Tracking of Tumor Cell–Derived Extracellular Vesicles In Vivo Reveals a Specific Distribution Pattern with Consecutive Biological Effects on Target Sites of Metastasis

In a first step, the isolation of extracellular vesicles from both, mouse serum and cell culture supernatant, was set up, adapting established protocols [10]. The isolated extracellular vesicles were first analyzed for their purity and subsequent exclusion of concomitantly isolated particles. The proteome content of extracellular vesicles from tumor cells as well as healthy mouse serum was analyzed using mass spectrometry. Before in vivo imaging experiments, feasibility and effectivity of fluorescence labelling of extracellular vesicles of different origin was confirmed. In first in vivo experiments, we assessed the dynamics of extracellular vesicle distribution and local accumulation, comparing two groups: the control group was injected with extracellular vesicles isolated from serum of healthy BALB/c mice and the experimental group received extracellular vesicles from the cell culture supernatant of highly malignant 4T1 tumor cells.

After the in vivo assessment, organs were harvested and subjected to ex vivo imaging for a biodistribution analysis. The lungs were subsequently prepared for flow cytrometric analysis of the cellular composition with regard to the immune cell infiltration. An overview over the experimental design of this study is given in Supplementary Figure 1.

All animal experiments in this study have been approved by the responsible authorities (reference of local government approval ID 81-02.04.2017-A431). All applicable institutional and/or national guidelines for the care and use of animals were followed. Forty-five female BALB/c mice (age 8–12 weeks), sourced from Charles River, were used for the experiments. Mice were dehaired using a razor and depilatory cream on back and bottom the day before start of the experiment to avoid artifacts in the fluorescence reflectance imaging (FRI) scan.

Murine breast cancer cell line 4T1 was cultured in RPMI medium (Sigma-Aldrich, St. Louis, Missouri), supplemented with XerumFree (TNCBio, Eindhoven, Netherlands) instead of extracellular vesicle-containing fetal bovine serum. Cells were grown until a confluency of approximately 80 % and cell culture supernatant was collected for extracellular vesicles isolation after 48 h of incubation.

Extracellular vesicles were isolated from the cell culture supernatant of 4T1 tumor cells (experimental group) as well as from the serum of healthy BALB/c mice (control group). Serum of healthy BALB/c mice was extracted from whole blood samples (1–1.5 ml), obtained via direct heart puncture of mice in deep prilocaine/ketamine anesthesia. Isolation of extracellular vesicles was performed as described previously for the isolation of extracellular vesicles from human plasma [10]. Briefly, the cell culture supernatant or murine serum was freed of debris with differential centrifugation, before ultrafiltration with a 0.22-μm filter. Size exclusion chromatography on a Sepharose 2B column was followed by ultracentrifugation at 105.000g for 2 h at 4 °C (see Fig. 2). The extracellular vesicle pellet was resuspended in PBS before further usage.

Quantification of extracellular vesicles based on proteins was performed in a 96-well plate using the Pierce Bicinchoninic acid assay (BCA) protein assay kit (Thermo Fisher, Waltham, Massachusetts) according to the manufacturer’s protocol. A Spark Tecan Reader (Tecan Trading AG, Maennedorf, Switzerland) was used for the read out of the plate. The results were used to determine the amount of extracellular vesicles needed for injection (100 μg). Quantification of extracellular vesicles based on the number of particles, as well as size determination, was achieved using a NanoSight system (Malvern Instruments, Malvern, UK) with a green 532-nm laser and an Andor CCD camera.

Western blotting with specific extracellular vesicles markers was used to ensure that the isolate indeed consisted of extracellular vesicles, whereas other cell organelles like golgi were not present. Western blotting was performed as described previously [11]. In brief, 20 μg of extracellular vesicles was mixed with 4-μl lane marker, boiled at 95 °C for 5 min. The samples were quickly cooled down by placing them on ice, spun down and subjected to SDS-PAGE. Proteins were transferred from the gel to a PVDF membrane (Carl Roth, Karlsruhe, Germany) using a semidry transfer method. Afterwards, the membrane was blocked with 5 % BSA-TBS-T (0.2 M Tris, 0.15 M NaCl and Tween20) and then exposed to primary antibody solution (GM130 6170823, BD Biosciences, Franklin Lakes, New Jersey; TSG101 sc-7964 / CD81 sc-23,962, Santa Cruz, Dallas, Texas), overnight at 4 °C. After a washing step, the membranes were incubated with the corresponding secondary antibody for 1 h at room temperature. After a second washing step, the resulting chemiluminescence signals of the substrate, converted by the bound HRP-conjugated secondary antibody, were detected with Pierce ECL Western Blotting Substrate (Thermo Fisher Scientific, Waltham, Massachusetts) using the FRI system.

Electron microscopy enabled visualization of the extracellular vesicles to ensure that they are of typical shape and size, and are intact after isolation. Negative staining was used to visualize the isolated extracellular vesicles. Extracellular vesicles (5 μl) were sedimented on Formvar-coated, carbon-sputtered cupper-grids.

After negative staining with 1 % phosphotungstic acid for 2 min, the samples were dried on filter paper and afterwards analyzed at 80 kV on a Tecnai 12 electron microscope (FEI, Hillsboro, Oregon). Images of selected areas were documented with Veleta 4 k CCD camera (emsis, Muenster, Germany).

Three replicate samples were prepared of each extracellular vesicles. Samples (50 μg protein) were lysed in 8 M urea, 100 mM Tris, 0.5 % (w/v) SDS, 10 mM Tris(2-carboxyethyl)phosphin (TCEP) with ultrasonic treatment (15 min) and centrifuged (4 °C, 30 min, 20.000×g). Proteins were prepared for mass spectrometry-based expression analysis with Synapt G2 Si coupled to M-Class nanoUPLC (Waters Corporation, Milford, Massachusetts) as previously described [12]. Briefly, proteins were reduced, alkylated, digested using trypsin, and injected into the instrument at 1 μg/μl (3 μl injections) after evaluating the signal intensities in diagnostic runs. Measurements were performed in replicate and data were analyzed using Progenesis QI for proteomics software (Nonlinear Diagnostics) and the UniProt database for Mus musculus. For subsequent considerations, the output was restricted to proteins with minimum two peptide hits, a fold value > 2 and an ANOVA value p ≤ 0.05. The heatmap was created using Heatmapper software [13]. Pathway and network analyses were performed using PantherDB (University of Southern California, USA) and Cytoscope software [14].

For labeling of the extracellular vesicles, 100 μg of extracellular vesicles was incubated with 2 μM of DiR (diluted with PBS) for 15 min at room temperature. Afterwards, ultracentrifugation for 90 min at 105.000g pelleted the labeled extracellular vesicles, before resuspension in PBS and preparation for further use. In order to secure a comparable signal intensity of the labeled extracellular vesicles between the tumor cell–derived extracellular vesicles and the control extracellular vesicles from healthy mice, in vitro dilution experiments were done. One hundred micrograms, 50 μg, 25 μg, and 10 μg of DiR-labeled exosomes were resuspended in PBS and pipetted into a 96-well plate, before measurement of signal intensities using the FRI system.

For the in vivo experiments, 100 μg of extracellular vesicles was labeled and resuspended in 100 μl of PBS. This amount was chosen to have a sufficient amount for the induction of changes in the immune cell composition of the lung. Injection in the tail vein was performed under isoflurane anesthesia (2 % isoflurane plus 0.9 l O₂), as were the in vivo scans.

Before injection of the extracellular vesicles, a baseline FRI scan was performed to check for individual autofluorescence signals. Imaging was then conducted at different time points after extracellular vesicle injection for the kinetics experiment (30 min, 3 h, 6 h, 12 h, 24 h, 48 h) and after 24 h to assess accumulation of the extracellular vesicles. For in vivo optical imaging, a fluorescence reflectance imaging (FRI) system (Bruker BioSpin, Billerica, Massachusetts) was used. Excitation wavelength was adapted to DiR (750 nm), and the resulting emission was recorded at 780 nm, using a filter equipped, high-sensitive charge-coupled device camera. Signal acquisition time was ten minutes for fluorescence images. White-light and X-ray images were acquired for anatomic correlation. After in vivo imaging, the mice were sacrificed and the parenchymatous organs harvested for ex vivo biodistribution analysis. The organs were placed on a petri dish and additional ex vivo FRI imaging performed similarly to the in vivo imaging.

For the analysis, a region of interest (ROI) was placed around the organs, similarly for the in vivo and ex vivo scans. FRI data were presented as mean intensity, organ intensity of ex vivo organs was normalized to the signal intensity of the muscle, resulting in an organ/muscle ratio. The fluorescence signal is measured in arbitrary units (a.u.). The lung was harvested and prepared for flow cytometry analysis.

For flow cytometric analyses, single-cell suspensions were produced from lung tissue. The cell content was determined and the cells stained with the antibodies and their corresponding isotype controls: Cd11b-PB (#101224, BioLegend, San Diego, California) and isotype control rat IgG2b,κ – PB (#400627, BioLegend, San Diego, California), Ly6C-PE (#128007, BioLegend, San Diego, California) and isotype control rat IgG2c, κ – PE (#400707, BioLegend, San Diego, California), F4/80-APC (#123116, BioLegend, San Diego, California) and isotype control rat IgG2a, κ – APC (#17–4321-81, eBioscience, San Diego, California), CD4-APC (#100412, BioLegend, San Diego, California) and isotype control rat IgG2b, κ – APC (#553991, BD Biosciences, Franklin Lakes, New Jersey), CD8a-FITC (#100706, BioLegend, San Diego, California) and isotype control rat IgG2a, κ – FITC (#553929, BD Biosciences, San Diego, California), CD3-PB (#100213, BioLegend, San Diego, California) and isotype control rat IgG2b, κ – PB (#400627, BioLegend, San Diego, California). Flow cytometry data was gated according to size and granularity to exclude cell detritus; isotype controls served for adjustment of individual flow cytometry measurements. All flow cytometry measurements were conducted using a fluorescence-activated cell sorting (FACS) Calibur system (Becton Dickinson, Franklin Lakes, New Jersey) and analyzed using the FlowJo software (FlowJo LLC, Ashland, Oregon). Data were presented as event frequency reduced by the individual isotype control to exclude unspecific staining.

Data were analyzed using the GraphPad Prism software (version 8; GraphPad Software Inc., San Diego, California). A Kolmogorow-Smirnow test ensured a normalized distribution pattern of our data, so differences in imaging between the results from the control (extracellular vesicles from the serum of healthy mice) and experimental group (tumor cell–derived extracellular vesicles) were evaluated using Student’s t test, for the fluorescent signal, as well as for differences in the immune cells. A p value below 0.05 was considered significant.

The isolation of extracellular vesicles was successful and the resulting suspension was virtually free of other organelles or cell debris, as confirmed by electron microscopy and western blotting (Fig. 1a, b). Analysis of the extracellular vesicle size with NanoSight revealed a mean size of 135.7 ± 50.6 nm for the tumor cell–derived extracellular vesicles from cell culture supernatant of 4T1 cells, whereas extracellular vesicles from the serum of healthy mice were smaller, with a mean size of 97.3 ± 36.2 nm (Fig. 1c). This is in line with studies that isolated extracellular vesicles from the serum of C57BL/6 mice with an average size of 91 nm [15] and extracellular vesicles from the cell culture supernatant of pancreas carcinoma cells (MIAPaCa) with a size range from 120.07 to 132.7 nm [16].

Proteome analysis was performed to identify differences in the extracellular vesicle cargo that might be responsible for potential differences in distribution and effects. The comparison of mass spectrometry–based protein analyses of tumor cell–derived and serum-derived extracellular vesicles revealed significant differences of their respective protein content (see heatmap in Fig. 2a). One biological replicate of serum-derived extracellular vesicles presented as outlier and was removed from statistical analysis. A total of 1762 proteins were detected which were short-listed based on their expression strength and ANOVA probability values (for data and principal component analysis, see Supplementary Tables S1 and S2, for gene ontology analysis see Supplementary Tables S4-S7). The known interactions of proteins, which were present 100-fold more in tumor cell extracellular vesicles as compared to extracellular vesicles from healthy serum, were visualized using network analysis (Fig. 2b, see Supplementary Table S8 for gene ontology analysis, see also analysis for 10-fold changes in Supplementary Table S9). Briefly, tumor cell–derived extracellular vesicles contained a substantially different set of proteins compared to extracellular vesicles isolated from healthy serum (Supplementary Tables S1 and S2). Representatives of more protein classes were present in tumor cell–derived extracellular vesicles at 100-fold higher concentration compared to control affecting about twice as many metabolic pathways as vice versa (Supplementary Table S8). Many proteins were, however, not present in such drastic concentration differences albeit at fold changes > 2. Among them were a number of highly interesting proteins in the context of cancer (Supplementary Table S3). Their network is illustrated in Fig. 2c. It shows the many interactions of heat shock factors and their direct involvement with Msn, Rab7a, Rras2, Sparc, and syndecan-4, whereas for Mib1 and Itgb2 no known connections to this network exist yet.

As a next step, after analysis of the differing cargo between the two groups of extracellular vesicles, their in vivo biodistribution was assessed. In vitro, equal amounts of extracellular vesicles from the cell culture supernatant of tumor cells and the serum of healthy mice yielded a comparable fluorescence signal (mean intensities serum–derived vs. tumor cell–derived: 320.9 ± 129 vs. 338.6 ± 149; 210.8 ± 54.1 vs. 205.1 ± 135; 98.5 ± 74 vs. 100.9 ± 23.7; 31.1 ± 17.1 vs. 27.6 ± 21.5; Fig. 1d).

In the kinetic study, DiR-labeled tumor cell–derived extracellular vesicles initially exhibited an unspecific perfusion of all organs after 30 min, before accumulation in target organs of metastasis. This accumulation was highest after six hours in the liver, before the signal intensity started to fade off again. In the lung, accumulation happened slower, with a steady increase of the signal intensity up to our last measurement at 48 h after injection (Fig. 3).

When compared to the distribution of extracellular vesicles from healthy BALB/c mice, accumulation of tumor cell–derived extracellular vesicles was significantly increased in lung, liver, spleen, and spine, whereas no differences were observed for the other analyzed organs (kidneys, pancreas, brain, and heart). Mean intensities [a.u.] of tumor extracellular vesicles vs. serum extracellular vesicles were lung 18.6 ± 6.6 vs. 10.4 ± 6.1, p = 0.01; liver 72.2 ± 14 vs. 56.5 ± 11.8, p = 0.02; spine 5.1 ± 1 vs. 3.5 ± 1.3, p < 0.01; spleen 53.4 ± 15.6 vs. 32.8 ± 13.9, p < 0.01; kidneys 3 ± 0.6 vs. 3 ± 0.3, p = 0.98 (Fig. 4).

Differences in the immune cell composition of the lungs 24 h after a single injection of 100 μg extracellular vesicles were identified between tumor cell–derived extracellular vesicles and those from the serum of healthy BALB/c mice. A trend towards a reduction in T cell abundance, with a lower percentage of CD4⁺ T cells after injection of tumor cell–derived extracellular vesicles (67 % vs. 44.5 % after injection of serum extracellular vesicles, p < 0.05) and a higher percentage of CD8⁺ T cells (35 % vs. 26 %, tumor cell–derived vs. serum-derived, p < 0.05) was observed. Furthermore, a trend towards decrease in monocytes (13 % vs. 6 %, tumor cell–derived vs. serum-derived, p = 0.09) and an increase in mature macrophages (7 % vs. 11 %, tumor cell–derived vs. serum-derived, p < 0.05) was detected (Fig. 5).