Comparison of separation methods for tissue‐derived extracellular vesicles in the liver, heart, and skeletal muscle

Extracellular vesicles (EVs), which are nanosized vesicles released by cells as intracellular messengers, have high potential as biomarkers. EVs are usually collected from in vitro sources, such as cell culture media or biofluids, and not from tissues. Techniques enabling direct collection of EVs from tissues will extend the applications of EVs. We compared methods for separating EVs from solid liver, heart, and skeletal muscle. Compared with a precipitation method, an ultracentrifugation‐based method for collection of EVs from solid tissues yielded a higher proportion of EVs positive for EV‐related markers, with minimum levels of intracellular organelle‐related markers. Some tissue‐specific modifications, such as a sucrose cushion step, may improve the yield and purity of the collected EVs.

intercellular communications [4], they have attracted high attention across many fields. Their potential as promising disease biomarkers relies on the characteristics of the cells that produce them, with the content changing under pathological or physiological conditions [5]. EVs are abundantly released following tissue damage [6,7] and affect tissue stem cell activity, disease remodeling, and tissue regeneration [8,9]. Most studies to date have evaluated EVs collected from in vitro sources, such as cell culture media or biofluids, and not tissues. Techniques enabling direct collection of EVs from tissues will be important for future applications as the study of tissue EVs is on the increase [10].
While polymer-based precipitation kits are easy-touse tools for separating the components of cell culture media and biofluids, ultracentrifugation remains the standard approach [9]. The multicellular nature of tissues produces different physical and chemical properties, and may require different technical concepts, particularly for tissue processing prior to EV collection. Several different methods were reported for the thymus and spleen [11], brain [12][13][14], muscle [15], apex of the heart [7], and adipose tissue [6]. The methods used in these studies were based on either (a) tissue mincing and/or enzymatic homogenization, or (b) combining differential centrifugation with other methods-triple sucrose cushion [13], sucrose gradient [12], repelleting [15], repelleting and sucrose gradient [6], or precipitation [7]. Additionally, the methods varied with respect to the applied centrifugation force (25 000 to 270 000 g) and time (45 min-16 h), repetitions , and tissue types. Methods based on sucrose gradients tend to be labor-intense, whereas precipitation kits are easy-to-use.
In the present study, we investigated various approaches for collecting EVs from three different solid tissues-liver, skeletal muscle, and heart.

Materials and methods
Animal damage models C57BL/6N mice (male, 10-12 weeks old, 21-26 g) were used for all the tissue damage models (three animals for acute liver damage, three for skeletal muscle degeneration, and 22 for acute cardiomyopathy). All animals were treated according to the regulations of the Standards for Humane Care and Use of Laboratory Animals of Tohoku University. The animal experiments were approved by the Animal Care and Experimentation Committee of Tohoku University, Graduate School of Medicine (Permission No. 103-2).
Because tissue damage is reported to trigger an abundant release of EVs compared with intact tissue [6,7], we induced tissue damage in the liver, muscle, and heart as described below.

Acute liver damage model
Damage was induced by a single intraperitoneal injection of carbon tetrachloride (CCL 4 ; 2 µLÁg À1 body weight) dissolved in olive oil (1 : 10 v/v), according to a previous report [16], and the liver was collected 24 h after injection.

Skeletal muscle degeneration model
Under deep anesthesia (intraperitoneal injection of medetomidine 0.3 mgÁkg À1 , midazolam 4 mgÁkg À1 , and butorphanol tartrate 5 mgÁkg À1 ), cardiotoxin (10 lM diluted in PBS) was injected into the thigh and calf muscles at multiple locations (100 µL per injection in three different areas). Muscles were collected 24 h after the injections.

Acute cardiomyopathy model
Damage was induced by a single intraperitoneal injection of doxorubicin hydrochloride (20 mgÁkg À1 ; #04021521; Fujifilm Wako, Tokyo, Japan) resuspended in saline, as described previously [17]. The whole heart was collected 5 days after the injection.

Collection of EVs from damaged liver
To collect EVs from the damaged liver, we applied three different protocols as described below (Fig. 1). One whole liver was used for each protocol, and each protocol was repeated three times. Therefore, three livers were used for each protocol. We applied either a precipitation kit or ultracentrifugation to determine the better protocol for collecting EVs. Before tissue collection, each animal was transcardially perfused with 20 mL of 19 PBS under deep anesthesia (intraperitoneal injection of medetomidine 0.3 mgÁkg À1 , midazolam 4 mgÁkg À1 , and butorphanol tartrate 5 mgÁkg À1 ). The liver was immediately stored in 20 mL RPMI 1640 medium (#11875093; Gibco, Thermo Scientific, Waltham, MA, USA), stored on ice, and processed according to the protocols shown in Fig. 1.

Protocol A
The whole liver (~0.970 g) with 20 mL RPMI 1640 medium was incubated in a 10-cm dish at 37°C in 5% CO 2 for 48 h [18]. The medium (18 mL) was then collected and subjected to differential centrifugation: 300 g for 10 min, 2000 g for 20 min, and 30 000 g for 20 min, all at 4°C (Kubota 7780, AG-508CA Rotor, Osaka, Japan). The supernatants were collected by decanting. The supernatant was filtered through a 0.22-lm membrane (Merck Millipore, Burlington, MA, USA) and incubated overnight with ExoQuick-TC (#EXOTC10A-1; System Biosciences, Palo Alto, CA, USA) at 4°C (2 mL mixed with 10 mL of supernatant). The next day, the sample was centrifuged at 1500 g for 30 min (room temperature) and the pellet was resuspended in 200 µL of filtered 19 PBS. An aliquot (40 µL) was stored at À80°C for later analysis by transmission electron microscopy (TEM). The remaining sample was immediately used for western blot and tunable resistive pulse sensing (TRPS) analysis.

Protocol B
The whole liver (~1.075 g) with 20 mL RPMI 1640 medium was transferred to a 10-cm dish, dissected into~1 9 1 mm pieces using a razor, and cultured as an explant for 48 h. The culture medium was then subjected to the same procedure as in Protocol A.

Protocol C
The sample was collected using a centrifugation-based protocol as described by Jang et al. [19] without purification of the EV pellet by isopycnic centrifugation with an iodixanol gradient. Briefly, the whole liver (~1.04 g) in 20 mL RPMI 1640 medium was transferred to 10-cm dish, cut into~1 9 1 mm pieces, and mixed with working solution (1 mL working solution/0.025 g tissue)-RPMI 1640 medium containing 2 mgÁmL À1 collagenase D (#11088858001; Roche, Basel, Switzerland) and 40 UÁmL À1 DNase I (#11284932001; Roche). After 30 min of incubation at 37°C, the sample (cut liver pieces in 40 mL working solution) was subjected to differential centrifugation: 300 g for 10 min, 2000 g for 20 min, and 30 000 g for 20 min, all at 4°C (Kubota 7780, AG-508CA Rotor), and filtered through a 0.22-lm membrane. The supernatant was centrifuged at 35 000 r.p.m. (210 053 g) for 1 h at 4°C (SW41 Rotor, Optima LX-80; Beckman Coulter, Brea, CA, USA). The pellet was resuspended in 200 µL filtered 19 PBS. An aliquot (40 µL) was stored at À80°C for later analysis by TEM. The remaining sample was immediately used for western blot and TRPS analysis.

Collection of damaged skeletal muscle-derived EVs
Muscle tissue from both legs (~1.4 g) was dissected out after transcardial perfusion with 20 mL of 19 PBS under deep anesthesia (intraperitoneal injection of medetomidine 0.3 mgÁkg À1 , midazolam 4 mgÁkg À1 , and butorphanol tartrate 5 mgÁkg À1 ) and immediately stored in 20 mL RPMI 1640 medium on ice until further processed according to Protocol C. This was repeated three times. The total amount of supernatant used for ultracentrifugation from a single collection was~40 mL. The supernatant was discarded, and the pellet was resuspended in 200 lL of filtered 19 PBS and an aliquot (40 lL) was stored at À80°C for TEM. The remaining sample was immediately used for western blot and TRPS.

Collection of damaged heart-derived EVs
Hearts were collected from mice under deep anesthesia (as described above), washed in 19 PBS to remove the blood, and immediately stored in RPMI 1640 medium on ice. A total of 22 hearts were processed using one of two protocols (11 hearts/protocol with a total weight of 1.45 and 1.39 g, respectively):

Protocol C method (ultracentrifugation-based)
Hearts were enzymatically digested and processed following Protocol C. The obtained pellet was resuspended in 200 µL filtered 19 PBS, an aliquot (40 µL) was stored at À80°C for TEM, and the remaining sample was analyzed by western blot and TRPS.
Protocol C + sucrose (combination of ultracentrifugation and sucrose cushion) Hearts were processed as in Protocol C with one modification -after the differential centrifugation and filtration, the supernatant was loaded on a 1-mL layer of 30% sucrose solution (called a sucrose cushion) according to Gupta et al. [20] (prepared in 19 PBS) and ultracentrifuged at 35 000 r.p.m. (210 053 g) for 1 h at 4°C (SW41 Rotor, Optima LX-80; Beckman Coulter). The sucrose fraction containing the EVs was then resuspended in 19 PBS and washed by ultracentrifugation again at 35 000 r.p.m. for 1 h at 4°C. The final pellet was resuspended in 200 µL of filtered 19 PBS, an aliquot (40 µL) was stored at À80°C for TEM, and the remaining sample was analyzed by western blot and TRPS.

Tunable resistive pulse sensing
The concentration and size distribution of the obtained samples were determined by qNano (Izon Science, Christchurch, New Zealand). First, all samples were serially diluted (1 : 100, 1 : 1000, 1 : 10 000) and analyzed starting at the lowest dilution by NP100 (size range: 50-330 nm) and NP400 (185-1100 nm) nanopores stretched between 45 and 48 mm. Only measurements with a particle count > 500 or a time period of 5 min, linear particle rate in time, and noise below 15 pA were recorded. The pressure was adjusted to achieve a particle flow rate > 100/s and a stable current between 120 and 150 nA. Calibration was performed using calibration beads of a known concentration and size [CPC100 (110 nm) and CPC400 (340 nm), both from Izon Science] diluted at 1 : 1000 according to the manufacturer's protocol. All samples, including calibration samples, were vortexed for 30 s before obtaining measurements. Data were analyzed by CONTROL SUITE V3.3 software for qNano (Izon Science).

Transmission electron microscopy
Samples were stored at À80°C prior to analysis. EV preparations were thawed on ice, and 25 µL from each sample was negatively stained. Prior to fixation, liver-and heartderived EVs were diluted at 1 : 100 in filtered 19 PBS. EVs were fixed with 2.5% glutaraldehyde in phosphate buffer [1 : 1 (v:v) ratio] overnight for more than 18 h. The sample (5 µL) was loaded on collodion-coated 150-mesh copper grids (#651; Nisshin EM Co., Ltd., Tokyo, Japan) and airdried for 40-60 min. Grids were then washed in 1 mL of ultrapure water (3 9 5 min), air-dried (20-30 min), and stained with 15 µL of 3% phosphotungstic acid (#162-02432; FUJIFILM Wako; dissolved in 19 PBS and filtered through a 0.22-lm membrane) for 5 min. Immediately after staining, the excessive staining solution was blotted, and the sample was washed in 1 mL of ultrapure water (3 9 5 min) and air-dried for 20-30 min. Four grids per each sample were prepared, and images were obtained by a transmission electron microscope (JEM-1011; JEOL, Tokyo, Japan) at 80 kV.

Results
While there is currently no consensus on definite specific markers for each EV subtype [21], characterization of the EVs collected in this study followed the criteria recommended by the International Society of EVs [21]. That is, small EVs were vesicles smaller than 150 nm and large EVs were vesicles larger than 150 nm.

Liver-derived EVs
Previous studies targeting tissue-derived EVs involved short-term culture of tissues explants or extraction from whole tissues [6,7,[11][12][13]15,22,23]. Tissue processing is particularly crucial for tissue extraction of EVs because mechanical disruption may release the intracellular content, leading to contamination. Jang et al. and Vella et al. [13,19] reported collecting exosomes from mechanically disrupted tissues. We used both intact and cut liver tissues in ex vivo tissue culture and compared the efficiency of the methods using a precipitation kit and ultracentrifugation.
To evaluate the size distribution of collected vesicles, freshly collected samples from the liver were subjected to TRPS analysis ( Fig. 2A). We used an NP100 nanopore membrane (pore size diameter: 50-330 nm) for the small EVs and NP400 (185-1100 nm) for the large EV measurements. Samples from Protocols A and B caused constant nanopore blocking, unstable current, and noise exceeding 15 pA, even at higher dilutions and thus could not be analyzed by NP100 membrane. Only Protocol C samples ( Fig. 2A; blue color graph) could be analyzed using the NP100 membrane (mean size: 117 AE 40.6 nm). In NP400 membrane analysis, we detected vesicles with the size > 150 nm in samples from all three protocols [Protocol A (yellow graph): 346 AE 152.9 nm, Protocol B (green graph): 540 AE 232.7 nm, Protocol C (red graph): 287 AE 65.9 nm]. The concentration of smaller EVs (4.89 9 10 11 particles per mL) was higher than that of larger EVs (9.63 9 10 10 particles per mL) in Protocol C.
We next analyzed several types of proteins associated with EVs. CD63, a tetraspanin protein highly expressed in EVs [24], was detected only in the sample from Protocol C (Fig. 2B). Regarding endosomal markers, Alix and TSG101, which are generally used to distinguish exosomes (small EVs) from other similar size vesicles for their role in exosomal biogenesis [24,25], TSG101 was expressed in the samples obtained from all the three protocols, whereas Alix was predominantly detected only in the Protocol C sample. HSP70, a commonly detected protein in EVs [25], was recognized in the samples from Protocols A and B, but to a lesser extent in the sample from Protocol C (Fig. 2B).
Samples were also examined for proteins thought to be relatively depleted in EVs versus cells. Their detection would indicate contamination with intracellular components that are unlikely to be packaged into EVs. Endoplasmic reticulum-related calnexin was detected in Protocol B, while it was under the detection limit in Protocol A and Protocol C samples. The protein RPL5, a protein that comprises the 60S ribosomal subunit [13], was under the detection limit in all three samples.
In TEM, EV-like vesicular structures were observed in the Protocol C sample, whereas the Protocol A and Protocol B samples mainly showed debris-like structures and EV-like structures were difficult to be detected (Fig. 2C).
Overall, the Protocol C sample seemed to contain a higher amount of smaller size EVs (mean vesicle size: 117 AE 40.6 nm) with an EV-like morphology. Those EVs were positive for EV-associated markers, CD63, Alix, TSG101, and HSP70, with a lesser extent of contamination on the basis of the calnexin and RPL5 expression. On the other hand, Protocols A and B yielded vesicles contaminated with non-EV content and some of the EV-associated markers were under detection limit.

Skeletal muscle-derived EVs
Based on the above result, we applied Protocol C to damaged skeletal muscle tissue (Fig. 3A). Similar to the liver, a heterogeneous mixture of large and small EVs was collected. TRPS analysis showed a proportion of large EVs (260 AE 80.8 nm) among the small EVs (88 AE 35.5 nm) with a concentration of 3.96 9 10 9 and 1.23 9 10 11 particles per mL, respectively (Fig. 3B). Western blot detected expression of TSG101 and CD63 in the EV preparation (Fig. 3C). These results suggest less contamination because the calnexin and RPL5 levels were under the detection limit (Fig. 3C). Similarly to Protocol C in the liver sample, TEM showed the presence of EV-like vesicular structures with sizes of~100 nm (Fig. 3D).

Heart-derived EVs
We first applied Protocol C to damaged hearts, although we could not detect CD63 in the EV preparation (Fig 4C). We then applied a simple modification of Protocol C by adding a sucrose cushion step (Fig. 4A) to determine whether CD63 is expressed in heart-derived EVs [7,20]. Heart-derived EV preparations collected by these protocols (Protocol C and the combination of Protocol C with sucrose cushion, called 'Protocol C + sucrose' in the following sentences) showed a size range of < 150 nm for small EVs (exosomes) with a mean vesicle size of 108 AE 33.5 and 125 AE 39.8 nm by Protocol C and Protocol C + sucrose, respectively (Fig. 4B). The concentration of small EVs was higher in the sample collected by Protocol C + sucrose compared with that collected by Protocol C (1.01 9 10 12 and 3.85 9 10 11 particles per mL, respectively). The size and concentration of large EVs collected by Protocol C and Protocol C + sucrose were 398 AE 167.5 nm with 1.27 9 10 9 particles per mL, and 302 AE 112.9 nm with 5.77 9 10 9 particles per mL, respectively. Total protein yield was higher in the EV preparation collected by Protocol C + sucrose; therefore, a different amount of maximum total protein was loaded for western blot in the samples acquired by both methods. HSP70, Alix, and TSG101 were detected in the samples from both methods except CD63, which was only detected in Protocol C + sucrose (Fig. 4C). Equal amounts of protein from both methods were analyzed by western blot for calnexin, which was expressed at higher levels in the sample collected by Protocol C than in that collected by Protocol C + sucrose (Fig. 4C). EV-like vesicular structures were observed in both preparations (Fig. 4D).

Discussion
In the present study, we collected EVs from solid tissues-the liver, muscle, and heart. Collected tissuederived EVs exhibited characteristics similar to those of EVs from cell culture or biofluids, in accordance with the minimal information for studies of extracellular vesicles (MISEV2018) recommendations by the International Society of EVs [21]. The tissue was either cultured ex vivo as a tissue explant or enzymatically and mechanistically disrupted prior to the centrifugation steps.
The precipitation method is an easy-to-use approach for collecting EVs, particularly in culture medium and some biofluids [9,26]. When applied to the liver, however, Protocols A and B yielded large EVs and non-EV contaminants, as suggested by constant nanopore blockage by TRPS, which was not observed in the sample from Protocol C. EV aggregates are a main cause of nanopore blockage that can be prevented by thorough vortexing before the measurement. We carefully vortexed the samples before every measurement in TRPS. These aggregates would usually be detected as vesicles with a size 2-3 times bigger than that of small EVs. We detected vesicles > 1 µm in size in both EV preparations using the precipitation kit in Protocols A and B. Additionally, precipitation kits are reported to co-isolate contaminating factors or nonexosomal impurities from the cell culture medium [27]. The expression of CD63 and Alix proteins was mainly observed in Protocol C, and not in Protocols A and B. Similarly, as in some other tissue-derived EVs, we observed EV-like structures by TEM in our preparations. Together, these findings suggested that EV preparations obtained by ultracentrifugation contain vesicles more comparable to endosome-derived exosomes with less contamination of intracellular components, compared with the ready-to-use precipitation kit. Similar results were observed by Van Deun et al. [27] in EVs from cell culture medium, when comparing a precipitation kit and ultracentrifugation method.
Interestingly, adding a sucrose cushion as a modification of Protocol C for the heart improved the overall yield of EVs, vesicle count, total EV protein yield, contamination, and CD63 expression (CD63 was not detected by basic ultracentrifugation). All of these parameters differ among cell types in cell culture or biofluids [28] and might also be applicable to tissuederived EVs. It may also depend on the tissue type or collection method, as was demonstrated by the lower detection of calnexin in heart tissue after adding the sucrose cushion. Depending on the downstream application and overall grade of EV purity, including an additional step such as a density gradient [13,19] may be beneficial.
Several protocols for tissue-derived EVs are available, targeting a specific single type of tissue [6,7,[11][12][13]15,19]. Our data show that a tissue-specific approach may be necessary and more suitable than a universal collection method. The starting weight of the tissue may significantly contribute to the overall yield of EVs and associated proteins. Perfect separation of EV subpopulations remains technically difficult, and precise specific markers for EVs are not yet defined at the cell culture level [21].
Current studies of EVs are mostly based on those collected from cell culture or biofluids [9]. While the method described here targeted the liver, skeletal muscle, and heart, it can likely be applied to other tissue types as well. We realize that modifications of ex vivo culture medium to preserve the natural condition and integrity of solid tissue or applying sucrose cushion in all tissues for comparison could affect the results. This was not part of the scope of the present study, but should be investigated in the future. EV content characterization such as the RNA profile or proteomics would also be beneficial and requires further investigation.