Brain metastatic cancer cells release microRNA-181c-containing extracellular vesicles capable of destructing blood-brain barrier

Abstract

Brain metastasis is an important cause of mortality in breast cancer patients. A key event during brain metastasis is the migration of cancer cells through blood-brain barrier (BBB). However, the molecular mechanism behind the passage through this natural barrier remains unclear. Here we show that cancer-derived extracellular vesicles (EVs), mediators of cell-cell communication via delivery of proteins and microRNAs (miRNAs), trigger the breakdown of BBB. Importantly, miR-181c promotes the destruction of BBB through the abnormal localization of actin via the downregulation of its target gene, PDPK1. PDPK1 degradation by miR-181c leads to the downregulation of phosphorylated cofilin and the resultant activated cofilin-induced modulation of actin dynamics. Furthermore, we demonstrate that systemic injection of brain metastatic cancer cell-derived EVs promoted brain metastasis of breast cancer cell lines and are preferentially incorporated into the brain in vivo. Taken together, these results indicate a novel mechanism of brain metastasis mediated by EVs that triggers the destruction of BBB.

ARTICLE

Received 2 Oct 2014 | Accepted 20 Feb 2015 | Published 01 Apr 2015

Naoomi Tominaga1,2,3, Nobuyoshi Kosaka1,4,5, Makiko Ono1, Takeshi Katsuda1, Yusuke Yoshioka1, Kenji Tamura6, Jan Ltvall7, Hitoshi Nakagama2,8 & Takahiro Ochiya1

Brain metastasis is an important cause of mortality in breast cancer patients. A key event during brain metastasis is the migration of cancer cells through bloodbrain barrier (BBB). However, the molecular mechanism behind the passage through this natural barrier remains unclear. Here we show that cancer-derived extracellular vesicles (EVs), mediators of cellcell communication via delivery of proteins and microRNAs (miRNAs), trigger the breakdown of BBB. Importantly, miR-181c promotes the destruction of BBB through the abnormal localization of actin via the downregulation of its target gene, PDPK1. PDPK1 degradation by miR-181c leads to the downregulation of phosphorylated colin and the resultant activated colin-induced modulation of actin dynamics. Furthermore, we demonstrate that systemic injection of brain metastatic cancer cell-derived EVs promoted brain metastasis of breast cancer cell lines and are preferentially incorporated into the brain in vivo. Taken together, these results indicate a novel mechanism of brain metastasis mediated by EVs that triggers the destruction of BBB.

1 Division of Molecular and Cellular Medicine, National Cancer Center Research Institute, 5-1-1 Tsukiji, Chuo-ku, Tokyo 104-0045, Japan. 2 Graduate School of Medicine, The University of Tokyo, 7-3-1 Hongo, Bunkyou-ku, Tokyo 113-0033, Japan. 3 Research Fellow of the Japan Society for the Promotion of Science (JSPS), Chiyoda-Ku, Tokyo 102-0083, Japan. 4 Department of Zoology, University of Oxford, The Tinbergen Building, South Parks Road, Oxford OX1 3PS, United Kingdom. 5 The Japan Society for the Promotion of Science (JSPS) Postdoctoral Fellow for Research Abroad, Chiyoda-Ku, Tokyo 102-0083, Japan.

6 Division of Breast and Medical Oncology, National Cancer Center Research Institute, 5-1-1 Tsukiji, Chuo-ku, Tokyo 104-0045, Japan. 7 Department of Internal Medicine and Department of Respiratory Medicine and Allergology, The Sahlgrenska Academy, University of Gteborg, Box 424, SE-405 30 Gothenburg, Sweden. 8 Division of Cancer Development System, National Cancer Center Research Institute, 5-1-1 Tsukiji, Chuo-ku, Tokyo 104-0045, Japan. Correspondence and requests for materials should be addressed to T.O. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 6:6716 | DOI: 10.1038/ncomms7716 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

& 2015 Macmillan Publishers Limited. All rights reserved.

DOI: 10.1038/ncomms7716 OPEN

Brain metastatic cancer cells release microRNA-181c-containing extracellular vesicles capable of destructing bloodbrain barrier

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7716

Brain metastasis is associated with a particularly poor prognosis for cancer patients. Therefore, novel insights into the brain metastatic process are needed. A key event

during brain metastasis is the migration of cancer cells through bloodbrain barrier (BBB)1,2, which consists of the endothelium and surrounding cells3,4. BBB limits the passive diffusion of molecules. One of the key features of brain metastasis is the destruction of BBB5. Tumour cells recognize and bind to components of the vascular membrane, thereby initiating extravasation, invasion of cancer cells through BBB and the beginning of new growth at secondary organ sites6,7. This process suggests that the intact endothelium can serve as a defensive barrier against the extravasation of tumour cells. Although these hypotheses are accepted, the exact molecular mechanisms that trigger brain metastasis are poorly understood. Humoral factors have been proposed to mediate the interaction between cancer cells and BBB-constructed cells, resulting in the destruction of BBB structure. However, the actions of these molecules do not fully explain the destruction of BBB during brain metastasis, given the complexity of the interactions among cells that comprise BBB8. Extracellular vesicles (EVs), including exosomes, are known to regulate multiple aspects of malignancy in cancer cells911. EVs are known to mediate cellcell communication via the delivery of their contents, including proteins, mRNAs and microRNAs (miRNAs)12. Although the contribution of EVs to cancer metastasis is evident, little is known regarding the roles of EVs in brain metastasis.

In the present study, we report that cancer-derived EVs break down BBB through the change in actin dynamics in vitro, and that they promote brain metastasis in vivo. Furthermore, in vitro study, miR-181c promotes the destruction of BBB through delocalization of actin bre via the downregulation of 3-phosphoinositide-dependent protein kinase-1 (PDPK1). PDPK1 degradation by miR-181c leads to the downregulation of phosphorylated colin and the resultant activated colin-induced modulation of actin dynamics. In addition, our clinical data suggest that cancer-derived EVs cause brain metastases in cancer patients.

ResultsEstablishment of brain metastasis breast cancer cell lines. To identify the inuence of EVs on BBB during breast cancer metastasis, we employed MDA-MB-231-luc-D3H2LN cells (D3H2LN), which are human mammary tumour cells with a high tumorigenic and metastatic capacity13,14 to produce a new brain high-metastatic cell line. D3H2LN cells were injected into immunodecient female mice by intracardiac (i.c.) injections to isolate populations of cells that colonized in the brain (Fig. 1a). Brain metastasis was monitored by in vivo imaging using intraperitoneal luciferin injections (Fig. 1b). Cancer cell colonization of the brain tissue was also conrmed by hematoxylin and eosin (HE) staining (Fig. 1c). After tumour dissociation and expansion in culture, the resulting cell populations (brain metastatic derivative 1a, BMD1a) were subjected to a second round of in vivo selection, yielding brain metastatic derivative cell populations 2a and 2b (BMD2a and BMD2b), which showed signicant increases in brain metastatic activity over the original cell line. When injected into the left ventricle, BMD1a cells metastasized to the brain in 60% (3 out of5) of the mice, whereas D3H2LN cells yielded 1 brain metastasis out of 15 injected mice (6.7%).

BBB in vitro model. To determine how the EVs from breast cancer cells that metastasize to the brain affect BBB, an in vitro BBB culture system that enables us to study the molecular and

cellular effects of the EVs is essential. To this end, recent studies have employed monolayer cell culture systems15. However, BBB consists of three different types of cell, and these cells cooperate with each other to maintain the structure of BBB. Therefore, we employed a new in vitro BBB model system that consists of primary cultures of brain capillary endothelial cells, brain pericytes and astrocytes (Fig. 1d). As shown in Fig. 1e, brain capillary endothelial cells, brain pericytes and astrocytes were assessed using Hoechst 33342 staining. Furthermore, tight junction formations and adherens junction formations were conrmed with immunouorescence analysis (Fig. 1f). This in vitro BBB model simulated BBB in vivo, allowing the detailed study of cellcell interactions in vitro. Moreover, transendothelial electrical resistance (TEER) was used to measure the formation of tight junctions by brain microvessel endothelial cells, which indicates the integrity of BBB. The TEER of this model was sufciently high to serve as a BBB in vivo (Fig. 1g)16,17.

Inhibition of EV secretion suppresses invasiveness through BBB. Because BBB consists of three different types of cells, knowing which cells incorporated EVs from cancer cells was essential for determining the precise mechanism of BBB destruction by EVs. The EVs from these cell lines were typical in size (B100 nm) and expressed conventional exosomal markers, such as CD63 and CD9 but not Cytochrome C (Fig. 2a, Supplementary Fig. 1a,b), which is a mitochondrial intermembrane-space protein known to be lacking in EVs18. The levels of EV secretion did not differ among the brain metastatic and D3H2LN cell lines (Supplementary Fig. 1c). To this end, we added PKH67-labelled EVs isolated from BMD2a, BMD2b and D3H2LN cells to the in vitro BBB model. As shown in Fig. 2b,c, EVs from all the cancer cells were incorporated into endothelial cells but not into pericytes or astrocytes. Interestingly, we observed higher uorescent intensity in the endothelial cells with BMD cell-derived EVs, suggesting that BMD cell-derived EVs undergo tropism in brain endothelial cells (Fig. 2b). To determine whether the EVs from brain metastatic cancer cells functionally affected the destruction of BBB, we added D3H2LN cell-, BMD2a cell-and BMD2b cell-derived EVs to the in vitro BBB model, and the TEER of each well was measured. As shown in Fig. 2d and Supplementary Fig. 1d, the TEER was signicantly reduced in wells containing the BMD2a- and BMD2b cell-derived EVs compared with those containing the D3H2LN cell-derived EVs (Po0.05) as a control or low-metastatic breast cancer cell line

MDA-MB-231-luc-D3H1 (D3H1) cell-derived EVs (Po0.01). Furthermore, we tested the permeability of BBB using sodium uorescein (NaF). Although the molecular weight of NaF is low (MW: 376.27), it cannot signicantly permeate the intact BBB19.

Therefore, NaF is used to assess the permeability of BBB by measuring its concentration with a uorescence monochromator. NaF showed a high apparent permeability coefcient (Papp) in wells containing BMD2a cell- (Po0.01) and BMD2b cell-(Po0.01) derived EVs as compared with those containing EVs derived from the D3H2LN cell line or the D3H1 cell line (Fig. 2e). Breast cancer cells are known to attach to microvessel endothelial cells to invade BBB (extravasation) during metastasis to the brain. Brain metastatic cancer cells are also considered to be highly invasive1. Matrigel invasion chamber assays were used to conrm the pathological implications of the metastatic potential of established cell lines. As shown in Supplementary Fig. 2a,b, BMD2a and BMD2b cells were more invasive than the D3H2LN cell line or a low-metastatic cell line D3H1. However, the morphology of BMD2a and BMD2b cells did not differ from that of the D3H2LN cell line (Supplementary Fig. 2c). To conrm the extravasation of BMD2a and BMD2b cells into the brain

2 NATURE COMMUNICATIONS | 6:6716 | DOI: 10.1038/ncomms7716 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7716 ARTICLE

Endothelial cells

Pericytes

Luminal side

Abluminal side

Intracardiac injection of D3H2LN cells

26 30 days

IVIS

2x105 cells per 100 l

Brain metastasis

Cell culture 30 days (selected by50 g ml1 zeocin

C.B-17/Icr-scid/ scid

7 weeks Two cycles

Brain metastasized cell lines

MDA-MB-231-D3H2LN

1.2

Astrocytes

Endothelial cells

Pericytes

Astrocytes

BMD1a

BMD2a

BMD2b

Luminescence Luminescence

1.0

1.5

Claudin-5

Endothelial cells

Occludin ZO-1 N-cadherin

0.8

x106

1.0

x105

0.6

0.4

0.2

0.5

Normal

Cerebral

cortex Midbrain

Metastasis

Higher magnification

TEER ( cm2 )

1,000

800

600

400

200

*

* up to 869.55 cm2

0 Day1 Day3 Day4

Figure 1 | Establishment of brain metastasis breast cancer cell lines and BBB in vitro model. (a) Schematic representation of the protocol for the in vivo-selected brain metastatic derivatives. MDA-MB-231-luc-D3H2LN breast cancer cell lines (2 105 cells) were injected intracardially into C.B-17 Icr-scid scid

mice. After 2630 days, the brain metastasis of cancer cells was monitored by in vivo imaging system (IVIS). The brain-metastasized cancer cells were recovered and cultured for B30 days in a culture medium containing 50 mg ml 1 Zeocin. This selection was performed twice, and we named the established cell lines BMD2a and BMD2b. (b) Bioluminescence image of a mouse with a BMD2a brain metastasis (left). Right image represents the bioluminescence image of a mouse brain with cancer cell metastasis. (c) Representative image of HE-stained sections from a mouse brain cerebral cortex and midbrain. Left upper and lower panels show the mouse cerebral cortex and midbrain, respectively, without metastasis of cancer cells. Middle upper and lower panels show the mouse cerebral cortex and midbrain, respectively, with metastasis of cancer cells. Arrow-head represent metastatic cancer cells. Right upper and lower panels show higher magnication. Scale bar, 100 mm. (d) The schematic representation of the in vitro model of BBB constructed from primary cultures of monkey brain capillary endothelial cells, brain pericytes and astrocytes. (e) Representative pictures of endothelial cells, pericytes and astrocytes are shown. Endothelial cells and pericytes were visualized using a confocal microscope. Astrocytes were visualized using a uorescence microscope. Scale bar, 20 mm. Scale bar in the panel of astrocytes represents 100 mm. (f) Immunouorescence of tight junction proteins (Claudin-5,

Occludin and ZO-1) and N-cadherin (red). Scale bar, 20 mm. (g) The transition of TEER after thawing until the start of the experiment. After thawing the BBB in vitro model, the value of TEER increased to a maximum of 869.55 O cm 2 (*mean maximum TEER.). Error bars represent s.d., n 12. Data are

representative of at least three independent experiments each. (Fig. e,f,g).

parenchyma side of this model, D3H1, D3H2LN, BMD2a and BMD2b cells were labelled for PKH26 to distinguish them in vitro and were added to the upper chamber of the in vitro BBB model. Cells that had inltrated the abluminal side were counted after 2 days (Fig. 2f). As shown in Fig. 2g, more BMD2a and BMD2b cells had extravasated compared with cells from the D3H2LN cell line or the D3H1 cell line, indicating that our established cell lines had a high potential for extravasation through BBB into the brain parenchymal side. To clarify the contribution of EVs to the extravasation of brain metastatic cells, we assessed the extravasation of BMD2a cells in the in vitro BBB model after the EV secretion in these cell lines was inhibited by siRNAs against EV secretion-related proteins, such as neutral

sphingomyelinase 2 (nSMase2) and RAB27B (Supplementary Fig. 2d,e)14,20. PKH26-labelled cells treated with control siRNA could still pass through BBB to the abluminal side, but cells in which the production of EVs was inhibited (Supplementary Fig. 2f) were not found on the abluminal side, even when those cells had the ability to pass through BBB (Fig. 2h). In addition, inhibiting the production of EVs only slightly reduced the invasiveness of these cell lines, as assessed by a Matrigel invasion assay (Supplementary Fig. 2g). This nding indicates that extravasation was not only a result of the cells invasive potential. Furthermore, to investigate whether the addition of EVs is sufcient to allow the extravasation of cancer cells to the parenchymal side of the brain, we examined the extravasation of

NATURE COMMUNICATIONS | 6:6716 | DOI: 10.1038/ncomms7716 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7716

Intake of EVs in endothelial cells** **

Intensity (folds)

Addition of EVs

50 N.C. D3H1 D3H2LN BMD2a BMD2b

**

5 4 3 2 1 0

D3H1

**

50

**

TEER ( cm2)

100

250

150

P app(106 cm s1)

8 7 6 5 4 3 2 1 0

200

**

D3H2LN BMD2a

Addition of EVs

BMD2b

300

400

350

N.C. D3H1 D3H2LN BMD2a BMD2b

* *

Add EVsfromD3H2LN, BMD2a, BMD2b. 1 day

PKH26-labelled cancer cells

Addition of EVs

Removed endothelial

cells

2 days

Count the invaded cells labelled by PKH26

**

1.2

0

** **

14 12 10

8 6 4 2 0

1.0

Addition of EVs

N.C. D3H1 D3H2LN BMD2a BMD2b

Endothelial cells Pericytes Astrocytes

Cell number (folds)

Cell number (folds)

Cell number (folds)

4.03.53.02.52.01.51.00.5 0

*

0.8

0.6

0.4

0.2

** ** **

N.C. D3H2LN BMD2a BMD2b

Addition of EVs

BMD2a BMD2b

D3H2LN

D3H1 N.C. RAB27B nSMase2 S + R

siRNA

D3H1 cells

D3H2LN cells

N.C.

N.C.

BMD2a cells BMD2b cells

RAB27B siRNA D3H2LN EVs

BMD2b EVs

nSMase2 siRNA

S + R siRNA

BMD2a EVs

Figure 2 | EVs from brain metastatic cancer cells were incorporated into endothelial cells, regulating the invasion through BBB of cancer cells.(a) Phase-contrast electron microscopy was used to visualize resuspended EVs pellets. Scale bar, 100 nm. (b) The intensity of PKH67-labelled EVs was measured by ImageJ. Error bars represent s.d., Students t-test, n 3. (**Po0.01). (c) EVs isolated from cancer cells were labelled using PKH67 and added

to the upper chamber. Representative pictures of endothelial cells, pericytes and astrocytes are shown. Negative control (N.C.), EVs from MDA-MB-231-luc-D3H2LN (D3H2LN), EVs from BMD2a and EVs from BMD2b are shown. Scale bar, 20 mm. Bar in the panel of astrocytes represents 100 mm. (d) The value of the TEER was monitored before (day 4) and after (day 5) the addition of EVs isolated from each cell line. EVs isolated from brain metastatic cancer cells were incubated in the in vitro BBB model for 24 h. Error bars represent s.d., Students t-test, n 3. (**Po0.01, *Po0.05). (e) Assessment of BBB

permeability determined by NaF (molecular weight 376.27). EVs from MDA-MB-231-luc-D3H1 (D3H1), D3H2LN, BMD2a or BMD2b cells and N.C. were

added to the in vitro BBB model. After 24 h, NaF was added. NaF that had passed through BBB was measured by a uorometer. Error bars represent s.d., Students t-test, n 3. (**Po0.01). (f) PKH26-labelled cancer cells (2 104 cells) were added to the in vitro BBB model. After a 48-h incubation,

endothelial cells were removed, and the invading cells were counted using a uorescence microscope. (g) In vitro BBB transmigration activity of D3H1, D3H2LN, BMD2a or BMD2b cells. The number of transmigrated cells relative to the D3H1 cell lines is plotted. Error bars represent s.d., Students t-test, n 3. (*Po0.05, **Po0.01). (h) In vitro BBB transmigration activity of the BMD2a treated with control siRNA (N.C.), RAB27B, nSMase2 and both siRNAs

(S R). The number of transmigrated cells relative to the control cells is plotted. Error bars represent s.d., Students t-test, n 3. (**Po0.01). (i) The

number of transmigrated D3H1 cells relative to the control cells is plotted. Error bars represent s.d., Students t-test, n 3 (**Po0.01). Data are

representative of at least three independent experiments each. (Fig. b,c,d,e,g,h,i).

low-metastatic D3H1 cells after the addition of D3H2LN cell- and BMD2a cell- and BMD2b cell-derived EVs into the in vitro BBB model. The EVs isolated from BMD2a, BMD2b and D3H2LN cells were added to each well and were incubated for 24 h before the addition of PKH26-labelled D3H1 cells. After 2 days, we counted the number of inltrated cells. As shown in Fig. 2i, D3H1 cells could not invade BBB without the addition of EVs. However, inltration of D3H1 cells to the abluminal side was signicantly increased by the addition of BMD2a cell- and BMD2b cell-derived EVs. By contrast, the addition of D3H2LN cell-derived EVs did not efciently promote the inltration of low-metastatic D3H1 cells through BBB compared with BMD2a- and BMD2b

cell-derived EVs. Taken together, these results suggest that the secretion of EVs by breast cancer cells affected the extravasation of metastatic cancer cells to the brain parenchyma side across BBB.

Cancer cell-derived EVs promote brain metastasis in vivo. To evaluate the effect of cancer-derived EVs on the brains of mice, in vivo permeability assay was performed. Puried EVs derived from D3H2LN and BMD2a cells were labelled with a XenoLight DiR uorescent dye before injection into the tail vein of mice. D3H2LN cell-derived EVs were used as a control. After 6 h,

4 NATURE COMMUNICATIONS | 6:6716 | DOI: 10.1038/ncomms7716 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7716 ARTICLE

Tracer-653 probe, an in vivo tracer dye for monitoring BBB disruption, was injected into the mice. The upper panels of Fig. 3a show the intake of EVs and the lower panels show the permeability of brain blood vessels. As shown in the upper panels of Fig. 3a, BMD2a cell-derived EVs were more incorporated within the brains of mice than were D3H2LN cell-derived EVs. As shown in the lower panels of Fig. 3a, BMD2a cell-derived EV-treated mice showed greater permeability of brain blood vessels as compared with D3H2LN cell-derived EV-treated mice in vivo.

Furthermore, to clarify the contribution of the EVs to brain metastasis in vivo, D3H2LN cell- or BMD2a cell-derived EVs were injected in the tail vein of mice. After 24 h, D3H2LN cells were injected in the left cardiac ventricle of mice; the mice were then observed after 18 days (Supplementary Fig. 3). Importantly, mice treated with the BMD2a cell-derived EVs (brain metastases: 5 out of 9) had more brain metastases as compared with mice treated with D3H2LN cell-derived EVs (brain metastases: 1 out of 9; Table 1). Moreover, mice injected with BMD2a cell-derived EVs had a greater metastatic burden in the brain (Po0.05, as measured by luciferase intensity, Fig. 3b) as compared with the mice treated with D3H2LN cell-derived EVs (Fig. 3b). However,

negative control-injected mice were not observed to exhibit brain metastasis (Table 1). In other words, mice treated with BMD2a cell-derived EVs had a higher rate of metastasis in the brain as compared with the mice treated with the negative control or mice treated with EVs derived from the D3H2LN cells (Table 1 and Fig. 3c, lower panels). Brain metastasis was conrmed by HE staining and immunouorescence staining against human vimentin (Fig. 3d). In mice treated with BMD2a cell-derived EVs, but not with D3H2LN cell-derived EVs, the presence of cancer cells in the brain was conrmed by HE staining or immunouorescence staining (Fig. 3d). These results show that BMD2a cell-derived EVs promoted brain metastasis by increasing the permeability of brain blood vessels, which resulted in BBB breakdown.

Disruption of intercellular junctions causes BBB breakdown. Tight junctions are known to regulate the low permeability of BBB and are formed by specic proteins in endothelial cells, such as Claudin-5, Occludin and ZO-1. On the other hand, N-cadherin, a calcium-dependent cellcell adhesion glycoprotein

**

*

D3H2LN EVs

BMD2a EVs

DiR (EVs) Tracer 653

2.4

2.0

1.6

15.0

14.0

13.0

12.0

1.5

1.0

0.5

1.4

1.2

Photon intensity (x105)

1.6

1.4

1.2

1.0

0.8

0.6

0.4

0.2

0

7

N.C. D3H2LN BMD2a

Injection of EVs

N.C. D3H2LN EVs BMD2a EVs

D3H2LN EVs

BMD2a EVs

HEHu-vimentin

x104

1.0

x106x107x10

Figure 3 | Cancer-derived EVs promoted the brain metastasis of breast cancer cells. (a) Fluorescence image of a mouse brain injected D3H2LN or BMD2a cell-derived EVs. The upper image represents the uptake DiR-labelled EVs of a mouse brain. The lower image represents the permeability of a mouse brain. D3H2LN cell-derived EVs were used as a control. This experiment was repeated twice. (b) Distribution of photon intensity in the brain, quantied by ImageJ analysis. P values were determined by MannWhitney one-tailed testing. Negative control (N.C.); n 9, D3H2LN; n 9, BMD2a;

n 9. (*Po0.05, **Po0.01). (c) Bioluminescence image of D3H2LN and BMD2a cell-derived EVs and N.C.-injected mice. The upper image represents the

bioluminescence whole-body image of mice. The lower image represents the bioluminescence image of a mouse brain with cancer cell metastasis. (d) Representative image of HE-stained sections from a mouse brain cerebral cortex (upper panels). The arrowhead shows the cancer cells. Scale bar, 100 mm. The lower panel shows a representative immunouorescence image of anti-human vimentin (Hu-vimentin). Scale bar, 20 mm. Data are representative of at least three independent experiments each (Fig. d).

NATURE COMMUNICATIONS | 6:6716 | DOI: 10.1038/ncomms7716 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7716

Table 1 | Brain metastasis rate after EVs treatment.

Treatment Brain metastasis, % (n 9)

N.C. 0 (0%) D3H2LN EVs 1 (11.1%) BMD2a EVs 5 (55.6%)

EVs: extracellular vesicles; N.C.: negative control.

D3H2LN: MDA-MB-231-D3H2LN-luc.

The brain metastasis rate after EV treatment. n 9 mice per group. D3H2LN cell-derived

EVs were used as a control.

composed of ve extracellular cadherin repeats that allow it to mediate strong cellcell adhesion, is mostly expressed on the apical and basal membranes. Tight junction proteins and N-cadherin regulate cell polarity through their intimate association with the actin cytoskeletal network. From these aspects, we hypothesized that the breakdown of BBB was caused by the dysregulation of these molecules. To prove this hypothesis, endothelial cells in the in vitro BBB model were co-immunouorescently stained against Claudin-5, Occludin, ZO-1 and N-cadherin with actin laments after the addition of EVs isolated from brain metastatic cancer cell lines or the D3H2LN cell line. Tight junction proteins were localized to the surface of the cell membrane in endothelial cells treated with PBS or EVs from D3H2LN cells; however, tight junction proteins and N-cadherin localized to the cytoplasm in cells that were treated with BMD2a cell- and BMD2b cell-derived EVs (Fig. 4a,b). Of note, we found that the expression of tight junction proteins, N-cadherin and actin was not affected by the addition of BMD2a cell- and BMD2b cell-derived EVs to brain blood vessel endothelial cells (Fig. 4c). These results strongly suggested that cancer-derived EVs changed the localization of tight junction proteins, N-cadherin and actin laments, but did not affect the expression levels of these proteins.

EV-contained miR-181c diminishes actin lament organization. To further investigate the molecular mechanisms by which EVs mediated the abnormal localization of tight junction proteins and adherence junction protein, we performed proteomic and miRNA microarray analysis of EVs isolated from BMD2a, BMD2b and D3H2LN cells. Although the proteomic analysis did not reveal EV protein candidates, we found that miR-181c was signicantly upregulated in BMD2a cell- and BMD2b cell-derived EVs as compared with those derived from D3H2LN cells (Fig. 5a,b and Supplementary Fig. 4a). On the contrary, there was not signicant expression of cellular miR-181c in D3H2LN, BMD2a or BMD2b cells (Supplementary Fig. 4b). Moreover, we assessed the expression of miR-181c in endothelial cells after the addition of EVs from breast cancer cells and found that its expression was signicantly increased by the addition of EVs isolated from BMD2a and BMD2b cells (Fig. 5c). To assess the direct effect of miR-181c on endothelial cells, synthetic miR-181c was transfected into endothelial cells from the in vitro BBB model. As shown in Fig. 5d, tight junction proteins, N-cadherin and actin localized to the cytoplasm in miR-181c-transfected cells, although localization of these molecules can be formed on membranes in negative control-treated cells. Indeed, transfection of miR-181c signicantly downregulated the value of TEER in the in vitro BBB model (Fig. 5e). Furthermore, we found that the expression of tight junction proteins, N-cadherin and actin was not affected by the transfection of miR-181c into brain blood vessel endothelial cells (Fig. 5f).

It is possible that miR-181c in EVs from the brain-metastasized cancer cells could be found in sera from breast cancer patients who have metastases in the brain due to the leakage of

circulating EVs from brain metastatic cancer. To clarify this hypothesis, we analysed the expression abundance of miR-181c in sera from breast cancer patients (n 56). As shown in

Fig. 5g, miR-181c in EVs collected from brain metastasis patients serum were signicantly higher compared with non-brain metastasis patients (brain metastasis; n 10, non-brain metas

tasis; n 46, Po0.05, t-test). As shown in Supplementary Fig. 4c,

the serum level of miR-181c was signicantly higher in the sera from brain metastasis patients (stage IV; n 10) as compared

with the sera from non-brain metastasis patients (stage III; n 13; Po0.05, stage IV; n 6; Po0.05, MannWhitney U-test).

Interestingly, in stage IV patients, the miR-181c level was higher in the sera from brain metastasis patients (Po0.05). This result emphasizes that secreting miR-181c is related to brain metastasis of breast cancer patients. Taken together, these results suggest that EVs from brain metastatic cancer cells induce the abnormal localization of tight junction proteins by transferring miR-181c into endothelial cells, which results in the destruction of the cell cell contact.

PDPK1-regulated actin localization is important to BBB. To understand in more detail the molecular mechanism of miR-181c in EV-mediated BBB destruction, we decided to identify the target gene of miR-181c in endothelial cells. We performed global gene expression analysis in endothelial cells after the transfection of negative control or miR-181c (Supplementary Fig. 5a and Fig. 6a) and or after the addition of EVs from BMD2a, BMD2b or D3H2LN cells (Supplementary Fig. 5b and Fig. 6d). We found that overexpression of miR-181c in brain endothelial cells exhibited the downregulation of PDPK1 as compared with the control cells by mRNA and protein expression (Fig. 6b,c). Furthermore, we also found that the expression of PDPK1 was downregulated in BMD2a cell- or BMD2b cell-derived EV-treated endothelial cells as compared with that in D3H2LN cells by mRNA and protein expression (Fig. 6df). These results suggest that miR-181c in EVs suppressed the expression of PDPK1 in brain endothelial cells.

Next, we analysed the effect of PDPK1 in the localization of tight junction-related proteins, N-cadherin and actin laments in brain endothelial cells after the treatment with PDPK1 siRNA. We showed that the expression of PDPK1 protein was downregulated by the transfection of PDPK1 siRNA (Supplementary Fig. 5c,d). As shown in Fig. 7a, we observed the localization of tight junction proteins and N-cadherin at the cellular membrane in control siRNA-treated cells; however, these localizations were dysregulated in PDPK1 siRNA-treated cells. Localization of tight junction proteins, N-cadherin and actin was found in cytoplasm, whose results were the same as those of BMD2a cell- or BMD2b cell-derived EV-treated cells or miR-181c-transfected cells. Interestingly, actin condensation was observed in the endothelial cells (Figs 4a,b,5d and 7a). We conrmed that the expression of tight junction proteins, N-cadherin and actin proteins expression was not changed with or without PDPK1 siRNA treatment (Fig. 7b). The TEER of the in vitro BBB model was signicantly downregulated by PDPK1 siRNA treatment (Fig. 7c). Furthermore, 30-untranslated region (30UTR) luciferase reporter assay was performed to analyse miR-181c and PDPK1 mRNA. This result showed that 30UTR of PDPK1 was a direct target of miR-181c (Supplementary Fig. 5e, and Fig. 7d). Previous reports have shown that PDPK1 is an upstream protein of colin phosphorylation21,22. Colin is a family of actin-binding proteins, which disassembles actin laments, that is activated with dephosphorylation. Finally, western blot analysis of phosphocolin was performed. Phosphorylation of colin in BMD2a cell- or BMD2b cell-

6 NATURE COMMUNICATIONS | 6:6716 | DOI: 10.1038/ncomms7716 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7716 ARTICLE

Claudin-5 Actin Merge Occludin Actin Merge ZO-1 Actin Merge

Addition of EVs

N.C. D3H2LN BMD2a BMD2b Addition of EVs

N.C. D3H2LN BMD2a BMD2b

N-cadherin

Actin Merge

Addition of EVs

BMD2b

N.C.

D3H2LN

BMD2a

250 kDa

50 kDa

150 kDa

50 kDa

37 kDa

ZO-1 (220 kDa)

Occludin (65 kDa)

Claudin-5

(23 kDa)

N-cadherin

(140 kDa)

Actin (43 kDa)

GAPDH (40 kDa)

20 kDa

Figure 4 | EVs from brain metastatic cancer cells promoted BBB breakdown. (a) Co-immunouorescence of tight junction proteins (Claudin-5, Occludin and ZO-1) (red) and actin laments (green) after the addition of EVs from D3H2LN, BMD2a or BMD2b cells. Scale bar, 20 mm. (b) Co-immunouorescence of N-cadherin (red) and actin laments (green) after the addition of EVs from D3H2LN, BMD2a or BMD2b cells. Scale bar, 20 mm.

(c) Western blot analysis of tight junction proteins, N-cadherin, Actin and GAPDH. Proteins from endothelial cells treated with negative control (N.C.) or EVs. This experiment was repeated twice. Data are representative of at least three independent experiments each (Fig. a,b).

derived EVs-treated brain endothelial cells was downregulated as compared with D3H2LN cell-derived EVs treatment (Fig. 7e). Furthermore, phosphorylation of colin in miR-181c or PDPK1 siRNA-treated brain endothelial cells was downregulated as compared with negative control siRNA treatment (Fig. 7f). Taken together, these results suggest that miR-181c in EVs modulates the actin dynamics through the downregulation of PDPK1 in brain endothelial cells (Supplementary Fig. 6).

DiscussionCirculating cancer cells have been shown to adhere to the brain blood vessel endothelium and, subsequently, to inltrate the brain parenchyma7. During this event, cancer cells secrete humoral factors to induce the destruction of BBB and allow extravasation. Previous reports have shown that vascular endothelial growth factor expression of cancer cells was necessary but insufcient for the production of brain metastasis8. Bos et al.13 showed that the

expression of ST6GALNAC5 in breast cancer cells enhances their adhesion to brain endothelial cells and their passage through the BBB. One of the key features of brain metastasis is the destruction of BBB5. In this article, we have clearly shown that brain metastatic cancer cell-derived EVs trigger BBB destruction to promote the extravasation of cancer cells through BBB. BBB is composed of tight junction proteins, such as Claudin-5, Occludin and ZO-1. The primary cytoskeletal protein, actin, has known binding sites on all ZO proteins and on claudins and Occludin. Actin laments serve both structural and dynamic roles in the cell1. Previous ndings have shown that PDPK1 protein has a reduced F-actin to G-actin ratio23 and is a critical regulator of actin polymerization24. Active colin is an actin-binding protein that severs laments. Phosphorylation of colin through PDPK1 is thought to inactivate colin in a spatial manner in which local activation occurs at the cell membrane. We have demonstrated that the actin dynamic in BBB was regulated with cancer-derived EVs in brain metastases. In this article, we have shown that

NATURE COMMUNICATIONS | 6:6716 | DOI: 10.1038/ncomms7716 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7716

EVs

Endothelial cells

0 D3H2LN BMD2a BMD2b

3.5

**

**

**

(D3H1-EVs)

(D3H2LN-EVs)

(BMD2a-EVs)

(BMD2b-EVs)

3.0

1.5

2.5

0 3 6

/total exosomal RNA (folds)

5

4

3

2

1

**

miR-181c

Colour range

2.0

1.0

hsa-miR-181b-5p hsa-miR-181a-5p hsa-miR-181d hsa-miR-181c-5p

0.5

0 D3H2LN BMD2a BMD2b

Incubated with EVs

Claudin-5 Actin Merge Occludin Actin Merge

miR-181c / U6 (folds)

Actin Merge

ZO-1 Actin Merge

N-cadherin

N.C. miR-181c

N.C.

miR-181c

EVs from breast cancer patients sera *

250 kDa

N.C.

miR-181c

**

TEER (cm2 )

ZO-1 (220 kDa)

Occludin (65 kDa)

Claudin-5

(23 kDa) N-cadherin

(140 kDa)

Actin (43 kDa)

GAPDH (40 kDa)

0

3.5

20 40 60 80 100

140 160 180 200

miR-181c / miR-16

50 kDa

50 kDa

37 kDa

3.0

1.5

2.5

2.0

20 kDa

120

150 kDa

1.0

0.5

0

Without brain metastasis

With brain metastasis

Figure 5 | miR-181c played a role in BBB breakdown and upregulation in brain metastasis patients sera. (a) Heat map showing expression levels of the miR-181c in cancer-derived EVs. (b) Amount of miR-181c in EVs isolated from D3H2LN, BMD2a and BMD2b cells. Error bars represent s.d., Students t-test, n 3. **Po0.01 as compared with EVs from D3H2LN cells. (c) Endothelial cells were incubated with EVs isolated from D3H2LN, BMD2a or BMD2b

cells for 24 h. RNA was isolated from the endothelial cells 24 h after the addition of EVs, and the expression of miR-181c in the endothelial cells was analysed by qRT-PCR. Each bar represents the mean s.d., Students t-test, n 3. **Po0.01 as compared with endothelial cells treated with EVs from

D3H2LN cells. (d) Co-immunostaining of Claudin-5, Occludin, ZO-1, N-cadherin (red) and actin laments (green) in endothelial cells after the addition of EVs from D3H2LN, BMD2a or BMD2b cells. Scale bar, 20 mm. These proteins localized to the cytoplasm in miR-181c-transfected cells. (e) The TEER value was monitored before (day 4) and after (day 5) the transfection of miR-181c or control siRNA. Transfected miR-181c or N.C. siRNA was incubated in the in vitro BBB model for 24 h. Error bars represent s.d., Students t-test, n 3. (**Po0.01). (f) Western blot analysis of tight junction proteins, N-cadherin,

actin and GAPDH. Proteins were from endothelial cells transfected with N.C. siRNA or miR-181c. This experiment was repeated twice. (g) Amount of miR-181c in EVs isolated from patients sera. Non-brain metastasis: n 46, brain metastasis: n 10 (*Po0.05). Associations between the miR-181c expression

levels of serum from breast cancer patients were assessed by Students t-test. Data are representative of at least three independent experiments each (Fig. b,c,d,e).

PDPK1 was downregulated by miR-181c in vitro BBB model. Decreased PDPK1 in brain endothelial cells resulted in the breakdown of BBB by activated colin. Many reports have shown that miR-181c is related to malignancies such as hepatocellular carcinoma25, basal cell carcinoma26 and breast cancer27. Our results support the possibility that miR-181c has an important role in malignancy.

Recently, Zhou et al.15 reported that breast cancer cell-derive EVs contributed to BBB breakdown. Despite the signicance of their ndings, we should emphasize that our study is novel in the following points. First, our study suggests the possibility of organ tropism in brain metastasis of breast cancer cells. Zhou et al. revealed the possibility of the breakdown of the junctions between endothelial cells throughout the body. In contrast, our study focuses the brain-oriented metastasis of breast cancer, by establishing brain-oriented metastatic cell lines. Of note, we

observed that EVs derived from these cell lines were more prone to accumulate in the brain than those from the parental cell line (Fig. 3a). Second, the mechanism described in the present study provides a novel miRNA-associated mechanism for EV-mediated BBB breakdown. Zhou et al.s work suggests that miR-105 suppressed the ZO-1 expression in endothelial cells, and the resulting loss of cellcell adhesion lead to the promotion of metastasis. However, considering the capacity of EVs to harbour a variety of miRNAs, it seems difcult to conclude that a single EV-miRNA can explain the entire mechanism of BBB breakdown. In this regard, we revealed another novel mechanism in which miR-181c contributed to actin degradation through the suppression of colin. Our ndings, along with those by Zhou et al., will help to understand the EV-mediated BBB breakdown.

Despite our observation that EVs accumulated preferentially in the brain, we have not yet identied the molecules responsible for

8 NATURE COMMUNICATIONS | 6:6716 | DOI: 10.1038/ncomms7716 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7716 ARTICLE

Microarrray analysis

1.2

1.0

0.8

0.6

0.4

0.2

0

PDPK1 (5868 kDa)

50 kDa

37 kDa

PDPK1 expression level (folds)

PDPK1 expression level (folds)

1.5

0.5

0.5

1.5

1.0

37 kDa

1.0

GAPDH (40 kDa)

N.C. miR-181c

PDPK1 mRNA /

PDPK1 / GAPDH

intensity (folds)

1.2

1.0

0.8

0.6

0.4

0.2

0

miR-181c

GAPDH mRNA (folds)

0

N.C.

N.C.

miR-181c

N.C.

miR-181c

Addition of EVs D3H2LN BMD2a BMD2b

N.C.

Microarray analysis

Endothelial cells

**

**

**

PDPK1 (5868 kDa)

50 kDa

GAPDH (40 kDa)

1.5

0.5

0.5

1.5

1.0

PDPK1 mRNA /

GAPDH mRNA (folds)

1.2

1.0

0.8

0.6

0.4

0.2

1.0

0 N.C. D3H2LN

BMD2a BMD2b

PDKK1 / GAPDH

intensity (folds)

1.2

1.0

0.8

0.6

0.4

0.2

0 N.C.

D3H2LN

BMD2a

BMD2b

Addition of EVs

0 N.C.

D3H2LN

BMD2a

BMD2b

Addition of EVs

Figure 6 | miR-181c regulates PDPK1 expression in brain endothelial cells. (a) Microarray analysis showing the expression levels of PDPK1 in brain endothelial cells after transfection of miR-181c. The data are represented as log2 value. (b) Expression level of PDPK1 mRNA in brain endothelial cells after the transfection of miR-181c. Error bars represent s.d., Students t-test, n 3. (**Po0.01). (c) Western blot analysis of PDPK1 and GAPDH. Proteins were

from brain endothelial cells transfected with miR-181c. The lower panel shows the intensity of PDPK1 obtained from the transfection of N.C. siRNA or miR-181c. This experiment was repeated twice. (d) Microarray analysis showing the expression levels of PDPK1 in brain endothelial cells after EV treatment. (e) Expression level of PDPK1 mRNA in brain endothelial cells after the addition of EVs from D3H2LN, BMD2a or BMD2b cells. Error bars represent s.d., Students t-test, n 3 (**Po0.01). (f) Western blot analysis of PDPK1 and GAPDH. Proteins were from endothelial cells treated with EVs from D3H2LN,

BMD2a or BMD2b cells. The lower panel shows the intensity of PDPK1 obtained from D3H2LN, BMD2a or BMD2b cells. This experiment was repeated twice. Data are representative of at least three independent experiments each (Fig. b,e).

this organ tropism. Molecular analysis of EV membrane components will be of great importance to elucidate the organ tropism of metastatic cancer-derived EVs. In addition, the degree of contribution of EV-delivered miR-181c to BBB breakdown in the brain is not yet clear. Nevertheless, we have detected an increased level of circulating miR-181c in breast cancer patients with brain metastasis, suggesting an important role of secretory miR-181c in brain metastasis.

EVs as therapeutic targets might aid the prevention of BBB destruction in brain metastatic cancer and other diseases. The phenomenon of BBB destruction is not limited to brain metastases. BBB destruction is also known to be involved in diabetes, stroke, trauma, Alzheimers disease, brain tumours and malaria28 that lead to disorders in brain function29,30. On the other hand, BBB destruction mechanism by cancer-derived EVs might be a useful drug-delivery system for brain through BBB. This BBB breakdown mechanism includes the possibility of developing to drug-delivery system using EVs. Furthermore, EVs from cells contain multiple molecules, such as mRNAs, miRNAs and proteins as well as membrane-associated molecules. These molecules might affect recipient cells in multiple ways and might contribute to target organ tropism of circulating EVs.

In conclusion, this study provides new insights into the mechanisms of brain metastatic cancer.

Methods

Isolation of brain metastatic cells. A cell suspension containing 2 105 MDA

MB-231-luc-D3H2LN breast cancer cells in a volume of 100 ml was injected into the left cardiac ventricle of anaesthetised 7-week-old C.B-17 Icr-scid scid female mice. Tumour development was monitored by weekly bioluminescence imaging using the IVIS Spectrum (Caliper Life Science, Hopkinton, MA). Brain metastatic lesions were conrmed by histological analysis after necropsy. Brain lesions were localized by ex vivo bioluminescence imaging and resected under sterile conditions. Half of the tissue was xed with 4% paraformaldehyde and processed for histo-logical analysis. The other half was minced and placed in RPMI1640 culture medium (Gibco) containing antibioticantimycotic agents and 10% fetal bovine serum (FBS). The cells were briey centrifuged, resuspended in 0.025% trypsin EDTA (Gibco), and incubated for an additional 10 min at 37 C. The cells were resuspended in a culture medium containing 50 mg ml 1 Zeocin (Gibco) and were allowed to grow to conuence on a 10-cm dish. BMD2a and BMD2b cells were sorted for further propagation in culture or inoculation in mice. All animal work followed a protocol approved by the NCC Institutional Animal Care and Use Committee (T12-005, T12-005-M01).

Cell culture. MDA-MB-231-luc-D3H1 (purchased from Xenogen Co., CA), MDA-MB-231-luc-D3H2LN (purchased from Xenogen Co., CA), BMD2a and BMD2b (established from MDA-MB-231-luc-D3H2LN) breast cancer cell lines were cultured in RPMI1640 medium and supplemented with 10% heat-inactivated FBS (Invitrogen) and antibioticantimycotic agents at 37 C in 5% CO2.

Transwell invasion assay. For invasion assays, 2 104 cells were plated in the top

chamber with a Matrigel-coated membrane (24-well insert, BD Biosciences, NJ, USA). Cancer cells were plated in RPMI1640 medium without FBS and RPMI1640

NATURE COMMUNICATIONS | 6:6716 | DOI: 10.1038/ncomms7716 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 9

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7716

Actin

Occludin Actin Merge Merge Actin Merge

N-cadherin

ZO-1

N.C. PDPK1 siRNA

Claudin-5 Actin Merge

ZO-1 (220 kDa)

N.C.

PDPK1 siRNA

N.C.

PDPK1 siRNA

Occludin (65 kDa) Claudin-5

(23 kDa) N-cadherin

(140 kDa)

Actin (43 kDa)

GAPDH (40 kDa)

250 kDa

50 kDa

20 kDa 150 kDa

50 kDa

37 kDa

TEER ( cm2 )

20

20 40 60 80 100 120 140

0

Relative luminescence

1.2

1.0

0.8

0.6

0.4

0.2

N.C. miR-181c

**

**

**

0 Human Macaca

Treated with EVs D3H2LN BMD2a BMD2b

PDPK1 siRNA

PDPK1 (5868 kDa)

Cofilin (19 kDa)

P-cofilin (19 kDa)

GAPDH (40 kDa)

50 kDa 50 kDa

20 kDa

20 kDa 20 kDa

PDPK1 (5868 kDa)

Cofilin (19 kDa)

P-cofilin (19 kDa)

GAPDH (40 kDa)

N.C. miR-181c

20 kDa

37 kDa 37 kDa

Figure 7 | The target of miR-181c, PDPK1, in endothelial cells regulates the localization of tight junction proteins, N-cadherin and actin. (a) Coimmunouorescence of tight junction proteins (Claudin-5, Occludin and ZO-1), N-cadherin (red) and actin lament (green) after the addition of EVs from D3H2LN, BMD2a or BMD2b cells. Scale bar. 20 mm. (b) Western blot analysis of tight junction proteins, N-cadherin, actin and GAPDH. Proteins were from brain endothelial cells treated with PDPK1 siRNA. This experiment was repeated twice. (c) The TEER value was monitored before (day 4) and after (day 5) the transfection of PDPK1 siRNA or negative control. Error bars represent s.d., Students t-test, n 3 (*Po0.01). (d) Luciferase activities measured by

cotransfecting miR-181c and the PDPK1 luciferase reporters. Error bars represent s.d., Students t-test, n 6 (**Po0.01). (e) Western blot analysis of PDPK1,

colin, phospho-colin (P-colin) and GAPDH. Proteins were from brain endothelial cells treated with EVs from D3H2LN, BMD2a or BMD2b cells. This experiment was repeated twice. (f) Western blot analysis of PDPK1, phospho-colin (P-colin), colin and GAPDH. Proteins were from brain endothelial cells treated with miR-181c or PDPK1 siRNA. This experiment was repeated twice. Data are representative of at least three independent experiments each (Fig. a,c,d).

medium supplemented with 10% FBS (for all other cells) was used as a chemoattractant in the lower chamber. The cells were incubated for 24 h, and cells that did not migrate or invade through the pores were removed using a cotton swab. Cells on the lower surface of the membrane were stained with the Diff-Quick Staining Set (Sysmex, Hyogo, Japan) and counted. All assays were performed in triplicate. The data are expressed as the invasion percentage through the Matrigel matrix and membrane relative to the migration through the control membrane, according to the manufacturers instructions.

Preparation of conditioned medium and EVs. Before the collection of cultured medium, MDA-MB-231-luc-D3H1, MDA-MB-231-luc-D3H2LN, BMD2a and BMD2b cells were washed three times with Advanced RPMI containing antibiotic antimycotic agents and 2 mM L-glutamine (medium A), and the medium was switched to fresh medium A. After incubation for 2 days, the medium was collected and centrifuged at 2000g for 10 min at 4 C. To thoroughly remove cellular debris, the supernatant was ltered with a 0.22-mm Stericup (Millipore, MA, USA). To prepare the EVs, the conditioned medium was ultracentrifuged at 110,000g for 70 min at 4 C. The pellets were washed with 11 ml of PBS and resuspended in PBS after ultra-centrifugation. The fraction containing the EVs was measured for its protein content using the Micro BCA protein assay kit (Thermo Scientic, MA, USA).

PKH67-labelled EV transfer. Puried EVs derived from MDA-MB-231-luc-D3H2LN, BMD2a and BMD2b cells were labelled with a PKH67 green uorescent labelling kit (Sigma-Aldrich, MO, USA). EVs were incubated with 2 mM of PKH67 for 5 min, washed ve times using a 100-kDa lter (Microcon YM-100, Millipore) to remove excess dye and incubated for 24 h in the in vitro BBB model at 37 C in 5% CO2.

In vitro BBB model. To investigate the function of EVs, the BBB kit (MBT24-H) was used as an in vitro bloodbrain barrier model (PharmaCo-Cell Co. Ltd, Nagasaki, Japan, http://www.pharmacocell.co.jp/

Web End =http://www.pharmacocell.co.jp/ ). The TEER value exceeded 150 O cm 2, which indicates that this system could be used as an in vitro BBB model.

TEER study. BBB function of the in vitro BBB model was quantied by its TEER. The resistance values (O) were measured with an ohmmeter (Millicell ERS-2, Millipore). The TEER values were calculated by means of the unit area resistance:

R (AB) 0.33 cm2

R TEER(O cm 2)

A measurement resistance value (O)

B blank resistance value (O)

Permeability assay. We used a method modied from a previously described permeability assay31. After the addition of 200 ml of NaF (10 mg ml 1, Sigma-Aldrich)

to the upper chamber, as well as the addition of 900 ml of DPBS-H (Dulbeccos PBS (Mg , Ca ) and 10 mM HEPES, 4.5 mg ml 1 D-glucose) to the lower chamber, the plate was incubated with shaking at 37 C. After 30 min, the DPBS-H of the lower chamber was dispensed into a black plate (n 8). These samples were measured with

a multi-detection monochrometer microplate reader (485/535 nm, SAFIRE, Tecan) and the apparent permeability coefcient (Papp) was calculated as follows:

Papp (VA [C]A) A 1 [C]Luminal 1. t 1

VA: volume of abluminal chamber (0.9 cm3)

A: membrane surface area (0.33 cm2)[C]Luminal: initial luminal tracer concentration (mg ml 1)[C]A: abluminal tracer concentration (mg ml 1) t: time of experiment (min)

10 NATURE COMMUNICATIONS | 6:6716 | DOI: 10.1038/ncomms7716 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7716 ARTICLE

In vivo permeability assay. Puried EVs derived from MDA-MB-231-luc-D3H2LN and BMD2a cells were labelled with a XenoLight DiR (#125964, Summit Pharmaceuticals International Co.). EVs were incubated with 10 mM of XenoLight DiR for 15 min and washed ve times using a 100-kDa lter (Microcon YM-100, Millipore) to remove excess dye. A XenoLight DiR-labelled EV, 5 mg in a volume of 100 ml, was injected into the tail vein of anaesthetised 7-week-old C.B-17 Icr-scid scid mice. After 6 h, 100 ml of Tracer-653 probe (TR-1001, Molecular Targeting Technologies Inc., PA, USA) was injected into the tail vein of mice. After 24 h, mice were reuxed to remove excess dye in the blood and were monitored by uorescence imaging using the IVIS Spectrum (Caliper Life Science, Hopkinton, MA).

Transwell invasion assay using the in vitro BBB model. Breast cancer cell invasion was assessed in an in vitro BBB model (PharmaCo-Cell Co. Ltd, Nagasaki, Japan). Cancer cells were trypsinized for the invasion assay and labelled with PKH26 (Sigma-Aldrich) when using the in vitro BBB model. Cancer cells (2 104

cells) were plated in Hams F-12 (Gibco) medium without serum, and the medium supplied by the in vitro BBB model kit containing 10% serum was used as the chemoattractant in the lower chamber. After 48 h, non-invading cells were removed with cotton swabs and the nuclei were stained with Hoechst 33342 (Dojindo, Kumamoto, Japan). The PKH26 uorescence of invading cells was subsequently counted. All assays were performed in triplicate.

siRNA transfection. RAB27B, nSMase2, and both siRNA transfections of BMD2a cells were performed using the DharmaFECT transfection reagent (Thermo Scientic) according to the manufacturers protocol. A total of 25 nM of siRNA was used for each transfection. After 24 h, the transfected cells were used for each assay. miR-181c (Ambion, ID: MC10181; AACAUUCAACCUGUCGGUGAGU) and PDPK1 (Ambion, ID: S10275; UUUCUCACAGCCUAACCGCT) siRNA transfection were performed using the DharmaFECT transfection reagent. A total of 25 nM siRNA was used for each transfection.

RNA isolation and detection of miRNA and mRNA. The total RNA from the cultured cells or patient sera, which included the efcient recovery of small RNAs, was isolated using the RNeasy Mini Kit (Qiagen). EVs isolated from patients sera were isolated using Total Exosome Isolation (from serum) (Invitrogen). The expression of mRNA and miRNA was assessed with quantitative reverse transcription PCR (qRT-PCR) as described previously14. PCR was carried out in 96-well plates using the 7300 Real-Time PCR System (Applied Biosystems, CA, USA). All reactions were performed in triplicate. All TaqMan MicroRNA assays were purchased from Applied Biosystems. RNU6 was used as an internal control. The expression levels of miR-181c and RNU6 were measured by qRT-PCR using a Universal PCR Master Mix (Applied Biosystems). Gene expression was analysed using TaqMan Gene Expression Assays (Applied Biosystems). The expression of RAB27B, nSMase2, b-actin, PDPK1 and GAPDH was measured by qRT-PCR using a Platinum Quantitative PCR SuperMix (Applied Biosystems). Primers and probes were dened as follows: RAB27B (Assay ID: Hs01072206_m1), nSMase2 (Assay ID: Hs00920354_m1), actin-beta (Assay ID: Hs01060665_g1), PDPK1 (Assay ID: Hs00176884_m1) and GAPDH (Assay ID: Hs02758991_g1). GAPDH was used as an internal control.

Western blot. Proteins were isolated from cells using M-PER (Thermo Scientic, MA, USA) separated in Mini-PROTEAN TGX Gel (412%, Bio-Rad) and electrotransferred onto a PVDF membrane (Millipore). After blocking in Blocking One (Nacalai Tesque, Kyoto, Japan), the membranes were incubated for 1 h at room temperature with primary antibodies, which included anti-CD63 (puried mouse anti-human CD63, H5C6, 1:200, BD), anti-CD9 (ALB6, 1:200, Santa Cruz Biotechnology Inc.), anti-cytochrome C (puried mouse anti-cytochrome C, 7H8.2C12, 1:200, BD), anti-Claudin-5 (Z43.JK, 1:200, Invitrogen), anti-Occludin (ZMD.481, 1:200, Invitrogen), anti-ZO-1 (H-300, 1:100, Santa Cruz Biotechnology), anti-N-cadherin (3B9, 1:500, Invitrogen), anti-PDPK1 (#3062, 1:500, Cell Signaling), anti-GAPDH (6C5, 1:1000, Millipore), anti-Colin (D3F9, 1:1000, Cell Signaling), anti-Phospho-Colin (Ser3) (#3311, 1:500, Cell Signaling) and anti-NSMase2 (H-195, 1:200, Santa Cruz Biotechnology Inc.). Secondary antibodies (HRP-linked anti-mouse IgG, NA931 or HRP-linked anti-rabbit IgG, NA934, GE Healthcare) were used at a dilution of 1:2,000. The membrane was then exposed to ImmunoStar LD (Wako, Osaka, Japan). Original scans of the cropped images in the gures are presented in Supplementary Figs 713.

Immunouorescence. Endothelial cells from the in vitro BBB model were xed with 3% paraformaldehyde in PBS for 10 min at room temperature and treated with PBS containing 0.1% Triton X-100 for 10 min after being washed with PBS containing Mg2 and Ca2 to permeabilize the cells. After xation, the cells were incubated with 3% BSA in PBS for 1 h to block the nonspecic binding of antibodies. Subsequently, the endothelial cells were incubated with rabbit polyclonal antibodies against Claudin-5 (Z43.JK, Invitrogen, CA, USA), Occludin (ZMD.467, Invitrogen), ZO-1 (ZMD.437, Invitrogen) and N-cadherin (3B9, Invitrogen) at37 C for 1 h. After washing with PBS containing Mg2 and Ca2, they were incubated with Alexa Fluor 594-conjugated anti-rabbit IgG (Invitrogen) for 1 h at

37 C. Actin was stained with ActinGreen 488 ReadyProbes Reagent (R37110, Molecular Probes). The stained cells were then washed in PBS without Mg2 and

Ca2 and were mounted in VECTASHIELD Mounting Medium (H-1200, Vector Laboratories, CA, USA) for observation under a confocal microscope (FluoView

FV1000, Olympus, Tokyo, Japan).

PDPK1 30UTR luciferase reporter assay. The 30UTR for PDPK1 was PCR amplied from total RNA extracted from brain endothelial cells of the in vitro BBB model. PCR primers used to amplify the 30UTR include forward: 50-AACTCG

AGAATGCTGGCTATTGTTGGCCTC-30 and reverse: 50-AAGCGGCCGCAAGA TTAAATCACTGACCCAATAG-30. The PCR products were cloned into a pGEMT Easy Vector (Promega, WI, USA). Amplied 30UTR were cloned downstream of the Renilla luciferase coding region in the psiCHECK-2 (Promega). Human umbilical vein endothelial cells (HUVEC) cells were seeded in 96-well plates 24 h before transfection. The following day, 100 ng of reporter plasmid along with 100 nM of pre-miR-181c was co-transfected using DharmaFECT Duo transfection reagent (Thermo Scientic). HEK293 cells were cultured at a density of 5 104

cells per well in 96-well tissue culture plates overnight. Cells were collected 24 h after transfection and assayed for luciferase activity using EnVision (PerkinElmer, MA, USA). To assess the precursor miRNAs effect on reporter activity, 100 ng of synthetic precursor miRNA (pre-miR; Ambion, Invitrogen) was co-transfected. All experiments were performed in triplicate.

Microarrays. miRNA expression of EVs. Total RNA was extracted from EVs using QIAzol reagent and miRNeasy Mini Kit (Qiagen). RNA quantity and quality were determined using NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientic Inc.) and Agilent Bioanalyzer (Agilent Technologies), as recommended. Total RNA was labelled with cyanine 3 (Cy3) using miRNA Complete Labeling and Hyb Kit (Agilent Technologies) following the manufacturers instructions. In brief, 100 ng of total RNA was dephosphorylated using Calf Intestinal Alkaline Phosphatase Master Mix incubated at 37 C for 30 min. Dephosphorylated RNA was denatured with dimethylsulphoxide incubated at 100 C for 5 min and then immediately transfers to on ice for 2 min. These products were mixed with a Ligation master mix for T4 RNA Ligase and Cy3-pCp (Cy3-Cytidine biphosphate) and incubated at 16 C for 2 h. Labelled RNA was dried using vacuum concentrator at 55 C for 1.5 h. Cy3-pCp-labelled RNA was hybridized on Agilent SurePrint G3 Human miRNA 8 60 K Rel.19 (design ID: 046064). After washing, microarray

were scanned using an Agilent DNA microarray scanner. Intensity values of each scanned feature were quantied using Agilent Feature Extraction software version10.7.3.1, which performs background subtractions. We only used features that were agged as no errors (Detected ags) and excluded features that were not positive, not signicant, not uniform, not above background, saturated and population outliers (not detected ags). This expression analysis was performed with Agilent GeneSpring GX version 12.6.1. There are a total of 2,006 miRNA probes on SurePrint G3 Human miRNA 8 60 K Rel.19 (design ID: 046064) without control

probes. We applied Z2-fold change in signal intensity to identify the signicant differences of gene expression in this study. Raw and normalized microarray data are available in the Gene Expression Omnibus database (accession numbers GSE63445).mRNA expression of cells. Total RNA was extracted from cultured brain endothelial cells before EVs or miR-181c treatment using the QIAzol reagent and the miRNeasy Mini Kit (Qiagen). RNA quantity and quality were determined using a NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientic Inc.) and an Agilent Bioanalyzer (Agilent Technologies), as recommended. Total RNA was amplied and labelled with Cy3 using Low Input Quick Amp Labeling Kit, onecolour (Agilent Technologies) following the manufacturers instructions. In brief, 100 ng of total RNA was reversed-transcribed to double-strand complementary DNA (cDNA) using a poly dT-T7 promoter primer. Primer, template RNA and quality-control transcripts of known concentration and quality were rst denatured at 65 C for 10 min and incubated for 2 h at 40 C with 5 rst strand buffer,

0.1 M dithiothreitol, 10 mM deoxynucleotide triphosphate mix and AfnityScript RNase Block Mix. The AfnityScript enzyme was inactivated at 70 C for 15 min. cDNA products were then used as templates for in vitro transcription to generate uorescent complementary RNA (cRNA). cDNA products were mixed with a transcription master mix in the presence of T7 RNA polymerase and Cy3-labelled CTP (cytidine 50-triphosphate) and incubated at 40 C for 2 h. Labelled cRNA was puried using RNeasy Mini Spin Columns (Qiagen) and eluted in 30 ml of nuclease-free water. After amplication and labelling, cRNA quantity and cyanine incorporation were determined using a NanoDrop ND-1000 spectrophotometer and an Agilent Bioanalyzer. For each hybridization, 0.60 mg of Cy3-labelled cRNA was fragmented and hybridized at 65 C for 17 h to an Agilent Cynomolgus macaque Gene Expression Proling Array (design ID: 028520). After washing, the microarrays were scanned using an Agilent DNA microarray scanner. Intensity values of each scanned feature were quantied using Agilent Feature Extraction software version 10.7.3.1, which performs background subtractions. We only used features that were agged as no errors (detected ags) and excluded features that were not positive, not signicant, not uniform, not above background, saturated and population outliers (compromised and not detected ags). Normalization was performed with Agilent GeneSpring GX version 12.6.1 (per chip: normalization to 75th percentile shift; per gene: normalization to median of all samples). There are a

NATURE COMMUNICATIONS | 6:6716 | DOI: 10.1038/ncomms7716 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 11

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7716

total of 12,243 probes on Agilent Cynomolgus macaque Gene Expression Proling Array (design ID: 028520) without control probes. The altered transcripts were quantied using the comparative method. We applied Z2-fold change in signal intensity to identify the signicant differences of gene expression in this study. Raw and normalized microarray data are available in the Gene Expression Omnibus database (accession numbers GSE63447).

Patient serum samples. Collection and usage of human serum from breast cancer patients (n 56) by the National Cancer Center Institute of Japan were approved

by the Institutional Review Board (No.2013-111). Serum was aliquoted and stored at 80 C until used, and freeze-thawing was avoided as much as possible after

that. Informed consent was obtained from all the patients.

Statistical analyses. The data are presented as the means.d. P values o0.05 were considered statistically signicant by Students t-test or MannWhitney U-test.

References

1. Arshad, F., Wang, L., Sy, C., Avraham, S. & Avraham, H. K. Blood-brain barrier integrity and breast cancer metastasis to the brain. Patholog. Res. Int. 2011, 920509 (2010).

2. Bos, P. D., Nguyen, D. X. & Massagu, J. Modeling metastasis in the mouse. Curr. Opin. Pharmacol. 10, 571577 (2010).

3. Winkler, E. A., Bell, R. D. & Zlokovic, B. V. Central nervous system pericytes in health and disease. Nat. Neurosci. 14, 13981405 (2011).

4. Ballabh, P., Braun, A. & Nedergaard, M. The blood-brain barrier: an overview: structure, regulation, and clinical implications. Neurobiol. Dis. 16, 113 (2004).

5. Lee, T. H., Avraham, H. K., Jiang, S. & Avraham, S. Vascular endothelial growth factor modulates the transendothelial migration of MDA-MB-231 breast cancer cells through regulation of brain microvascular endothelial cell permeability. J. Biol. Chem. 278, 52775284 (2003).

6. Nicolson, G. L. Organ specicity of tumor metastasis: role of preferential adhesion, invasion and growth of malignant cells at specic secondary sites. Cancer. Metastasis Rev. 7, 143188 (1988).

7. Orr, F. W., Wang, H. H., Lafrenie, R. M., Scherbarth, S. & Nance, D. M. Interactions between cancer cells and the endothelium in metastasis. J. Pathol. 190, 310329 (2000).

8. Yano, S. et al. Expression of vascular endothelial growth factor is necessary but not sufcient for production and growth of brain metastasis. Cancer Res. 60, 49594967 (2000).

9. Yang, C. & Robbins, P. D. The roles of tumor-derived exosomes in cancer pathogenesis. Clin. Dev. Immunol. 2011, 842849 (2011).

10. Bobrie, A. et al. Rab27a supports exosome-dependent and -independent mechanisms that modify the tumor microenvironment and can promote tumor progression. Cancer Res. 72, 49204930 (2012).

11. Peinado, H. et al. Melanoma exosomes educate bone marrow progenitor cells toward a pro-metastatic phenotype through MET. Nat. Med. 18, 883891 (2012).

12. Valadi, H. et al. Exosome-mediated transfer of mRNAs and microRNAs is a novel mechanism of genetic exchange between cells. Nat. Cell Biol. 9, 654659 (2007).

13. Bos, P. D. et al. Genes that mediate breast cancer metastasis to the brain. Nature 459, 10051009 (2009).

14. Kosaka, N. et al. Neutral sphingomyelinase 2 (nSMase2)-dependent exosomal transfer of angiogenic microRNAs regulate cancer cell metastasis. J. Biol. Chem. 288, 1084910859 (2013).

15. Zhou, W. et al. Cancer-secreted miR-105 destroys vascular endothelial barriers to promote metastasis. Cancer Cell 25, 501515 (2014).

16. Gaillard, P. J. et al. Establishment and functional characterization of an in vitro model of the blood-brain barrier, comprising a co-culture of brain capillary endothelial cells and astrocytes. Eur. J. Pharm. Sci. 12, 215222 (2001).17. Wilhelm, I., Fazakas, C. & Krizbai, I. A. In vitro models of the blood-brain barrier. Acta Neurobiol. Exp. (Wars) 71, 113128 (2011).

18. Caby, M. P., Lankar, D., Vincendeau-Scherrer, C., Raposo, G. & Bonnerot, C. Exosomal-like vesicles are present in human blood plasma. Int. Immunol. 17, 879887 (2005).

19. Nakagawa, S. et al. A new blood-brain barrier model using primary rat brain endothelial cells, pericytes and astrocytes. Neurochem. Int. 54, 253263 (2009).

20. Ostrowski, M. et al. Rab27a and Rab27b control different steps of the exosome secretion pathway. Nat. Cell Biol. 12, 1930 sup pp 11-13 (2010).

21. Lyle, A. N. & Griendling, K. K. Modulation of vascular smooth muscle signaling by reactive oxygen species. Physiology (Bethesda) 21, 269280 (2006).

22. Higuchi, M., Onishi, K., Kikuchi, C. & Gotoh, Y. Scaffolding function of PAK in the PDK1-Akt pathway. Nat. Cell Biol. 10, 13561364 (2008).

23. Zaru, R., Mollahan, P. & Watts, C. 3-phosphoinositide-dependent kinase 1 deciency perturbs Toll-like receptor signaling events and actin cytoskeleton dynamics in dendritic cells. J. Biol. Chem. 283, 929939 (2008).

24. Lu, Z., Cox-Hipkin, M. A., Windsor, W. T. & Boyapati, A. 3-phosphoinositide-dependent protein kinase-1 regulates proliferation and survival of cancer cells with an activated mitogen-activated protein kinase pathway. Mol. Cancer Res. 8, 421432 (2010).

25. Ji, J. et al. Identication of microRNA-181 by genome-wide screening as a critical player in EpCAM-positive hepatic cancer stem cells. Hepatology 50, 472480 (2009).

26. Sand, M. et al. Expression of microRNAs in basal cell carcinoma. Br. J. Dermatol. 167, 847855 (2012).

27. Lowery, A. J. et al. MicroRNA signatures predict oestrogen receptor, progesterone receptor and HER2/neu receptor status in breast cancer. Breast Cancer Res. 11, R27 (2009).

28. Brown, H. et al. Evidence of blood-brain barrier dysfunction in human cerebral malaria. Neuropathol. Appl. Neurobiol. 25, 331340 (1999).

29. Abbott, N. J., Rnnback, L. & Hansson, E. Astrocyte-endothelial interactions at the blood-brain barrier. Nat. Rev. Neurosci. 7, 4153 (2006).

30. Polimeni, M. & Prato, M. Host matrix metalloproteinases in cerebral malaria: new kids on the block against blood-brain barrier integrity? Fluids Barriers CNS 11, 1 (2014).

31. Kis, B. et al. Adrenomedullin regulates blood-brain barrier functions in vitro. Neuroreport 12, 41394142 (2001).

Acknowledgements

This work was supported in part by a Grant-in-Aid for the third term of a comprehensive 10-year strategy for cancer control; a Grant-in-Aid for scientic research on priority areas in cancer from the Ministry of Education, Culture, Sports, Science and Technology; the Program for Promotion of Fundamental Studies in Health Sciences of the National Institute of Biomedical Innovation (NiBio); the Japan Society for the Promotion of Science through the Funding Program for World-Leading Innovative R&D on Science and Technology (FIRST Program) initiated by the Council for Science and Technology Policy; and a Grant-in-Aid for Scientic Research on Innovative Areas (functional machinery for non-coding RNAs) from the Japanese Ministry of Education, Culture, Sports, Science and Technology. N.K. is supported by the JSPS Postdoctoral Fellowship for Research Abroad. We thank Dr. Kou Furuta for providing patient serum samples. We thank Ayako Inoue for excellent technical assistance.

Author contributions

N.T. performed the experimental work, analysed the data and wrote the manuscript. N.K. assisted in the writing of the manuscript and provided helpful discussion. K.T. provided the patients serum samples. M.O., T.K., Y.Y., H.N. and J.L. provided helpful discussion. The manuscript was nalized by T.O. with the assistance of all of the authors.

Additional information

Accession codes. Raw and normalized microarray data are available in the Gene Expression Omnibus (GEO) database (accession numbers GSE63445, and GSE63447).

Supplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Web End =http://www.nature.com/ http://www.nature.com/naturecommunications

Web End =naturecommunications

Competing nancial interests: The authors declare no competing nancial interests.

Reprints and permission information is available at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/ http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Tominaga, N. et al. Brain metastatic cancer cells release microRNA-181c-containing EVs capable of destructing the bloodbrain barrier in vitro. Nat. Commun. 6:6716 doi: 10.1038/ncomms7716 (2015).

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

12 NATURE COMMUNICATIONS | 6:6716 | DOI: 10.1038/ncomms7716 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.