:9 and sonicated at room temperature for 10 min to form CLP-MBs. At a lipids to MBs mass ratio of 12.5:1, the encapsulation efficiency of MBs was >98%; therefore, CLP-MBs were used directly in the following studies without the separation of free MBs.

Human Serum Samples.

Deidentified human serum samples from 5 normal controls and 10 treatment naїve NSCLC patients were obtained from the Data Bank and BioRepository Shared Resource (DBBR) at the Roswell Park Comprehensive Cancer Center (Buffalo, NY). The clinical data of all human subjects are summarized in Table S1. This study was approved by the Institutional Review Boards at both the University at Buffalo and the Roswell Park Comprehensive Cancer Center.

Isolation of Exosomes from Human Serum Samples.

To isolate exosomes from human sera, a total exosome isolation (from serum) kit (4478360, Thermo Fisher Scientific) was added in the serum at an 1:5 volume ratio. The mixture was incubated at 4 °C for 30 min and then centrifuged at 10 000g for 10 min at room temperature to precipitate exosomes. The exosome pellets were resuspended in equal volume of 1× PBS as the volume of input serum for further analysis.

Characterization of Exosomes and CLP-MBs.

The size, size distribution, and number concentration of exosomes and CLP-MBs were measured by nanoparticle tracking analysis (NTA, Nanosight, LM10, Malvern Instruments Ltd., Worcestershire, UK). The exosomes and CLP-MBs were diluted in 1× PBS so that 50–100 nanoparticles were tracked in the field of view of the NTA system. The camera level was set at 14 under the view-capturing mode, the detection threshold was set at 6, and the screen gain was set at 8 for postcapturing video processing. These parameters were kept identical for all measurements. The ζ potentials of exosomes and CLP-MBs were measured by a NanoBrook 90Plus PALS instrument (Brookhaven Instruments Corporation, Holtsville, NY). Exosomes and CLP-MBs at a concentration of 10⁸ particles/mL were used for ζ potential measurements. Cryo-transmission electron microscopy (cryo-TEM) was used to observe the morphology of exosomes from a NSCLC patient serum. Exosomes (5 μL) were added onto lacey carbon coated copper grids (400 mesh, Pacific Grid Tech, San Francisco, CA). After the exosomes were snap-frozen in liquid ethane, the morphology of exosomes was examined using an FEI Tecnai G2 F20 ST TEM (FEI, Hillsboro, OR). The quality of the exosomes was evaluated using Exo-Check exosome antibody array (EXORAY200A; SBI System Biosciences, Palo Alto, CA). The Exo-Check exosome antibody array included nine blotting spots targeting eight common exosomal markers (CD63, CD81, ALIX, FLOT1, ICAM1, EpCam, ANXA5, and TSG101) and one intercellular cis-Golgi marker (GM130) to confirm the existence of exosome and monitor potential cellular contamination.

Detection of Exosomal RNAs by the mCLN Assay.

Exosomes (100 μL) and CLP-MBs (100 μL) were fed into the micromixer biochip through two inlets at a flow rate of 6 mL/h. The mixture of exosomes and CLP-MBs was collected at the outlet. The fluorescence intensity of the mixture (I_{exosome/CLP-MB}) was measured immediately via a microplate reader (Infinite F200, Tecan Group Ltd., Männedorf, Switzerland). PBS solution containing no exosomes was also flowed through the micromixer biochip with CLP-MBs, and the background fluorescence signal (I_PBS/CLP-MB) was used to calculate the expression of exosomal RNAs, as shown in eq 1.

Detection of Exosomal RNAs by RNA Isolation–qRT-PCR Workflow.

Exosomes were isolated from 100 μL serum samples using the total exosome isolation (from serum) kit as described above. Total RNA was isolated from exosomes using a total RNA purification kit (17200, Norgen Biotek, Thorold, ON, Canada) following the manufacturer’s protocol. Then, a universal cDNA synthesis kit II (203301, Exiqon) was used to reverse transcribe total RNA into cDNA. The qPCR amplification of cDNA was performed using the hsa-miR-21–5p LNA PCR primer set (204230, Exiqon, Woburn, MA) and ExiLENT SYBR Green master mix (203421, Exiqon) on a Light Cycler 480 instrument (Roche, Indianapolis, IN). For all measurements, miR-191 (hsa-miR-191–5p LNA PCR primer set, 204306, Exiqon) was used as the endogenous control in the corresponding samples.^57–59 The expression of miR-21 was normalized to miR-191 and reported by miR-191-normalized Cq values.