Exodisc for Rapid, Size-Selective, and Efficient Isolation and Analysis of Nanoscale Extracellular Vesicles from Biological Samples Hyun-Kyung Woo,†,‡ Vijaya Sunkara,†,‡ Juhee Park,⊥ Tae-Hyeong Kim,⊥ Ja-Ryoung Han,† Chi-Ju Kim,†,⊥ Hyun-Il Choi,§ Yoon-Keun Kim,§,∥ and Yoon-Kyoung Cho*,†,⊥ †
Department of Biomedical Engineering, School of Life Sciences, Ulsan National Institute of Science and Technology (UNIST), Ulsan 44919, Republic of Korea § Department of Life Sciences, Pohang University of Science and Technology (POSTECH), Pohang 37673, Republic of Korea ∥ Institute of MD Healthcare, Seoul 03923, Republic of Korea ⊥ Center for Soft and Living Matter, Institute for Basic Science (IBS), Ulsan 44919, Republic of Korea S Supporting Information *
ABSTRACT: Extracellular vesicles (EVs) are cell-derived, nanoscale vesicles that carry nucleic acids and proteins from their cells of origin and show great potential as biomarkers for many diseases, including cancer. Efficient isolation and detection methods are prerequisites for exploiting their use in clinical settings and understanding their physiological functions. Here, we presented a rapid, label-free, and highly sensitive method for EV isolation and quantification using a lab-on-a-disc integrated with two nanofilters (Exodisc). Starting from raw biological samples, such as cell-culture supernatant (CCS) or cancerpatient urine, fully automated enrichment of EVs in the size range of 20−600 nm was achieved within 30 min using a tabletop-sized centrifugal microfluidic system. Quantitative tests using nanoparticle-tracking analysis confirmed that the Exodisc enabled >95% recovery of EVs from CCS. Additionally, analysis of mRNA retrieved from EVs revealed that the Exodisc provided >100-fold higher concentration of mRNA as compared with the gold-standard ultracentrifugation method. Furthermore, on-disc enzyme-linked immunosorbent assay using urinary EVs isolated from bladder cancer patients showed high levels of CD9 and CD81 expression, suggesting that this method may be potentially useful in clinical settings to test urinary EV-based biomarkers for cancer diagnostics. KEYWORDS: extracellular vesicles, lab-on-a-disc, size-based filtration, ELISA, bladder cancer
E
pivotal roles of EVs in intercellular communication. Therefore, facile isolation and analysis of EVs offer great potential for understanding their physiological functions and determining possible roles in disease diagnosis, prognosis, and monitoring of therapeutic responses by means of noninvasive liquid biopsies. Despite the increasing clinical importance of EVs as potential biomarkers, current methods of EV isolation and analysis suffer from complicated procedures with long processing times.17,18 For example, ultracentrifugation (UC) is the most commonly used method for EV isolation and requires time-consuming steps involving centrifugation and acquisition of large sample volumes, and the results suffer from low yield and purity.19−21 To improve the purity of isolated EVs, an additional
xtracellular vesicles (EVs) are lipid-bilayer-enclosed vesicles from 40 to 1000 nm in size that are excreted by most cell types and play diverse roles in intercellular communications.1−5 They are involved in the regulation of normal physiological processes as well as in tumorigenesis and the spread of pathogenic agents.6−10 EVs are prevalent in bodily fluids, such as blood, urine, saliva, cerebrospinal fluid, and ascites, and carry important genetic information stored in nucleic acids, which remain stable inside the vesicles and are protected from degradation due to the lipid bilayer. Additionally, EVs carry a variety of membrane and cytoplasmic proteins, making them attractive potential biomarkers and therapeutic agents for many diseases, including carcinomas, diabetes, and renal diseases.10−12 Elevated EV levels are observed in several human disorders,13,14 and their potential for diagnosing cancer and utility in targeted molecular therapies have been explored.15,16 This has led to active fundamental research to understand the © 2017 American Chemical Society
Received: September 11, 2016 Accepted: January 9, 2017 Published: January 9, 2017 1360
DOI: 10.1021/acsnano.6b06131 ACS Nano 2017, 11, 1360−1370
Article
www.acsnano.org
Article
ACS Nano
Figure 1. Schematic diagram showing the detailed microfluidic features of (A) the device for EV isolation and detection containing a sampleloading chamber, two filtration chambers, two waste chambers, two washing buffer storage chambers, an elution buffer storage chamber, a collection chamber, and connecting channels with valves.28 (B) Cross-sectional view of the filters highlighted in (A), showing fluidics for the size-selective isolation of EVs. (C) A photograph of the Exodisc. A disc has two identical units for the analysis of two samples. (D) SEM images of filter I (pore diameter = 600 nm) and filter II (pore diameter = 20 nm).
purification step involving a sucrose gradient was utilized;20 however, this increased the total processing time and required specialized and expensive equipment for gradient UC. To overcome these limitations, precipitation reagents were introduced that allow EV sedimentation into a pellet at lower gforces; however, most of these reagents require long incubation times of up to 12 h, and the chemicals used for EV precipitation were not disclosed, resulting in ambiguities in the analysis of isolated EVs.19 Moreover, some of these methods isolate nonvesicular protein complexes in addition to EVs,22,23 which may lead to incorrect conclusions involving downstream molecular analysis. Recently, significant progress was made in the sensitive detection of EVs using various kinds of microfluidic devices, including those integrated with on-chip lysis, extraction of nucleic acids, or direct detection of proteins.24−27 Most of these devices utilize immunoaffinity-based isolation methods using magnetic beads or microfluidic chips coated with EV-specific antibodies such as CD9, 63, and 81, which isolate EVs with relatively high purity. However, considering tumor heterogeneity, all EVs may not have high expression of the surface proteins, which could cause marked variability in the results of subsequent analyses. Furthermore, the volume of biological samples available for processing in microfluidic chips is relatively small. For this reason, although they are advantageous for diagnostic purposes, isolated vesicles may not be adequate for various kinds of downstream analyses. Another promising technology for EV isolation from biological samples is size-based filtration. Although it is used in conjunction with other isolation methods, such as UC, to eliminate larger particles during the initial steps, filtration cannot be utilized exclusively for the isolation of nanosized EVs due to potential limitations, including the lack of filters and nonspecific binding of EVs to the filters, resulting in low recovery levels and concerns regarding EV stability under the applied pressure necessary for nanofiltration.
Many new technologies and devices continue to be developed based on the importance of the field, and there are several challenges that still need to be addressed, including improvements in yield, purity, reproducibility of methods involving EV isolation from biological samples, automation of the EV-enrichment process enabling its rapid application to clinical settings, and facile retrieval of bioactive EVs compatible with downstream molecular analysis. Here, we present an integrated centrifugal microfluidic platform (Exodisc) that can be used for the automatic enrichment and isolation of EVs from biological samples along with subsequent analysis by in situ enzyme-linked immunosorbent assay (ELISA) or fluorescence labeling and/ or retrieving EVs for downstream molecular analysis. Upon spinning the disc at relatively low g-forces (100-fold higher concentrations of mRNA as compared with gold-standard UC methods), (4) highly sensitive protein detection within 30 min via on-disc ELISA using the enriched EVs on the filter, and (5) utilization of captured EVs for additional biochemical or physical characterization. This study also examined clinical samples by analyzing EVs from bladder cancer patient urine and showed that urinary EVs could be potentially useful biomarkers for cancer diagnostics. 1361
DOI: 10.1021/acsnano.6b06131 ACS Nano 2017, 11, 1360−1370
Article
ACS Nano
Figure 2. Schematic diagram and CCD images of the Exodisc, showing the operational steps of EV isolation and recovery. Left and right columns show the schematics of the top and cross-sectional views of the disc to illustrate the device operation and fluidics, with the corresponding pictures of the spinning device shown in the middle panel. (A) Precipitation of large debris. (B) Transfer of clear supernatant to the filtration chambers, resulting in large particles being captured on filter I and EV enrichment on filter II. Blue solid lines and red dotted lines indicate fluid transfer on the top and bottom of the disc, respectively. (C) Washing of enriched EVs. (D) Removal of the solution under the filter by spinning at a lower spin speed (1500 rpm). As shown in the CCD image in the middle column, the solution under the membrane (meniscus indicated by the green arrow) is transferred to waste chamber II, and (E) recovery of enriched EVs is accomplished by transferring the solution on the top of filter II (meniscus indicated by the orange arrow) to the collection chamber. The pink arrow indicates EV-enriched solution. Numbers in blue and red circles indicate the open and closed states of the valves, respectively.
sided adhesive tape (Figure S1), and fluid transfer was controlled by the actuation of reversible diaphragm valves (Figure S2).28 After sedimentation of large debris by spinning the disc at 3000 rpm (1 h, the pluronic solution was removed, and the device was washed with phosphate-buffered saline (PBS). After complete removal of the washing solution and drying, all valves, except valve #3, were closed, making the Exodisc ready for use. First, the sample (1 mL), washing buffer (600 + 600 μL), and elution buffer (100 μL) were loaded into the corresponding chambers. As the Exodisc spun at a rotation speed of 3000 rpm, the debris from the sample was precipitated into the tilted chamber29 (Figure 2A), followed by transfer of the clear supernatant to waste chamber I through two filtration chambers by opening valve #2 (Figure 2B). During filtration, large particles were captured by filter I with a pore diameter of 600 nm, and nonvascular proteins were removed through filter II with a pore diameter of 20 nm, resulting in EV enrichment on filter II. After filtration, valve #3 was closed, valves #4 and #5 were sequentially opened, and the washing buffer was transferred twice through filter II (Figure 2C). Next, valve #6 was opened, and the solution remaining under filter II was transferred to waste chamber II by spinning at a speed of 1500 rpm (Figure 2D). Then, valve #8 was opened, and the enriched-EV solution (100 μL) on the filter II was transferred to the collection chamber by spinning the disc at 1500 rpm. For elution, valve #8 was closed, and #7 was opened, and the elution buffer was transferred to filter II at a rotation speed of 2000 rpm. Finally, valve #8 was reopened, and the remaining
EV solution was transferred to the collection chamber by spinning the disc at 1500 rpm (Figure 2E). The EVs retrieved from this collection chamber were subjected to further analysis, including NTA, ELISA, and reverse transcription (RT)-PCR. After loading a biological sample, such as cell-culture supernatant (CCS) or a raw urine sample, the enriched EV solution was prepared in a fully automated manner within 30 min using a custom-designed tabletop spinning system, with the detailed spin program summarized in Table S1. Filter Selection and EV Enrichment. To determine a filter combination capable of yielding large amounts of EVs, different combinations of filters, including those with 200:20 nm, 600:100 nm, or 600:20 nm pore diameters (filter I:filter II, respectively), were examined by filtering 100 nm polystyrene (PS) nanoparticles. Based on our results, 600:20 nm filters were selected as exhibiting an efficient capture and high recovery of EVs. When 200:20 nm filters were used, most of the 100 nm PS nanoparticles remained in the pores of the 200 nm filter (Figure 3A). The fraction of PS nanoparticles that remained in the pores of the filter was evaluated by filtering a known amount of particles and analyzing the recovered particles from both filters. When 6.7 × 108 particles were filtered, only 13.6 × 106 particles (2.03% of the total input) were recovered, whereas ∼98% of the particles remained stuck to the filter. Of the recovered particles, 8.6 × 106 (1.3% of the total input) were collected by filter I, and 5 × 106 (0.75% of the total input) were collected by filter II. Similarly, for the 600:100 nm filters, the particles passed through the 600 nm filter, but a large number remained on the 100 nm filter and were not recovered (Figure 3A). Our results showed that the 600:20 nm filters resulted in the highest recovery of PS nanoparticles (Figure 3B). Next, a mixture of 800 nm (1.25 × 1010 particles) and 100 nm (1.16 × 109 particles) PS nanoparticles in 1 mL of solution was loaded onto the Exodisc, and size-selective separation was 1363
DOI: 10.1021/acsnano.6b06131 ACS Nano 2017, 11, 1360−1370
Article
ACS Nano
Figure 4. Characterization of EVs prepared by Exodisc. EV concentration before and after filtration from (A) LNCaP CCS and (B) bladder cancer patient urine as measured by NTA. (C) SEM images of the filters showing large debris on filter I and EVs on filter II after filtration of bladder cancer patient urine. (D) TEM and (E) SR-SIM images of EVs recovered from filter II.
evaluated. After filtration, the membranes were removed from the Exodisc, and SEM images were recorded. As shown in Figure 3C, size-selective sorting was possible, with 800 nmsized particles captured on filter I and only the 100 nm-sized particles enriched on filter II. NTA analysis of the particle mixture before filtration (1.37 × 1010 particles) and the solution retrieved from the collection chamber after filtration (8.48 × 108 100 nm particles) confirmed that only the 100 nm particles were enriched on filter II (Figure 3D), with a recovery of 73.1%. Addition of a 10-fold greater number of larger (800 nm) particles was intentional to assess filtration efficiency, finding that these were efficiently filtered, with only the 100 nm-sized particles passing through the 600 nm filter. After confirming the efficient separation of differently sized PS nanoparticles, we evaluated the performance of EV enrichment from 1 mL samples of CCS (Figure 4A) and urine (Figure 4B). NTA analysis of the EVs separated from the CCS samples showed a recovery rate of >95 ± 1%, and SEM images of the filters showed large particles captured on filter I and urinary EVs on filter II (Figure 4C), whereas transmission electron microscopy (TEM) images of recovered EVs revealed vesicles exhibiting a round and intact morphology (Figure 4D). EVs were also analyzed by super-resolution fluorescence imaging following EV labeling on the disc by incubation with biotinylated antibodies, followed by streptavidin-Alexa Flour conjugation. Excess antibodies and dye molecules were removed by washing the labeled EVs from filter II, followed by their recovery and fluorescence-image acquisition by superresolution structured-illumination microscopy (SR-SIM) (Figure 4E). The SEM, TEM, and fluorescence images revealed that the Exodisc filtered EVs were intact, even after undergoing several washing procedures to remove proteins and unbound dye molecules. Comparison with Conventional EV-Isolation Methods. Exodisc enrichment efficiency was compared with two commonly used EV-isolation methods: UC and Exospin. Starting with 1 mL of LNCaP CCS, EVs were isolated using
three different protocols as summarized in Tables 1 and S2. Briefly, process times of 6 h, 4 h, and 30 min and maximum gTable 1. Comparison of the Operation Conditions for UC, Exospin, and Exodisc Isolation of EVs maximum g-force (g) protein detection (ELISA) sample input volume (mL) elution volume (μL) total time (h) EVs isolation time (h) EVs detection time (h)
UC
Exospin
Exodisc
150,000 96-well plate 1 200 11 6 5
16,000 96-well plate 1 200 9 4 5
500 on-disc 1 200 1 0.5 0.5
forces of 150,000g, 16,000g, and 500g were required for EV isolation by UC, Exospin, and Exodisc, respectively. In this study, we used the same initial CCS volume for each method, and subsequent analyses were performed using an equal volume of retrieved EVs. As shown in Figure 5, EVs enriched by the three different methods were characterized by NTA, ELISA, RNA electrophoresis, RT-PCR, and bicinchoninic acid (BCA) protein assay. Our results demonstrated that the Exodisc provided a higher yield of enriched EVs with higher purity. NTA results showed that the concentrations of retrieved EVs by UC, Exospin, and Exodisc were 6.5 ± 0.1 × 109 particles/mL, 9.1 ± 0.8 × 109 particles/mL, and 25.2 ± 0.2 × 109 particles/mL, respectively, indicating that the Exodisc yielded a 3.9-fold higher number of particles as compared with those yielded by the other methods (Figure 5A). Additionally, CD9/CD81 sandwich ELISA results indicated that the highest optical density (OD) for the EVs samples that were isolated by the Exodisc as compared with ODs from samples obtained using the other two methods (Figure 5B). Furthermore, the CD9/CD81 sandwich ELISA results following their conversion to EV concentration using the 1364
DOI: 10.1021/acsnano.6b06131 ACS Nano 2017, 11, 1360−1370
Article
ACS Nano
Figure 5. Comparison of EV-isolation results between conventional methods. (A) NTA results showing the ability of the Exodisc to enrich 3.9fold higher numbers of EVs as compared with UC-based isolation methods. (B) ELISA results demonstrating that EVs retrieved from the Exodisc exhibited higher concentrations of CD9/CD81-specific EVs as compared with UC and Exospin results. (C) Size ranges and concentration of RNA extracted from isolated EVs showing identical size range from all three methods, with a higher RNA concentration derived from the Exodisc results. EVs isolated by the Exodisc contained a 16.5-fold higher concentration of RNA as compared with those isolated by UC. (D) RT-PCR results showing >100-fold higher levels of GAPDH, CD9, PSA, and PSMA mRNA from EVs isolated by Exodisc as compared with conventional UC or Exospin methods. (E) The effect of the number of washing steps on Exodisc operation and EV purity as defined by the number of EVs normalized by total protein concentration. All measurements were performed in triplicate, and the mean value is plotted with the standard deviation represented by error bars.
Our results indicated that Exodisc isolation yielded highly pure EVs, with 50.5 ± 16.3 × 107 particles/μg protein, whereas UC and Exospin yielded 2.0 ± 0.1 × 107 particles/μg protein and 0.4 ± 0.9 × 107 particles/μg protein, respectively (Table S3). The number of washing steps required for removing protein contaminants on the Exodisc was determined to be two based on the results of BCA analysis, which indicated removal of >95% of protein contaminants after the second wash (Figure 5E). Notably, when the number of washing steps was increased up to eight, during the use of the Exodisc, the level of protein contamination decreased dramatically after the second washing step. However, additional washing after this point had only minor effects on protein removal, with no significant change in the number of particles. These results indicated that EV purity was enhanced by including additional washing steps during the isolation process on the Exodisc (Figure 5E). On the other hand, both total protein content and the number of particles decreased with increasing washing steps in EVs isolated by UC, indicating no significant effect of additional washing steps on the purity of EVs. In the case of EVs isolated by Exospin, we followed a recommended protocol without including additional washing steps, and the purity of isolated EVs is less compared to UC and Exodisc. Overall, our results indicated that EVs isolation by Exodisc provided a significantly higher yield and purity (Table S3), with a shorter processing time and without the requirement for high-speed centrifugation. On-Disc EV Detection by ELISA. After confirming the efficient size-selective enrichment and retrieval of EVs using Exodisc, we explored the on-disc detection of proteins using ELISA. As schematically shown in Figure 6A, the enriched EVs on filter II were incubated with biotinylated antibodies for tetraspanin proteins (e.g., CD9 or CD81) for 15 min, followed by a 5 min incubation with horseradish peroxidase (HRP)-
calibration curves (Figure S4) confirmed a 13-fold greater concentration of CD9/CD81 expressing EVs using the Exodisc relative to the concentrations recovered by the other methods (Table S3). Moreover, we examined the yield and size distribution of RNAs from EVs isolated by the three methods. The structural integrity of the RNAs over a broad range of sizes was determined by electrophoresis (Labchip GX bioanalyzer; PerkinElmer, Waltham, MA, USA). The size range and yield of RNA are highly dependent upon the isolation method;30 therefore, we used the same kit (miRNeasy; Qiagen, Hilden, Germany) and protocol for extracting RNAs from EVs obtained by each of the three methods. Figure 5C shows the size ranges and concentrations of extracted RNAs. Interestingly, all three methods yielded RNAs exhibiting an identical size range based on their superimposed peak distributions, with all extracted samples also showing abundant small RNAs (100-fold higher mRNA concentrations as compared to that observed in EVs enriched by the other two methods (Figure S5 and Table S3). The trend in relative expression of CD9, PSA, and PSMA mRNA with respect to that of GAPDH was similar in all samples prepared by the three methods. EV purity, defined here as the number of particles per microgram of total protein, was calculated from the NTA and BCA data obtained for the EVs isolated by the three methods. 1365
DOI: 10.1021/acsnano.6b06131 ACS Nano 2017, 11, 1360−1370
Article
ACS Nano
Figure 6. On-disc protein detection by ELISA. (A) Schematics showing the process of on-disc ELISA. Biotinylated anti-CD9 antibodies were incubated with EVs enriched on filter II, followed by washing and reaction with HRP-conjugated streptavidin. After washing, the TMB solution was introduced. (B) ELISA results demonstrating the capture efficiency of EVs dependent upon the selection of filter combination. The combination of 600 and 20 nm filters as filters I and II, respectively, returned the highest signal. All measurements were performed in triplicate, and the mean value is plotted, with the standard deviation representing the error bars. (C) ELISA results with EVs enriched from various volumes of LNCaP CCS solution. (D) ELISA results demonstrating high retrieval efficiency. ELISA experiments were performed on filter II before and after retrieval. (E) Photographs of filter II after on-disc CD9-ELISA and the addition of TMB solution using enriched EVs from the urine of healthy control (N) and bladder cancer patient (B).
Figure 7. On-disc ELISA for the detection of CD9- and CD81-expressing EVs isolated from clinical samples by three different methods. Panels represent data from healthy controls (N; n = 5) and bladder cancer (BC) patients (B; n = 5).
through different combinations of filters. A combination of 600 and 20 nm filters resulted in a higher OD as well as a higher signal-to-noise ratio as compared with other combinations, such as 600 and 100 nm or 200 and 20 nm filters. To assess the linearity of on-disc detection, different volumes ranging from 200 μL to 1 mL of LNCaP CCS were used, and ELISA was performed using anti-CD9 antibodies. Figure 6C shows the OD measured as a function of the input volume of the CCS sample loaded on the Exodisc. Notably, the on-disc ELISA showed a higher signal as compared with conventional 96-well plates, even when the same amount of EVs was analyzed. One explanation might be the high accessibility of the
conjugated streptavidin. After each immunoreaction, EVs were washed twice with washing buffer, and excess antibodies and buffer in each step were removed by spinning the disc at 3000 rpm (∼500g). Finally, the buffer solution was completely removed, and 3,3′,5,5′-tetramethylbenzidine (TMB) solution was added (Figure 6A). After a 1 min of incubation, the bluecolored solution was transferred to the detection chamber, which was preloaded with a stop solution, and the OD was measured at 450 nm using a custom-built detection system.32 On-disc ELISA confirmed that the combination of filters selected was the best among the tested combinations. Figure 6B shows the OD values for on-disc ELISA of LNCaP CCS filtered 1366
DOI: 10.1021/acsnano.6b06131 ACS Nano 2017, 11, 1360−1370
Article
ACS Nano
this device is that intact EVs captured on the filter can be easily retrieved for downstream molecular analysis and/or be directly analyzed for surface−protein markers using on-disc ELISA, which can be easily adapted for clinical settings. This method exhibited enhanced nanoscale-filtration efficiency through the incorporation of highly porous membranes on the centrifugal microfluidic platform. The soft spin-through, surface-modified, nanoporous membrane expedited clog-free filtration, allowing the two filters to sort EVs of desired size ranges while filtering out nonvesicular proteins. Here, we demonstrated the following: (1) isolation and purification of EVs from raw samples within 30 min; (2) high recovery rates of enriched EVs, with >95% recovery of EVs and >100-fold higher mRNA concentrations of GAPDH, CD9, PSA and PSMA as compared with UC results; (3) direct detection of enriched EVs via the same platform within 30 min using on-disc ELISA; and (4) enrichment of urinary EVs from bladder cancer patients and on-disc ELISA for protein quantification. This system is also capable of enriching EVs exhibiting different size ranges by altering the desired pore size of membranes. The use of nanoporous filters allows several advantages. First, a highly porous and hydrophilic anodisc inorganic membrane with a pore size of 20 nm facilitated efficient capture and retrieval of EVs. Second, compared to immunoaffinity-based capture using specific antibodies, size-based isolation is capable of unbiased capture of all EVs within the size range, and with further protein analysis by ELISA, it is possible to explore surface proteins from the captured EVs. Third, compared to the immobilized EVs on a 96-well plate from ELISA, the nanoporous membrane was capable of capturing EVs in a chamber without immobilization, thereby facilitating ELISA levels of sensitivity with reduced steric hindrance for immunoreactions. Finally, the method was capable of labeling EVs along with efficient washing of unbound dyes or antibodies and without the necessity for complicated purification processes. In summary, the Exodisc allowed for efficient EV isolation from both CCS and bladder cancer patient urine samples, followed by on-chip protein detection using ELISA. The sizeselective filtration at a low g-force (
