An Immuno-Biochip Selectively Captures Tumor-Derived Exosomes and Detects Exosomal RNAs for Cancer Diagnosis

Abstract

Tumor-derived exosomes (TEXs) play instrumental roles in tumor growth, angiogenesis, immune modulation, metastasis, and drug resistance. TEX RNAs are a new class of noninvasive biomarkers for cancer. Neither current techniques, such as quantitative reverse transcription polymerase chain reaction (qRT-PCR) and next-generation sequencing, nor new ones, such as electrochemical or surface plasmon resonance-based biosensors, are able to selectively capture and separate TEXs from normal cell-derived exosomes, making TEX RNAs potentially less sensitive biomarkers. We developed an immuno-biochip that selectively captures TEXs using antibodies against tumor-associated proteins and quantifies in situ TEX RNAs using cationic lipoplexes containing molecular beacons. We used the immuno-biochip to measure the expression of miR-21 microRNA and TTF-1 mRNA in EGFR- or PD-L1-bearing exosomes from human sera and achieved absolute sensitivity and specificity in distinguishing normal controls from non-small cell lung cancer patients. Our results demonstrated that the effective separation of TEXs from other exosomes greatly improved the detection sensitivity and specificity. Compared with the traditional immunomagnetic separation–RNA isolation–qRT-PCR workflow, the immuno-biochip showed superior lung cancer diagnostic performance, consumed less samples (~30 μL), and shortened assay time from ~24 to 4 h.

Graphical Abtract

INTRODUCTION

Exosomes are nanovesicles (diameter ranging from 50 to 150 nm) secreted by cells into extracellular environments during cellular exocytosis.^1,2 They are present in all types of body fluids, such as blood, urine, breast milk, and saliva. Exosomes play important roles in cell–cell communication by transferring many types of biomolecules (protein, DNA, RNA, and lipid) between cells.³ Emerging evidence has shown that tumor-derived exosomes (TEXs) contribute to tumorigenesis and regulate tumor growth, angiogenesis, immune modulation, metastasis, and drug resistance.^4–6 They therefore represent new and potent biomarkers for cancer liquid biopsy. Among the various types of cargoes carried by exosomes, RNAs, especially microRNAs (miRs), have been demonstrated as promising and new biomarkers in cancer screening, diagnosis, and prognosis.^7–11 For example, exosomal miR-21, miR-24, let-7b, and let-7e distinguished non-small cell lung cancer (NSCLC) patients from healthy controls with a sensitivity of 80.25% and a specificity of 92.31%.¹² Exosomal miR-17–5p, miR-93–5p, miR-130a-3p, and miR-340–5p were potential biomarkers to detect breast cancer recurrence.¹³

Exosomal RNAs are typically quantified by conventional methods such as quantitative reverse transcription polymerase chain reaction (qRT-PCR), microarray, and next-generation sequencing. These methods rely on an RNA isolation procedure that extracts all RNAs from a sample of body fluid. Exosomal RNA measurements obtained by the conventional methods are not specific for TEX RNAs because the same RNAs are also physiologically released into body fluids by many nontumor cells. This makes the traditional methods potentially less sensitive for TEX RNA biomarker applications. Moreover, these methods are tedious, time-consuming, expensive, and thus difficult to use in clinical settings. To overcome these limitations, many biosensors and analytical approaches have been developed for simple, rapid, sensitive, and cost-effective characterization of exosomal RNAs. For example, electrochemical biosensors have been developed to quantify exosomal miR-143, miR-146a, and miR-21 for early detection of cancer.^14–16 Molecular beacon (MB)-based fluorescence detection methods have been used to characterize exosomal miRs and mRNAs for the diagnosis of breast cancer,^17,18 lung cancer,^19,20 and pancreatic cancer.²¹ Joshi et al. have developed a localized surface plasmon resonance (LSRP) biosensor to detect exosomal miR-10b for pancreatic cancer diagnosis.²² A microfluidic biochip that lyses exosomes by surface acoustic waves and quantifies exosomal miRs by an ion-exchange nanomembrane has been described by Taller et al.²³ Although these biosensors have simplified the detection procedures, many of them still require the RNA extraction process, and none of them is able to selectively capture TEXs from normal cell-derived exosomes.

In this study, we have developed a novel and simple immuno-biochip that enriches TEXs by targeting tumor-associated protein biomarkers via surface-tethered antibodies and then in situ detects exosomal RNAs using cationic lipoplexes containing RNA target-sensing molecular beacons (CLP-MBs). The exosomal RNA detection relies on the fusion of positively charged CLP-MBs with negatively charged exosomes through electrostatic interaction, which allows the mixing of MBs with RNAs and restores fluorescence signals from the MBs. Total internal reflection fluorescence (TIRF) microscopy is used to measure the MB fluorescence intensity, which indicates the exosomal level of a target RNA.

We have selected lung cancer as the disease model to demonstrate the clinical utility of the immuno-biochip for early detection of cancer. Lung cancer holds high mortality and high prevalence among all cancers. The 5 year survival rate of lung cancer patients plummets from 55% for early stage lung cancer to 4% for late stage lung cancer.²⁴ Early detection is the only known intervention to significantly reduce the mortality and improve outcomes for patients with lung cancer. Annual low-dose computed tomography (CT) is the recommended screening test for high-risk individuals.^25–27 Although it reduces lung cancer mortality by 20% in high-risk smokers, the test has a false positive rate of >95%, causing unnecessary invasive diagnostic biopsy and repeated irradiation exposure.^28–30 Detection of exosomal RNAs via our immuno-biochip may be used as a patient-friendly, minimally invasive confirmatory test to reduce the false positive rate of low-dose CT and to justify its benefit. We have chosen the epidermal growth factor receptor (EGFR) and programmed death ligand-1 (PD-L1), two proteins overexpressed in lung cancer tumors,^31–35 as the selection markers to capture lung cancer TEXs. We have then quantified the expression of miR-21 and thyroid transcription factor-1 (TTF-1) mRNA in the EGFR- or PD-L1-bearing serum exosomes to examine the value of these biomarkers for early detection of lung cancer. We have shown that the immuno-biochip-based TEX RNA measurements had much higher sensitivity and specificity for lung cancer diagnosis than their whole serum measurements. We have also demonstrated that the immuno-biochip has a better sensing performance in detecting TEX RNAs than the existing technology, that is, immunomagnetic separation–RNA isolation–qRT-PCR (IMS-PCR) workflow.

RESULTS AND DISCUSSION

Detection Mechanism of Immuno-Biochip.

The immuno-biochip is able to selectively capture TEXs based on tumor-associated proteins and characterize in situ the intravesicular RNA biomarkers of captured TEXs. As shown in Figure 1, to prepare the biochip, a cover glass was first coated with a 15 nm gold film in sensing areas. A poly-(dimethylsiloxane) (PDMS) layer was put on the cover glass to form the sample wells (diameter of 4 mm). Then, a mixture of PEG200-SH and biotin-PEG1000-SH was applied to generate the poly(ethylene glycol) (PEG) layer on the surface. After washing off unbound PEG, avidin and biotinylated antibodies were sequentially applied to modify the biochip surface with antibodies that recognize tumor-associated proteins through biotin–avidin interaction. In this study, anti-EGFR or anti-PD-L1 antibodies were tethered on the biochip surface to capture EGFR- or PD-L1-bearing exosomes, that is, EGFR+ or PD-L1+ exosomes, respectively. After washing off unbound exosomes, positively charged CLP-MBs were added to fuse with negatively charged exosomes through electrostatic interaction, which allowed the binding of MBs with exosomal RNAs. MBs are hairpin-shaped oligonucleotides with a fluorophore initially quenched by a quencher. The binding of target RNAs to MBs results in the separation of the fluorophore from the quencher and thus restores the fluorescence from MBs. The fluorescence signals, recorded by TIRF microscopy, provided the expression levels of target RNAs in exosomes (Figure 1 and Figure S1 in Supporting Information).

Figure 1.

Schematics of detection procedures of immuno-biochip and conventional IMS-PCR workflow. In the immuno-biochip, tumor-derived exosomes (TEXs) are captured by antibodies against tumor-associated proteins. CLP-MBs are fused with TEXs through electrostatic interaction, allowing the binding of MBs to exosomal RNAs. The restored fluorescence from MBs is analyzed to determine the levels of TEX RNA biomarkers. For the IMS-PCR workflow, TEXs are isolated by magnetic beads conjugated with antibodies against tumor-associated proteins. Then, RNAs are extracted from TEXs, and the expression of TEX RNA biomarkers is quantified by qRT-PCR. The immuno-biochip offers a simpler detection procedure, shorter assay time (4 vs 24 h), and less sample consumption (30 vs 100 μL) than IMS-PCR.

The expression of RNAs in EGFR+ or PD-L1+ exosomes can also be measured using the existing IMS-PCR workflow. Briefly, EGFR+ or PD-L1+ exosomes are first separated from all exosomes using magnetic beads conjugated with anti-EGFR or anti-PD-L1 antibodies. The total RNA is then extracted from the captured exosomes. Finally, the expression of exosomal RNAs is quantified by qRT-PCR. Compared with the IMS-PCR workflow, the immuno-biochip consumes less amount of sample (30 vs 100 μL) and shortens assay time from about 24 to 4 h.

Characterization of Exosomes by Nanoparticle Tracking Analysis (NTA).

The size, size distribution, and number concentration of exosomes isolated from cell culture-conditioned medium and human serum samples were characterized by NTA. Figure 2 shows representative size distributions of exosomes isolated from the cell culture medium of A549 human NSCLC cells and from the serum of an early stage NSCLC patient. The size of exosomes in both cell culture medium and human serum samples ranged from 50 to 150 nm. The serum exosome number concentration ranged from 10¹² to 10¹³ exosomes/mL (Table S1 in Supporting Information).

Figure 2.

Size distributions and images of exosomes. Exosomes isolated from (a) A549 cell culture-conditioned medium (100× dilution) and (b) serum of an early stage lung cancer patient (10,000× dilution) were examined by NTA.

Characterization of Exosome Capture and Exosome-CLP-MB Fusion by Scanning Electron Microscopy (SEM).

SEM was used to visualize the capture of exosomes by antibodies and the fusion of exosomes with CLP-MBs. The surface of the immuno-biochip was first modified with anti-EGFR antibodies. Then, the exosomes isolated from the A549 cell culture-conditioned medium were applied at a concentration of 10¹⁰ exosomes/mL. After washing off unbound exosomes, CLP-MBs were added to detect the levels of EGFR+ exosomal miR-21. SEM was used to visualize the EGFR+ exosomes captured on the biochip before and after the fusion of CLP-MBs. As shown in Figure 3a, the EGFR+ exosomes showed a single-particle morphology on the biochip surface. The measured particle sizes were between 50 and 100 nm, which agreed well with the diameter range typically observed for exosomes. After the addition of CLP-MBs, we observed nanoparticles with larger sizes (ranging from 100 to 150 nm) and clusters of nanoparticles on the biochip surface, indicating the successful fusion of CLP-MBs with EGFR+ exosomes (Figure 3b).

Figure 3.

SEM characterization of exosomes on the immuno-biochip. (a) EGFR+ exosome capture by anti-EGFR antibodies and (b) the fusion of EGFR+ exosomes with CLP-MBs on the immuno-biochip.

Immuno-Biochip Quantifies EGFR+ or PD-L1+ Exosomal RNAs in Cell Culture-Conditioned Medium.

Cell-derived exosomes were first used to examine the capability of the immuno-biochip in capturing EGFR+ and PD-L1+ exosomes and quantifying the intravesicular RNA levels in these exosome subpopulations. The immuno-biochip was modified with anti-EGFR or anti-PD-L1 antibodies. Exosomes from conditioned culture media of A549 NSCLC cells or BEAS-2B human normal bronchial epithelial cells were used on the biochip at 10¹⁰ exosomes/mL. After the unbound exosomes were removed by phosphate buffered saline (PBS) washing, CLPs containing both miR-21-sensing MBs (MB-miR-21) and TTF-1-mRNA-sensing MBs (MB-TTF-1) were applied and incubated with captured exosomes at room temperature (RT) for 1 h. TIRF microscopy was used to record the restored fluorescence signals from MB-miR-21 and MB-TTF-1. The immuno-biochip successfully detected miR-21 and TTF-1 mRNA in both EGFR+ exosomes and PD-L1+ exosomes (Figure 4a). Significantly higher levels of miR-21 and TTF-1 mRNA were observed in EGFR+ or PD-L1+ exosomes from A549 cells than those from BEAS-2B cells, demonstrating that EGFR+ or PD-L1+ exosomal miR-21 and TTF-1 mRNA are promising biomarkers to distinguish lung cancer cells from normal cells (Figure 4b, c). In A549 EGFR+ exosomes, the levels of miR-21 and TTF-1 mRNA were 1.6-fold and 2.8-fold higher than those in BEAS-2B EGFR+ exosomes, respectively. Meanwhile, PD-L1+ exosomes derived from A549 cells had 5.3-fold and 5.9-fold higher levels of miR-21 and TTF-1 mRNA, respectively, than those from BEAS-2B cells. These results suggested that PD-L1+ exosomal miR-21 and TTF-1 mRNA have better cancer-distinguishing performance than EGFR+ exosomes. In addition, for both A549 cells and BEAS-2B cells, we observed higher levels of miR-21 but lower levels of TTF-1 mRNA in EGFR+ exosomes than PD-L1+ exosomes, suggesting that the RNA cargoes may have different distribution profiles in different exosome subpopulations. Recent studies indicated that exosome subpopulations have distinct morphologies and compositions of lipids, RNAs, and proteins.^36–38 Future studies may be required to investigate exosome heterogeneity and how it influences the diagnostic value of exosome-based biomarkers.

Figure 4.

Quantification of EGFR+ or PD-L1+ exosomal miR-21 and TTF-1 mRNA expression in cell culture-conditioned media by immuno-biochip. (a) A representative set of TIRF microscopy images of EGFR+ or PD-L1+ exosomal miR-21 and TTF-1 mRNA from A549 and BEAS-2B cells. (b, c) Higher levels of miR-21 and TTF-1 mRNA were observed in both EGFR+ exosomes and PD-L1+ exosomes from A549 cancer cells than those from BEAS-2B normal cells. (d) Expression of EGFR+ exosomal miR-21 from A549 cells at exosome concentrations of 10⁵, 10⁶, 10⁷, 10⁸, 10⁹, and 10¹⁰ exosomes/mL. The detection sensitivity of the immuno-biochip was 10⁶ exosomes/mL. (e) Immuno-biochip showed higher sensitivity than the conventional IMS-PCR workflow in detecting EGFR+ exosomal miR-21 from A549 cells (n = 3; *: p < 0.05; **: p < 0.01; ***: p < 0.001).

To investigate the detection specificity of the immuno-biochip, IgG control antibodies and MBs with scrambled sequences were used as negative controls. Little fluorescence was observed with these negative controls (Figure S2 in Supporting Information), demonstrating the good sensing specificity of the immuno-biochip. To investigate the selectivity of the immuno-biochip in detecting other miRs, we measured the expression levels of two additional miRs, miR-155 and miR-210, in EGFR+ exosomes and PD-L1+ exosomes derived from A549 cells and BEAS-2B cells (Figure S3 in Supporting Information). For EGFR+ exosomes, 2.97-fold higher miR-155 expression was detected in A549 cell-derived exosomes than in BEAS-2B cell-derived exosomes; however, similar miR-210 expression was observed in exosomes from both A549 cells and BEAS-2B cells. For PD-L1+ exosomes, 3.09-fold and 2.66-fold higher levels of miR-155 and miR-210 were observed in exosomes from A549 cells than those from BEAS-2B cells, respectively. Besides, the immuno-biochip detected different expression levels of these RNAs. For examples, for EGFR+ exosomes derived from A549 cells, miR-21 had the highest expression level (~35,000), followed by miR-155 and miR-210 (~12,000), and TTF-1 mRNA showed the lowest expression level (~5000). For PD-L1+ exosomes released by A549 cells, miR-210 showed the highest expression level (~27,000), followed by miR-21 and TTF-1 mRNA (~23,000), and miR-155 had the lowest expression level (~15,000). These results together demonstrated that the immuno-biochip showed good selectivity in measuring the expression of various exosomal RNAs.

To compare the detection sensitivity of immuno-biochip and IMS-PCR, the expression of EGFR+ exosomal miR-21 from A549 cells was measured by both methods. Six exosome concentrations were used (10⁵, 10⁶, 10⁷, 10⁸, 10⁹, and 10¹⁰ exosomes/mL). With the immuno-biochip, no significant difference was observed in the expression of EGFR+ exosomal miR-21 when the exosome concentrations were 10⁵ and 10⁶ exosomes/mL (Figure 4d and Figure S4). Increased EGFR+ exosomal miR-21 levels were observed with increasing exosome concentrations from 10⁶ to 10¹⁰ exosomes/mL. These results indicated that the detection sensitivity of immuno-biochip was 10⁶ exosomes/mL bulk concentration, which corresponded to ~2000 exosomes/mm² surface sensitivity. The IMS-PCR workflow was not able to detect EGFR+ exosomal miR-21 at lower exosome concentrations (10⁵, 10⁶, and 10⁷ exosomes/mL). At higher exosome concentrations, although IMS-PCR detected EGFR+ exosomal miR-21 levels, the quantification cycle (C_q) values were relatively high, that is, C_q = 34.42 for 10⁸ exosomes/mL, C_q = 33.30 for 10⁹ exosomes/mL, and C_q = 31.33 for 10¹⁰ exosomes/mL. To allow the comparison of the immuno-biochip and IMS-PCR workflow, results from the concentration of 10⁸ exosomes/mL were used as the normalization factor because these two assays used different units. As shown in Figure 4e, the immuno-biochip detected higher levels of EGFR+ exosomal miR-21 than the IMS-PCR workflow with increasing exosome concentration. These results demonstrated that the immuno-biochip had 100-fold higher sensing sensitivity than the IMS-PCR workflow.

Diagnostic Value of Serum EGFR+ and PD-L1+ Exosomal RNAs Quantified by Immuno-Biochip.

To demonstrate the potential clinical utility of the immuno-biochip in cancer diagnosis, human serum samples from 10 normal controls, 13 early stage (stages I and II), and 7 late stage (stages III and IV) NSCLC patients were analyzed (Table S1 in Supporting Information). Exosomes were pre-enriched from 30 μL serum samples using the total exosome isolation kit and characterized by NTA. We did not observe any significant difference in the size and number concentration of exosomes between normal controls and NSCLC patients (Figure 5a). The expression levels of EGFR+ or PD-L1+ exosomal miR-21 and TTF-1 mRNA were measured using the immuno-biochip. Figure 5b shows representative TIRF microscopy images for one normal control and one NSCLC patient. Significantly higher levels of miR-21 and TTF-1 mRNA were observed in EGFR+ and PD-L1+ exosomes from NSCLC patients’ sera than those from normal controls’ sera, indicating that EGFR+ and PD-L1+ exosomal miR-21 and TTF-1 mRNA are potent serum biomarkers to discriminate NSCLC patients from healthy controls (Figure 5c). Similar levels of EGFR+ exosomal miR-21, PD-L1+ exosomal miR-21, and PD-L1+ exosomal TTF-1 mRNA were observed between early and late stage NSCLC patients. However, higher levels of EGFR+ exosomal TTF-1 mRNA were detected in sera from late stage NSCLC patients than those from early stage patients, indicating that EGFR+ exosomal TTF-1 mRNA may be a biomarker for lung cancer staging.

Figure 5.

Quantification of EGFR+ or PD-L1+ exosomal miR-21 and TTF-1 mRNA expression in human sera by immuno-biochip. (a) No significant difference was found in the size and number concentration of serum-derived exosomes between NSCLC patients and normal controls. (b) Representative sets of TIRF microscopy images generated from human serum-derived exosomes, detecting miR-21 and TTF-1 mRNA in EGFR+ and PD-L1+ exosomes. (c) Results from the immuno-biochip showed elevated levels of miR-21 and TTF-1 mRNA in EGFR+ or PD-L1+ exosomes from NSCLC patient serum samples compared with normal controls (serum sample volume: 30 μL. *: p < 0.05; **: p < 0.01; ***: p < 0.001).

Conventional IMS-PCR workflow was also used to measure the expression of EGFR+ and PD-L1+ exosomal miR-21 and TTF-1 mRNA in the same set of human serum samples (Figure 6). The IMS-PCR workflow used 100 μL serum samples and successfully detected exosomal miR-21 but was not able to detect exosomal TTF-1 mRNA, which may be due to the limited amount of TTF-1 mRNA in EGFR+ or PD-L1+ exosomes. For EGFR+ exosomal miR-21, significantly higher levels were detected in serum samples from early stage NSCLC patients than those of normal controls. However, no significant difference was observed between late stage NSCLC patients and normal controls. For PD-L1+ exosomal miR-21, no significant difference was observed among normal controls, early stage, and late stage NSCLC patients. We also measured the miR-21 expression in whole serum, without isolating exosomes. As shown in Figure 6c, no significant difference in the serum miR-21 expression was observed among normal controls, early stage, and late stage NSCLC patients. On the other hand, EGFR+ exosomal miR-21 levels were significantly different between normal controls and early stage NSCLC patients (p < 0.05), suggesting that the separation of TEXs from other exosomes and the expression of TEX-associated miR-21 instead of whole serum miR-21 may greatly improve the detection sensitivity.

Figure 6.

Quantification of miR-21 in exosomes and whole serum by IMS-PCR workflow. The levels of (a) EGFR+ exosomal miR-21, (b) PD-L1+ exosomal miR-21, and (c) whole serum miR-21 are compared between NSCLC patients and normal controls. EGFR+ exosomal miR-21, instead of whole serum miR-21, distinguished early stage NSCLC patients from normal controls, suggesting that the selection of EGFR+ exosomes improved the detection sensitivity. However, PD-L1+ exosomal miR-21 and serum miR-21 did not discriminate NSCLC patients from normal controls. (serum sample volume: 100 μL. *: p < 0.05).

Receiver operating characteristic (ROC) curves were plotted, and the area under the curve (AUC) values were calculated to investigate the diagnostic accuracy of each biomarker measured using both the immuno-biochip (Figure S5 in Supporting Information) and the IMS-PCR workflow (Figure S6 in Supporting Information). As summarized in Table 1, for each biomarker, the immuno-biochip showed overall higher AUC values and better sensitivity and specificity than the IMS-PCR workflow in distinguishing normal controls from early stage and late stage NSCLC patients. For example, the AUC for EGFR+ exosomal miR-21 measured by immuno-biochip was 0.92 (sensitivity = 0.75, specificity = 1.00), higher than that measured by IMS-PCR (AUC = 0.87, sensitivity = 0.63, specificity = 1.00) (Figure 7).

Table 1.

Sensitivity (SEN), Specificity (SPE), and Area under Curve (AUC) of Each Biomarker and Combined Biomarkers in Lung Cancer Diagnosis

		normal vs NSCLC			normal vs early stage NSCLC			normal vs late stage NSCLC			early stage NSCLC vs late stage NSCLC
assay	marker	SPE	SEN	AUC	SPE	SEN	AUC	SPE	SEN	AUC	SPE	SEN	AUC
immuno-biochip	EGFR+ exosomal miR-21	1.000	0.750	0.920	0.800	1.000	0.954	1.000	0.714	0.857	0.923	0.429	0.571
	EGFR+ exosomal TTF-1	0.900	0.850	0.890	0.900	0.846	0.877	1.000	0.857	0.914	1.000	0.857	0.857
	PD-L1+ exosomal miR-21	0.900	0.900	0.915	0.900	0.846	0.869	1.000	1.000	1.000	0.308	1.000	0.626
	PD-L1+ exosomal TTF-1	0.900	0.8S0	0.920	0.900	0.769	0.885	0.900	1.000	0.986	0.923	0.429	0.681
	biomarker signature	EGFR+ exosomal miR-21			EGFR+ exosomal miR-21			PD-L1+ exosomal miR-21			EGFR+ exosomal TTF-1
		EGFR+ exosomal TTF-1			PD-L1+ exosomal miR-21
		PD-L1+ exosomal miR-21
		1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000	0.857	0.857
	all markers	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000	0.923	0.857	0.890
IMS-PCR	EGFR+ exosomal miR-21	1.000	0.632	0.868	0.875	0.833	0.875	0.750	0.857	0.857	0.417	0.857	0.619
	PD-L1+ exosomal miR-21	0.750	0.737	0.724	0.875	0.750	0.710	0.750	0.571	0.554	0.583	1.000	0.815
	both markers	1.000	0.737	0.928	1.000	1.000	1.000	0.875	0.857	0.893	0.917	0.857	0.857
PCR	serum miR-21	0.900	0.421	0.611	0.900	0.500	0.692	0.700	0.714	0.529	0.833	0.714	0.667

Figure 7.

Diagnostic value of EGFR+ exosomal and whole serum miR-21. ROC curves for EGFR+ exosomal miR-21 measured by (a) the immuno-biochip or (b) IMS-PCR, and whole serum miR-21 measured by (c) PCR for lung cancer diagnosis. The AUC, the detection sensitivity, and the specificity were used to evaluate the diagnostic performance.

We also performed a best subset logistic regression analysis where the Akaike information criterion was used to identify the best biomarker subset, that is, biomarker signatures, for lung cancer. As shown in Table 1 and Figure S5 in Supporting Information, for the immuno-biochip, the biomarker signature that distinguishes normal controls from all NSCLC patients is EGFR+ exosomal miR-21, EGFR+ exosomal TTF-1 mRNA, and PD-L1+ exosomal miR-21 combined biomarkers (AUC = 1.00, sensitivity = 1.00, specificity = 1.00). The biomarker signature that discriminates normal controls from early stage NSCLC patients is EGFR+ exosomal miR-21 and PD-L1+ exosomal miR-21 combined biomarkers (AUC = 1.00, sensitivity = 1.00, specificity = 1.00). The biomarker signature for differing normal controls from late stage NSCLC patients is PD-L1+ exosomal miR-21 alone (AUC = 1.00, sensitivity = 1.00, specificity = 1.00). The biomarker signature that distinguishes early stage NSCLC from late stage NSCLC is EGFR+ exosomal TTF-1 mRNA (AUC = 0.86, sensitivity = 0.86, specificity = 1.00). Results from the immuno-biochip also agreed well with those from the IMS-PCR workflow. When the EGFR+ exosomal miR-21 and PD-L1+ exosomal miR-21 combined biomarkers were used, both assays achieved the same detection sensitivity (1.000) and specificity (1.000) in distinguishing normal controls from early stage NSCLC patients. However, the IMS-PCR workflow is much more labor-intensive and time-consuming.

Finally, we note that results from ROC analysis also support that the quantification of TEX-associated miR-21 instead of whole serum miR-21 further improves the diagnostic accuracy (Table 1, Figure 7, and Figure S7 in Supporting Information). Compared with whole serum miR-21, EGFR+ and PD-L1+ exosomal miR-21 measured by both the immuno-biochip and IMS-PCR workflow showed higher AUC values in differing both early and late stage NSCLC patients from normal controls.

TEXs play active and instrumental roles in cancer progression, immune regulation, metastasis, and drug resistance.^4–6 They are present in body fluids in relatively large quantities and remain remarkably stable for months and even years when stored at −80 °C.^39–41 These merits make TEXs promising biomarkers for cancer diagnosis. However, a major challenge for developing exosomal cancer biomarkers is the separation of TEXs from those released by non-tumor cells, the latter of which can render the test less sensitive or incapable of detecting TEX biomarkers. To overcome these challenges, we have developed the immuno-biochip, which captures and separates TEXs using tumor-associated protein biomarkers and then uses CLP-MBs to quantify the expression of TEX RNAs on the same platform (Figure 1). Although many biochips have been developed to detect exosomal RNAs, such as electrochemical biosensors,^14–16 LSRP biosensors,²² tethered cationic nanoparticle biosensors,^20,21 and microfluidic chips,²³ none of them are able to effectively enrich TEXs from non-TEXs.

In this study, we have demonstrated the potential clinical application of the immuno-biochip in lung cancer diagnosis. EGFR and PD-L1 were used as the selection biomarkers to enrich lung cancer TEXs over non-TEXs. EGFR is overexpressed in >50% lung cancer and plays significant roles in regulating cell proliferation and survival.^31–33 Exosomal EGFR has been identified as a promising biomarker in lung cancer diagnosis^42,43 and prognosis.⁴⁴ PD-L1 is a key surface ligand in the programmed cell death-1/programmed death ligand-1 (PD-1/PD-L1) signaling pathway, which is a typical pathway for immune system activation against tumorigenesis. The overexpression of PD-L1 in tumor cells could suppress the function of T lymphocytes, and high prevalence of PD-L1 expression has been found among lung cancer patients.^34,35 Exosomal PD-L1 mRNA levels are found to be correlated with the response to anti-PD-1 immune checkpoint inhibitors, nivolumab and pembrolizumab, in NSCLC patients and melanoma patients.⁴⁵ By targeting these two protein markers, we aimed to capture TEXs from the total exosomes. The RNA markers we measured in this study were miR-21 and TTF-1 mRNA. Recent meta-analysis reported that miR-21 is one of the potentially effective biomarkers for lung cancer diagnosis, which holds high distinguishing sensitivity and specificity.^9,46–50 TTF-1 mRNA is selectively expressed in bronchiolar epithelial cells and has been proved to be related with tumorigeneses in lung tissue.^51,52 By measuring these two RNA markers in EGFR+ exosomes and PD-L1+ exosomes, we expected to detect lung cancer with high sensitivity and specificity.

We first used SEM to confirm the capture of TEXs by antibodies (Figure 3a) and the fusion of TEXs with CLP-MBs (Figure 3b). Then, we evaluated the diagnostic performance of the immuno-biochip using exosomes derived from A549 NSCLC cells and BEAS-2B normal lung bronchial epithelial cells. The immuno-biochip detected significantly higher EGFR + and PD-L1+ exosomal miR-21 and TTF-1 mRNA in A549 cell-derived exosomes than in BEAS-2B cell-derived exosomes, demonstrating its capability in distinguishing lung cancer from normal control (Figure 4a,b,c). Compared with the IMS-PCR workflow, the immuno-biochip also showed 100-fold higher detection sensitivity (Figure 4d,e).

Finally, we investigated the clinical utility of the immuno-biochip in lung cancer diagnosis using human serum samples from normal controls, early stage, and late stage NSCLC patients. All four biomarkers, that is, EGFR+ exosomal miR-21, EGFR+ exosomal TTF-1, PD-L1+ exosomal miR-21, and PD-L1+ exosomal TTF-1, showed very high sensitivity and specificity in distinguishing normal controls from early stage and late stage NSCLC patients (Figure 5 and Table 1). The EGFR+ exosomal miR-21 and PD-L1+ exosomal miR-21 combined biomarkers distinguished normal controls from early stage NSCLC patients with absolute sensitivity and specificity. PD-L1+ exosomal miR-21 alone could be used to discriminate normal controls from late stage NSCLC patients with absolute sensitivity and specificity. The four biomarkers were also characterized by the IMS-PCR workflow, but they showed lower sensitivity and specificity in lung cancer detection than the immuno-biochip (Figure 6 and Table 1). The poor diagnostic performance of the IMS-PCR workflow may be caused by the low abundance of miR-21 in EGFR+ exosomes and PD-L1+ exosomes, which was indicated by the high qRT-PCR C_q values (~30–35). However, in the immuno-biochip, TIRF microscopy, which is able to realize single-molecule sensitivity, was used to measure the exosomal RNA signals from captured EGFR+ and PD-L1+ exosomes. Even though the amount of miR-21 and TTF-1 mRNA in EGFR+ and PD-L1+ exosomes is low, TIRF microscopy is still capable of highly sensitive detection of these RNAs, as shown in Figure 5b. Besides, although we selected cancer-overexpressed proteins, that is, EGFR and PD-L1, to selectively capture TEXs from all exosomes, we recognize that normal cell-derived exosomes may also carry EGFR and PD-L1 on the surface and affect the detection sensitivity and specificity. With the immuno-biochip, we can carefully set the cutoff level during the image analysis, effectively remove the weak singles from normal-cell-derived EGFR+ and PD-L1+ exosomes, reduce the noise from non-TEXs, and thus achieve better distinguishing performance. Unfortunately, the IMS-PCR workflow is not able to remove interference from normal-cell-derived EGFR+ and PD-L1+ exosomes, which leads to inferior sensitivity and specificity in lung cancer detection.

CONCLUSIONS

In summary, we have developed a highly sensitive, simple, fast, and cost-effective immuno-biochip to capture TEXs and quantify RNA levels in TEXs for lung cancer diagnosis. The immuno-biochip has shown high detection sensitivity and specificity in distinguishing NSCLC patients from normal controls, which is better than the conventional IMS-PCR workflow. In the future, we will further validate the clinical utility of the immuno-biochip in large cohorts of normal controls, smokers at high risk of lung cancer, and early stage and late stage lung cancer patients. We will confirm the superior diagnostic performance of the immuno-biochip over the IMS-PCR workflow with a large sample size. The immuno-biochip has a user-friendly design, which can be quickly adapted to clinical settings. It is also a universal platform that can be easily modified with different antibodies and molecular beacons for the diagnosis of other types of cancer and diseases, such as cardiovascular diseases and infectious diseases. We expect to develop the immuno-biochip into an accurate, exosome-based in vitro test assisting in medical diagnosis.

EXPERIMENTAL SECTION

Materials.

1,2-Di-O-octadecenyl-3-trimethylammonium propane (DOTMA, 890898) was purchased from Avanti Polar Lipids (Alabaster, AL). 3-Mercaptopropyl trimethoxysilane (3-MPS, 175617–100G), cholesterol (C3045–5G), d-α-tocopherol poly(ethylene glycol) 1000 succinate (TPGS, 57668–5G), and methyl-(PEG)₄-thiol (PEG200, MW 224Da, 26132) were purchased from Sigma-Aldrich (St. Louis, MO). HS-PEG1000-biotin (PEG1000, MW 1000Da, PG2-BNTH-1k) was purchased from Nanocs (Boston, MA). NeutrAvidin protein (31000) was purchased from Thermo Fisher Scientific (Rockford, IL). Biotinylated anti-EGFR antibodies (ab24293) and anti-PD-L1 antibodies (13-5983-82) were purchased from Abcam (Cambridge, MA) and Thermo Fisher Scientific, respectively. Biotinylated mouse IgG1 control antibodies (400103) were purchased from BioLegend (San Diego, CA). MBs were synthesized by Sigma-Aldrich. The sequence of miR-21-sensing MBs (MB-miR-21) is 5′-[6FAM]-CGCGATC-[+T]CA[+A]CA[+T]CA[+G]TC[+T]GA[+T]AA[+G]CTA-GATCGCG-[BHQ1]-3′. The sequence of scrambled MBs for miR-21 sensing (MB-miR-21-scramble) is 5′-[6FAM]-CGCGATC-[+T]CT[+A]CT[+T]CT[+G]TG[+T]GT[+T]AT[+G]CAA-GATCGCG-[BHQ1]-3′. The sequence of TTF-1 mRNA-sensing MBs (MB-TTF-1) is 5′-[Cyanine5]-CGCGATC-[+C]TA[+G]GC[+A]TT[+T]AG[+T]CC[+A]AC[+T]TT-GATCGCG-[BHQ3]-3′. The sequence of scrambled MBs for TTF-1 mRNA sensing (MB-TTF-1-scramble) is 5′-[Cyanine5]-CGCGATC-[+C]TT[+G]GG[+A]TA[+T]AC[+T]CG[+A]AG[+T]TA-GATCGCG-[BHQ3]-3′. The sequence of miR-155 sensing-MBs (MB-miR-155) is 5′-[6FAM]-CGCGATC-[+A]CC[+C]CT[+A]TC[+A]CG[+A]TT[+A]GC[+A]TTAAGATCGCG-[BHQ1]-3′. The sequence of miR-210-sensing MBs (MB-miR-210) is 5′-[Cyanine5]-CGCGATC-[+T]CA[+G]CC[+G]CT[+G]TC[+A]CA[+C]GC[+A]CAG-GATCGCG-[BHQ3]-3′. [+A], [+G], [+C], and [+T] represent locked nucleic acid bases. A SYLGARD 184 silicone elastomer kit (PDMS, 2065622) was purchased from Dow Corning Corp. (Midland, MI).

Cell Culture.

A549 human NSCLC cells and BEAS-2B human normal bronchial epithelial cells were acquired from American Type Culture Collection (ATCC, Manassas, VA). Both cells were cultured in RPMI 1640 medium (11875093, Thermo Fisher Scientific) supplemented with 10% v/v exosome-depleted fetal bovine serum (FBS, 26140079, Thermo Fisher Scientific) and 1× penicillin-streptomycin (PS, 15140122, Thermo Fisher Scientific). To prepare exosome-depleted FBS, the FBS was centrifuged at 100,000g at 4 °C for 3 h to remove the exosomes and then sterilized by filtration through 0.22 μm filters. The A549 and BEAS-2B cells were subcultured every 2–3 days.

Preparation of CLP-MBs.

A lipid mixture was first prepared by mixing DOTMA, cholesterol, and TPGS in ethanol (49:49:2 molar ratio). Separately, MB-miR-21 and MB-TTF-1 were mixed at a mass ratio of 1:1. Then, two mixtures were combined at a lipid/MB mass ratio of 12.5:1 and an ethanol/water volume ratio of 2:3. Finally, one part of the lipid/MB mixture was quickly injected into nine parts of PBS to form CLPs containing both MB-miR-21 and MB-TTF-1 (CLP-MBs). Same method was used to prepare CLPs containing MB-miR-155 and MB-miR-210.

Fabrication of Immuno-Biochip.

A cover glass (12544D, Fisher Scientific, Pittsburgh, PA) was first cleaned with water and 200 proof ethanol through 10 min sonication. The cover glass was dried and placed in 3-MPS vapor for 30 min to form the adhesion layer on the glass surface. A 15 nm Au layer was then deposited on the cover glass surface by electron-beam evaporation (Indel system, Switzerland). The deposition rate was controlled at 0.5 Å/s. Finally, a piece of PDMS mold with wells 4 mm in diameter for sample loading was treated with oxygen plasma and bound onto the Au-coated cover glass.

Surface Modification of Immuno-Biochip.

The PEG mixture was prepared by mixing biotinylated PEG1000 with PEG200 at a molar ratio of 1:3 in PBS. The Au-coated surface of the immuno-biochip was incubated with 10 mM PEG mixture for 1 h at RT. Unbound PEG was carefully washed off with PBS. Then, 50 μg/mL NeutrAvidin was added and incubated for 1 h at RT. After excess NeutrAvidin was washed off with PBS, biotinylated anti-EGFR or anti-PD-L1 antibodies (50 μg/mL in PBS) were added and incubated at 4 °C overnight to conjugate antibodies on the surface of the biochip through the interaction between biotin and avidin. Excess antibodies were removed carefully by PBS washing, and the biochip was stored at 4 °C for further use.

Isolation of Exosomes from Cell Culture-Conditioned Media.

A549 and BEAS-2B cells were cultured for 2 days in RPMI 1640 medium with 10% v/v exosome-depleted FBS and 1× PS. The cell culture medium was collected, and the exosomes were isolated by precipitation using the total exosome isolation (from cell culture medium) kit (4478359, Thermo Fisher Scientific) following manufacturer’s protocol with a minor revision. Briefly, the cell culture-conditioned medium was first centrifuged at 4000g for 30 min to remove all cells. The supernate was collected and centrifuged at 10,000g for 1 h to remove cell debris. Then, the total exosome isolation kit was mixed thoroughly with cell culture medium at a 1:2 volume ratio. After incubation at 4 °C overnight, the mixture was centrifuged at 10,000g for 1 h at 4 °C to precipitate the exosomes. Exosome pellets were resuspended in PBS for future analysis. In this study, exosomes from 1 mL cell culture medium were resuspended in 20 μL PBS.

Human Serum Samples.

Serum samples and relevant clinical data of healthy controls and NSCLC patients before treatment were obtained from the Data Bank and BioRepository Shared Resource (DBBR) at Roswell Park Comprehensive Cancer Center (Buffalo, NY). The NCI-supported DBBR collects and provides de-identified, high-quality serum samples, and associated epidemiological and clinical data. Specimens are procured by trained phlebotomists and centrifuged within 1 h after procurement. The supernate is automatically aliquoted using a MAPI robot from Cryobiosystem and stored in liquid nitrogen. After the serum samples were transferred to the University at Buffalo, they were stored at −80 °C. Approval from Institutional Review Boards (IRBs) of Roswell Park Comprehensive Cancer Center and of the University at Buffalo was obtained for the use of human serum samples for this study. All serum samples had a clear yellow color and did not have distinct oxyhemoglobin absorbance peak at 414 nm, indicating that the serum samples did not have hemolysis.⁵³

Isolation of Exosomes from Human Sera.

Exosomes were isolated from sera using the total exosome isolation (from serum) kit (4478360, Thermo Fisher Scientific) following manufacturer’s protocol. Briefly, the total exosome isolation kit was mixed thoroughly with serum at a 1:5 volume ratio and incubated at 4 °C for 30 min. The exosomes were collected by centrifuging the mixture at 10,000g for 10 min at RT and then resuspended in PBS at a volume equal to that of the input serum.

Characterization of Exosomes by NTA.

The NTA system (Nanosight, LM10, Malvern Instruments, Worcestershire, UK) was used to measure the size, size distribution, and number concentration of exosomes. To maintain the accuracy and consistency of the measurements, exosomes were first diluted with PBS until 50–100 nanoparticles can be tracked in the field of view. The setting of measurement parameters was also identical for all measurements. The camera level was set at 14 during the view-capturing process. The detection threshold was set at 6, and the screen gain was set at 8 during the video processing process.

Detection of Exosomal RNAs by Immuno-Biochip.

Exosomes were applied on the immuno-biochip modified with anti-EGFR or anti-PD-L1 antibodies and incubated at RT for 2 h. Unbounded exosomes were carefully washed off with PBS. The CLP-MBs were then added to the immuno-biochip at MB concentration of 16 μg/mL and incubated with captured exosomes for 1 h at RT. After washing off excess CLP-MBs by PBS, TIRF microscopy (Eclipse Ti-E, Nikon Instruments, Melville, NY) was used to detect the restored fluorescence signals from MBs. The fluorescence signals of miR-21 and miR-155 were observed in the FAM channel (λ_ex = 488 nm and λ_em = 535 nm). The fluorescence signals of TTF-1 mRNA and miR-210 were observed in the Cy5 channel (λ_ex = 640 nm and λ_em = 670 nm). Images were collected for each biomarker on an Andor iXon EMCCD camera (Andor, UK) with a 100× lens, 20% laser power, and exposure time of 200 ms. A total of 100 images were taken for each biomarker, and the levels of exosomal miR-21 and TTF-1 mRNA were determined by the following image analysis method.

Image Analysis.

ImageJ and MATLAB (Natick, MA) were used to measure the fluorescence intensities of FAM and Cy5 from MBs, respectively. Briefly, a signal cutoff level was identified by Image J, and pixels with fluorescence intensity higher than the cutoff level were considered as signals from tumor-derived exosomal RNA. The sum intensity of these pixels, I_signal, was calculated by MATLAB for each image. ImageJ was also used to determine the highest pixel intensity in the dark areas of images as the background signal level. The mean intensity of all pixels with an intensity smaller than the background level, I_background, was calculated as a normalization factor to overcome chip-to-chip variation. The expression of exosomal RNA was calculated as I_signal/I_background. For each biomarker, at least 100 images were collected and analyzed. The mean expression value calculated across the 100 images was used for plotting data and statistical analysis. More details are provided in Figure S1 in Supporting Information.

SEM.

The binding of exosomes with antibodies and the fusion of exosomes with CLP-MBs on the immuno-biochip were characterized using a field emission scanning electron microscope (SU-70, Hitachi High Technologies, Schaumburg, IL) with an energy-dispersive X-ray spectrometer (Oxford Instruments, UK). Briefly, after the capture of exosomes by antibodies and after the fusion of exosomes with CLP-MBs, the biochip was carefully washed with PBS, dried at 4 °C, and surface-coated with carbon. The exosomes and the exosome-CLP-MB complexes were characterized by SEM.

Immuno-Separation of Exosomes by Antibody-Conjugated Magnetic Beads.

Exosomes that bear EGFR and PD-L1, that is, EGFR+ exosomes and PD-L1+ exosomes, were isolated using the total exosome isolation kit and the exosome–streptavidin isolation/detection reagent (Dynabeads magnetic beads, 10608D, Thermo Fisher Scientific) following manufacturer’s protocols. Briefly, the exosomes were first pre-enriched from the cell culture medium or human serum samples using the total exosome isolation kit as described above. A549 exosomes were resuspended in 100 μL of PBS at concentrations ranging from 10⁵ to 10¹⁰ exosomes/mL. Exosomes from 100 μL of serum samples were resuspended in 100 μL PBS. Biotinylated anti-EGFR or anti-PD-L1 antibodies were conjugated onto Dynabeads magnetic beads by incubating 4 μg of biotinylated antibodies with 1 mL bead solution containing 10⁷ Dynabeads magnetic beads. Then 2 × 10⁶ antibody-conjugated magnetic beads were incubated with 100 μL of exosomes at 4 °C overnight with gentle mixing. The bead-bound exosomes were separated from the unbound exosomes with the magnetic separator and resuspended in 100 μL of PBS for further analysis.

RNA Isolation from Sera and Exosomes.

The total RNA was isolated from exosomes using the total RNA purification kit (17200, Norgen Biotek, Canada). To 100 μL of exosomes, 250 μL of the lysis buffer was used. RNA was isolated from the lysate as per manufacturer’s protocol. The total RNA from 0.2 mL of serum was isolated using miRCURY RNA isolation biofluids kit (300112, Exiqon, Woburn, MA) as per manufacturer’s protocol. RNA preparations were 50 μL.

MiR qRT-PCR.

A universal cDNA synthesis II kit (203301, Exiqon) was used to reverse transcribe 15 μL RNA into cDNA. For qPCR of cDNA (3 μL of 1:40 dilution) in duplicate in a volume of 15 μL with thermocycling as per Exiqon’s recommendation, an ExiLENT SYBR Green qPCR master mix (203421, Exiqon) and hsa-miR-21–5p (20423, Exiqon) or hsa-miR-191–5p (204306, Exiqon) LNA PCR primer sets were used in a Light Cycler 480 instrument (Roche, Indianapolis, IN). Using miR-191–5p as the normalizer,^54–56 the average C_q values for miR-21 and miR-191 as determined by the instrument’s second derivative-maximum method were subtracted to obtain miR-21 expression levels in a C_q unit. C_q values are inversely proportional to base-2 logarithm of RNA analyte concentrations.

Statistical Analysis.

For each comparison (normal controls vs early stage NSCLC patients, normal controls vs late stage NSCLC patients, and early stage NSCLC vs late stage NSCLC), binary logistic regression analyses using different biomarkers separately and jointly were performed. Furthermore, the best subset model selection approach and the Akaike information criterion were used to find the best biomarker subset (biomarker signature). For ROC analyses, the pROC package in R software was used.⁵⁷ Cutoffs with the highest sum of sensitivity and specificity were first identified, and AUC, sensitivity, and specificity were determined with these cutoffs. For two-group comparison, standard two-tailed t tests were used, and a p value of <0.05 was used to deem significance.