Tumor-Derived Extracellular Vesicles Breach the Intact Blood-Brain Barrier via Transcytosis

Br-EVs Cross the Brain Endothelium via Transcytosis.

To determine the mechanism(s) by which Br-EVs breach the brain endothelium, we first developed an in vitro static BBB model. Primary human brain endothelial cells rapidly lose their junctional features in culture and therefore cannot recapitulate the integrity of the in vivo BBB.²² It has been shown that with an increase in internal cAMP, these cells can regain their barrier characteristics.²³ Accordingly, in our BBB model, we treated human brain endothelial cells cultured on transwell filters with a combination of CPT-cAMP and Ro20–1724, an inhibitor of cAMP degradation,²³ to enhance the expression of junctional proteins such as ZO-1 (Supplementary Figure S2a,b). This treatment resulted in an approximately 50% and 80% reduction in the permeability coefficient of the endothelial monolayer to 10 kDa Alexa 647 (post-treatment P_app 2.15 × 10⁻⁶ ± 4.964 × 10⁻⁷ cm/s) and 70 kDa fluorescein isothiocyanate (FITC) dextrans (post-treatment P_app 1.933 ×10⁻⁷ ± 6.26 × 10⁻⁸ cm/s), respectively (Supplementary Figure S2c).

To determine whether the transfer of EVs across the brain endothelial monolayer is through an active or passive mechanism, Gaussia luciferase-labeled Br-EVs were incubated with brain endothelial cells in the luminal (top) chamber of the transwell filters for 2 h at 37 or 4 °C. The amount of luminescent signal detected in the abluminal (lower) chamber was significantly lower when the filters were incubated at 4 °C (Figure 2a), suggesting that a mechanism that is active in physiological conditions is involved in the transport of Br-EVs across the brain endothelial monolayer. Moreover, treatment of cells with Dynasore, an inhibitor of endocytosis,²⁴ also resulted in a dose-dependent decrease in the abluminal signal (Figure 2b). The permeability of the barrier to 10 kDa and 70 kDa dextran was not changed by Br-EVs during this incubation (Figure 2c). To verify that the source of the detected signal in the abluminal chamber is the luciferase associated with intact EVs as opposed to free luciferase, we ultracentrifuged the media from the lower chamber on an iodixanol OptiPrep density gradient. As a positive control, Gaussia luciferase- labeled Br-EVs were added directly to the top of a gradient for ultracentrifugation. Consistent with our previous findings, in the positive control group, luminescent signal was detected at low- and high-density fractions, corresponding to EV density of 1.105–1.184 g/mL (Supplementary Figure S 2d). In the fractions collected from the lower chamber media, luminescent signal was detected in the high-density fraction, corresponding to EV density of 1.184 g/mL (Figure 2d), confirming the EV source of the signal. It should be noted that some signal was also detected in the supernatant, indicative of free luciferase that could have been released during the processing and degradation of a subpopulation of EVs.

Interestingly, electron microscopy analysis of the low- and high-density fractions of EVs showed that high-density Br-EVs generally had a smaller size with 68% being below 70 nm, whereas this percentage was 34% in the low-density EVs (Supplementary Figure S2e,f). This finding suggests that a high-density subpopulation of EVs that are smaller in size can undergo a transcellular transport. Previous studies have attempted to isolate EV subpopulations with different densities.^25,26 Consistent with our findings, one study found two distinct subpopulations of EVs with low and high density and showed that the high-density EVs were smaller in size.²⁵

This study also demonstrated that the two subpopulations had distinct protein and RNA profiles. More recently, using an asymmetric flow field-flow fractionation method, a study isolated a subpopulation of extracellular vesicles smaller than 50 nm that were introduced as exomeres.²⁷ Whether the subpopulation of EVs for which transcytosis was observed in our studies have similar characteristics to those of exomeres remains a matter of future investigation.

To confirm that the static incubation of EVs with endothelial cells as well as the increase in internal cAMP in endothelial cells do not act as confounding factors on the mechanism of EV transport, we verified these findings in a microfluidic organ- on-a-chip model of the BBB (BBB chip).²⁸ The BBB chip is a 2-channel microfluidic culture device that contains a vascular channel lined by induced pluripotent stem cell-derived human microvascular endothelial cells, which is separated by a porous extracellular matrix-coated membrane from an abluminal channel containing primary human astrocytes and pericytes.²⁸ TdTom-Br-EVs were flowed through the lumen of the vascular channel for 5 h. Fluorescent signal was detected in the abluminal chamber at 3 h and increased significantly over time (Figure 2e), under conditions in which permeability of the barrier to 10 kDa and 70 kDa dextran did not change (Figure 2f). Fluorescence microscopy analyses also revealed the presence of Br-EVs that were taken up by astrocytes in the abluminal chamber (Figure 2g). Overall, these findings demonstrated that Br-EVs can interact with endothelial cells under flow conditions and continuously cross the endothelial monolayer through transcytosis.

We next explored the transcytosis of Br-EVs in vivo, using a zebrafish model. Zebrafish develop a mature BBB at 3 days postfertilization (dpf) and serve as a suitable model for BBB studies.²⁹ We conducted intracardiac injection of TdTom-Br- EVs in 6–7 dpf Tg(kdrl:GFP) zebrafish embryos and monitored the distribution of Br-EVs in the brain through live imaging. At the time of imaging, Br-EVs were taken up by a number of cells within the brain parenchyma, demonstrating their ability to go beyond the BBB in vivo (Figure 2h). Moreover, with time-lapse imaging, movement of EV-containing endocytic vesicles within endothelial cells could be observed. As shown in Figure 2h, a number of these vesicles moved toward the plasma membrane and fused with the membrane, suggestive of a transcytosis process. The integrity of the BBB remained intact throughout the duration of these experiments, as treatment with Br-EVs did not increase the permeability of the BBB to 10 kDa and 70 kDa dextran in zebrafish (Figure 2i,j).

Together, these in vitro and in vivo findings demonstrate that a subpopulation of Br-EVs can breach the brain endothelial barrier through transcytosis, in a manner that does not compromise junctional permeability. This finding provides an opportunity to develop efficient approaches to address the long-standing clinical challenge of delivering therapeutics to the brain. The previous attempts for brain delivery of a variety of antibody-based drugs have been associated with delivery rates of ~0.1%.^30,31 Comparable to these reports, in our BBB chip model, we found the efficiency of EV transfer across the endothelial barrier to be 0.16% ± 0.0003%, including the technical loss of EVs that usually occurs due to the entrapment of vesicles on the membrane between chambers or their uptake and processing by cells in the barrier. In the field of drug delivery, several advantages have been suggested for EVs as drug delivery vehicles compared to other currently available nanomedicines, for example, lack of immunogenicity and enhanced distribution.⁶ Our discovery of the inherent ability of tumor-derived EVs to breach an intact BBB via transcytosis provides an additional advantage to EV-based drug delivery to the brain and can potentially be generalized to a variety of EV types for such applications. Moreover, the observed association between the physical characteristics of EVs and their ability to breach the BBB underscores an important factor that should be taken into consideration in the development of future brain drug delivery approaches.

Br-EV Transcytosis Involves Caveolin-Independent Endocytosis, Recycling Endosomes, and Basolateral SNAREs.

We further explored the mechanistic details of Br-EV transport through brain endothelial cells by focusing on the three major steps of transcytosis: (1) endocytosis through the apical (luminal) membrane of brain endothelial cells, (2) intracellular trafficking, and (3) release into the extracellular environment from the basolateral (abluminal) membrane. To evaluate the mechanism(s) of endocytosis, we pretreated brain endothelial cells with chemical inhibitors of the different endocytosis pathways and measured the uptake of TdTom-Br-EVs via flow cytometry. Inhibition of clathrin-dependent endocytosis by chlorpromazine³² and a Cdc42/Rac1 GTPase inhibitor, ML141,³³ significantly decreased the uptake of Br- EVs (Figure 3a). Inhibition of macropinocytosis by 5-(N-ethyl- N-isopropyl) amiloride (EIPA) and cytochalasin D³² also led to a significant decrease in the uptake of Br-EVs. Further confirming these findings, Br-EVs partially colocalized with transferrin and 70 kDa dextran, markers of clathrin-dependent endocytosis³⁴ and macropinocytosis,³⁵ respectively (Figure 3b). Filipin, an inhibitor of caveolin-dependent endocytosis,³² showed no effect on Br-EV uptake by endothelial cells (Figure 3a). A lack of colocalization of Br-EVs with caveolin also indicated that caveolin-dependent endocytosis is not involved in the uptake of Br-EVs by brain endothelial cells (Supplementary Figure S3a). At the BBB, both clathrin- and caveolin-dependent endocytosis pathways have been shown to initiate the transcytosis of different macromolecules.³⁶ Consistent with these well-defined mechanisms, our findings suggest the involvement of the clathrin-dependent pathway but not the caveolin-dependent pathway in the transcytosis of Br-EV across the BBB. Our findings also implicate the involvement of macropinocytosis in the uptake of Br-EVs by brain endothelial cells. However, the association of macropinocytosis and transcytosis of macromolecules is yet to be elucidated.

Next, we evaluated the intracellular trafficking of Br-EVs in brain endothelial cells. Upon endocytosis, the majority of molecules are transferred to early endosomes, where they are sorted to different routes. Molecules sorted into late endosomes are eventually transferred to lysosomes for degradation, whereas molecules sorted into recycling endosomes will be transferred to the plasma membrane.³⁷ Rab11 recycling endosomes have been shown to be involved in the transcytosis of macromolecules.^38,39 As expected, following endocytosis, Br-EVs colocalized with EEA1, a marker of early endosomes⁴⁰ (Supplementary Figure S3b). To examine whether Br-EVs can proceed through the recycling route in the endocytic pathway for transcytosis, we evaluated the colocalization of Br-EVs with rab11, a marker of recycling endosomes.⁴⁰ We found that Br- EV-containing vesicles colocalized with rab11 in the perinuclear region (62.9% ± 1.27% colocalization of red and green fluorescent signal, Figure 3c,e). We also evaluated the trafficking of Br-EVs to the degradation route using BODIPYconjugated DQ-ovalbumin as a marker of endolysosomal structures.^41,42 DQ-ovalbumin is a self-quenched marker that only fluoresces upon the release of quenching following degradation in late endosomal structures and lysosomes.⁴² As expected, colocalization of a subset of Br-EVs with DQ-ovalbumin was also observed in the perinuclear region (61.1% ± 6.4% colocalization of red and green fluorescent signal, Figure 3d,e). The trans-Golgi network can also be involved in the transfer of molecules back to the plasma membrane.⁴³ We found no colocalization of Br-EVs with TGN46, a marker of the trans-Golgi network,⁴³ suggesting that this network is not involved in the intracellular trafficking of Br-EVs in brain endothelial cells (Supplementary Figure S3c). These findings demonstrate that different subpopulations of Br-EVs can be sorted into different endocytic pathways that would lead to their recycling and transcytosis or degradation.

Rab11 recycling endosomes can transfer their contents both to the apical membrane for recycling and exocytosis and to the basolateral membrane for recycling and transcytosis of molecules.^38,39,44 To evaluate the last step of EV transcytosis (i.e., the release of EVs on the abluminal side), we studied the factors involved in the interaction of EV-containing endosomes with the basolateral membrane. Soluble NSF attachment protein receptors (SNAREs) are known to be involved in vesicle fusion with the target membrane and include vesicle SNAREs (v-SNAREs) and target SNAREs (t-SNAREs).⁴⁵ Among the different types of v-SNAREs, vesicle associated membrane protein (VAMP)-3 is associated with recycling endosomes and is involved in exocytosis, whereas VAMP-7 is involved in the fusion of late endosomes with lysosomes.⁴⁶ Our microscopy analyses demonstrated that Br-EV-containing vesicles colocalized with both VAMP-3 and VAMP-7 (Figure 3f–h). However, colocalization with VAMP-3 was significantly higher than that with VAMP-7, suggesting that recycling of Br-EVs was a prominent event in this case. The fusion of recycling endosomes with the basolateral plasma membrane, required for transcytosis of macromolecules, can occur through the interaction of VAMP-3 with SNAP23/syntaxin 4, a t-SNARE complex localized on the basolateral membrane.^47,48 Here, we found that Br-EV-containing endosomes colocalized with both SNAP23 and syntaxin 4 (Figure 3i,j), demonstrating the involvement of these SNARE complexes in the fusion of these vesicles with the basolateral membrane. It is important to note that our findings do not exclude the possibility of the transfer of Br-EV-containing recycling endosomes to the apical membrane through other mechanisms, a matter of further investigation.

Taken together, these findings demonstrate that while a subpopulation of Br-EVs are sorted into late endosomes for degradation, a large subset of these EVs can be sorted into rab11⁺ recycling endosomes, which could lead to the VAMP3/ Snap23/syntaxin 4-dependent release of these EVs at the basolateral membrane.

Br-EVs Downregulate Endothelial Rab7 to Facilitate Their Transport.

Under physiologic conditions, the rate of transcytosis in the BBB is maintained at a low level through different regulatory mechanisms.² This characteristic of the BBB raised the question as to whether Br-EVs can facilitate their transcellular transport. We hypothesized that Br-EVs can specifically modulate the endocytic pathway in brain endothelial cells to increase their transport efficiency. We evaluated the effect of EV treatment on the two major routes in the endocytic pathway, degradation and recycling. Interestingly, we found that Br-EV treatment of brain endothelial cells significantly decreased the expression of the late endosomal marker rab7, whereas the expression of rab11, marker of recycling endosomes, was not changed (Figure 4a–c). This finding demonstrated that Br-EVs exhibit the ability to modulate the degradation pathway in brain endothelial cells.

Rab7 is involved in the transfer of early endosomes to late endosomes and late endosomes to lysosomes.⁴⁹ To determine whether the decrease in endothelial rab7 can lead to a decrease in the transfer of molecules to lysosomes, we used siRNA to knockdown the expression of rab7 in brain endothelial cells and treated the cells with DQ-Ovalbumin, which fluoresces upon being processed in late endolysosomal structures.⁴² Rab7 KD decreased the fluorescent signal from DQ-Ovalbumin, suggesting a decrease in the transfer of this molecule to late endosomes and lysosomes (Figure 4d,e). The total number of lysosomes as measured by the lysosomal marker, LAMP1, was not changed by Rab7 KD.

Rab7 can also interact with and increase the activity of rac1, a small GTPase protein that acts as a central regulator of actin remodeling.^50–52 Increased rac1 activity has been associated with an increase in the rate of macropinocytosis⁵² and a decrease in clathrin-mediated uptake of molecules such as epidermal growth factor and transferring.⁵³ Accordingly, we hypothesized that the Br-EV-driven decrease in rab7 can indirectly affect the rate of EV endocytosis. Our flow cytometry studies demonstrated that rab7 KD significantly increased the uptake of Br-EVs by brain endothelial cells (Figure 4f,g). Fluorescent microscopy of Br-EV uptake by endothelial cells demonstrated that the pattern and the size of Br-EVscontaining endosomes were not different between Rab7 and control siRNA-treated cells (Figure 4h,i). This result confirms that increased signal detected by flow cytometry is due to increased uptake of Br-EVs rather than the accumulation of Br-EVs in late endosomal structures. Whether the observed increase in uptake of Br-EVs is specific to the subpopulation that is taken up through clathrin-dependent endocytosis and can be subject to future transcytosis is a possibility that requires further investigation.

The observed evidence of the disruption of the degradation pathway and the lack of accumulation of EVs in endosomal structures suggests the possibility that the EV-derived downregulation of rab7 could trigger the endocytic pathway to switch to the recycling track. A supporting mechanism has been described recently in a study that showed that knockdown of the NBEAL2 gene in megakaryocytes can disrupt the transport of fibrinogen to rab7 late endosomes and lysosomes.⁵⁴ This disruption of degradation increased the transfer of fibrinogen to rab11 recycling endosomes.

Taken together, these findings provide insight into the previously unknown mechanisms by which tumor-derived EVs can manipulate the brain endothelium to facilitate their transfer into brain parenchyma. This mechanism can be further applied to brain drug delivery approaches to compensate for the low transcytosis rates in the BBB and increase the efficiency of drug delivery to the brain.

EV Labeling.

231P and 231Br breast cancer cells were transduced with lentiviral vectors to express palmitoylated TdTomato (Palmtd-Tomato)⁵⁶ or membrane-bound Gaussia luciferase.⁵⁷ Both DNA constructs were gifts of Dr. X. Breakefield, Massachusetts General Hospital (MGH), and the lentivirus vectors were made at the MGH Vector Core, Boston, MA. EVs were isolated from stable clones as described above. The presence of labels on EVs was confirmed via fluorescent microscopy for tdTomato and via luciferase assay for Gaussia luciferase. Briefly, a 20 μM concentration Gaussia luciferase substrate, native coelenterazine (Prolume Ltd. cat. no. 303), was prepared and incubated for 30 min at room temperature. The substrate (50 μL) was added to each well containing the samples, and luminescence intensity was measured immediately using a SpectraMax M2 plate reader (Molecular Devices, Inc.).

In Vitro EV Uptake Studies. To evaluate the uptake of EVs by brain ECs, astrocytes, and pericytes, cells grown to confluence in 96-well plates were incubated with 2 × 10⁹ particle/well of tdTomato P-and Br-EVs. After 2 h of incubation, cells were washed 3 times and were fixed for imaging. Images from four different fields were taken using a Zeiss fluorescent microscope, and the level of fluorescence intensity was analyzed using the ImageJ software. To eliminate the confounding effect of cell size on the uptake level, fluorescent intensity for tdTom-EVs was measured per unit of cell surface area. For endocytosis inhibition studies, brain ECs cultured in 12-well plates were treated with chlorpromazine hydrochloride (Millipore Sigma, cat. no. C8138, 20 μM), ML141 (100 μM, Millipore Sigma, cat. no. 217708), 5-(N-ethyl-N-isopropyl) amiloride (EIPA) (100 μM, Tocris, cat. no. 3378), cytochalasin D (500 nM, Tocris, cat. no. 1233), and filipin III (10 μM, Millipore Sigma, cat. no. F4767) for 30 min prior to addition of TdTom-Br-EVs (10¹⁰ particle/well). Following 3 h of incubation with EVs, cells were washed and EV uptake was measured by flow cytometry using a BD FACSCalibur flow cytometer (BD Biosciences, San Jose, CA).

In Vitro Transcytosis Studies.

In Vitro Colocalization Studies.

Human brain endothelial cells were cultured on fibronectin-coated glass-bottom microslides (Ibidi, cat. no. 80827). Confluent cells were used in these studies. Cells were co-incubated with TdTom-Br-EVs (8 × 10⁹ particles/well) and 70 kDa FITC-dextran (0.5 mg/mL), Alexa Fluor 647-conjugated transferrin (50 μg/mL, Thermo Fisher Scientific, cat. no. T23366), or DQ ovalbumin (200 μg/mL, Thermo Fisher Scientific, cat. no. D12053) for 30 min. Subsequently, cells were washed 4 times with PBS and fixed with 4% formaldehyde for 10 min. For evaluation of colocalization with EEA1 and caveolin-1 (15 min incubation) or SNARE complexes (30 min incubation), cells were incubated with TdTom-Br-EVs (8 × 10⁹ particles/well) and then washed and fixed for staining with anti-EEA1 (1:100, Cell Signaling Technologies, cat. no. 3288), anti-caveolin-1 (1:100, Cell Signaling Technologies, cat. no. 3267), anti-VAMP-3 (1:100, Abcam, cat. no. ab200657), anti-TGN46 (1:50, Abcam, cat. no. ab2809), anti-VAMP-7 (1:100, Cell Signaling Technologies, cat. no. 13786), anti-syntaxin 4 (1:50, R&D Systems Inc., cat. no. MAB7894), and anti-snap23(1:100, R&D Systems Inc., cat. no. AF6306) antibodies. For colocalization studies with Rab11, cells were initially transfected with GFP-rab11 plasmid using Lipofectamine 3000 reagent (Thermo Fisher Scientific). GFP- rab11 WT plasmid was a gift from Richard Pagano (Addgene plasmid no. 12674; http://n2t.net/addgene:12674; RRID, Addgene_12674).⁵⁹ Transfected cells were then cultured on microslides for incubation with TdTom-Br-EVs as described.

Epifluorescence microscopy was performed on a Leica microscope coupled to high-resolution objectives and a Hamamatsu Orca CCD (Japan). To quantify the colocalization of EVs with different markers, at least 10 different fields were evaluated for each experiment. Colocalization with rab11 and DQ ovalbumin was quantified using a plugin for ImageJ developed by Jaskolski et al.⁶⁰ Colocalization with VAMP-3 and VAMP-7 was quantified through manual counting of the percentage of the colocalized EV-containing endosomes.

Rab7 siRNA Studies.

Human brain endothelial cells (at 30% confluency) were treated with a pool of Rab7A siRNAs or nontargeting siRNAs (100 nM, Dharmacon, siGENOME SMARTpool), using DharmaFECT 4 transfection reagent. Experiments were performed 72 h after transfection. For imaging, cells cultured on microslides, were incubated with DQ ovalbumin (200 μg/mL) or TdTom-Br-EVs (8 × 10⁹ particles/well) for 30 min. Subsequently, cells were washed 4 times with PBS and fixed for staining with anti- Rab7 antibody (1:100, Abcam, cat. no. 137029) and anti-LAMP-1 antibody (1:100, Cell Signaling Technologies, cat. no. 9091). Epifluorescence microscopy was performed as described previously. Using ImageJ, fluorescence intensity was measured in 6 fields for each condition and was normalized to autofluorescence intensity captured from empty microslides. For flow cytometry, transfected cells cultured in 12-well plates were incubated with TdTom-Br-EVs (10¹⁰ particles/ well) for 3 h, and cell uptake was quantified through flow cytometry as described above.

In Vivo Experiments.

All animal experiments were conducted in accordance with the Institutional Animal Care and Use Committee (IACUC) guidelines of the Boston Children’s Hospital, Boston, MA.

Western Blot Analyses and ELISA.

Cells were lysed with lysis buffer (Cell Signaling Technology, Danvers, MA) supplemented with phenylmethylsulfonyl fluoride protease inhibitor. Following centrifugation of the lysates at 14000g for 10 min at 4 °C, the supernatant was collected for Western blot (30 μg of total protein/lane). EV samples resuspended in PBS were directly used for Western blot (15 μg of total protein/lane). Protein concentration was measured using the Bradford method (Biorad laboratories, CA). Immunoblotting was conducted as reported previously.⁶¹ Antibodies against the following proteins were used for immunoblotting: CD63 (1:500, Abcam, cat. no. 59479), CD9 (1:500, Cell Signaling Technologies, cat. no. 13174), Alix (1:1000, Cell Signaling Technology, cat. no. 2171S), GM130 (1:1000, Cell Signaling Technologies, cat. no. 12480), Rab 7 (1:1000, Abcam, cat. no. 137029), and Rab 11 (1:250, Cell Signaling Technology, cat. no. 5589S).

Immunocytochemistry and Immunohistochemistry.

For immunocytochemistry, cells were fixed with 10% formalin for 10 min and then permeabilized with 0.1% triton X-100 for 5 min. For immunohistochemical staining, frozen sections was were fixed with ice-cold acetone for 10 min. Blocking was performed using 3% bovine serum albumin for 30 min. Cells or tissue sections were incubated with the primary antibody for 1 h at room temperature or overnight at 4 °C, respectively. Following washes, cells or tissue sections were incubated with the relevant secondary antibody (1:200) for 1 h. Sections were washed and mounted with Fluoro-gel mounting medium (Electron Microscopy Sciences). Images were taken using a Zeiss Axiocam fluorescent microscope.

Statistical Analyses.

All quantified data are presented as mean ± SD from 3 independent experiments. For animal experiments, the minimum number of animals required to obtain data amenable to statistical analysis was used. Animals were randomly divided into groups. Blinded analyses were only conducted to evaluate the presence of brain metastasis.

All statistical analyses were performed using the GraphPad Prism software. Statistical significance was considered at P values lower than 0.05. P values are shown as *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001. No outliers were excluded in this study. The methods of statistical analyses have been indicated in figure legends. All comparisons between two experimental groups were performed by unpaired two-tailed Student’s t test. Comparisons between more than 2 groups were performed by one-way ANOVA with Tukey’s correction for multiple comparisons. Groups of data involving more than one variable were analyzed by two-way ANOVA with Sidak’s correction for multiple comparisons. All mouse experiments were evaluated using the Mann-Whitney test.