M-Trap: Exosome-Based Capture of Tumor Cells as a New Technology in Peritoneal Metastasis

Exosome Purification From Ovarian Cancer Patients’ Ascites

Ascites fluid from advanced stage III/IV ovarian cancer patients (n = 9) was collected in sterile conditions at the Medical Oncology Department at the University Hospital of Santiago de Compostela (Spain) under fully informed consent and ethical approval by the Galician Ethical Committee (reference: 2014/309). Ascites samples were sequentially centrifuged (300g, 10 minutes; 800g, 15 minutes; 10.000g, 30 minutes) and filtered (0.22 μm) prior to exosome purification by ultracentrifugation as described (8). Briefly, cleared ascites samples were centrifuged at 100.000g for one hour, the pellet was resuspended in phosphate-buffered solution (PBS), and centrifuged at 100.000g for one hour. Exosomes were resuspended in PBS and stored at -80°C. Reproducibility among batches of exosomes was assessed by morphology/size (Zeta sizer device; Malvern instrument Ltd. Malvern, Worcestershire, UK), protein content by nanodrop spectrophotometer (Thermo Fisher Scientific Inc., Wilmington, MA) and BCA assay (Thermo Fisher Scientific Inc., Wilmington, MA), and adhesion assays. Pretreatment of polystyrene wells with 50 μl of total cell-free ascites was performed overnight at 37ºC in humidified CO₂ conditions before the PBS wash and cell seed. Ascites fractionation and mass spectrometry are described in detail in the Supplementary Methods (available online).

Cell Lines

Human ovarian cancer SKOV3, OV90, and TOV112 cells, human endometrial cancer Ishikawa cells, and human colorectal cancer HCT116 cells were purchased from ATCC and authenticated by short tandem repeat DNA fingerprint. Cells were transfected with the luciferase reporter pEGFPLuc by Lipofectamine 2000 (Invitrogen, Carlsbad, CA) according to manufacturer conditions, selected with Geneticin (InvivoGen, San Diego, CA), and maintained in McKoy’s complete medium (10% fetal bovine serum 2mM glutamine, 100U/mL penicillin and 100 ug/mL streptomycin).

Cell Adhesion Assay

1x10⁵ human ovarian carcinoma SKOV3 cells labeled with calcein (4mM; Life Technologies, Thermo Fisher Scientific Inc., Waltham, MA) for 30 minutes were seeded onto 96 microtiter black polystyrene 96-well plates and allowed to adhere for 60 minutes. Upon supernatant removal and PBS wash, adhered cells were quantified in a luminometer (FLUOstar OPTIMA, BMG Labtech GMBH, Ortenberg, Germany). The human phospho-kinase assay is described in detail in the Supplementary Methods (available online).

M-TRAP Device

Exosomes (25 µl at 2 µg/µl) were embedded into 3D Insert scaffold (polystyrene polycaprolactone; 3D Biotech LLC, North Brunswick, NJ) and incubated overnight at 37ºC in humidified CO₂ conditions. Alternatively, purified exosomes were labeled with 1mM Vibrant DiD cell-labeling solution (Molecular Probes, Life Technologies, Thermo Fisher Scientific Inc., Waltham, MA) following manufacturer’s instructions before decoration of 3D scaffolds. The M-Trap dynamic cell adhesion assay and electron microscopy methods are described in detail in the Supplementary Methods (available online).

Ovarian Cancer Mouse Model for Peritoneal Dissemination and Surgical M-Trap Implantation

Female SCID Beige mice (Barcelona Biomedical Research Park, Barcelona, Spain) aged eight weeks were injected intraperitoneally with 100 μL PBS containing 1x10⁶ luciferase-expressing SKOV3 cells, and the pattern of peritoneal metastasis was analyzed one week later by bioluminescence in an in vivo image system Series Pre-clinical In Vivo Imaging System (Xenogen Corp., Caliper Life Sciences Inc., PerkinElmer), by peritoneal injection of 10 μL/g of mice D-luciferine substrate (P/N 122796, PerkinElmer, Waltham, MA), in the presence (n = 24) and absence (n = 34) of M-Trap devices. Additionally the pattern of metastasis was evaluated with the bare scaffold (n = 15), with Poly-Hema pretreated bare scaffold (n = 8), and with bare scaffold decorated with the adhesive tetraspanins CD9 (n = 3) and CD81 (n = 3). Alternatively, 1x10⁵ luciferase-expressing SKOV3 cells embedded in matrigel were directly injected into one of the ovaries of mice to generate an orthotopic model of peritoneal dissemination (14) and evaluate the efficacy of M-Trap in the presence (n = 5) or absence (n = 5) of the device. M-Trap devices were surgically implanted one week before SKOV3 cell injection to avoid any impact from the inflammatory reaction to surgery. Mice were anesthetized with Ketamin (Ketolar, Pfizer, New York, NY) at 75mg/kg and Metomidin (Domtor, Orion Pharma Corp., Orionintie, Finland) as analgesic at 1mg/kg. M-TRAP was implanted after laparotomy at the inner wall of the peritoneum using surgical glue (Glubran2, GEM S.R.L. ref. G-ND-2, Italy) to fix the device to the mesothelium on the side opposite the pancreas (see Figure 3B). The wound was closed (Ethicon VICRIL suture; Johnson & Johnson, New Brunswick, NJ), and mice reverted from anesthesia with Atipemazol (1mg/kg; Antisedan, Orion pharma Corp.). Survival was assessed as indicated in Figure 5A, with in vivo bioluminescence imaging and monitoring of weight of mice as follow-up upon intraperitoneal injection of 2.5x10⁶ luciferase-expressing SKOV3 cells in the presence (n = 7) or absence (n = 7) of the M-Trap device, with a third arm including M-Trap removal one month after SKOV3 cell injection (n = 4). The endpoint was defined as a drop maximum weight of 25%; mice were killed in a CO₂ chamber and by cervical dislocation. Mice were housed and maintained under specific pathogen-free conditions and used in accordance with institutional guidelines approved by the Use Committee for Animal Care.

Histology Analysis

Tissue samples were immersion fixed in 10% buffered formalin for 24 hours, dehydrated in ethanol series, embedded in paraffin, sectioned (4 μm thick), mounted on microslides, and, after drying, dewaxing, and rehydratation, tissues were stained with haematoxylin-eosin and examined using an Olympus BX51 microscope equipped with an Olympus DP70 camera (Olympus, Tokyo, Japan).

Immunoblot

Immuno blot analysis was performed using 40 μg of total protein extracted from exosomes purified from ovarian cancer patients’ ascitic fluid and incubated overnight with mouse monoclonal antibodies directed against human CD9 (1:1000; Santa Cruz Biotechnology, Santa Cruz, CA) and CD81 (1:1000; Santa Cruz Biotechnology). A secondary antimouse monoclonal antibody (1:1000; Santa Cruz Biotechnology) was used to visualize proteins using an AmershamTM ECLTM immuno blotting analysis system (GE Healthcare, Amersham Biotechnology, UK).

Statistical Analysis

All experiments were repeated at least three times, and data are represented as mean values with 95% confidence intervals. The Student’s t test was used to compare the differences between two groups, and analysis of variance was used to compare the differences among multiple groups. Kaplan-Meier survival curves were used to estimate cumulative survival probabilities using IBM SPSS statistics. Overall P value was calculated using a log-rank test. Statistical significance was defined as a P value less than .05. All statistical tests were two-sided.

Exosomes as Key Components in Ovarian Cancer Communication

We evaluated the role of ovarian cancer ascites and its exosomal component in the crosstalk between metastatic tumor cells and the environment and in the promotion of tumor cell attachment to the peritoneum. Cellular-free ascitic fluid from advanced ovarian cancer patients efficiently promoted in vitro adhesion of fluorescence-labeled ovarian cancer SKOV3 cells (6.43-fold, P = .0012, Student’s t test) (Figure 1A). High-Performance Liquid Chromatography (HPLC) fractionation of the ascitic fluid (upper panel in Figure 1B) showed adhesive components concentrated in the first fractions corresponding to 670-44kDa proteins (black bars in histogram of Figure 1B). The depletion of multivesicular bodies including exosomes from ascites by ultracentrifugation (middle panel in Figure 1B) resulted in a statistically significantly 0.21-fold reduced adhesive potential of SKOV3 cells mediated by the first HPLC fraction of ascites (P = .0035, Student’s t test) (gray bars in histogram of Figure 1B). Conversely, exosomes purified from ascites of ovarian cancer patients exhibited the capacity to promote in vitro adhesion of SKOV3 cells compared to the untreated control (4.34-fold, P < .001, Student’s t test) (Figure 1C). Purified exosomes were characterized in terms of morphological and biomarker features. Exosomes were shown to be a homogeneous population of vesicles approximately 100nm in diameter by Scanning Electron Microscope (SEM) and Transmission Electron Microscope (TEM) (Figure 1D), with the presence of CD9/CD81 exosome markers (Figure 1E). Mass spectrometry was used to further characterize the proteome of exosomes purified from ovarian cancer ascites (Supplementary Table 1, available online). One hundred eighty-three proteins were identified, and comprehensive pathway and network analysis was employed to accurately interpret the biological significance of these proteins. This analysis indicated that the principal biological role of exosomes is to mediate cell communication and interaction with the environment (Figure 1F). Phospho-kinase assay and adhesion impairment with specific inhibitors further highlighted the ERK1/2 and Akt pathways as the underlying mechanism in exosome-mediated ovarian tumor cell attachment (Supplementary Figure 1, available online). Collectively, these findings demonstrate the important role of ascites-derived exosomes in ovarian cancer cell communication with the environment. As critical intermediaries of tumor cell attachment, exosomes offer the potential to function as a specific modulator of metastatic tumor cell attachment and homing in ovarian cancer.

Design and Characterization of M-Trap Device

A novel tumor cell capture device (M-Trap) was constructed by embedding exosomes purified from ovarian cancer patient ascites onto 3D-Polystyrene/Polycaprolactone inserts. Electron microscopy showed exosomes attached to the fibers of the 3D scaffold (Figure 2A, compared with the empty scaffold without exosomes, shown in the insert). Immunogold staining with antibodies against two membrane proteins used as markers of exosomes, tetraspanins CD9 (left panel in Figure 2B) and CD81 (right panel in Figure 2B), confirmed the adherence of the exosomes to the fibers of the scaffolds.

Release of exosomes from the 3D scaffold was assessed to verify the stability of the M-Trap exosome coating. No substantial in vitro release of exosomes decorated with the fluorescent marker DiD was observed for up to five weeks (Figure 2C). Concordantly, mass spectrometry analysis with nano-HPLC (Supplementary Table 2, available online) and short-term adhesion experiments (Supplementary Figure 2A, available online) also demonstrated that there was no substantial release of exosomes from the M-Trap device and also no sustained release of protein components from the exosomes to the environment, indicative of the nonpharmacological mode of action of the device.

The efficacy of SKOV3 cell attachment to M-Trap in a dynamic in vitro environment was assessed by subjecting M-Trap devices to orbital movement designed to mimic the natural flow of peritoneal fluid within the abdominal cavity (15). In the absence of an active pharmacological chemotaxis attraction of metastatic SKOV3 cells, the adhesive capacity of the M-Trap technology behaves as the principal mode of action to capture tumor cells in this dynamic orbital assay. The M-Trap device showed a 3.4-fold enhanced SKOV3 cell attachment compared with bare scaffold (P = .001), while decoration of the bare scaffold with CD9 improved adhesion (2.5-fold, P = .003), and conversely pretreatment of M-Trap with blocking antibody against CD9 reduced its SKOV3 cell capture efficacy (0.47-fold, P = .005) (Figure 2, D-E; Supplementary Figure 2, B-G, available online). In summary, these findings characterize M-Trap as a device comprised of exosomes stably adhered to the fibers of a scaffold with substrate-bound exosomes promoting tumor cell attachment via a nonpharmacological, adhesive mode of action.

Efficacy of M-Trap Technology to Capture Metastatic Tumor Cells in an In Vivo Model of Ovarian Cancer Dissemination

An in vivo murine model of ovarian cancer peritoneal dissemination and natural metastasis was created by intraperitoneally injecting 1x10⁶ SKOV3 cells stably expressing the luciferase reporter gene (16). One week later, the pattern of natural peritoneal dissemination was evaluated by bioluminescence. Results showed the pancreas and gonadal fat pad as preferential sites of tumor cell attachment and homing (left panel in Figure 3A). Haematoxylin-eosin histological examination further confirmed the presence of tumor implants in the pancreas and gonadal fat pad (right panels in Figure 3A). Remarkably, when an M-Trap device was surgically implanted at the inner wall of the peritoneum opposite to the pancreas and the gonadal fat pad (scheme in Figure 3B), the pattern of dissemination of SKOV3 metastatic ovarian tumor cells was completely remodeled, with eradication of the regular sites of metastasis and the focalization of metastasis in a new, unique focus within the M-Trap device (left panel in Figure 3C). Confocal microscopy of the M-Trap device immediately upon removal from the peritoneum of the killed mice showed DiD-labeled SKOV3 cells captured by M-Trap, decorating the fibers of the 3D scaffold (upper right panel in Figure 3C). Histological examination of the pancreas and gonadal fat pad confirmed the absence of tumor implants at these natural sites of metastasis (middle and lower right panels in Figure 3C). Moreover, quantification of the bioluminescence signal in a broad series of control mice (0.16 mean % of signal at area of M-Trap implantation, 95% CI = 0.05 to 0.37, n = 34) and M-Trap mice (98.36 mean % of signal, 95% CI = 96.82 to 99.89, n = 24) confirmed the capacity of M-Trap to capture SKOV3 tumor cells disseminating within the peritoneal cavity and to completely remodel the pattern of metastasis in a murine model of ovarian cancer dissemination (P < .0001, Student’s Test) (Figure 3D). Similar to in vitro data, an empty scaffold without exosomes presented an intermediate pattern of metastasis with the presence of SKOV3 implants within the natural sites of metastasis (pancreas and gonadal fat pad) with an additional focus of SKOV3 cells within the empty scaffold (48.04 mean % of signal at area of implantation, 95% CI = 42.75 to 53.33, n = 15) (Figure 3, E-F). Moreover, when the adhesive capacity of the scaffold was limited by pretreatment with Poly-Hema, a reduced ability to capture metastatic SKOV3 cells was observed (11.71 mean % of signal at area of implantation, 95% CI = 3.01 to 20.40, P < .0001, one-way analysis of variance [ANOVA], n = 8) (Figure 3F). Conversely, when the adhesive capacity of the 3D scaffold was enhanced by decoration with tetraspanin CD9, an increased ability to capture metastatic SKOV3 cells was seen (79.76 mean % of signal at area of implantation, 95% CI = 61.65 to 97.87, P = .003, one-way ANOVA, n = 3) (Figure 3F). Similar results were obtained with CD81 (P = .004, n = 3) (Figure 3F). These results indicate that M-Trap tumor cell capture efficacy is a function of the adhesive capacity of the scaffold coating for the attachment and homing of metastatic cells, with the exosome-based M-Trap device showing the optimal adhesive capacity as evidence of a complete remodeling of the pattern of metastasis.

Similar results were obtained with clinically representative primary cultures isolated from ascites of ovarian cancer patients, cell lines of papillary serous and endometrioid ovarian cancer histologies (OV90 and TOV112 cell lines, respectively), and colorectal and endometrial cancer derived cell lines (n = 3) (Supplementary Figure 3, available online). These results support the broad applicability of M-Trap to a range of gynecologic and gastrointestinal cancers exhibiting peritoneal metastasis.

Efficacy of M-Trap Device in Sustained Peritoneal Dissemination and Orthotopic Models

The efficacy of M-Trap was evaluated in a clinically relevant model to assess the ability of the device to capture a sustained release of ovarian tumor cells metastasizing from primary lesions within the peritoneal cavity. To create this model, luciferase-expressing SKOV3 cells were intraperitoneally injected to generate primary lesions at the pancreas and gonadal fat pad (Figure 4A). The primary lesions were allowed to progress prior to implantation of an M-Trap device two months after tumor cell injection (Figure 4A). At four months after tumor cell injection, SKOV3 cells released into the peritoneal cavity from the primary lesions at the pancreas and gonadal fat pad were captured by M-Trap technology (Figure 4A). At death, the pattern of metastasis in this sustained release model was distributed at the two sites of primary implantation at the pancreas and gonadal fat pad, plus a unique focus of metastasis implantation within the M-Trap device (n = 5), compared with the peritoneal carcinomatosis observed in control mice in the absence of the M-Trap device (n = 5) (Figure 4A). Quantification of bioluminescence at death demonstrated a similar overall tumor burden in the control and M-Trap groups, but a complete remodeling of the pattern of metastasis (histogram in Figure 4A).

M-Trap efficacy was further demonstrated in an orthotopic model created by direct injection of SKOV3 cells into the ovary (Figure 4B). Orthotopic tumors were allowed to progress for two weeks prior to implantation of an M-Trap device decorated with fluorescent exosomes (Figure 4B). In contrast to the massive peritoneal carcinomatosis present in control mice in the absence of M-Trap (n = 5) (Figure 4B), use of M-Trap resulted in a focalized disease at the primary ovary lesion and within the M-Trap device, as a result of the capture of metastatic cells sustainably released from the ovarian tumor (n = 5) (Figure 4B).

Finally, the ability of M-Trap to capture tumor cells disseminating throughout the peritoneal cavity in sequential injections of SKOV3 cells (Supplementary Figure 4, available online) further validated the efficacy of M-Trap in another clinically relevant model of sustained ovarian metastasis. From all these results we conclude that M-Trap acts as a potent artificial premetastatic niche that preferentially captures metastatic cells disseminating in the peritoneum in competition with natural sites of homing and metastasis.

Impact of M-Trap on Survival in the Ovarian Metastasis Model

The impact of M-Trap technology on survival outcomes was assessed in the murine model of ovarian cancer peritoneal dissemination. This preclinical trial included three arms simulating the following scenarios: 1) the natural pattern of massive peritoneal dissemination in ovarian cancer (control group; n = 7); 2) the transformation of massive peritoneal dissemination into a focalized disease by the M-Trap technology (M-Trap group; n = 7); and 3) the surgical removal of M-Trap upon focalization of the metastatic disease (re-operated group; n = 4) (see scheme in Figure 5A).

Kaplan-Meyer survival curves illustrated that in absence of M-Trap the control group reached the study endpoint at four months after intraperitoneal injection of 2.5x10⁶ SKOV3 cells (mean survival = 117.5 days, 95% CI = 107.6 to 127.4) (Figure 5A) because of massive peritoneal carcinomatosis (Figure 5, B and C). The control group was designed to mimic the clinical scenario today in advanced ovarian cancer. Remarkably, focalization of the disease by the M-Trap technology resulted in a tumor mass contained within the device, thereby preventing further dissemination (Figure 5D) and resulting in a statistically significant cumulative survival (mean survival = 198.8 days, 95% CI = 170.4 to 209.2, P < .001, log-rank Test) (Figure 5A). Of note, the M-Trap group exhibited 100% survival at five months when all mice included in the control group reached the endpoint of the study (Figure 5A). Histological evaluation of the explanted M-Trap device showed an infiltrative component of fibroblasts associated with a foreign body reaction (17) (Figure 5E). Finally, surgical removal of M-Trap upon focalization of the disease resulted in the eradication of large metastasis with a residual luciferase signal at the site of M-Trap implantation (re-operated group) (Figure 5A). This scenario represents the expected clinical use of the M-Trap technology and led to an even more marked survival benefit, with three out of four mice alive one year after SKOV3 tumor cell injection (mean survival = 309.4, 95% CI = 249.4 to 369.4) (Figure 5A). These results clearly demonstrate the substantial survival benefits attributable to M-Trap. In summary, M-Trap technology is a compelling approach, disrupting the process of metastasis that can effectively transform a systemic disease into a focalized disease where proven therapeutic approaches such as surgery can extend survival.