Cancer Cell Biology
Use of a novel 3D culture model to elucidate the role of mesothelial cells, fibroblasts and extra-cellular matrices on adhesion and invasion of ovarian cancer cells to the omentum
Article first published online: 1 JUN 2007
Copyright © 2007 Wiley-Liss, Inc.
International Journal of Cancer
Volume 121, Issue 7, pages 1463–1472, 1 October 2007
How to Cite
Kenny, H. A., Krausz, T., Yamada, S. D. and Lengyel, E. (2007), Use of a novel 3D culture model to elucidate the role of mesothelial cells, fibroblasts and extra-cellular matrices on adhesion and invasion of ovarian cancer cells to the omentum. Int. J. Cancer, 121: 1463–1472. doi: 10.1002/ijc.22874
- Issue published online: 24 JUL 2007
- Article first published online: 1 JUN 2007
- Manuscript Accepted: 18 APR 2007
- Manuscript Received: 11 OCT 2006
- National Cancer Institute. Grant Number: R01 CA111882
- Gynecologic Cancer Foundation (2005-2006 GCF/Molly Cade Ovarian Cancer Research Grant), Ovarian Cancer Research Fund
- 3-dimensional culture;
- tumor-stroma interaction;
- ovarian cancer
The omentum is a major site of ovarian cancer metastasis. Our goal was to establish a three-dimensional (3D) model of the key components of the omental microenvironment (mesothelial cells, fibroblasts and extracellular matrices) to study ovarian cancer cell adhesion and invasion. The 3D model comprised of primary human fibroblasts extracted from normal human omentum, mixed with ECM and covered by a layer of primary human mesothelial cells, also from normal human omentum. After addition of ovarian cancer cells, the histological appearance of the 3D culture mimicked microscopic metastases to the omentum from patients with ovarian cancer. When ovarian cancer cells, SKOV3ip.1 and HeyA8, were applied to the 3D omental culture, 60% and 68% of all cells attached, respectively, but only 18% and 25% were able to invade. Ovarian cancer cells preferentially adhered to and invaded collagen I, followed by binding to collagen IV, fibronectin, vitronectin, laminin 10 and 1. Omental mesothelial cells significantly inhibited ovarian cancer cell adhesion and invasion, while omental fibroblasts induced adhesion and invasion. This effect is clearly mediated by direct cell–cell contact, since conditioned medium from mesothelial cells or fibroblasts has a minimal, or no, effect on ovarian cancer cell adhesion and invasion. In summary, we have established a 3D model to study the early steps of ovarian cancer metastasis to the human omentum, and found that omental mesothelial cells inhibit, while omental fibroblasts and the ECM enhance, the attachment and invasion of ovarian cancer cells. © 2007 Wiley-Liss, Inc.
Ovarian cancer is typically diagnosed at an advanced stage, because most patients develop clinical symptoms late in the course of the disease, after the tumor has metastasized.1 The most common sites of metastasis from the primary ovarian tumor are the abdominal peritoneum and the omentum. Often the pliable, semitransparent omental tissue is transformed into a dense omental “cake” that impinges on the stomach and the transverse colon, causing a bowel obstruction. Even after infragastric removal of the omentum, which is regularly performed for ovarian cancer staging or debulking at the time of surgery, recurrent tumor cells home to residual areas of the omentum at the splenic hilum, the stomach and the transverse colon causing the reappearance of significant clinical symptoms.
The majority of ovarian cancers are of epithelial origin, arising from the single layer of cells that covers the ovary, which is separated by a basement membrane from the underlying ovarian stroma.2, 3 There are features unique to ovarian cancer that might facilitate its intraperitonel (ip) spread. The ovarian epithelium can adopt a mesenchymal phenotype characterized by cells that lack tight junctions, which makes them prone to exfoliation.4 Once an ovarian epithelial cell undergoes transformation it is free to disseminate throughout the peritoneal cavity, carried by the flow of the peritoneal fluid. Subsequent implants are characterized by the adhesion, migration and invasion of the tumor cells into the omentum and peritoneum.3
The primary microenvironment for ovarian cancer cells is the mesothelial cell. These cells cover the peritoneum, the small and large bowel serosa and the omentum. They form a low-friction, nonadhesive surface and selective barrier, which regulates the transport of fluid and solutes between the circulation, the interstitial space and the body cavities.5 Binding of ovarian cancer cells to mesothelial cells are mediated, at least in part, by integrins and by CD44, a cell surface receptor for hyaluronic acid.6, 7, 8, 9 Mesothelial cells secrete soluble fibronectin, which may play a role in invasion of ovarian cancer cell lines.9, 10 At the level of the peritoneum, the mesothelial cell layer is attached to a basement membrane (BM) predominantly composed of collagen types I and IV, fibronectin and laminin.11 Primary ovarian cancer cells adhere preferentially to type I collagen, a process that can be blocked with an α2β1-integrin antibody that inhibits the interaction with collagen.12
While recent studies examine interactions between ovarian cancer cells, the ECM and mesothelial cells have been informative, they have a number of limitations. Most studies have been done with two-dimensional monolayer cultures composed only of ECM or mesothelial cells. Because it has been assumed that all mesothelial cells are the same, the cells used have either been immortalized (e.g. LP-9) or isolated from malignant ascites.7, 8, 10, 13 However, it has been determined that there are differences in the expression of mesothelial specific markers depending on the origin of the cells.14 Current studies also do not address the role of fibroblasts in the adhesion and invasion of ovarian cancer cells. In addition, most published studies have only investigated metastasis of ovarian cancer cells to the abdominal peritoneum, and inferred from these observations that the mechanisms involved in metastasis to the omentum are similar. However, ovarian cancer metastasis to the omentum is more common (80% for serous papillary ovarian cancers15) than to any other site and, as yet, we do not understand why ovarian cancer cells have a clear predilection for the omentum.
We reasoned that a three-dimensional (3D) cell culture model, which would emulate a more physiologically relevant microenvironment, would be a useful tool for understanding the adhesion and invasion of ovarian cancer cells to the omentum. We have developed a model that incorporates ECM, primary human mesothelial cells and primary human omental fibroblasts to study the role of both the stromal cells and ECM in the metastatic process. We report our findings that human mesothelial cells inhibit, while human fibroblasts enhance, the adhesion and invasion of ovarian cancer cells to the omentum. We believe that our 3D model will complement current animal and cell culture models used for the study of ovarian cancer, and has the potential to enhance our overall understanding of the very early steps of ovarian cancer metastasis.
Materials and methods
Collagen I (rat tail), matrigel, vitronectin, fibronectin and laminin 1 were purchased from BD Biosciences (Franklin Lakes, NJ). Laminin 10 and 11 was from Millipore (Billerica, MA). Vimentin, prolyl-hydroxylase and CA-125 antibodies were purchased from Dako Cytomation (Carpinteria, CA). CAM 5.2 antibody against cytokeratin 8 was from Becton Dickinson (Mountain View, CA). The fibronectin antibody was purchased from Sigma-Aldrich (St. Louis, MO), and the α6-integrin (clone NKI-GoH3) antibody was from Chemicon (Temecula, CA). The antibodies against GM130 (clone 35) and β-catenin (#14) were purchased from BD Biosciences Pharmingen (San Jose, CA). The human ovarian cancer cell lines, HeyA8 and SKOV3ip.1, were provided by Dr. Gordon B. Mills (M.D. Anderson Cancer Center, Houston, TX). These cell lines were chosen because they reflect the 2 growth patterns of ovarian cancer: growth as a few solid tumors (HeyA8) and disseminated growth with miliary disease (SKOV3ip.1). The IOSE 29, immortalized ovarian epithelial cell line was provided by Dr. Nelly Auersperg (University of British Columbia, Vancouver, Canada).
Isolation and culture of primary human mesothelial, fibroblast and ovarian cancer cells
Specimens of human omentum were acquired from patients undergoing surgery for benign conditions and human ascites were obtained from patients with newly diagnosed ovarian cancer treated by a gynecologic oncologist (DY, EL) at the University of Chicago Hospitals/Section of Gynecologic Oncology. Benign omental specimens were collected from female patients free of pathologic conditions such as cancer, inflammation or endometriosis. Informed consent was obtained before surgery and the study was approved by the IRB of the University of Chicago. After several washings with sterile phosphate-buffered saline (PBS), a 2–3 cm2 piece of omentum was minced with scissors and incubated on an orbital shaker with 10 ml of PBS and 10 ml of 0.25% trypsin/25 mM EDTA at 37°C for 30 min.16, 17 The solution containing cells in suspension was centrifuged at 1,500 rpm for 5 min, and the pellet was washed twice with 20 ml of RPMI 1640, 20% FBS, 100 U/ml penicillin and 100 μg/ml streptomycin. Mesothelial cells were plated with 20 ml of RPMI 1640, 20% FBS and 100 U/ml penicillin/100 μg/ml streptomycin. A confluent layer of mesothelial cells has a cobblestone appearance (Fig. 1h) as previously described.16 To isolate fibroblasts, the tissue was further digested on an orbital shaker with 100 U of hyaluronidase and 1,000 U of collagenase type 3 in 100 ml PBS at 37°C for 6 hr.18 The tissue was discarded and solution containing cells in suspension was centrifuged at 1,500 rpm for 5 min, and the pellet was washed twice with 20 ml of RPMI 1640, 20% FBS and 100 U/ml penicillin/100 μg/ml streptomycin. Fibroblasts were plated with RPMI 1640, 20% FBS and 100 U/ml penicillin/100 μg/ml streptomycin. Purification of primary human mesothelial cells was verified by vimentin and cytokeratin 8 positive and prolyl-hydroxylase negative immunohistochemical staining. Fibroblast purification was confirmed by prolyl-hydroxylase positive immunohistochemical staining. Mesothelial and fibroblast cells at early passages (1–3) were used for the experiments to minimize dedifferentiation and modification of the original phenotype as described.16, 17, 18 Mesothelial or fibroblast cells were incubated in serum-free media for 24 hr and conditioned media collected.
Primary ovarian cancer cells were isolated from ascites by plating equal volumes of ascites fluid and RPMI 1640, 20% FBS and 100 U/ml penicillin/100 μg/ml streptomycin. Cells at passage 10–15 were used for the experiments.19 Primary ovarian cancer cell purification was verified by positive immunohistochemical staining for CA-125.
3D omental culture gel construction
For the adhesion assays, the 3D omental culture was assembled by plating 2,000 primary fibroblast cells mixed with 0.5μg of collagen I in 100 μl of growth media (RPMI 1640, 20% FBS and 100 U/ml penicillin/100 μg/ml streptomycin) into a 96-well culture plate. The collagen solidified at 37°C for 2–4 hr. After solidification, 10,000 primary mesothelial cells were added to the culture and incubated at 37°C until a confluent layer of mesothelial cells formed (18–38 hr). For the invasion assays, a 24-well transwell plate with 8 μm pore size, was coated with collagen I mixed with 4,000 fibroblasts in 100 μl of growth media, and incubated at 37°C for 2–4 hr. Mesothelial cells (20,000) in 100 μl of growth media were added to collagen I (15 μg) alone or collagen I with fibroblasts, and incubated at 37°C for 18–38 hr. For the 3D culture that was cross-sectioned in Figure 1, 1 ml of a collagen mixture (7:1:1:1 solution of collagen I: fetal bovine serum: 0.34N sodium hydroxide: 10× RPMI growth media) was plated in the top well of a 6-well transwell plate (24 mm) with a 3 μm pore size.20 After solidification for 2 hr, 1 ml of collagen mixture and 100,000 fibroblasts were plated. After solidification and culture for 24 hr at 37°C, 1 × 106 mesothelial cells in 1 ml of growth media were plated to complete the 3D omental culture gel for 18 hr. 1.5 × 106 SKOV3ip.1 were added and cultured for 4–24 hr. The 3D omental culture gel was fixed in 10% formalin. Gels were paraffin embedded, sliced at 5 μm, and standard hematoxylin and eosin staining performed.
Adhesion assay to ECM, primary omental cells, 3D omental culture and full human omentum
IOSE, Hey A8 and SKOV3ip.1 cells were fluorescently-labeled with CMFDA (Invitrogen, Carlsbad, CA) at 37°C for 30 min in serum-free media, recovered for 30 min in serum-containing media, and washed twice with serum-free media.21 Fifty-thousand cells were plated in a 96-well plate which had been precoated with 0.5 μg of the indicated ECM, mesothelial cells (10,000, confluent layer), fibroblasts (2,000), or the 3D omental culture. After incubation at 37°C for the indicated times, cells were washed 3 times in PBS and fixed with 10% formalin. The number of adhesive cells were quantified by measuring the fluorescence intensity (Ex = 488 nm, Em = 528 nm) with a fluorescence spectrophotometer (Synergy HT) and use of a standard curve. Adhesion assays were run in triplicate. In addition, adherent cells were visualized by fluorescent microscopy at 100× magnification on a Leica Axiovert 200 microscope equipped with a Leica digital camera.
For conditioned media experiments, fluorescently-labeled SKOV3ip.1, Hey A8 or IOSE cells were pretreated with 1 ml of mesothelial cell conditioned media in a 6-well tissue culture plate prior to conducting an adhesion assay to a confluent layer of mesothelial cells on collagen I or fibroblasts embedded in collagen I. In addition, SKOV3ip.1, Hey A8 or IOSE cells were pretreated with 1 ml of fibroblast conditioned media in a 6-well plate prior to conducting an adhesion assay to a confluent layer of mesothelial cells on collagen I or fibroblasts embedded in collagen I.
For adhesion assays to omentum, a fresh piece of full human omentum was cut into 6 mm circular biopsies. The omental discs were placed in growth media or a 1:1 mixture of trypsin and PBS for enzymatic removal of mesothelial cells and incubated at 37°C for 20 min. The digested discs were washed twice in growth media. All omental discs were placed in a 96-well dish and 50,000 fluorescently-labeled SKOV3ip.1, Hey A8 or IOSE cells were plated in each well on top of the omentum. After incubation at 37°C for 4 hr, omental discs were washed 3 times in PBS, digested in 5% nonidet P40 for 30 min at 37°C and scraped with a metal spatula. All cells removed during digestion were placed in a 24-well plate and the total fluorescent intensity per well was quantified. Adhesion assays were run in quadruplicate.
Invasion assay through ECM, primary omental cells and 3D omental culture
For invasion assays through different ECMs, a 24-well trans-well plate with 8 μm pore size was coated with 15 μg of ECM. For invasion assays through different cellular components, a 24-well trans-well plate with 8 μm pore size was coated with 15 μg of collagen I or collagen I mixed with fibroblasts (4,000) in 100 μl of growth media and incubated at 37°C for 2–4 hr. Mesothelial cells (20,000) in 100 μl of growth media were added to collagen I (with or without fibroblasts) and incubated at 37°C for 18–38 hr. For migration assays, the trans-well plates were not coated. In all invasion assays, each well was gently washed 3 times with serum-free media, and 50,000 fluorescently-labeled IOSE 29, Hey A8 or SKOV3ip.1 cells in 100 μl of serum-free media were plated on top. The bottom of the transwell was filled with 400 μL of full growth media containing 20% fetal bovine serum (the chemoattractant). After incubation for the indicated times at 37°C, membranes were washed in PBS and a Q-tip was used to scrape all cells and matrix off the top of the membrane. The membranes were fixed in 10% formalin. The number of invasive or migratory cells were quantified in five 100× magnification fields and averaged for each well by fluorescent microscopy on a Leica Axiovert 100 microscope equipped with a Leica digital camera. Invasion assays were run in triplicate.
For conditioned media experiments, fluorescently-labeled SKOV3ip.1, Hey A8 or IOSE cells were pretreated with 1 ml of mesothelial cell conditioned media in a 6-well tissue culture plate prior to conducting an invasion assay through a confluent layer of mesothelial cells on collagen I or fibroblasts embedded in collagen I. In addition, SKOV3ip.1, Hey A8 or IOSE cells were pretreated with 1 ml of fibroblast conditioned media in a 6-well plate prior to conducting an invasion assay through a confluent layer of mesothelial cells on collagen I or fibroblasts embedded in collagen I.
For immunohistochemical experiments, human omentum was fixed in 10% formalin for 18 hr and paraffin embedded. The paraffin blocks were cut onto Superfrost Plus charged slides (Thermo Fisher Scientific, Waltham, MA), deparaffinized in xylene and hydrated with alcohol. The peroxidase activity was quenched with 3% H2O2/methanol blocking solution for 30 min. Slides were boiled in 0.01 M sodium citrate pH 5.0 for 20 min to retrieve antigens. Cells cultured on chamber slides were fixed in acetone:methanol (1:1). All slides were blocked in avidin and biotin blocking solutions (Vector Laboratories, Burlingame, CA). The slides were incubated with the primary antibodies against cytokeratin 8 (prediluted), vimentin (388 μg/l), prolyl-hydroxylase (1:100), fibronectin (1:50) or CA-125 (1:200 dilution) for 1 hr. After 3 washes in TBS, the slides were incubated with a rabbit-anti mouse biotinylated secondary antibody (1:200). Again the slides were washed with TBS then incubated with peroxidase-linked avidin using the Vectastain ABC kit (Vector Laboratories, Burlingame, CA) for 30 min. The slides were rinsed in TBS and stained with 3-3′-diaminobenzidine chromogen, then counterstained with hematoxylin. Appropriate negative controls for the immuonstaining were prepared by omitting the primary antibody.
For immunofluorescence analysis, the 3D omental culture was plated on a polystyrene SlideFlask (Nalge Nunc International, Roskiida, Denmark). The 3D omental culture was assembled by plating 60,000 primary fibroblast cells mixed with 15 μg of collagen I in 3 ml of growth media (RPMI 1640, 20% FBS and 100 U/ml penicillin/100 μg/ml streptomycin) into the SlideFlask. The collagen solidified at 37°C for 2–4 hr. After solidification, 300,000 primary mesothelial cells were added to the culture and incubated at 37°C until a confluent layer of mesothelial cells formed (18–38 hr). The slides were fixed for 20 min in 2% paraformaldehyde pH7.4, permeabilized in 0.5% Triton X-100 in PBS for 20 min and incubated in PBS:glycine (130 mM NaCl, 7 mM NaHPO4, 100 mM glycine) 3 times for 15 min. The slides were blocked for 2 hr in blocking buffer (130 mM NaCl, 7 mM Na2HPO4, 3.5 mM NaH2PO4, 7.7 mM NaN3, 0.1% BSA, 0.2% Triton X-100, 0.05% Tween 20 and 10% serum of host of secondary antibody), and incubated overnight at 4°C with primary antibodies against β-catenin (1:50), GM130 (1:50) or α6-integrin (1:100) diluted in blocking buffer. Slides were washed 3 times for 15 min shaking at 100 rpm with blocking buffer without serum, and incubated for 1 hr with Alexa Fluor 488-labeled secondary antibodies (1:1,000, Invitrogen) diluted in blocking buffer. After washing 3 times for 15 min shaking at 100 rpm with blocking buffer without serum a cover slip with mounting media containing DAPI (Vector Laboratories) was placed on the slide. Localization of staining was imaged by fluorescent microscopy on a Leica Axiovert 100 microscope equipped with Leica digital camera. Trichrome staining to detect collagen fibers was conducted as previously published.22
Cell viability assay
Fibroblasts were fluorescently-labeled with calcein AM fluorogenic esterase substrate (Invitrogen) at 37°C for 30 min in growth media containing 20% fetal bovine serum, and washed twice with growth media. The 3D omental culture was constructed with the fluorescently labeled fibroblasts as described previously in a 96-well plate, and 50,000 SKOV3ip.1 cells were added to the culture for 24 hr. The calcein AM staining was visualized by fluorescent microscopy at 100× magnification on a Leica Axiovert 100 microscope equipped with a Leica digital camera over the course of the experiment.
Adhesion assays were performed in triplicate and at least 3 independent experiments conducted with different ECMs, mesothelial cells and fibroblasts. The mean and standard deviations are reported. Significant changes (*p < 0.05, **p < 0.01) were determined by two-sided unpaired t-tests.
Establishment of a 3D model of the human omentum to study ovarian cancer adhesion and invasion
We reasoned that a 3D culture model could emulate the microenvironment in which early ovarian cancer metastasis occurs. Such a model would allow investigators to study and understand the role that the ECM and stromal/mesothelial cells play in the initial adhesion, migration and invasion of ovarian cancer cells. To verify the appearance of micro-metastases for comparison to our model, we reviewed the histological appearance of omental biopsies from patients with FIGO stage IIIA ovarian cancer. These patients have microscopic metastatic disease to the upper abdomen including the omentum.1 Microscopic omental metastases smaller than 0.1 mm in diameter are characterized by tumor proliferation on top of the mesothelial cell layer without invasion (Fig. 1a and left side of the dotted line in Fig. 1b). As the omental implant enlarges, the stroma exhibits reactive changes with increased numbers of fibroblasts, capillaries and a thickened basement membrane. The tumor cells invade the omental adipose tissue (Fig. 1b right side of the dotted line) forming a solid, invading tumor cell nest. In contrast, the basement membrane of normal human omentum (Figs. 1c–1g) is covered by a continuous single layer of mesothelial cells, which serves as the anchoring substratum for the mesothelium. Underneath are submesothelial fibroblasts within a dense matrix that overlies vascularized adipose tissue with fibrous septae. The mesothelial cells covering normal omentum express the epithelial cell marker, cytokeratin 8, as well as the mesenchymal marker vimentin (Figs. 1c and 1d) as previously reported.17 The spindle shaped submesothelial fibroblasts and the adipocytes express vimentin, confirming their mesenchymal origin (Fig. 1d). The basement membrane, on which the mesothelial cells rest, is composed of a dense matrix of collagen fibrils, as is documented by trichrome staining (Fig 1f), and also contains fibronectin (Fig. 1g).
To construct the 3D model (Fig. 1r), primary human mesothelial cells and fibroblasts were extracted from fresh biopsies of human omentum removed at surgery. As described before,17 the cultured mesothelial cells showed a classic cobblestone appearance and expressed cytokeratin 8 and vimentin, but not prolyl-hydroxylase (Figs. 1h–1k), and therefore closely mimicked mesothelial cells in omentum (Figs. 1c and 1d). The extracted fibroblasts maintained their spindle cell appearance in culture and were positive for prolyl-hydroxylase, a fibroblast marker.23 (Figs. 1m and 1n). Both primary human mesothelial cells and fibroblasts secreted a modest amount of collagen (blue) when cultured in a monolayer (Figs. 1l and 1o). Primary human ovarian cancer cells (Figs. 1p and 1q), which were isolated from ascites, expressed the surface cell marker CA-125.24 and were used at early passages to validate the results obtained with the 2 invasive ovarian cancer cell lines SKOV3ip.1 and HeyA8.
To recreate, as accurately as possible, the construction of the normal omentum, we reviewed H&E stains from 8 normal omental biopsies. Careful analysis revealed that the ratio of fibroblasts to mesothelial cells is 1:5 to 1:8. Therefore the 3D culture was assembled (Fig. 1r) by first mixing primary human fibroblasts with ECM and then overlying this culture with 5 times the number of mesothelial cells, yielding a uniform layer on top of the ECM/fibroblast mix. Eighteen-hours after attachment, the primary mesothelial cells had a uniform, cobblestone appearance under phase contrast microscopy (Fig. 1s), and intercellular junctions were present between the mesothelial cells as shown by immunostaining for β-catenin (Fig. 1t). In many epithelia, including breast cancer, epithelial cells are polarized as determined by apical expression of the cis-golgi protein GM130 and basal expression of α6-integrin.25 Interestingly, the confluent layer of primary human mesothelial cells lacked apicobasal polarization in the 3D omental culture. GM130 was localized throughout the cells and α6-integrin was located in the cytosol as well as basal surface of the mesothelial cell layer (Fig. 1u and 1v). In polarized epithelia, α6-integrin is located on the basal surface, whereas GM130 assumes an apical location.25 At this point, ovarian cancer cells were added to study adhesion and invasion. After 4 hr, SKOV3ip.1 cells adhered to the 3D culture and retraction of mesothelial cells was evident under phase contrast microscopy (Fig. 1w). We confirmed the viability of fibroblasts within the collagen gel throughout the 3D omental culture before (data not shown) and after addition of SKOV3ip.1 cells with calcein AM staining (Fig. 1x). The 3D culture was fixed and stained by H&E. After 4 hr, a dense and adherent layer of ovarian cancer cells coated the mesothelial cells, while after 24 hr the first ovarian cancer cells had invaded (Fig. 1y and 1z).
Ovarian cancer cells adhere and invade the 3D omental culture more efficiently than normal ovarian surface epithelial cells
To understand the time frame of ovarian cancer cell adhesion to the omentum, 50,000 immortalized ovarian surface epithelial (IOSE) or ovarian cancer cells were fluorescently labeled and adhesion to the 3D model investigated at the indicated time points. After 30 min, only 742 ± 68 (2%) of the immortalized IOSE cells were attached enough to withstand the 3 PBS washing steps in the adhesion assay, while 14,854 ± 533 (30%) SKOV3ip.1 cells and 15,623 ± 459 (31%) HeyA8 cells were sufficiently attached (Fig. 2a). Over the next 8 hr, the number of attached SKOV3ip.1 (21,346 ± 996) and Hey A8 (20,529 ± 1,002) ovarian cancer cells increased only marginally. The number of ovarian cancer cells that attached over the first 8 hr was significantly greater than the number of IOSE cells that attached to the 3D omental culture (p < 0.01). After 12 hr the number of IOSE cells adhering was comparable to the number of cancer cells that attached, but attachment of the benign cells was significantly slower, with most of the benign cells adhering between 8 and 12 hr after the cells were added. A significant and reproducible increase in attachment of an additional 14,954 ± 1,204 (30%, p < 0.01) SKOV3ip.1 and 18,399 ± 1,026 (37%, p < 0.01) Hey A8 cells were observed between 8 and 12 hr after the cells were added. In total, 60% SKOV3ip.1 and 68% Hey A8 cells attached to the 3D omental culture.
In stark contrast to the adhesion assay, the results from the invasion assay showed that only 2,636 ± 137 (6%) of 50,000 IOSE cells were able to invade the 3D culture of primary mesothelial cells, fibroblasts and collagen after a 72 hr incubation period compared to 9,064 ± 431 (18%) SKOV3ip.1 and 12,428 ± 509 (25%) Hey A8 (p < 0.01). The results did not change with a longer incubation time of 5 days (Fig. 2b). All experiments were performed with IOSE, SKOV3ip.1 and HeyA8 cells (as shown in Fig. 2), but given space constraints only the results for IOSE and SKOV3ip.1 are shown in Figures 3–5, since the 2 invasive ovarian cancer cell lines behaved very similarly in both the adhesion and invasion assays.
Ovarian cancer cells preferentially adhere to and invade through collagen
Since the ECM is a physical anchor point and major barrier to cancer cells, attachment and invasion into various ECM matrices were investigated. The benign IOSE cells that successfully attached, bound to each matrix equally (Fig. 3a). In contrast, there was a marked difference in the attachment of the ovarian cancer cells to various ECMs. SKOV3ip.1 cells (Fig. 3a) showed maximum attachment to collagen I and IV: 66% of all SKOV3ip.1 cells (33,182 ± 1,556) attached to collagen I and 65% (33,182 ± 1,556) attached to collagen IV. In comparison, 35% of these cells attached to vitronectin (21,339 ± 2,068) and 43% to fibronectin (17,668 ± 2,432). Even fewer cells attached to laminin-1 (4%, 3,076 ± 971) and laminin-10 (17%, 8,535 ± 764), which is an efficient ECM for the attachment of colon cancer cells.26 A combination of matrix proteins present in matrigel supported adhesion almost as efficiently as collagen. However, it is evident that some of the matrigel effect was produced by growth factors, since significantly fewer SKOV3ip.1 cells (37% versus 53% of the 50,000 cells plated) attached to growth factor reduced matrigel. Consistent results were obtained with HeyA8 cells (data not shown).
Only a fraction of SKOV3ip.1 cells that adhere to the various ECM's are able to invade that particular ECM, suggesting that invasion is a relatively inefficient process. Collagen I was associated with the highest SKOV3ip.1 cell invasion (6,696 ± 505, Fig. 3b), but this number only represents 20% of all cells that had attached to this substratum. A very small number of IOSE cells invaded through the various ECM's (Fig. 3b), and only collagen I was associated with a slight increase in invasion (244 ± 59, 1%). However, while IOSE cells are not invasive, they are mobile, since 10 % (3,086 ± 933 cells) of the IOSE cells could efficiently migrate across the 8 μ m porous filter towards the serum-containing media. Only 3% (916 ± 454) of all SKOV3ip.1 cells that bound to collagen IV invaded, implying that invasion is an even less efficient process when the ECM is rich in collagen IV. Invasion through vitronectin and fibronectin followed a similar trend with only 8% (2,304 ± 153) and 4% (1,276 ± 224) of SKOV3ip.1 cells invading, which represents 11 and 7% of the cells that had attached to these matrices. While significantly fewer cells adhered to laminin-1 and -10 (3,076 ± 971 and 8,456 ± 1,429, Fig. 3a), 94% and 51% of these cells were able to invade through the matrix to the underside of the porous filter. Invasion through growth factor reduced matrigel was intermediate (1,972 ± 464, 7%) between collagen I and fibronectin, probably reflecting the ECM composition of matrigel that contains these two matrices.27 Similar results for the invasion through several ECM were obtained with HeyA8 cells (data not shown).
Both mesothelial cells and fibroblasts play key roles in ovarian cancer cell adhesion and invasion to the omentum
Since adhesion and invasion of cancer cells can be influenced by stromal host cells,23 we considered the possibility that fibroblasts and/or mesothelial cells affect the early steps of ovarian cancer metastasis to the omentum. To answer this question, a 3D culture that mimics human omentum was assembled by culturing primary human fibroblasts in a collagen matrix and overlying this culture with primary human mesothelial cells until they reached a confluent layer (Fig. 1s). For the subsequent experiments, collagen I was used as the principal ECM in light of the dense collagen matrix present in the omental basement membrane (Fig. 1f). In addition, collagen is secreted by mesothelial cells and fibroblasts (Figs. 1l and 1o), and ovarian cancer cells significantly adhere to and invade collagen I (Fig. 3a).12
When an adhesion assay was performed with 50,000 SKOV3ip.1 cells, a comparable number of cells adhered to the 3D culture (23,527 ± 997, 47%) and a coculture of mesothelial cells and fibroblasts without collagen I (22,198±1,898, 44%), indicating that the stromal cells may have a stronger influence on adhesion than the ECM (Fig. 4a). When an adhesion assay was conducted with SKOV3ip.1 cells on collagen that was covered by mesothelial cells or on mesothelial cells alone, adhesion was inhibited by 59% (5,733 ± 357 cells adhered versus 23,527 ± 997 to the full 3D culture, p < 0.01). In contrast, fibroblasts plated alone or in collagen I without mesothelial cells correlated with increased adhesion of SKOV3ip.1 cells (34,830 ± 1,780 cells adhered versus 5,733 ± 357 to mesothelial cells alone, p < 0.01). Consistent results were obtained with HeyA8 cells (data not shown). While significantly fewer IOSE cells attached to the 3D model, when compared to the SKOV3ip.1 cells (6,982 ± 1,127 versus 23,527 ± 997, p < 0.01) the influence of mesothelial cells and fibroblasts was similar, with mesothelial cells inhibiting adhesion and fibroblasts inducing it (Fig. 4a). To validate the results obtained with the 2 cultured ovarian cancer cell lines, 3 primary ovarian cancer cell cultures were established and added to the 3D culture at an early passage. Again, fewer cells adhered to mesothelial cells alone (11,850 ± 1,150) than to any other culture containing fibroblasts (fibroblasts alone 19,840 ± 550; fibroblasts and mesothelial cells 15,121 ± 1,533; fibroblasts, mesothelial cells and collagen 20,388 ± 1,902, p < 0.05, Fig. 4b). In addition, a significantly greater number of primary ovarian cancer cells bound to fibroblasts alone when directly compared to the number of cells that bound to the coculture of fibroblasts and mesothelial cells (p < 0.05).
Invasion followed the same pattern seen with adhesion, but significantly fewer cells invaded the culture than adhered to it. When fibroblasts are present significantly more ovarian cancer cells invade than when mesothelial cells are alone. 2,444 ± 262 (5%) SKOV3ip.1 cells invaded through the full 3D culture, while in the absence of fibroblasts, SKOV3ip.1 cells showed reduced invasion through mesothelial cells plated on collagen (1,340 ± 157, 3%, p < 0.01). In contrast, a significantly greater number of SKOV3ip.1 cells invaded fibroblasts in collagen alone (2,952 ± 136, 6%, Fig. 4c) or fibroblasts in collagen and cocultured with mesothelial cells (5%) than the number of ovarian cancer cells that invaded mesothelial cells on collagen alone (3%, p < 0.01). The same pattern was also seen with a primary ovarian cancer cell line (p < 0.01, Fig. 4d) and the HeyA8 ovarian cancer cell line (not shown) indicating that the response of ovarian surface derived epithelial cells to mesothelial cells is very similar and independent of passage number or culture conditions of the ovarian cancer cell. Invasion of IOSE cells through the 3D culture was minimal (Fig. 4c), which was very similar to what was observed with different ECMs (Fig. 3b).
We then directly compared and investigated the effect of primary mesothelial cells on ovarian cancer cell attachment to, and invasion through, the collagen I ECM. An adhesion and invasion assay was performed with fluorescently-labeled SKOV3ip.1 cells plated on collagen I alone or on a collagen matrix covered by mesothelial cells (Figs. 5a and 5b). The mesothelial cells markedly inhibited attachment (5,732 ± 357 cells versus 35,166 ± 1,298 cells, p < 0.05) and invasion (1,340 ± 157 cells versus 4,588 ± 247, p < 0.01) to collagen I independent of fibroblasts. Next, we explored whether mesothelial cells inhibit ovarian cancer cell attachment in the context of their intact microenvironment. Full human omentum was digested with trypsin to remove the mesothelial cells, and an adhesion assay was performed with fluorescently-labeled SKOV3ip.1 cells (Fig. 5c). Significantly more cells attached to the omentum when the mesothelial cells were removed, while intact omentum with at least a partial layer of mesothelial cells (some float off during the transport to the laboratory) inhibited attachment of cancer cells (p < 0.01, Fig. 5c).
We then studied whether direct cell–cell contact between the stromal cells and the cancer cells is necessary for the effects observed, or if fibroblasts or mesothelial cells secrete a soluble factor(s) that affects adhesion or invasion (Figs. 5d and 5e). Adhesion to a culture of collagen and fibroblasts was inhibited by 25% when SKOV3ip.1 cells were pretreated with mesothelial cell conditioned media (compared to serum-free media) (p < 0.05, Fig. 5d). However, adhesion of untreated SKOV3ip.1 cells to mesothelial cells alone is 79% lower than adhesion to fibroblasts alone (p < 0.01, Fig. 5d). The same pattern was observed with respect to invasion: the presence of mesothelial cells inhibited invasion by 73% (p < 0.01), while conditioned medium inhibited invasion by 29% (p < 0.05, Fig. 5e). In contrast, both fibroblast conditioned media and direct contact increased the adhesion of SKOV3ip.1 cells (p < 0.01, Fig. 5d), while only direct contact with omental fibroblasts increased invasion when compared to mesothelial cells alone (p < 0.01, Fig. 5e). Consistent results were obtained with HeyA8 cells (data not shown). Pretreatment of SKOV3ip.1 cells with mesothelial cell conditioned media had no significant effect on adhesion (Fig. 5d) to mesothelial cells or invasion (Fig. 5e) through mesothelial cells and collagen. In addition, pretreatment of SKOV3ip.1 cells with fibroblast conditioned media had no significant effect on adhesion to fibroblasts or invasion through fibroblasts and collagen. These data suggest that direct cell–cell contact with fibroblasts and mesothelial cells has a greater effect on ovarian cancer cell adhesion and invasion, than pretreatment with mesothelial cell or fibroblast conditioned media.
Little is known about the role that the microenvironment plays in metastasis of ovarian cancer cells to the omentum. One reason for this paucity of data is the lack of an adequate model system with which to study interactions of the omentum with ovarian cancer cells. Due to species specific differences that cannot be eliminated in an animal model, one cannot always extrapolate findings from studies using an ovarian cancer mouse xenograft model to human tumor biology. Human cancer cells are injected into an immunocompromised mouse host, which lacks factors that are present in humans. In addition, the mouse omentum is anatomically adjacent to the pancreas and has a different histologic appearance than human omentum.28 Coculture models also have limitations, since they involve the interaction of ovarian cancer cells with only 1 ECM or 1 stromal cell type, which is often taken from a human site other than the omentum (e.g. foreskin fibroblasts) or from a different species.29, 30, 31 Furthermore, ovarian cancer cells grown in two-dimensional (2D) monolayer cultures are morphologically diverse and express dissimilar markers than cancer cells grown in 3D. Ovarian cancer cells grown in 2D also lack the microenvironmental context to reproduce the earliest stages of metastasis as they occur in vivo.25, 32
After careful study of the histological appearance of microscopic ovarian cancer metastasis in patients with FIGO IIIA ovarian cancer, we developed a 3D omental culture in order to examine the early steps of ovarian cancer metastasis. The histology of the superficial layer of normal omentum includes mesothelial cells on top of a matrix containing collagen fibers and primary fibroblasts. To mimic the in vivo situation we extracted primary human fibroblasts and mesothelial cells from nondiseased human omentum. We embedded early passage fibroblasts in various ECMs and then plated a confluent layer of primary human mesothelial cells on the surface. Certain mesothelial cells are polarized in vivo. The confluent primary human omental mesothelial cell layer in the 3D omental culture lacks apicobasal polarization according to α6-intergrin and GM130 localization. The cells may not have established polarity in the time the experiments were conducted or the cells may be polarized and additional markers could demonstrate this polarity. By adding ovarian cancer cells to the 3D omental culture and analyzing their behavior at different time points, we were able to study the contribution of various cell types and ECMs to the adhesion and invasion of ovarian cancer cells. Through its modular concept, the presented 3D culture (Fig. 1r) allows for the manipulation of individual culture conditions and may, in the future, provide the foundation for the inclusion of other cell types (endothelial cells, adipocytes, inflammatory cells) that are present in human omentum and that in other cancer types have been shown to contribute to the establishment of metastasis. A recent study has analyzed the role of adipocytes in the progression of breast cancer. The authors found that adipocytes secrete soluble factors that induce cancer cell proliferation, invasion and inhibit apoptosis.33 In view of the fact that the omentum, like the breast, consists of significant adipose tissue, we speculate that adipocyte will also turn out to play an important role in ovarian cancer metastasis. Ongoing studies investigate this possibility. While our 3D-model attempts to mimic the omental surface, it is still not comprehensive, and is only an approximation of the in vivo situation. We did not find apicobasal polarization of mesothelial cells with molecular markers (α6-intergrin and GM130) previously used to detect polarization in epithelial cells.25 However, we can not exclude that in the 3D culture mesothelial cells become polarized, because electron microscopy showed that the luminal side of abdominal mesothelial cells have microvilli.34
Recently, another elegant method of studying ovarian cancer invasion has been described.35 A monolayer of the LP-9 immortalized mesothelial cell line13 is permeabilized and fixed on to plastic with DMSO, and then stained with trypan blue. OVCAR-5 cells, which are viable and therefore do not take up the blue stain, are added on top of the mesothelial cells and then tracked as they invade through the mesothelial cell carpet. Since unstained areas indicate the displacement of mesothelial cells by ovarian cancer cells, one can study the factors involved in attachment and invasion. Our model expands on this dye based method and may mimic the architecture of the omental basement membrane more accurately, since it incorporates viable primary stromal lines and includes various ECMs. Staining of our 3D culture with a polyanionic fluorescein derivative (calcein), which identifies viable cells, showed that both fibroblasts and mesothelial cells were viable before we added the ovarian cancer cells.
Our studies indicate that primary human mesothelial cells inhibit, at the very least, the initial adhesion and invasion of 2 ovarian cancer cell lines and 3 different early passage human ovarian cancer cell cultures. During the course of this study we extracted mesothelial cells from at least 60 patients, and found that, when tested, almost all mesothelial cell preparations behaved similarly. The effect of mesothelial cells on ovarian cancer cells is reproducible and almost independent of demographic factors or preparations. We conclude that the principal inhibitory effect of mesothelial cells on ovarian cancer cell adhesion and invasion is mediated by direct cancer cell—mesothelial cell contact because pretreatment of cancer cells with mesothelial cell conditioned media minimally inhibited adhesion and invasion. Our results are consistent with findings that indicate that the role of other epithelial cell layers in the human body is to function as a protective barrier (e.g. cervical epithelium). In the peritoneum, mesothelial cells act to limit the adhesion of cancer cells in order to protect the underlying tissue and limit access to the retroperitoneum. Indeed, ovarian cancer tumors are rarely seen retroperitoneally unless they originated as lymph node metastases. We consistently observed that, while the 3D culture was covered by a confluent layer of mesothelial cells, a few hours after the attachment of ovarian cancer cells, mesothelial cells started to focally detach and partially retract in a manner very similar to that reported by Niedbala et al. He observed that ovarian tumor cells exhibit strong attachment to the ECM once mesothelial cells are removed.36 We investigated the possibility that mesothelial cells inhibit adhesion/invasion through TGF- β, which the mesothelial cells produce in abundance, but we did not find any effect of TGF- β on early adhesion/invasion (data not shown). The detachment of mesothelial cells might be explained by a recent report using a coculture of colon cancer and primary mesothelial cells. When the peritoneal mesothelial cells were cultured with the SW 480 colon cancer cells, the mesothelial cell retracted and underwent apoptosis. This was induced by Fas ligand secreted by the cancer cells, which bind to the Fas receptor on mesothelial cells. It is possible that FasL/Fas receptor interactions mediate detachment of omental mesothelial cells in ovarian cancer metastasis to the omentum.
To further understand the role of omental fibroblasts in omental metastasis, we studied the adhesion and invasion of cancer cells in both the presence and absence of fibroblasts. Our results support emerging data showing that stromal fibroblasts play an important role in cancer cell progression (reviewed in Refs.37 and38). Ovarian cancer cell lines, SKOV-3ip1 and HeyA8, as well as the primary cell cultures, were significantly more adhesive and invasive when cultured on fibroblasts embedded in ECM. This increase involved both soluble mediators and direct interaction(s) between tumor cells and adjacent fibroblasts. Activated fibroblasts often secrete growth factors, including HGF/SF, CXCL12 and FGF, which can induce invasion of adjacent epithelial cells by activation of proteases.39, 40, 41, 42 Analysis of MMP-9 in an elegant organotypic skin cancer model revealed that fibroblasts induce the expression of MMP-9 protein and mRNA at the tumor–stroma interface43 and we have shown that coculture of fibroblasts with SCC cells induces MMP-9 transcription.44 In the present study, it is conceivable that fibroblast mediated stimulation of proteases were responsible for the increased adhesion and invasion of the ovarian cancer cells.
Adhesion of cancer cells to laminin 1 and 10 was found to be very low, despite the fact that both the SKOV3ip.1 and the HeyA8 cells express the laminin receptor α6β1-integrin in abundance (data not shown). Yet, we were surprised to see that while only 17% of cells adhered to laminin-1 and -10, the majority (82%) of these cells were able to invade through the matrix. This might be explained by a recent finding that the binding of the Hey ovarian cancer cell line to laminin, but not to collagen and vitronectin, induces LPA production and activates PI3-kinase signaling that subsequently enhances migration45 and proliferation.26 The SKOV3ip.1 and HeyA8 cells showed the strongest adhesion and invasion into collagen I, which is consistent with a strong expression of α3β1-integrin on the cell surfaces of both cell lines (data not shown) and, as we show, the abundant expression of collagen in the submesothelial omental basement membrane. It is also consistent with reports that many ovarian cancer cell lines have a predisposition to adhere to collagen.12, 46, 47 However, we cannot exclude the possibility that this effect is, at least in part, mediated by other ECM proteins that are laid down by the mesothelial cells or the fibroblasts during the time the 3D culture is assembled. Certainly, mesothelial cells secrete fibronectin,47 which as we show, is also part of the omental basement membrane underlying the mesothelial cells. Still, our findings imply that the stromal cells are more important than the ECM in determining if ovarian cancer cells attach and invade. The stimulation of invasion by collagen I can be abrogated by mesothelial cells or further induced by fibroblasts, suggesting that, at least in the early phases of omental metastases, the stromal cells can modulate the proinvasive signals of the ECM.
In summary, we have established a 3D model to study the early steps of ovarian cancer metastasis to the human omentum, and found that omental mesothelial cells inhibit, while omental fibroblasts and the ECM enhance, the attachment and invasion of ovarian cancer cells. This data, together with that from other studies,29, 36, 48 might help to explain why intraperitoneal tumor dissemination is higher after an open laparotomy, during which the surgeon's hand often wipes off the mesothelial cell layer, than after a laparoscopy, which leaves most of the peritoneal surface untouched.49 The mechanical removal of the protective mesothelial cell layer probably promotes the attachment of cancer cells by exposing the omental basement membrane and fibroblasts. Our future studies are aimed at elucidating what factors are responsible for the inhibition of ovarian cancer cell adhesion and invasion by mesothelial cells.
Abbreviations: 3D, three-dimensional; CI, collagen I; ECM, extracellular matrix; F, fibroblasts; HPF, primary human fibroblasts; MC, mesothelial cells; HPMC, primary human mesothelial cells.
We thank Prof. N. Topley, Cardiff University, UK for providing us with the protocol to establish mesothelial cells from omentum. We want also to thank Dr. Nelly Auersperg (University of British Columbia, Vancouver, Canada) for providing us with several IOSE cell lines and Dr. Gordon B. Mills (MD Anderson Cancer Center, Houston, TX) for the SKOV3ip.1 and HEYA8 ovarian cancer cell lines. This work was supported by a Penny Severns Breast, Cervical, and Ovarian Cancer Research postdoctoral fellowship from the Illinois Department of Public Health (to H.K.) and by grants from the Gynecologic Cancer Foundation (2005-2006 GCF/Molly Cade Ovarian Cancer Research Grant), the Ovarian Cancer Research Fund (Liz Tilberis Scholars Program) and the National Cancer Institute to E.L. We are very grateful to Mrs. Gail Isenberg for critical review of the manuscript.
- 3Tumors of the ovary and peritoneum. In: KleihuesP,SobinL, eds. Tumors of the breast and genital organs. Lyon, France: IARC Press, 2003. 113–203., .
- 22Connective tissues and stains. In: BancroftJD,GambleM, eds. Theory and practice of histological techniques. New York: Churchill Livingstone,2002;218–242..