Comparative Evaluation of Methods for Isolating Small Extracellular Vesicles Derived from Pancreatic Cells

Background: Small extracellular vesicles (sEVs) are nanosized vesicles involved in cell-to-cell communication. sEVs have been widely studied for clinical applications such as early detection of diseases and as therapeutics. Various methods for sEVs isolation have been using, but different methods may result in different qualities of sEVs and impact downstream analysis and applications. Here, we compared current isolation methods and performed a comparative analysis of sEVs derived from pancreatic cancer cells. Results: Ultracentrifugation, ultraltration and co-precipitation as concentration methods were rstly evaluated for yield, size, morphology and protein level of pellets. Then, isolate sEVs obtained by four different purication methods: size exclusion chromatography, density gradient ultracentrifugation, ultracentrifugation, and immunoanity capturing, were analysed and compared. For the concentration process, ultracentrifugation method obtained high quality and concentration pellets. For the purication process, immunoanity capturing method obtained the purest sEVs with less contaminants, while density gradient ultracentrifugation-based method obtained sEVs with the smallest size. Proteomic analysis revealed distinct protein contents of puried sEVs. Conclusions: For isolating sEVs derived from pancreatic cancer cells, ultracentrifugation-based method is recommended for concentration of sEVs, density gradient ultracentrifugation-based method may be suitable for isolation of sEVs for therapeutic study, immunoanity capturing may be applied for studies exploring sEVs as biomarkers.


Introduction
Extracellular vesicles (EVs) are biological vesicles released by almost all types of cells. EVs have gained increasing interests over the last decade for their cell-to-cell communication properties. EVs have been emerging as attractive therapeutic tools for their content and their natural carrier role. Small EVs are able to be engineered as nano drug delivery vehicles due to their relatively small size and properties such as crossing the biological barrier, circulation stability and inherent targeting.
The methods for isolation of sEVs has been extensively studied. Conventional differential ultracentrifugation has been widely used, but may not remove all contaminants 1,2 . To obtain sEVs with small size, 0.22 µm ltration membranes were used in some studies to remove the microvesicles 3,4 , but the ltration membrane could not remove protein contaminants, which may mislead the study results 5,6 .
The International Society for Extracellular Vesicles (ISEV) advised that the qualities of EV from different isolation methods were different 7 . Besides, each isolation method may have disadvantages. Recently, it was reported that ultracentrifugation could not obtain EVs with high purity 1,2 , and may have problems such as clogging and trapping of vesicles 8 . Polymer co-precipitation-based methods are simple to perform but will precipitate large vesicles and contaminant proteins in the sample 9,10 . Size exclusion chromatography-based method and density gradient ultracentrifugation-based may be effective but Puri cation of sEVs Based on the results of evaluation of pellets after concentrating sEVs, ultracentrifugation was used for concentrating sEVs for further puri cation of sEVs. Four methods were used for purifying crude sEVs ( Fig. 1B): (1) Density gradient ultracentrifugation (DGUC): 200 µL of crude sEVs was loaded onto the top of a 12 mL discontinuous sucrose (Solarbio, China) gradient solution (15%, 20%, 25%, 30%, 40%, 60% sucrose in PBS, 2 mL for each gradient solution) and then centrifuged at 100,000 g for 16 h at 4 °C. 12 fractions (1 ml) were collected for each gradient 21  Then the solution was centrifuged at 10,000 g, 4 °C for 70 min. The pellet was re-suspend by 200 µL of PBS. (4) Immunoa nity capture (IAC): Magnetic beads (BeaverBeadsTM Protein A/G immunoprecipitation kit, Beaver, China) were washed and then activated by incubating with 100 µL of anti-CD63 antibody (50 µg/mL, ab134331, abcam, UK) for 15 min. After magnetic separation, the anti-CD63-conjugated beads were incubated with 200 µL of crude sEVs at 25 °C for 1 h. The sEVs-beads complexes were separated by a magnet and eluted and re-suspended in 40 µL of PBS. Puri ed sEVs were stored at -80 °C within two days before analysis.

Analysis of fraction from DGUC and SEC
Total protein level in each fraction was determined by using a BCA Protein Assay Kit (MultiSciences Biotech Co., China). The level of CD63 in each fraction was determined by using an enzyme-linked immuno-sorbent assay (ELISA) kit (CSB-E14107h, CUSABIO Biotech Co. Ltd., China). After analysis, fractions 6, 7, 8, 9, 10 from DGUC were mixed used for further analysis, while fractions 9, 10 from SEC were mixed and used for further analysis.
Characterization and analysis of crude sEVs and puri ed sEVs Nanoparticle Tracking Analysis (NTA) For crude sEVs, samples were diluted 100 times with PBS before NTA. For puri ed sEVs, samples obtained via IAC was diluted 125 times, samples obtained via SEC was diluted 2.5 times, samples obtained via UC was diluted 25 times, while samples obtained via DGUC was not diluted before NTA (NS300, Malvern, UK). The RR of particles was calculated as the following formula. The assay was repeated 3 times. (1)

Transmission Electron Microscopy
Page 5/14 10 µL of samples was dropped onto an ultrathin carbon lm-coated 400 mesh copper grid and washed with PBS for two times. After drying excess liquid, the EVs-coated grid was stained by phosphotungstic acid (1%) and then washed with PBS for two times and then dried and imaged with a multipurpose eld emission transmission electron microscope (TEM, JEM-1200EX, JEOL Ltd., Japan).

Protein level
The protein level was determined by using a BCA Protein Assay Kit. The Protein Recovery rate (RR) of the puri cation process was calculated as the following formula. The assay was repeated 3 times.
(2) Western blotting 5 × SDS-PAGE Loading Buffer (New cell and Molecular Biotech Co., Ltd, China) was added into the sample. The sample was kept at 100 °C for 10 min. 6.02 × 10 9 EVs were loaded on each well in 12% SDS-PAGE (Lianke Bio, China). After the electrophoresis, the proteins were transferred to PVDF membrane (Millipore, USA). The membrane was blocked with 5% milk solution for 1. Coomassie brilliant blue staining 5 × SDS-PAGE Loading Buffer (New cell and Molecular Biotech Co., Ltd, China) was added into the sample. The sample was kept at 100 °C for 10 min. 6.02 × 10 9 EVs were loaded on each well in 12% SDS-PAGE (Lianke Bio, China). After the electrophoresis, the gel was incubated with 20 mL of working solution (0.0025% Coomassie brilliant blue, 45% methanol, 10% glacial acetic acid) (Solar bio, China) for 1 h, and then washed by elution solution (25% methanol, 8% glacial acetic acid) for 4 h.

ELISA
The level of CD63, CD81, TSG101, beta-actin, GAPDH, CD47 (CUSABIO Biotech Co. Ltd., China) and ago-1 (MyBiosource, Canada) in puri ed samples and in crude sEVs were determined by ELISA kits. The protein per EV was calculated by the following formula. The assay was repeated 3 times. (3)

Digestion of proteins
SDT solution (4% SDS, 100 mM Tris-HCl, pH 7.6) was added into the puri ed sEVs. The sample was incubated under boiling water for 15 min, followed by centrifugation at 14,000 g for 15 min. The supernatant was collected as protein sample. DTT (Sigma, USA, 43819-5G) was added in the protein sample to 100 mM. The sample was incubated under boiling water for 5 min, and then cooled to room temperature. 200 µL of UA buffer (8M Urea, 150 mM Tris-HCl, pH 8.5) was added to the sample, followed by centrifuging at 12,500 g for 15 min using a 30 kDa ultra ltration tube and centrifuged at 12,500 g for 15 min. Then, 100 µL of iodoacetamide (IAA) buffer (100 mM IAA in UA) was added and kept at room temperature in darkness for 30 min. The sample was centrifuged at 12,500 g for 15 min. 100 µL of UA buffer was added to the supernatant and then centrifuged at 12,500 g for 15 min again. 100 µL of 40 mM NH 4 HCO 3 solution was added to the sample followed by centrifugation at 12,500 g for 15 min. Then

Statistical analysis
Data were presented as mean values ± SD. One-way analysis of variance (ANOVA) and students' t test were performed at the signi cance level α = 0.05.

Analysis of crude sEVs
Same volume (50 ml) of the supernatant was processed to compare concentration methods. Size distribution of crude sEVs obtained via UF, UC and Co-P were presented in Fig. 2A, 2B and 2C, respectively. sEVs obtained via UC showed a smaller size distribution than other two methods. All samples showed plenty of big particles, indicating that neither single method (UC, Co-P or UF) could obtain pure sEVs. Crude sEVs obtained via UC showed signi cantly more total particle number and small particle (30-150 nm) number than UF and Co-P (Fig. 2D). sEVs obtained via UF showed signi cantly high protein levels than UC and Co-P as evaluated by BCA assay test (Fig. 2E) and coomassie brilliant blue staining (Fig. 2F), but the protein level in the control (fresh medium after UC) was also very high. TEM images for crude sEVs were presented in Fig. 2G. Big and small EVs could be observed for all crude sEVs groups. Besides, aggregation of unknown small particles (red frame) was found in crude sEVs obtained via Co-P. Based on the quality of crude sEVs, UC was used for crude sEVs collection for further puri cations.

Analysis of fraction from DGUC and SEC
Total protein and CD63 levels in fractions from DGUC and SEC were shown in Fig. 3. Fractions with a relatively high level of CD63 was collected as puri ed sEVs 22,23 . Hence, fractions 6, 7, 8, 9 and 10 from DGUC were mixed as the puri ed sample (Fig. 3A, 3B), fractions 9 and 10 from SEC were mixed as the puri ed sample (Fig. 3C, 3D).

Analysis of puri ed sEVs
The size distribution of puri ed samples via UC, SEC, IAC and DGUC were shown in Fig. 4A, 4B, 4C and 4D, respectively. The big particles in sEVs for all samples were removed substantially. The size of the sample via DGUC showed the smallest size, and the sample via IAC showed a relatively small size with less microvesicles. The sample via UC showed more particle and sEVs numbers than other methods (Fig. 4E). The sample via SEC showed the least particles and sEVs. Besides, samples via DGUC or IAC showed a relatively high proportion of sEVs in all particles (Fig. 4E). The sample via UC showed the highest RR (Fig. 4F). TEM images of puri ed sEVs were shown in Fig. 4G

Protein evaluation
Puri ed sEVs showed lower total protein levels for all methods than crude sEVs (Fig. 5). The protein level of the sample via UC was higher than other methods as evaluated by BCA assay (Fig. 5C) and coomassie brilliant blue staining (Fig. 5A). The sample puri ed by IAC showed the least total protein level and recover rate (Fig. 5C, 5D).
WB results were shown in Fig. 5B. Three EVs marker proteins (CD81, CD63, CD9) and CD47 were detected in the puri ed samples. Compared to the crude sEVs, ago-1 as contaminant protein was signi cantly decreased in puri ed samples, which was consistent with the WB results (Fig. 5E). Results of ELISA were summarized in Fig. 5F. Samples puri ed by UC showed less CD47, CD81, GAPDH and β-actin. The sample via IAC showed the least TSG101. The sample via IAC had a higher CD63 level as the sample was isolated by anti-CD63-conjugated beads. RR of CD63 was almost 100% for IAC method, indicating that almost all particles expressing CD63 were extracted. For the samples puri ed by SEC and DGUC, levels of most proteins tested were similar.

Proteomics
A total of 817 proteins were detected in proteomic study. There were 631 proteins in the crude sEVs sample via UC, 383 proteins in the puri ed sEVs sample via UC, 78 proteins in the sample via SEC, 154 proteins in the sample via DGUC and 76 proteins in the sample via IAC, 25 proteins were identi ed for all ve groups (Fig. 6A). Heat map analysis of all proteins was summarized in Fig. 6B. Puri ed sEVs obtained via UC showed signi cantly more protein content, which was consistent with our results of protein evaluation (Fig. 5). Reported potential protein biomarkers for cancer development 24 , metastasis 25 and drug resistance 26 were also analysed in our proteomic study. Further analysis revealed distinct protein contents in samples (Fig. 6). High contents of both overexpressed and downexpressed proteins associated with metastasis and drug-resistance were detected in crude sEVs (Fig. 6C to 6H). Puri ed sEVs via IAC retained most pancreatic cancer-overexpressed (Fig. 6C), but not downexpressed (Fig. 6D), proteins (compared to normal adjacent pancreatic tissue). sEVs obtained via SEC showed the least contents of proteins associated with metastasis (Fig. 6E) and drug resistance (Fig. 6G).

Discussion
Despite progress in technique, isolation of sEVs has been challenging. It would be wise to choose a strategy and develop isolation protocols depending on experiment purpose. Three concentration methods and four puri cation methods were evaluated in our study. A summarization of advantages, disadvantages and possible suggestions for isolation methods used in this study was shown in Table 1. For the concentration of sEVs, UC-based method showed high yield and purity of sEVs and was used to concentrate the sample before puri cation. For the puri cation of sEVs, IAC was effective and timesaving and yield purest sEVs among all methods evaluated, thus was recommended to be suitable for biomarker study 13,14 . DGUC method effectively produced the smallest and also puri ed sEVs with relatively high yield, thus may be suitable for EV therapeutic study as pure and large number of sEVs are required for therapeutic application 15,27−29 . More importantly, sEVs obtained via DGUC method could avoid aggregation and precipitation 30 and were superior for cellular uptake 31 .
In the presented study, the size, yield, morphology and protein were assessed for quality of sEVs. The size and yield of sEVs were crucial for therapeutic use, especially for drug delivery. It has been reported that smaller EVs could be uptaken by cells more e ciently 31 . The morphology showing the presence of EV was observed, sEVs obtained in this study was saucer-shaped under TEM images as reported 32 . CD63 was a marker for EVs and was used in IAC for puri cation 9,33,34 . Besides, in our fraction analysis, CD63 level was used to re ect EV-containing fractions, this method was also reported in previous studies 21,23,35 . However, for SEC, CD63 may not fully represent the sEVs fractions as expected. Possibly, CD63 was not exclusive to the sEVs. Recent studies reported that microvesicles also expressed CD63 36,37 . A combined strategy of several sEVs marker proteins such as CD63, TSG101 and CD9 may re ect the sEVscontaining fractions more accurately.
Our study included the majority of current sEVs isolation methods, except commercial kits and micro uidic-based methods. Because of unstable quality of the extracted EVs, high price and unknown solutions 38 , commercial kits were not included in our study. The micro uidic technology was popular but not included in our study as the technique was mostly used for methodological studies, such as sEVs detection, instead of therapeutic application study, even if it was potentially available for isolation 39,40 .
In the presented study, an equal number of EVs particles was evaluated for each method in case of contaminant protein in uencing the results. The level of target protein could be weak as detected if equal protein, with a high amount of contaminant protein, was loaded onto each well in the gel. This may explain why the level of EV marker protein was weak in study 41 .
The UC-based technique remains the most common method for sEVs isolation [42][43][44] . However, our study results demonstrated that DGUC and IAC methods produced sEVs with better purity than UC. Besides, the interpretation of results should be cautious when commercial sEVs isolation kits based on the Co-P method was used 38,45 . The combination of several isolation methods may produce purer sEVs. Jeppesen et al. used ultracentrifugation-based technique, density gradient ultracentrifugation-based technique and immunoa nity capture-based technique to purify sEVs 46 . But more isolation steps may produce fewer sEVs, and combined isolation protocols were often hard to follow as the strategy could be complicated and fussy. The technique for evaluation of sEVs has been advancing. Tian et al. splendidly used nano ow cytometry to evaluate the quality of sEVs 38 . But nano ow cytometry-based analysis of sEVs needs further re nement of methods. For better comparisons between groups and future replication, we applied basic characterization of sEVs isolated by different methods 23,35,47 . sEVs are a heterogenous group of vesicles 48 . Single EV may carry distinct proteins, isolation method based on EV marker may lose EV subtypes of potential interest. For example, anti-CD63 conjugatedbeads could be used to obtain CD63-enriched EVs, but effects of other vesicles may be neglected during subsequent experiments. This may explain why several metastasis and drug resistance-associated proteins were not highly expressed in puri ed sEVs obtained via IAC. However, future basic and clinical studies are likely to provide valuable information regarding their heterogeneity and advance our understanding of biological functions, thus reveal and harness their potentials for disease detection and therapy.

Conclusions
Page 10/14 Current methods are useful for isolating sEVs. Recommendations for choosing sEVs isolating method depend on the aim of study. For isolation of Panc-1 cell-derived sEVs, UC was advised to concentrate the medium. IAC method is effective for isolating sEVs with high purity, thus was suitable for biomarker study. Meanwhile, DGUC was recommended for EV therapeutic study due to high yield and small size distribution.  Table   Table 1. Summary of concentration and puri cation methods for isolating sEVs from pancreatic cancer cells