Exosomes Promote Ovarian Cancer Cell Invasion through Transfer of CD44 to Peritoneal Mesothelial Cells

Ovarian cancer, which has a cure rate of only 30%, is generally characterized by widespread peritoneal dissemination and ascites (1). During ovarian cancer dissemination, cells detach from the primary site of origin, which was initially thought to be the ovary but is now believed to be the fallopian tube. Subsequently, the cancer cells float in the peritoneal cavity and attach to secondary sites of implantation, most notably the omentum, which is the most common site of ovarian cancer metastasis (2). All of the organs within the peritoneal cavity are lined with a single layer of mesothelial cells which cover an underlying stroma composed of extracellular matrices (ECM) and stromal cells (3). Once ovarian cancer cells attach to this layer, they invade through the barrier of mesothelial cells into the peritoneum, omentum, or bowel serosa. Therefore, inhibiting the initial interaction of cancer with mesothelial cells could be a means of preventing ovarian cancer progression and metastasis.

Mesothelial cells were first thought to function only as a mechanical barrier. This view was supported by histologic studies of ovarian cancer nodules attached to peritoneal organs that showed an absence of mesothelial cells under the tumor, suggesting that the mesothelial cells had failed as a protective barrier and contributed to successful tumor cell implantation (4–7). Recent reports suggest that mesothelial cells actively participate in establishing an ovarian cancer metastatic niche and that they are reprogrammed by ovarian cancer cells to facilitate tumor growth becoming cancer-associated mesothelial cells (2, 8, 9).

Exosomes are 30- to 100-nm membrane vesicles of endocytic origin, mediating diverse biologic functions, including tumor cell invasion, cell–cell communication, and antigen presentation through transfer of proteins, mRNAs, and microRNAs (10). The very recently characterized mechanism of cell–cell communication through exosomes raises the possibility that the transfer of genetic information or proteins via exosomes modulates cellular activities in recipient cells (10).

Given that the role of mesothelial cells during ovarian cancer metastasis is unclear and that exosomes are able to transfer information between different cell types, we set out to determine whether exomes play a role in communication between mesothelial cells and ovarian cancer cells.

GW4869 (#92613), 5-(N-ethyl-N-isopropyl)-amiloride (EIPA; #A3058), PKH67 (MIDI67), and Ponceau-S (P3045-10G) were purchased from Sigma-Aldrich. E-Cadherin antibody (#610181) and the integrin antibodies Sampler Kit (#611435; integrin-α2, -α5, -αV, -β1, and -β3) were purchased from BD Biosciences. Antibodies against integrin-α1 (sc-10728), cytokeratin-18 (sc-6259), and Oct-3/4 (sc-9018) were obtained from Santa Cruz Biotechnology. Antibodies against β-actin (#4967) and CD44 (156-3C11) were obtained from Cell Signaling. Alexa Fluor 633 Goat Anti-Mouse IgG (A21050), Lipofectamine 2000 (#11668019), TRIzol (#15596-018), and CellTracker Green CMFDA (C2925) were obtained from Life Technologies. Donkey Anti-Mouse IgG 10 nm gold (ab39593), Alexa Fluor 488 Goat Anti-Mouse IgG H&L (ab150113), Alexa Fluor 555 Donkey Anti-Rabbit IgG H&L (ab150074), antibody against MMP9 (ab53296), and antibody against CD63 (ab134045) were from Abcam. Antibody against CD63 (EXOAB-CD63A-1) was purchased from SBI. Antibody against CD63 (#11-343-C025) and CD81 (#11-558-C025) was purchased from EXBIO. Antibody against vimentin (M0725) was purchased from DAKO. Antibody against calretinin (RB-9002-P0) was purchased from Thermo Fisher Scientific Inc. Antibody against MMP9 (GTX-100458) was purchased from Genetex. Recombinant human TGFβ1 (#100-21) was purchased from PeproTech. LY2157299 (#120-05991) was purchased from Wako Pure Chemical Industries.

The HeyA8 cell line was generously provided by Dr. Anil Sood (MD Anderson Cancer Center, Houston, TX). The IOSE cell line was generously provided by Dr. Masaki Mandai (Kinki University, Osaka, Japan). Briefly, ovarian surface epithelial cells were purified from normal ovary and transfected with SV40 large T antigen and the human telomerase reverse transcriptase (hTERT; ref. 11). TYK-nu cell lines were purchased from Health Science Research Resources Bank (Osaka, Japan). RMG-1, RMUG-S, OVSAHO, and OVKATE cell lines were purchased from Health Science Research Resources Bank. OVCAR-3 cell line was provided by the Cell Resource Center for Biomedical Research Institute of Development, Aging and Cancer of Tohoku University (Sendai, Japan). ES-2 and Caov-3 cell lines were purchased from ATCC. Cells were authenticated by short tandem repeat DNA profiling at Takara-Bio Inc. and were used for this study within 6 months of resuscitation.

Exosomes were isolated from cell culture medium as previously described (12) with minor modifications. Conditioned medium (CM) containing exosome-depleted FBS (prepared by overnight ultracentrifugation at 100,000 × g at 4ºC) was prepared by incubating cells grown at subconfluence for 48 hours. CM was centrifuged at 2,000 × g for 20 minutes at room temperature. Supernatant fractions were filtered using 200-nm pore size filters. The resulting cell-free medium was ultracentrifuged at 100,000 × g for 70 minutes to generate an exosome pellet that was washed once with PBS using the same ultracentrifugation conditions, followed by resuspension in PBS. The amount of exosomal protein was assessed by the Lowry method (Bio-Rad).

Electron microscopy was performed as described (12) using a transmission electron microscope (H-7650; Hitachi, Ltd.).

Western blotting was performed as described (13). The primary antibodies were used in the following dilutions: CD44 (1:1,000), CD63 (1:250), CD81 (1:250), E-cadherin (1:1,000), β-actin (1:2,000), integrin-α1 (1:200), integrin-α2 (1:1,000), integrin-α5 (1:2,500), integrin-αV (1:2,500), integrin-β1 (1:2,500), integrin-β3 (1:1,000), Oct-3/4 (1:200), and calretinin (1:500).

Immunohistochemistry of ovarian cancer paraffin-embedded tissue was performed as previously described (14). The slides were incubated with the primary antibody against CD44 (#156-3C11, 1:200), calretinin (1:200), or MMP9 (1:50). For human peritoneal mesothelial cell (HPMC) staining, cells were plated on 8-well chamber slides and were fixed with 4% paraformaldehyde. Slides were stained with an antibody against cytokeratin-18 (1:50), vimentin (1:50), or calretinin (1:100).

HPMCs were fixed with 4% paraformaldehyde and stained with anti-CD44 antibody (#156-3C11, 1:200) at room temperature for 1 hour (15). After washing, samples were incubated with Alexa Fluor 633–labeled goat anti-mouse IgG (1:100) and stained with DAPI. The samples were imaged using an FV1000-D Laser Scanning Confocal microscope (Olympus).

Overexpression of CD44s in HPMCs was performed by transient transfection using the CD44s pBabe-puro retroviral vector (Plasmid#19127, Addgene) at a concentration of 1 μg/mL. Twenty-four hours after transfection, HPMCs were used for experiments. The retroviral vector pBABEpuro gateway (Plasmid#51570, Addgene) served as a control. CD44 expression in HeyA8 cells was transiently knocked down using a predesigned siRNA duplex directed against CD44 at a concentration of 200 nmol/L. The sequences of the siRNA duplexes (Japan Bioservice) were as follows: CD44 siRNA #1, 5′-AAAUGGUCGCUACAGCAUCTT-3′ (sense) and 5′-GAUGCUGUAGCGACCAUUUTT-3′ (antisense); CD44 siRNA #2, 5′-GUAUGACACAUAUUGCUUCTT-3′ (sense) and 5′-GAAGCAAUAUGUGUCAUACTT-3′ (antisense). MISSION siRNA Universal Negative Control #1 (SCI001, Sigma) was used as a control.

Total RNA was extracted using TRIzol. RNA isolated from cells and exosomes were analyzed using the Agilent 2100 Bioanalyzer (Agilent Technologies, Inc.).

For exosome tracking, exosomes were fluorescently labeled using the PKH67 dye according to manufacturer's protocol.

HPMCs (2 × 10⁴) were plated onto 25-μg Matrigel-coated 24-well FluoroBlok cell culture inserts with an 8-μm pore membrane (351152 Corning Incorporated). Because FluoroBlok membranes almost completely block the passage of light from 490 to 700 nm, only fluorescently labeled cells that have invaded are detected by bottom-reading of fluorescence (16). After HPMCs grew to confluence, 5 × 10⁴ fluorescently (CMFDA)-labeled epithelial ovarian cancer (EOC) cells (HeyA8 and TYK-nu) cultured in 0.1% BSA/DMEM were incubated on the monolayer of HPMCs and allowed to invade for 48 hours. Cell culture medium consisting of 10%FBS/DMEM was used as a chemoattractant and was kept in the lower chamber. Fluorescence of invaded cells was read at 494/517 nm (Ex/Em) on a bottom-reading fluorescent plate reader (SH-9000Lab, Corona Electric Co., Ltd.), and the number of invaded cells was quantitated by imaging an average of 25 random areas per well. In experiments using exosomes, culture media were changed to 0.1% BSA/DMEM, once HPMCs grew to confluence, and isolated exosomes were added 24 hours before the treatment with cancer cells.

CMFDA-labeled EOC cells were directly plated onto Matrigel-coated inserts. When exosomes were used, isolated exosomes in 0.1% BSA/DMEM were added to the Matrigel-coated inserts 24 hours before treatment with cancer cells.

EOC cells were plated onto 96-well plates (5 × 10³ cells/well) in 10%FBS/DMEM for 24 hours and incubated in 0.1% BSA/DMEM with or without exosomes (100 μg/mL) or GW4869 for 48 hours. Primary HPMCs were plated onto 96-well plates (1 × 10⁴ cells/well) in 20%FBS/RPMI for 24 hours and incubated in 0.1% BSA/RPMI with or without exosomes (1, 10, 100 μg/mL) for 24 hours. Cell proliferation was evaluated by a modified MTS assay as previously described (15).

Primary human HPMCs (5 × 10⁴) were plated on a 24-well plate and incubated until confluent (7). Culture media were switched to 0.1%BSA/DMEM, and 100 μg/mL of isolated exosomes was added. Twenty-four hours following the incubation, the cells were fixed and stained with Acti-Stain 555–conjugated phalloidin (#PHDH1, Cytoskeleton). Mesothelial clearance area was defined as cell-free area, and mesothelial clearance was calculated as the ratio of cell-free area to total area using the ImageJ Software (NIH, Bethesda, MD).

Serum-free conditioned medium was collected from 1 × 10⁵ HPMCs with or without 100 μg/mL of EOC-derived exosomes, and a gelatin zymogram was performed as previously described (17).

Total RNA was extracted using TRIzol, and cDNA was synthesized using ReverTra Ace qPCR RT Master Mix (FSQ-201; TOYOBO). qRT-PCR was performed using the StepOnePlus Real-Time PCR System. The following probes were used: human matrix metalloproteinase-9 (MMP9: Hs00234579_m1) and human GAPDH (4326317E). Relative expression of CD44 was assayed using the SYBR Green assay (Applied Biosystems: #4309155). The designed primers in this experiment were as follows: CD44: forward primer: 5′-AGTCACAGACCTGCCCAATGCCTTT-3, reverse primer: 5′-TTTGCTCCACCTTCTTGACTCCCATG-3′; GAPDH: forward primer: 5′-CATGAGAAGTATGACAACAGCCT-3′, reverse primer: 5′-AGTCCTTCCACGATACCAAAGT-3′ (Applied Biosystems). PCR was performed in triplicate, and the relative levels of CD44 and MMP9 expression were calculated using the 2^−ΔΔCt method.

The concentration of TGFβ in exosomes was analyzed using Human/Mouse TGFβ 1 ELISA Ready-SET-Go! (#88-8350; eBioscience).

GW4869 (18) was stored at −20°C as a 3 mmol/L stock suspension in DMSO. Right before use, the suspension was solubilized by the addition of 5% methane sulfonic acid (MSA; Wako).

Nanoparticles in exosome suspensions were analyzed using a NanoSight LM10V-HS (Malvern Instruments Ltd.).

Specimens of human omentum were obtained following written informed consent from patients undergoing gynecological surgeries at Osaka University Hospital (Osaka, Japan). Samples were obtained in accordance with the Declaration of Helsinki (Institutional Review Board of Osaka University Graduate School of Medicine approval no.: 10064). HPMCs were isolated as previously described (19). Briefly, benign omental specimens were minced and incubated with 1:1 solution of PBS and 0.25% trypsin/25 mmol/L EDTA at 37°C for 30 minutes. After the solution containing cells was centrifuged, the pellet was suspended and cultured in 20% FBS/RPMI-1640 with 100 U/mL penicillin/streptomycin. Cells at passage 1 were used for the subsequent experiments. EOC samples were collected from 15 patients with FIGO stage III–IV epithelial serous ovarian cancer consecutively treated at Osaka University Hospital between 2011 and 2015. All cases were histologically confirmed by a gynecologic pathologist as being high-grade serous carcinomas.

Female athymic BALB/c nude mice were purchased from CLEA Japan, Inc. All animal experiments were approved by the Institutional Animal Care and Use Committee of Osaka University. HeyA8 cells indicated in the figures were injected intraperitoneally. Seven days after the inoculation, the mice were sacrificed and omental tissues were carefully dissected. No statistical methods were used to predetermine sample size. The experiments were not randomized, and the investigators were not blinded to allocation during experiments and outcome assessment.

JMP software version 10.0.2 (SAS Institute Japan Ltd.) was used for statistical analysis. All data had normal distribution and equal variances were tested by F-test. Data are expressed as mean ± SEM. Differences were analyzed using one-way ANOVA and Bonferroni correction for multiple comparisons. Differences were considered statistically significant at P < 0.05.

To elucidate the effects of exosomes derived from EOC cells on peritoneal dissemination, exosomes were characterized by electron microscopy, Western blotting, and capillary electrophoresis of RNA. Electron microscopy revealed that exosomes had a round-shaped morphology and were approximately 50 to 100 nm in size and stained positive for CD63, an exosome marker (ref. 20; Fig. 1A). Exosome secretion was measured in several EOC cells, and HeyA8 and TYK-nu were found to produce the highest quantity of exosomes (Fig. 1B) and were used in subsequent experiments. As a nonmalignant control, exosomes from immortalized ovarian surface epithelial cells (IOSE) were used. Immunoblotting showed that CD63 (21) was seen in exosomes from all cell lines used in these experiments. CD81 (21), another exosome marker, was also enriched in exosomes from HeyA8 cells and IOSEs, whereas whole-cell lysates expressed minimal CD63 and CD81 (Fig. 1C). In contrast, β-actin was expressed in whole-cell lysates, whereas exosomes showed low expression (Fig. 1C). Total RNAs extracted from whole cells using capillary electrophoresis expressed the 18S and 28S ribosomal subunits, whereas they were not detected in exosomes which expressed small RNAs, such as miRNAs (ref. 22; Fig. 1D).

We then studied the internalization of EOC-derived exosomes by HPMCs. Primary HPMCs displayed a cobblestone-like morphology and expressed cytokeratin-18, vimentin (19), and calretinin, a discriminatory marker for mesothelial cells (Fig., 1E; ref. 23). Exosomes from EOC cell lines and IOSE were internalized by HPMC and detected in the cytoplasm of the HPMCs using confocal laser microscopy (Fig. 1F), indicating that EOC-derived exosomes can be internalized by HPMCs.

The role of EOC-derived exosomes in ovarian cancer invasion was studied using an in vitro invasion assay. To mimic the mesothelial layer covering the omentum and other peritoneal organs, a monolayer of HPMCs was plated onto growth factor–reduced Matrigel (Fig. 2A). Pretreatment with HeyA8- or TYK-nu–derived exosomes on HPMCs significantly stimulated the invasion of ovarian cancer cells, whereas pretreatment with IOSE-derived exosomes did not affect EOC invasion. EOC exosomes did not affect proliferation of HPMCs (Fig. 2B and C). In the absence of HPMCs (Matrigel-only coating), exosomes derived from either HeyA8 or TYK-nu cells did not promote invasion (Fig. 2D), nor did treatment with exosomes alone promote ovarian cancer cell proliferation (Fig. 2E), indicating that exogenous treatment of EOC-derived exosomes affected HPMCs but not cancer cells.

Next, we aimed to identify the changes in HPMC surface receptors induced by exosomes (Fig. 3A). The expression pattern of key proteins related to tumor cell adhesion or invasion (24) was examined by Western blotting (Fig. 3B). Uptake of EOC-derived exosomes into HPMCs induced expression of the standard form of CD44 (CD44s) and downregulation of E-cadherin, whereas the expression patterns of integrin-α1, -α2, -α5, -αv, -β1, -β3 were unchanged (Fig. 3B). CD44 has several isoforms generated by alternative mRNA splicing (25). Only the expression of CD44s (80 kDa) was increased in HPMCs (but not in IOSE), whereas the expression of other CD44 variants (CD44v, 120–200 kDa) was very low (Fig. 3B). Western blotting as well as electron microscopy showed that CD44s was expressed by ovarian cancer–derived exosomes but not IOSE-derived exosomes (Fig. 3C and D). Three minutes after the uptake of PKH67-labeled exosomes, HPMCs were immunostained with Alexa Fluor 555–labeled CD44 antibody. CD44 was colocalized with exosomes, indicating that CD44 derived from ovarian cancer cells was actually internalized in HPMCs via exosomes (Fig. 3E). There was no significant change in CD44 mRNA levels in HPMCs following treatment with EOC-derived exosomes, indicating that exosomes do not regulate CD44 expression on the transcriptional level (Fig. 3F). Given that E-cadherin expression was decreased in HPMCs, we considered the possibility that mesothelial cells undergo epithelial–mesenchymal transition (EMT; ref. 2). Confocal microscopy revealed that HPMCs treated with EOC-derived exosomes become elongated and spindle-shaped, whereas HPMCs treated with IOSE-derived exosomes retained their cobblestone-like morphology (Fig. 1F). Because EOC-derived exosomes induced the morphologic transition to a mesenchymal phenotype, we determined HPMC cell clearance (6). Twenty-four hours after treatment with EOC-derived exosomes, cell–cell contact between HPMCs was reduced and the cell-free areas (i.e., mesothelial clearance) were significantly greater than those found with mesothelial cells treated with IOSE-derived exosomes (Fig. 3G). Because CD44 is associated with a proteolytic form of the MMP9 on the cancer cell membrane and CD44-associated cell surface MMP9 is known to promote cell-mediated degradation of ECM (26), MMP9 expression in HPMCs was assessed after exosome uptake. Quantitative real-time PCR confirmed that MMP9 mRNA expression was significantly increased in HPMCs treated with EOC-derived exosomes as compared with those treated with PBS only or IOSE-derived exosomes (HeyA-8; 5-fold, TYK-nu; 6.3-fold; Fig. 3H). Gelatin zymography was performed to analyze whether EOC-derived exosomes induce HPMCs to secrete MMP. Conditioned medium from HPMCs treated with EOC-derived exosomes increased the gelatinolytic activity of pro-MMP9, whereas they did not affect gelatinolytic activity of MMP-2 (Fig. 3I). The involvement of MMP9 activity in EOC invasion was analyzed using an anti-MMP9 antibody after its inhibitory effect against MMP9 activity was confirmed by gelatin zymography (Supplementary Fig. S1A). EOC invasion was significantly decreased by MMP9 inhibition, especially when HPMCs were pretreated with EOC-derived exosomes (Supplementary Fig. S1B). Collectively, these data show that EOC-derived exosomes promote ovarian cancer invasion by affecting HPMCs through the increase of mesothelial clearance and the induction of CD44-mediated MMP9 secretion by HPMCs.

GW4869, a neutral sphingomyelinase inhibitor, inhibits exosome secretion (18). Addition of GW4869 to HeyA8 (Fig. 4A and B) and TYK-nu (Supplementary Fig. S2A) ovarian cancer cells blocked exosome secretion in a dose-dependent manner, without affecting cell viability (Fig. 4C, Supplementary Fig. S2B). GW4869 treatment significantly inhibited (49%) the invasion of both cell lines in the presence of HMPC coating, whereas in the absence of HPMCs, invasion was unchanged (Fig. 4D, Supplementary Fig. S2C). Consistent with a regulation of CD44 by exosomes, GW4869 attenuated CD44 uptake in HPMCs cocultured with EOC cells in a dose-dependent manner. E-cadherin expression was inversely correlated with CD44, suggesting that the inhibitor blocked EMT in the HPMCs (Fig. 4E, Supplementary Fig. S2D) which was also confirmed by phase-contrast microscopy (Fig. 4F; Supplementary Fig. S2E). When EIPA, a known exosome uptake inhibitor that blocks macropinocytosis (27), was treated with HPMCs cocultured with EOC cells (Supplementary Fig. S3), CD44 expression in HPMCs was similarly inhibited.

In addition, GW4869 restored the epithelial appearance of HPMC in a dose-dependent manner (Fig. 4F). These results indicate that the inhibition of exosomes released from EOC cells inhibit ovarian cancer invasion by attenuating CD44 expression and reversing EMT in HPMCs.

Because TGFβ signaling is one of the key regulators of EMT (28) and is also correlated with enhanced CD44 expression (29), we studied the involvement of TGFβ in the acquisition of CD44 by HPMCs. At 10 ng/mL, TGFβ induced CD44 expression and downregulated E-cadherin expression, effects which were abrogated by use of the potent TGFβ receptor I kinase inhibitor LY2157299 (Fig. 5A). In addition, TGFβ induced the morphologic change of HPMCs to a spindle shape, an effect which was also inhibited by LY2157299 (Fig. 5B). The TGFβ receptor I kinase inhibitor did not interfere with the ability of EOC-derived exosomes to upregulate CD44s and downregulate E-cadherin in HPMCs (Fig. 5C). ELISA assays revealed that each exosome contains a low level of TGFβ (6–162 pg/100 μg exosomes) and that this concentration was similar between EOC cells and IOSEs (Fig. 5D). Coculture with EOC cells not only induced CD44 expression in HPMCs but also resulted in acquisition of a spindle cell, EMT-type morphology, despite LY2157299 treatment (Fig. 5E and F). These experiments suggest that CD44 acquisition by HPMCs via exosomes is independent of TGFβ signaling.

To further understand the role of exosomal CD44 in ovarian cancer progression, CD44 expression in HeyA8 cells was knocked down using siRNAs (Fig. 6A and B), which resulted in a reduction of CD44 expression in the HeyA8-derived exosomes (Fig. 6C). Treatment of HPMCs with exosomes derived from HeyA8 CD44 knockdown (KD) cells abrogated the upregulation of CD44s and the downregulation of E-cadherin in HPMCs (Fig. 6D). The pretreatment of HPMCs with exosomes from HeyA8 CD44 KD cells significantly blocked invasion (Fig. 6E) and mesothelial cell clearance (Fig. 6F) and reduced the gelatinolytic activity of MMP9 (Fig. 6G).

In the opposite experiment, HPMCs were transfected with a CD44 expression vector and their effect on EOC invasion was determined. Immunoblotting showed that CD44 was overexpressed in HPMC, leading to the downregulation of E-cadherin expression (Fig. 6H). CD44-overexpressing HPMCs displayed a spindle cell morphology and lost cell–cell contact, whereas control vector–transfected HPMCs retained a cobblestone-like appearance (Fig. 6I). CD44-overexpressing HPMCs promoted invasion of EOC cells when compared with control vector–transfected HPMCs (Fig. 6J). The mesothelial clearance area was significantly increased in CD44-overexpressing HPMCs (Fig. 6K), and the conditioned medium of CD44-overexpressing HPMCs increased the gelatinolytic activity of MMP9 (Fig. 6L). Taken together, the data in Fig. 6 show that exogenous expression of CD44 in HPMCs promotes ovarian cancer invasion by increasing mesothelial clearance and inducing MMP9 secretion.

Next, we investigated whether the results found using cell lines have correlates in human tissue from patients with EOC; specifically whether CD44 expression is increased in HPMCs from omental metastasis. Omental tissue harboring tumor was obtained from 15 patients and immunostained with anti-CD44 antibody (Supplementary Table S1). Normal HPMCs lining the omentum surface were identified using positive calretinin staining (Fig. 7A). In HPMCs from normal omentum, CD44 expression was mostly negative (Fig. 7A). In contrast, in 9 of 15 (60%) tumors, strong CD44 expression was seen in the primary ovarian high-grade serous performed (Supplementary Fig. S4) and showed CD44 to be colocalized with CD63 carcinomas (Fig. 7B). To confirm exosomes contain CD44 in these primary ovarian high-grade serous carcinomas, double immunostaining with CD44 and CD63, a representative exosome marker, was conducted. In omental tissue harboring disseminated ovarian cancer cells, CD44 was strongly expressed in metastatic implants including HPMCs. However, HPMCs did not express CD44 in areas further away from the tumor cells (Fig. 7C). HPMCs intertwined with metastatic cancer cells displayed spindle-like morphology, and the mesothelial cell barrier was disrupted (Fig. 7C). In serial sections, CD44-positive HPMCs harboring tumors expressed MMP9, whereas normal HPMCs were mostly MMP9-negative (Supplementary Fig. S5).

CD44 expression in mesothelial cells was also studied using an ovarian cancer xenograft model (30). To mimic early ovarian cancer metastasis to omental tissue, HeyA8 cells transfected with control siRNA or CD44 siRNA were injected intraperitoneally into female BALB/c nu/nu mice (31). One week after inoculation, mice showed microscopic tumor dissemination on the omental surface. Human CD44 expression was seen at the leading edge of invading HeyA8 cells transfected with control siRNA and mouse mesothelial cells harboring cancer cells displayed positive human CD44 expression (Fig. 7D, top). However, HeyA8 cells transfected with CD44 siRNA showed faint CD44 expression and CD44 was negative in mouse omental tissues harboring cancer cells (Fig. 7D, bottom). To further quantify the induction of CD44 expression, we injected either 1 × 10⁷ or 1 × 10⁶ HeyA8 cells intraperitoneally into female athymic BALB/c nude mice and compared human CD44 expression in omental tissues from the 2 groups of mice (Supplementary Fig. S6A–S6C). In the tissue of mice injected with 1 × 10⁷ HeyA8 cells, strong CD44 expression was seen (Supplementary Fig. S6D) as compared with tissue from mice injected with 1 × 10⁶ HeyA8 cells. These data support our in vitro findings indicating that CD44 acquisition by mesothelial cells is induced by direct contact with EOC cells.

Cancer-derived exosomes can act as mediators of metastasis, and, as a result, a minority population of cancer cells can reprogram stromal cells through the intercellular transfer of proteins, mRNAs, and miRNAs (32). Here, we show that EOC-derived exosomes promoted ovarian cancer invasion by transferring exosomes containing CD 44 to HPMC which reprograms them to a more EMT phenotype, thereby promoting ovarian cancer cell invasion and metastasis. EOC cells, by releasing exosomes, reprogram HPMCs, which then facilitate EOC invasion supporting a model of bidirectional communication (Fig. 7E).

Several studies have shown that cancer cell–derived exosomes reprogram or educate other cells to aid tumor survival and promote metastasis (33); however, to date, only a limited number of reports have investigated the cancer-promoting effect of a particular protein or miRNA in cancer-derived exosomes. For instance, melanoma-derived exosomes reprogram bone marrow progenitor cells to support tumor growth and metastasis through MET (34) and breast cancer–derived exosomal miR-105 promotes metastasis by destroying endothelial cell barriers (35). Recently, Singh and colleagues revealed that integrin-αvβ3 is transferred from tumorigenic to nontumorigenic cancer cells via exosomes and its de novo expression in recipient cells promotes cell migration on its ligand (36).

It is well-known that the tumor microenvironment, including stromal cells, plays an important role in cancer progression (37). Under normal conditions, mesothelial cells act as a mechanical barrier to protect intra-abdominal organs (3). However, reports have shown that mesothelial cells can also serve as a metastatic niche, promoting cancer cell adhesion and invasion, which are essential steps of the peritoneal dissemination of ovarian cancer (2, 38). Since several previous reports have focused on TGFβ signaling as a means of communication between EOC cells and HPMCs, we inhibited TGFβ signaling using a potent inhibitor, LY2157299. While this TGFβ inhibitor completely abrogated the morphologic changes to a mesenchymal phenotype and the downregulation of E-cadherin induced by TGFβ, treatment with EOC-derived exosomes induced these phenomena in the presence of LY2157299, indicating that the mechanism is TGFβ-independent.

Several adhesion molecules such as integrins or ECM components have been identified as important mediators of cancer metastasis to the mesothelium (39). Among these, hyaluronan (HA) and its receptor, CD44, have been shown to play critical roles in ovarian cancer metastasis (40). It is reported that ovarian cancer cell adhesion to mesothelial cell monolayers is mediated, at least in part, by the interaction between HA and CD44, which further mediates ovarian cancer progression (41). A recent immunohistochemical study showed that CD44 expression was seen in 42 of 72 (58.3%) ovarian tumors and that CD44 (+)/E-cadherin (−) ovarian tumors have a more aggressive phenotype and are associated with poor patient survival (42). We found a similar strong expression of CD44 in tumors analyzed. CD44 is also known to be a key regulator of EMT in several cancer types (43–45). Consistent with these reports, our findings show that HPMCs treated with EOC-derived exosomes displayed a mesenchymal phenotype and reduced E-cadherin expression. CD44 has also been shown to recruit MMP9 to the cell surface and mediate tumor cell invasion (25), which is consistent with our zymographic analyses (Fig. 6). Given the important role of CD44 in cancer invasion, the molecule holds promise as a therapeutic target in ovarian cancer treatment. However, clinical trials of systemic treatment with CD44 neutralizing antibodies have been terminated because of unacceptable levels of toxicity (46). Nontoxic alternative therapies need to be developed for future trials. Capturing CD44-expressing exosomes in body fluid appears to be less toxic and has the potential to be a possible anticancer therapy.

Interestingly, inhibition of exosomes released from EOC cells using GW4869 attenuated CD44 acquisition in HPMCs and accordingly suppressed ovarian cancer invasion. Kosaka and colleagues suggested that cancer cell–derived exosomes circulating in the blood not only promote cancer malignancy through the formation of premetastatic niches but also interfere with treatment (47). Indeed, Ciravolo and colleagues reported that the exosomes released by the HER2-overexpressing breast cancer cell lines express a full-length HER2 molecule and that these exosomes bound to trastuzumab and inhibited its anticancer cell proliferative activity (48). On the basis of this preclinical data, Aethlon Medical Inc. has developed HER2osome, as a therapeutic strategy to combat HER2-positive breast cancer through the capture of circulating HER2-positive exosomes (49).

There are some limitations to this study. First, as a large amount of exosomes were needed to stably reproduce the experimental data, HeyA8 and TYK-nu cells, which express exosomes most abundantly, were used. However, a recent genomic profiling of ovarian cancer cell lines (50) suggests that HeyA8 seems not to be typical of high-grade serous ovarian cancer (HGSC). While TYK-nu conforms more closely to HGSC, this cell line was not transplantable to mice and, therefore, we could use it in our in vivo experiments. Second, it was not conclusively demonstrated that EOC-derived exosomes contain enriched CD44. Ultracentrifugation-based exosome isolation is commonly used; however, it is not possible to completely eliminate contamination pulled down from other sources in the culture medium (51).

In conclusion, we propose a novel role for EOC-derived exosomes in targeting adjacent peritoneal mesothelial cells (Fig. 7E). Our data identified exosome-mediated transfer of CD44 as a key regulator of mesothelial cell reprogramming and the metastatic progression of ovarian cancer. In the light of several publications highlighting the importance of exosomes in cancer biology and the results described here, we hypothesize that targeting exosomes might have a therapeutic application in ovarian cancer treatment and recommend that it should be explored further for future clinical use.

No potential conflicts of interest were disclosed.

Conception and design: K. Sawada, T. Kimura

Development of methodology: K. Nakamura, K. Sawada, K. Hashimoto

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Kinose, A. Yoshimura, A. Toda, E. Nakatsuka, K.-i. Morishige

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Nakamura, E. Nakatsuka, E. Lengyel, T. Kimura

Writing, review, and/or revision of the manuscript: K. Nakamura, K. Sawada, H. Kurachi, E. Lengyel

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Nakatsuka, K.-i. Morishige, H. Kurachi

Study supervision: K. Sawada, S. Mabuchi, T. Kimura

We thank Mami Morikawa for her secretarial assistance and Ayako Okamura for her technical assistance. We also thank Gail Isenberg (University of Chicago, Chicago, IL) for critical editing of this article.

This work was supported by a Grant-in-Aid for Scientific Research from the Ministry of Education, Science, Sports and Culture of Japan (24592515, 26670725, and 26293360 to K. Sawada and 24249080 to T. Kimura). This work was also supported by Takeda Science Foundation(to K. Sawada) and Osaka Medical Research Foundationfor Intractable diseases (to K. Nakamura).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.