Transcytosis of Bacillus subtilis extracellular vesicles through an in vitro intestinal epithelial cell model

Bacterial EVs have been related to inter-kingdom communication between probiotic/pathogenic bacteria and their hosts. Our aim was to investigate the transcytosis process of B. subtilis EVs using an in vitro intestinal epithelial cell model. In this study, using Confocal Laser Scanning Microscopy, we report that uptake and internalization of CFSE-labeled B. subtilis EVs (115 nm ± 27 nm) by Caco-2 cells are time-dependent. To study the transcytosis process we used a transwell system and EVs were quantified in the lower chamber by Fluorescence and Nanoparticle Tracking Analysis measurements. Intact EVs are transported across a polarized cell monolayer at 60–120 min and increased after 240 min with an estimated average uptake efficiency of 30% and this process is dose-dependent. EVs movement into intestinal epithelial cells was mainly through Z axis and scarcely on X and Y axis. This work demonstrates that EVs could be transported across the gastrointestinal epithelium. We speculate this mechanism could be the first step allowing EVs to reach the bloodstream for further delivery up to extraintestinal tissues and organs. The expression and further encapsulation of bioactive molecules into natural nanoparticles produced by probiotic bacteria could have practical implications in food, nutraceuticals and clinical therapies.

Functional foods benefit human health beyond their basic nutritional properties 1 . These types of food contain chemical or biological compounds that provide health benefits with the concomitant reduction of disease risk. Functional food and drinks are products enriched with nutrients or substances such as probiotics, prebiotics or omega 3 fatty acids 2 . Probiotics are live microorganisms, which when administered in adequate amounts confer a health benefit to the host 3 .
How food products modulate host physiology remain mostly unknown 4 . An interesting approach for the prevention and treatment of various diseases is by dietary interventions, which alter the gut microbiota. Gut microbiota can impact on different systems, i.e. digestive, immune and neuroendocrine. The link between changes in the gut microbiota to gastrointestinal and extraintestinal disorders has recently begun to emerge 5,6 . In this line, probiotics act through a wide repertoire of mechanisms that positively affect composition and/or function of the gut microbiota 7 . However, the way that probiotics exert their benefic effect must still be elucidated 8 .
There have been reports about the use of Bacillus spp. as probiotics for 50 years and one of the species that has been studied is Bacillus subtilis 9,10 . Endospore formers, such as B. subtilis, are interesting because they produce heat-stable spores. This could be a great advantage for the formulation of probiotic products that can be stored in a desiccated form without deleterious effects on bacterial viability 10 . Bacillus subtilis spores can germinate in the GIT and this differentiation occurs shortly in comparison to laboratory conditions 11 . B. subtilis is the dominant functional bacterium in all naturally fermented soy-based foods in Asian countries 12 and cause a transient alteration of human intestinal microbiota 13,14 . In the case of pathogens belonging to this genera, such as B. anthracis and B. cereus, the entry into the GIT is an essential part of their virulent life cycle 15,16 .
The gastrointestinal mucosal surface is the primary interface between probiotics living in the intestine and host organs and systems 17 . The small intestine consists of heterogeneous cell populations, and absorptive enterocytes represent approximately 90% of the cells. In vitro models have been used to predict human intestinal absorption [18][19][20] . In this line, Caco-2 human cell line has been extensively used as a model for intestinal barrier studies, as a result of their physiological and morphological differentiation to enterocytes 19 . Differentiated Caco-2 cell model shows a good correlation between transport and bioavailability for lipid-based nanoparticles [21][22][23][24][25][26][27] .
The communication between mammalian gastrointestinal tract (GIT) and probiotics could in part occur through bacterial EVs and other mediators 28,29 . These led to the hypothesis that EVs could cross the epithelium of gastrointestinal mucosa and reach the bloodstream to be delivered to extraintestinal tissues and organs 30 .
Extracellular vesicles (EVs) seem to be an universal way of communication between cells as its production was reported for archaea, bacteria and eukarya 31 . These nanostructures contain biomolecules as proteins, lipids, nucleic acids and bioactive compounds. Even more, EVs can be applied in different fields as biomedicine and nanotechnology 32 . EVs are considered natural nanoparticles with potential application as delivery systems 33,34 . Several reports describing loading cargos in eukaryotic EVs such as siRNA 35 , mRNA 36 , proteins 37 and chemotherapeutics 38 have shown that EVs can function as nanocarriers. From a nanostructural point of view, EVs are membrane vesicle units constituted by lipids and surface proteins that offer advantages for cellular uptake over existing lipid-based delivery systems 39,40 . Advantages of nanocarriers include an increased solubility, higher colloidal stability and circulation time in the body as well. Additionally, overcoming the natural barriers of the body results crucial for oral delivery and to reach the blood for distribution to organs and tissues 41 .
EVs have been reported in Gram-positive bacteria, with a size between 10 and 400 nm, and have been related to bacterial physiology and its probiotic and pathogenic effects 42 . The diversity in cargo molecules contained in EVs might explain their variety of described roles ranging from bacterial competition, genetic material exchange, survival, host immune modulation and evasion, as well as infection 43 . As examples, Lactobacilli EVs are able to dampen pro-inflammatory cytokine response 44 , Bifidobacterium longum EVs effectively alleviated food allergy response in a mouse model 45 , and Lactobacillus EVs enhance host immune responses against vancomycin-resistant enterococci 42 .
The isolation and characterization of EVs from B. subtilis was reported for the first time by Brown et al. (2014), and several authors have further deepened these studies [46][47][48][49] . EVs formation occurs over the entire cellular lifespan of these spore forming bacteria and this probably does not escape what happens in GIT 14 . In this context, it is important to notice that the proteome of EVs formed during sporulation or vegetative growth differs 49 .
It has been previously demonstrated the internalization of Group B Streptococcus EVs by HeLa cells and from other Gram-positive bacteria 43,50 . But so far, no information is available on the internalization of intact Bacillus spp. EVs or their transport across human polarized epithelial cell monolayer.
The aim of this work was to explore the transcytosis process of EVs derived from the probiotic strain B. subtilis. To this end we used an in vitro transwell model of intestinal epithelial cell barrier. The transport of fluorescent-labeled EVs was visualized by Confocal Laser Scanning Microscopy (CLSM), and in combination with Confocal Time Lapse Imaging, we were able to determine the EVs trajectory within the Caco-2 cells. Transcytosis of EVs was also estimated using fluorescent measurements and Nanoparticle Tracking Analysis (NTA).

Results and Discussion
EVs uptake and internalization in Caco-2 cells determined by confocal Laser scanning microscopy analysis. We utilized our established protocol for the specific isolation of B. subtilis 168 EVs from stationary phase cultures 51 . EVs were quantified by NTA and revealed a unique population of 115 ± 27 nm (Fig. 1). The peak obtained ranged from 71 nm to 230 nm (n = 3). The EVs size obtained from NTA data was consistent with measurements previously obtained from Dynamic Light Scattering (DLS) 46,51 .
The formation of a monolayer of Caco-2 cells was confirmed by the analysis of CLSM images. These cells presented a polarized morphology, with the apical face, rich in F-actin, constituting the brush border (red), and the basal region containing the aligned cell nuclei (blue) (Supplementary Fig. S1 and Fig. 2a).  Supplementary Fig. S2). At the beginning of the incubation, the majority of CFSE-EVs (green) were found on the brush border of the enterocyte-like cells. At longer incubation time points most of the fluorescent signal measured was intracellular, suggesting active uptake and internalization of intact EVs (Fig. 2a). CFSE labels EVs protein cargo as well as transmembrane proteins via the action of intravesicular esterases, allowing only the detection of intact EVs 52,53 . The amount of EVs incorporated into Caco-2 cells seemed to be time-dependent. The maximum EVs uptake by cells occurred at 120 min, and decreased at 240 min ( Fig. 2a,b). At 240 min, most of the CFSE-EVs observed were located on the basal face ( Fig. 2a and Supplementary Fig. S2). As expected, CFSE fluorescence levels inside Caco-2 had a similar pattern to that obtained from the quantification of EVs per cell (Fig. 2b,c). It should be noted that the measured areas of the fluorescent EVs could not reflect their actual size because the point-spread function contributes to some degree of spreading (i.e., blurring) 54 . In this case, the theoretical lateral resolution for the fluorescent microscope is 183 nm (diameter=λem/2 N.A.; λem = 521 nm and N.A. = 1.42) 51 . The reduction of signal inside the cells, at 240 min in comparison to 120 min (n = 3, p < 0.05), could be explained by the release of the EVs and/or degradation of the CFSE dye in lysosomes (Fig. 2b,c). For that reason, we decided to carry out the transcytosis assays with a maximum incubation period of 240 min.
Our results are in accordance with Li et al. (2017), which described that Lactobacillus plantarum EVs could be internalized into Caco-2 cells 42 . Further research would be necessary to elucidate whether this is a cell type-specific process. Results from other studies showed that fluorescently labeled EVs can be taken up by virtually all cell types tested, whereas others suggest that vesicular uptake is a highly specific process which can only occur when cells and EVs share the right combination of ligand and receptor 55 . For example, ex vivo bone marrow-derived mast cells have a higher uptake ability compared to other cell types, and can efficiently take up EVs from Enterococcus faecalis as well as from Bifidobacterium longum, during an incubation period of 120 min 45 . Taking into account these data, Caco-2 cells would be capable of internalizing EVs from different probiotics strains.
Importantly, we tested that EVs from B. subtilis 168 collected from stationary phase cultures had no effect on Caco-2 monolayer cellular proliferation, viability or cytotoxicity at the evaluated incubation time points (n = 8, P > 0.05) (Fig. 3). In line with this observation, EVs from the probiotic strain L. plantarum do not cause any significant toxicity to Caco-2 cells after 24 h incubation 42 . However, EVs from Lactobacillus rhamnosus were shown to have a significant cytotoxic effect on a human liver epithelial cell line (HepG2) when incubated at an intermediate concentration for 24 h 56 . Concerning to the pathogenic bacteria Staphylococcus aureus, EVs of different strains show different cytotoxic activity toward host cells. The presence or absence of cytotoxicity may result from proteome differences in S. aureus-secreted EVs, but further analysis would be necessary to identify cytotoxic factors in these EVs 57,58 .

Bacterial EVs transcytosis across Caco-2 cells as determined by fluorescence and nanoparticle
tracking analysis measurements. In general terms, EVs transport across cellular barriers could be achieved either by a paracellular route or a transcellular route. The paracellular route corresponds to the passage between epithelial cells via transient disruption of the tight junctions that interlock adjacent cells. Instead, the transcellular route corresponds to the internalization, endocytosis, transport across the cell body and secretion at the opposite cell surface (transcytosis). TEER is an accepted technique to measure quantitatively the integrity of epithelial cells monolayer, which is determined by the tight junction dynamics. This approach is used to evaluate drugs or lipid-based nanoparticles transport 25,59 .
In our assays, Caco-2 cells were differentiated on transwells and TEER was measured every 2-3 days. On day 14 after cell seeding, TEER measurement stabilized at 1714 Ω*cm 2 ± 72 Ω*cm 2 , and Caco-2 cells monolayers were then used for transcytosis assays (Fig. 4a). Validation of Caco-2 permeability barrier in this model was also done using immunostaining of tight junctions with anti-occludin primary antibodies and Alexa 488 labeled secondary antibodies on days 3 and 14, showing no disruption of the barrier ( Supplementary Fig. S3).
To carry out the transcytosis assays, CFSE-labeled EVs were incubated with Caco-2 cells monolayers in the upper chamber. After different time points, fluorescence was measured in the medium collected from the lower chamber. Fluorescence was first detected at 120 min and the intensity of the signal increased 43% after 240 min of incubation (Fig. 4b). Transwells without cells were used as a control of passive diffusion. In this case, just after 5 min of incubation fluorescence levels were detected in the medium collected from the lower chamber and increased until 60 min, remaining stable for the rest of the incubation time. www.nature.com/scientificreports www.nature.com/scientificreports/ To estimate the uptake efficiency of EVs, we compared the fluorescence units recovered in the lower chamber to the fluorescence from CFSE-labeled EVs seeded on the upper chamber (see Mat & Met). Considering that 6212 fluorescence arbitrary units (a.u.) were seeded (approximately 5 × 10 9 EVs), EVs uptake was 20% ± 2% at 120 min www.nature.com/scientificreports www.nature.com/scientificreports/ and 29% ± 13% at 240 min. At both time points, the fluorescence units remaining in the upper chamber plus the fluorescence units recovered in the lower chamber equaled 95% of the total fluorescence seeded in the transwell. We speculate that the remaining 5% of the total fluorescence might be inside the cells within the compartments of endocytosis and/or transcytosis pathways and/or degradation path through lysosomes 60,61 .
In order to quantify the number of EVs transcytosed through Caco-2 monolayers we analyzed the samples by NTA. As mentioned before, about 5 × 10 9 EVs were seeded in the upper chamber. EVs were first detected in the lower chamber after 60 min of incubation (Fig. 4b). The number of EVs measured in the lower chamber increased 38% at 120 min and up to 76% at 240 min. Based on the measurement of fluorescence, the appearance of EVs only after 60 min of incubation was unexpected but might be explained by a higher sensitivity of NTA compared to fluorescence to detect the presence of nanoparticles in the sample 62 .
We further estimated the efficiency of uptake using the values obtained by NTA. In this case, 27% ± 11% of the seeded EVs were obtained in the lower chamber at 60 min, 38% ± 14% at 120 min and 46% ± 14% at 240 min. Even though for quantification of EVs by NTA we selected those particles in a range from 71 nm to 230 nm, we could not discard the presence of EVs produced by Caco-2 cells (n = 3).
In order to corroborate that the measured particles corresponded to the transcytosed B. subtilis EVs, we performed a dose-dependent curve using an incubation period of 240 min (Supplementary Fig. S4). As seen in Supplementary Fig. S4 the number of recovered EVs in the lower chamber increased concordantly with and increment of the seeded EVs. An efficiency of uptake of 19% ± 5% was observed for the more concentrated sample that was seeded with 5 × 10 9 EVs, the same amount used in our previous experiment. Moreover, non-significant differences in the efficiency of uptake were observed when half of the EVs were seeded in the transwell, indicating that the number of EVs used in our experiments was in excess for this process (n = 3, P > 0.05). Furthermore, we only detected a small number of particles of the analyzed size in the lower chamber when no B. subtilis EVs were seeded ( Supplementary Fig. S4), supporting that the number of particles released by Caco-2 cells did not interfere with our results. Overall, considering the data of both experiments measured by NTA, the calculated efficiency of uptake of B. subtilis EVs after 240 min of incubation ranged from 19% to 46%. This is also in agreement with the 29% obtained from fluorescence measurements.
To additionally verify the presence of intact transcytosed EVs, aliquots collected from the lower chamber were examined by Epi-fluorescence Microscopy. We observed intact EVs with a spherical shape in the aliquots collected from the lower chamber at 120 min (Fig. 4c). At that time, EVs showed an identical spherical shape compared to the EVs analyzed from aliquots collected from the upper chamber (data not shown) and from previously purified EVs 51 .
In our assays, the TEER value was constant during 240 min of EVs incubation (1730 Ω*cm 2 ± 84 Ω*cm 2 ). Therefore, it can be concluded that EVs transport across the Caco-2 cells exerted no effects on integrity and permeability of cell monolayers. Cell-free supernatant of the probiotic Bifidobacterium lactis increases tight junction integrity, while that from the pathogenic bacteria Escherichia coli O157:H7 (EHEC) induces damage on tight junction integrity. Furthermore, pre-administered B. lactis cell-free supernatant protects the tight junctions from EHEC-induced damage 63 . In this same work, the authors suggested the beneficial effects of specific probiotic bacteria in protecting intestinal epithelial cells from the deleterious effects of pathogenic bacteria was mediated by bioactive metabolites present in the culture supernatant from B. lactis. The authors concluded that further studies would be required to identify the active compounds and structural components of the supernatant that make possible the probiotic effect. Undoubtedly, it would be of great interest to elucidate whether EVs play a role in exerting directly these probiotic beneficial effects or are acting as delivery systems of bioactive compounds.
The concept of prokaryotic EVs transcytosis across intestinal epithelial cell has been previously proposed in the scientific literature but not demonstrated. Zhang et al. (2013) showed, in an in vitro model of intestinal porcine enterocytes, that EVs of Helicobacter suis were internalized, and the activity of cargo enzymes was detected in the lower chamber in a transwell system 64 56 . On the other hand, B. anthracis produces toxin-laden EVs allowing a concentrated delivery of toxin components to target host cells, a mechanism that may increase toxin potency. These EVs are toxic for macrophages, and are capable of inducing a protective response in immunized mice 65 . In line with this, it would be interesting to study whether EVs from pathogens such as B. anthracis and B. cereus cross human intestinal epithelial cells, and check whether this step is an essential part of their virulent life cycle. If it is the case, the transcytosis of EVs across the epithelial cells of the GIT might be a universal mechanism shared by probiotic and pathogenic bacteria 64 .

EVs internalization and movement in Caco-2 cells determined by confocal time lapse imaging.
In order to monitor the movement of EVs through Caco-2 cell monolayers, we carried out internalization assays and measured the transit of EVs by Confocal Time Lapse Imaging. An imaging area (XY) was fixed and consecutives slices obtained from Z-stacks were collected using 1.4 µm intervals. Series of time-lapse scans were taken at intervals of 10 min during the 120 min incubation with EVs. To infer EVs movement over time from these four coordinates (X, Y, Z and time) we assigned individual colours to each time point with an image processor (FIJI software). Then images were merged to observe the sequence of colours (time) through the XYZ space; i.e. we expected to find in the highest Z-stacks (apical face of Caco-2 cells) colours assigned to shorter time points, and in the lower Z-stacks (basolateral face of the cell) colours assigned to longer incubation time points. Figure 5a shows a panoramic view of different EVs where the sequence of colours corresponds to red (0 min), green (60 min) and blue (120 min) through the Z axis. As a representative example, Fig. 5b shows a zoom area where the sequence red, green and blue, corresponding from highest to lowest values of the Z axis, can clearly be seen. When this sequence of images was analysed through a 3D reconstruction software tool, the typical spherical Scientific RepoRtS | (2020) 10:3120 | https://doi.org/10.1038/s41598-020-60077-4 www.nature.com/scientificreports www.nature.com/scientificreports/ shape of EVs and their movement through Z axis could be elucidated (Fig. 5c). Such a typical spherical shape from the 3D reconstruction came from 2-3 Z-stacks. Supplementary Video S1 shows the images superposition of 0, 60 and 120 min (Huygens Professional software).
If we consider the number of Z-stacks and the time required by EVs to pass through them, it is possible to calculate an approximate velocity of EVs transportation inside the Caco-2 cells. To do this, the number of Z-stacks crossed by one EV in the first hour can be calculated as the number of Z-stacks between the first red Z-stack and the first green Z-stack. In this case, the average of Z-stacks crossed by eight EVs was 2.69 Z-stacks ± 0.41 Z-stacks during the first hour of incubation. If we take into account the distance between Z-stacks (1.4 µm), the height of the monolayer of Caco-2 cells through the images (9.6 μm), and assuming that this speed is constant through the cells, we can estimate the time required by EVs to pass through the Caco-2 monolayers. This estimated time was 152 ± 28 min (n = 8), which is in accordance with our previous fluorescence results (Figs. 2  and 4). Supplementary Video S2 shows all Z-stacks of CFSE-EVs (green) internalized into the Caco-2 cells after 60 min of incubation (FIJI software).
On the other hand, when this image was analysed with a Z-slices projection, the merge of the 3 colours was obtained. This would indicate scarce movements on the XY plane (Fig. 5d). To measure the movement in X and Y axis, the coordinates in X and Y axis at 0, 60 and 120 min for 9 EVs were obtained using FIJI software (Supplementary Fig S5a). To quantify the movement, the mean value for the difference between coordinates at 0-60 and 60-120 min was calculated for these 9 EVs (Supplementary Fig. S5b). No statistically significant differences were found between movement on X and Y axis for the examined time periods (n = 9, p > 0.05) (Supplementary Fig. S5b). Based on these results, the movement on the X axis (1.5 μm ± 0.5 μm) and Y axis (1 μm ± 0.2 μm) resulted 70% lower compared to the travelled distance calculated on Z axis during the first hour (3.8 μm ± 0.6 μm). This last value was calculated as mentioned above considering the average of Z-stacks crossed by EVs in the first hour and the distance between Z-stacks.
There is an increasing amount of evidence showing that bacterial EVs can enter into eukaryotic cells and deliver their cargo 43,55 . Cells appear to take up EVs by a variety of endocytic pathways, including clathrin-dependent endocytosis, and clathrin-independent pathways such as caveolae-mediated uptake, macropinocytosis, phagocytosis, lipid raft-mediated internalization and direct membrane fusion 55 . Interestingly, pathogenic Gram-positive bacteria EVs seem to be internalized by eukaryotic cells through endocytic mechanism in several epithelial and macrophage cell lines 50,66,67 . Moreover, EVs could contact the host cells via cell membrane receptors to further initiate a cascade of intracellular signaling paths triggering the inflammatory response 66,68 . Thay et al. (2013) reported that the fusion of S. aureus EVs with the plasma membrane represents a route for delivery of a key virulence factor to human cells 69 . Except for membrane fusion, where the EVs content is released into the cytoplasm directly, whole EVs can enter into the recipient cell, whereby cargo release is mediated by fusion of the EVs membrane with the endosome membrane as described by Tian et al. (2013) 70,71 . Recently, a transport pathway for molecules from blood to brain cells was described by Villaseñor et al. (2017), involving dynamic intracellular tubules 72 . This transcytosis process regulated by sorting tubules is an additional pathway for receptor-mediated transcytosis that is distinct from non-selective transport like the caveolae class and trans-endothelial channels.
Our results suggest that Caco-2 cells were able to take up intact EVs from B. subtilis and, at least part of them, are not fused with cell membranes and are transported and secreted at the opposite cell surface side. Possibly, EVs populations, heterogeneous in terms of content, could enter into the cell by more than one way. Further studies should elucidate how bacterial EVs are transported across the intestinal epithelial cell barrier.

conclusions
In this work, we demonstrated that intact EVs from B. subtilis could be transported across Caco-2 cells differentiated to enterocyte-like cells. Images from CLSM demonstrated that B. subtilis EVs uptake and internalization by Caco-2 cells were time-dependent. Using a combination of Fluorescence and NTA measurements in a transwell system, we were able to show that intact EVs could be translocated from the upper to the lower chamber with an average efficiency of uptake of 30% after a 240 min incubation period and this process was also dose-dependent. Using a Time Lapse Microscopy technique we elucidated that EVs movement was mainly through Z axis and slightly on X and Y axis. In conclusion, our studies could partially explain one of the possible mechanisms of communication between probiotics and host. This mechanism could be a first step that allows EVs to reach the bloodstream and consequent delivery to extraintestinal tissues and organs. The encapsulation of proteins, siRNA, mRNA, and chemotherapeutics into EVs could have practical implications in food, nutraceuticals and clinical therapies. www.nature.com/scientificreports www.nature.com/scientificreports/ cells, cultures were centrifuged for 25 min at 4000 g, at 25 °C. Supernatant was filtered through 0.8, 0.65 and 0.45 μm to remove cellular debris. Centricon ultrafiltration system (cut off = 100 kDa) was used for concentration the filtrated supernatant. Then, the concentrated supernatant was filtered again through 0.45 μm filter to remove www.nature.com/scientificreports www.nature.com/scientificreports/ the aggregated material. This filtered supernatant was centrifuged 120 min at 110,000 g, 4 °C using a SW 41 Ti rotor (Beckman Optima L-80). The resulting pellet was washed with a phosphate buffered saline solution (PBS). EVs contained in the pellet were labeled, resuspended in PBS or Dulbecco's modified Eagle's medium (DMEM), filtrated by a 0.45 mm filter and used in further determinations.

Materials. Bacillus subtilis
Nanoparticle tracking analysis (NTA). Particle number and size distribution was obtained using a Nanosight NS3000 System equipped with a 488 nm laser. To this end, EVs resuspended in PBS were injected in the device. Three videos of 10 seconds were obtained at camera level 9, at 25 °C for each sample. NTA 3.2 software was used for data analysis with a detection threshold 3 and maintaining other settings at default. PBS was completely particle-free at these settings.
EVs labeling with 5-(6) carboxyfluorescein diacetate N-succinimidyl ester (CFSE). EVs were stained with CFSE at 10 mM final concentration in filtered PBS for 30 min at 37 °C and washed twice with filtered PBS 51,73 . This non-fluorescent dye diffuses across EVs cell membrane where esterases of the plasmatic membrane cleave its acetate group forming fluorescent CFSE. This fluorogenic dye is now not permeable and, consequently, is confined into EVs.

EVs uptake and internalization in Caco-2 cells by Confocal Laser Scanning Microscopy (CLSM).
Differentiated Caco-2 monolayers were obtained as specified above, washed with PBS and incubated with 1.3 × 10 9 labeled EVs resuspended in basal DMEM at different times (0, 5, 15, 30 45, 60, 120, and 240 min) at 37 °C in a humidified atmosphere containing 5% CO 2 64 . After incubation, Caco-2 cells were fixed with paraformaldehyde 4% (w/v) in PBS for 20 min at 25 °C. After several washes with PBS, cells were permeabilized for 10 min with Triton X-100 dissolved in PBS (0.05%, v/v) at 25 °C. Following washes, non-specific binding sites were blocked with 3% (w/v) BSA in PBS 1 h at RT. After that, cells were incubated with Rhodamine Phalloidin 0.1% (w/v) and To-Pro3 Iodide 0.1% (w/v), both in PBS 1 h at RT. Finally, coverslips were mounted with Mowiol and examined by CLSM (Olympus FV 1000 module). CFSE dye has a λexc = 494 nm and λem = 521 nm, which was measured using 488 nm laser excitation and 505/525 band pass filter for detection 73 . Rhodamine Phalloidin has a λexc = 540 nm and λem = 565 nm, which is measured using 543 nm laser excitation and 560/620 band pass filter for detection 74 . To-Pro3 Iodide has a λexc = 642 nm and λem = 661 nm, which is measured using 633 nm laser excitation and 655/755 band pass filter for detection 75  www.nature.com/scientificreports www.nature.com/scientificreports/ = × Uptake efficiency (fluorescence lower chamber/fluorescence upper chamber) 100 Transwells without cells were used as control of passive diffusion 76 . TEER was measured during the incubation with EVs to verify the barrier integrity of the differentiated Caco-2 cell monolayer 77 . The experiment was performed in triplicate.
confocal time lapse imaging. Labeled EVs were resuspended in basal DMEM. Caco-2 monolayers were obtained as specified above, washed with PBS and incubated with 115 µg of labeled EVs at 37 °C in a humidified atmosphere containing 5% CO 2 (live chamber controller) 78 . An imaging area (XY) was fixed and 20 slices and Z-stacks were collected using 1.4 µm intervals. Series of time-lapse scans were taken at intervals of 10 min during 120 min. The fluorescent images were taken using an Olympus FV 1000 module using the same microscopy settings described above. Digital pictures were processed with FIJI software (National Institutes of Health, Baltimore, USA) and Huygens Professional software (Scientific Volume Imaging, Netherlands).
Mtt assay. This colorimetric assay measures the reduction of yellow 3-(4,5-dimethylthiazol-2-yl)−2,5diphenyltetrazolium bromide (MTT) by mitochondrial succinate to purple colored formazan crystals. Since reduction of MTT can only occur in metabolically active cells the level of activity is a measure of cellular proliferation, viability and cytotoxicity 79 . Briefly, cells were grown in 96 well plates and incubated with approximately 5 × 10 9 EVs suspended in basal DMEM at 37 °C for 0, 5, 15, 30 45, 60, 120, and 240 min. Then, cells were incubated with 200 µl MTT solution (5 mg/ml MTT in basal DMEM) for 120 min at 37 °C. The medium was removed and the crystals formed were dissolved in 200 µl of DMSO per well. Absorbance was measured at 570 nm and 690 nm (background) using POLARstar Omega microplate reader. The relative cell viability was calculated as Abs570nm-Abs 690 nm. The assay was performed in octuplicate.
Statistical analysis. The number of internalized EVs and fluorescence units per cell analyzed by CLSM, the data obtained from MTT assay and the number of EVs measured by NTA were analyzed by means of a one-way ANOVA in a completely randomized design. The data obtained from EVs movement on X and Y axis analyzed by Confocal Time Lapse Imaging were analyzed by means of a 2-way ANOVA with interactions (axis and time) in a completely randomized design. Tukey's test was used for post hoc comparisons. The assumptions of normality and homoscedasticity were studied analytically by the Shapiro-Wilks and the Levene tests, respectively. Results were considered significant at p < 0.05. Statistical analyses were performed using the statistical program Infostat (Universidad Nacional de Córdoba, Argentina).