Elevation of seprase expression and promotion of an invasive phenotype by collagenous matrices in ovarian tumor cells

Abstract

Tumor cells do not constitutively exhibit invasive activity, but rather, can be transiently induced to adhere and form lesions. We report here that the expression of seprase, a dominant EDTA-resistant gelatinase in malignant tumors, is dependent on tumor cell exposure to type I collagen gel (TICg). The induced seprase expression of ovarian tumor cells influences their collagen contraction and invasion capability. Importantly, tumor cells with reduced seprase expression, due to manipulation by RNA interference, showed a reduction of TICg contraction in the gel contractility assay, inhibition of tumor cell invasion through TICg as shown by a transwell migration assay, and inhibition of peritoneal membrane tumor lesion in a mouse model. In addition, mAb C27, an antibody against β1 integrin, which blocks cellular avidity to TICg, can induce seprase RNA expression and promote the invasive phenotype and metastatic potential of ovarian tumor cells. Thus, collagenous matrices in the tumor cell niche induce the expression of seprase and initiate tumor invasion and metastatic cascades.

Introduction

Epithelial ovarian cancer progresses by increased tumor cell proliferation and proteolytic activity that degrades the extracellular matrix (ECM) of the peritoneum.^1,2 Although blocking tumor progression by inhibiting protease activity is well-known, it is also possible that tumor progression could be suppressed by hindering the initial induction of these proteases.

Type I collagen (TIC) is comprised of three strands of gelatin each of which contain adjoining glycine and proline residues.³ It is this bond that is the known substrate for seprase.⁴ Seprase was first described in the LOX human malignant melanoma cell line and is a type II transmembrane, 760aa glycoprotein whose 97kDa monomer homodimerizes to form a catalytically active complex of 170kDa.^5-7 Molecular cloning of seprase revealed it to be identical to that of fibroblast activation protein α (FAP-α) in gene sequence, protein form, as well as dipeptidyl peptidase and gelatinase activities, even though FAP-α was independently identified in reactive stromal fibroblasts but not in tumor or endothelial cells.^6-11 In fact, seprase was found to be present on tumor cells of gastric, colon, melanoma, ovarian and breast cancers.^5-7,12-15 It is likely that activation, which involves the seprase dimer being shorn from the cell membrane to allow greater accessibility of its TICg substrate to the catalytic pore, is required prior to gelatinase activity.^16,17 Seprase is not only present on activated fibroblast during wound healing and active endothelial cells during angiogenesis, but has been found to be constitutively expressed at high levels in some malignant melanoma, glioma and carcinoma cells; however, most epithelial tumor cell lines express little or no seprase (see gene expression profiles of NCI60 tumor cell lines http://cgap.nci.nih.gov/Microarray/FNResults?ORG=Hs&ACCESSION=W45237_0&SRC=NCI60_STANFORD&SHOW=1). Seprase appears to be inducible in epithelial tumor cells and therefore has great biological relevance in the prevention of epithelial cancer progression.

Free collagen and gelatin in the peritoneal fluid are produced by mesothelial cells that line the peritoneal cavity.^18,19 A collagen form has even been used as a prognostic biomarker in ascitic fluid for progression-free and overall survival in women with ovarian cancer.²⁰ In this study we aimed to determine if TICg, the known substrate of seprase, triggers seprase induction in the SB247 monoclonal cell line. In addition, we tested if the invasive propensity of the SB247 cells is triggered by TICg, but not other ECM molecules found in Matrigel, induced seprase expression and β1 integrin clustering. A reduction in seprase expression via RNAi has revealed a corresponding reduction in TICg contraction capability and a reduction in tumor progression on the peritoneal membrane of a late stage ovarian cancer mouse model. As the perturbation of β1 integrin function in cell adhesion ^21,22,23 using the mAb C27 against β1 integrin could affect cellular invasion into TICg, we tested if the induced seprase expression, gel contractility, peritoneal invasion and metastasis are altered by treatment of tumor cells with mAb C27. Overall, we show that the SB247 ovarian carcinoma cell line interaction with TIC induces seprase expression, which facilitates tumor cell invasion to the mesothelial layer lining the peritoneal membranes and leads to tumor progression. This investigation supports the notion that seprase plays an important role in tumor progression on the peritoneal membrane and that prevention of β1 integrin interaction with TIC on actively invading tumor cells may inhibit tumor progression. Therefore, targeting TIC induction of seprase expression and β1 integrin interaction may be a useful tool in the treatment of late stage ovarian cancer.

Materials and Methods

Cell lines

The human SB247 cell line was derived from an ascitic fluid sample obtained from an ovarian cancer patient with stage IIIB serous adenocarcinoma. Mixed cell populations in the ascitic fluid of a 74-year-old patient were cultured and cloned by serial dilution over the course of two months. In immunocytochemistry experiments, this cell line was positive for common epithelial antigens, including pan-cytokeratins, epithelial surface antigen (ESA), and epithelial cell adhesion molecule (EpCAM). Cells were cultured in a 1:1 mixture of Dulbecco’s modified Eagle’s medium (DMEM, Gibco, Grand Island, New York) and RPMI 1640 (Gibco) supplemented with 10% calf serum (Gibco), 5% Nu-serum (Collaborative Research, Inc., Bedford, MA), 2 mM L-glutamine (Gibco), 1 unit/ml penicillin and 10 μg/ml streptomycin (Gibco) together called Cancer Cell Culture (CCC) media at 37°C under 5% CO₂ in a humidified incubator.

SB247 cells were used to generate several monoclonal subpopulations by transfection via Lipofectamine™ 2000 (Invitrogen, Carlsbad, CA) using the pGUS vector with different RNAi oligonucleotides. The oligonucelotide hairpins used were 1384 (forward: accGAAAGGTGCCAATATTACACaagcttgGTGTAATATTGGCACCTTTCtttttt-3′; reverse: 5′-catgaaaaaaGAAAGGTGCCAATATTACACcaagcttGTGTAATATTGGCACCTTT-3′, or a no target control hairpin that targets no sequence in the human genome (forward 5′-accGATTCGCGGGACCGATCGTGaagcttgCACGATCGGTCCCGCGAATCttttttt-3′; reverse: 5′-catgaaaaaaGATTCGCGGGACCGATCGTGcaagcttCACGATCGGTCCCGCGAAT-3′. SB247 cells are resistant to multiple antibiotics, which made selection for stable transfectants difficult. The antibiotic resistance is possibly due to that antibiotics were used in cytoreduction therapy of the patient that further facilitated antibiotic resistance of the cell line derived from the patient’s ascitic fluid. A subpopulation of stably transfected GFP-expressing cells then underwent a limited dilution to isolate monoclonal populations.

RT-PCR analysis

Seprase expression in cells was measured with RT-PCR. In all cases, RNA extraction was conducted using the RNeasy Mini Kit (Qiagen, Valencia, CA) according to manufacturer’s instruction. Immediately after RNA extraction, samples were converted to cDNA using 1^st Strand cDNA synthesis for RT-PCR (AMV, Roche Applied Science, Indianapolis, IN) according to manufacturer’s protocol and then stored at -20 °C for later RT-PCR application. Briefly, the cDNA of each sample was amplified using each pair of primers for seprase (forward: 5′-TACCCAAAGGCTCCAGCTAA-3′ and reverse: 5′-ACAGGACCGAAACATTCTGG-3′) and the endogenous control β-actin (forward: 5′-AGATGACCCAGATCATGTTTGA-3′ and reverse: 5′-GCACAGCTTCTCCTTAATGTCA-3′) under the following amplification scheme: 10 min at 95°C; 40 cycles of 30 sec at 94°C, 30 sec at 60°C and 1 min at 72°C; 10 min at 72 °C. Quantitech™ SYBR Green PCR Kit (Qiagen) was used to prepare PCR reactions according to manufacturer’s specifications. Real-time RT-PCR was conducted on the Opticon II (MJ Research, Watertown, MA) and analysis was facilitated by OpticonMONITOR™ Analysis Software Version 2.02 (MJ Research). All samples were normalized to β-actin and fold-change was calculated over the base line set to 1.

Transwell Migration Assay

Millipore Multiscreen MIC 96-well plate (Millipore, Billerica, MA) 8-μm membranes were coated with TIC (20 μg/ml); control wells were uncoated. Cell lines were suspended in DMEM and added to the upper chamber at 5 × 10³ cells/well. CCC media containing 10% serum was added to the lower chamber. After 48 hrs, a cotton-tipped swab removed all non-migratory cells and only cells that had migrated through the membrane remained. The membrane was washed 2X with PBS, fixed and stained with Hema 3 Stat Pack cell stain according to manufacturer instruction (Fisher Scientific, Kalamazoo, MI). Membranes were then detached, air dried on a glass slide, and mounted with Permount (Sigma-Aldrich, St. Louis, MO). Cells were counted at 400X magnification. Background values determined by counting cells that migrated in wells containing no TIC were subtracted from those values that reflected migration through TIC.

TICg contractility assay

Gel contractility assay (GCA) was performed to examine the ability of cells to reorganize and contract TICg. TIC diluted in DMEM to 220 μg/ml, 360 μg/ml, or 570 μg/ml was added to the three SB274 subclones to a final cell concentration of 1 × 10⁵ cells/well. Cells and TIC dilutions were mixed and plated into 12-well tissue culture plates. After 1 hr at 37°C under 5% CO₂ to insure gel formation, the perimeter of each disk was freed via pipette tip and CCC was added. An additional GCA examined the influence of mAb C27, the mAb against β1 integrin, and the control IgG-matched mAb C37 against glycoprotein gp90.^21,22,23 For each of the three SB274 subclones, 360 μg/ml of TIC diluted in DMEM was added to 200 μg/ml mAb C27 or control mAb C37 to a final cell concentration of 1 × 10⁵ cells/well. In both experiments the diameters of the major and minor axes were taken daily and averaged. Each condition was evaluated in triplicate for three experiments.

An ovarian cancer peritoneal metastasis model

The experimental protocol for this study was approved by the Institutional Animal Care and Use Committee of Stony Brook University. Each monoclonal cell line derived from SB247 cells were stably transfected with the pGUS plasmid, which contains a seprase RNAi target (1384 #1, 1384 #4) or the plasmid containing a hairpin with no target (NT). For seprase RNAi experiments, 1 × 10⁶ cells in a 0.2 ml volume of PBS, were i.p. injected into 6 female, 4-6 weeks old, Balb/c ^nu/nu mice (Charles River Laboratories, Wilmington, MA). Every four days from Day 4 through Day 36 isofluorane sedated mice were imaged using the multispectral Maestro™ in vivo imaging system (CRI, Woburn, MA). Images were captured at 2 × 2 binning, 500-720 nm spectral emission range and 300 sec exposure to visualize the GFP labeled lesions forming in the abdominal cavity. All images were processed using the associated software for spectral analysis.

In addition, approximately half of a million SB247 cells formed peritoneal xenografts and metastases when inoculated i.p. into female FOX Chase SCID mice (Taconic, Hudson, NY). SCID mice were used to evaluate the effect of mAbs on peritoneal metastasis. To recover ascitic fluid samples, a 1 ml syringe fixed to a 22-gauge needle was coated with a 10X solution of Anticoagulant Citrate Detxtrose (Baxter Fenwal®, Deerfield, IL,) and 200 units/ml Spectrum® Heparin Sodium solution so that the void volume of the syringe and the bore of the needle were filled with 50 μl of anti-coagulant mixture. Mice were anesthetized with isofluorane and using reverse i.p. injection we recovered ascitic fluid that was placed on ice until processing. Ascitic fluid or tumor tissues from mice were evaluated for the presence of collagen and tumor cells by SDS-PAGE and Western Blot analysis to identify species of TIC present and density of GFP. The polyclonal anti-human TIC primary antibody (Rockland, Gilbertsville, PA) was diluted 1:10,000 and the polyclonal antibody against CNBr induced fragments of TIC (Calbiochem, San Diego, CA) was diluted 1:5,000. The intact collagen and gelatin strands were compared to control dilutions of purified TIC (BD Biosciences, Bedford, MA) for molecular size. Western immuno-blotting quantification of lung and liver metastases involved the use of antibodies against GFP (Sigma-Aldrich, St. Louis, MO).

Immunohistochemistry (IHC)

Paraffin embedded sections were cut to 4 μm. Paraffin was removed and the slides were rehydrated in an ethanol series, boiled in Antigen Unmasking Solution (Vector Laboratories, Burlingame, CA) for 20 min, treated with 1% H₂O₂ for peroxidase activity and blocked with 4% BSA in TT buffer (500 mM NaCl, 10 mM Trizma and 0.05% Tween-20 solution) for 20 min. A rabbit polyclonal antibody against Ki67 (abcam Inc., Cambridge, MA), which targets a cell cycle dependent nuclear protein that is absent in G0 but present in all other phases of the cell cycle, was used as an indicator of proliferation via conventional IHC. The Ki67 primary antibody was diluted 1:100 and was incubated overnight at 4°C. Biotinylated goat anti-rabbit IgG secondary antibody BA-1000 (Vector Laboratories) was applied at a 1:400 dilution and incubated for 40 min. ABC-HRP and DAB solution (Vector Laboratories SK-4100) was prepared according to manufacturer’s instruction. Gill’s Hematoxylin (Fischer Scientific, Fairlawn, NJ) was used for counterstain. Ki67 positive cells were counted as a percentage of all cells in three random fields.

Statistical Analysis

Statistical analyses were performed using GraphPad Prism® Software (GraphPad Software, San Diego CA). Student’s T-test was used to determine statistical significance. The values presented are the averages ± SEM unless otherwise indicated. For analysis of the gel contractility assay, ANOVA was corrected for repeated measures and coupled with a post-hoc analysis between the three cell types using Tukey’s multiple comparison test. Findings that were P ≤0.05 in Students T-test were regarded as significant.

Results

TICg increases seprase RNA expression in ovarian tumor cells

SB247 cells in culture expressed low levels of seprase RNA. A 3.8-fold increase in seprase RNA expression was produced in the SB247 cell line when cells were incubated for 24 hrs on a rehydrated TIC coated plate; however, Growth Factor Reduced Matrigel did not elicit the same response and levels of seprase remained comparable to the very low levels produced by cells plated on a plastic (Fig. 1a). In addition, already adherent SB247 cells exposed to 50 μg/ml of TICg in media for 1 hr yielded a seprase expression response of 2.5-fold higher than media alone and, after 4 hrs of exposure at that same concentration, seprase expression was four-fold greater than media alone (Fig. 1b). Furthermore, when the non-adherent ascitic fluid environment was simulated through culture in a bacterial dish, the SB247 cells expressed seprase in amounts 2.7- and 2.9-fold greater when stimulated with 30 and 40 μg/ml dilutions of TICg in media, respectively, as compared to non-stimulated controls after incubation for 2 hrs (Fig. 1c). While unstimulated SB247 cells maintain very low levels of seprase expression, interaction with TIC, either as a substrate or in the media, regardless of cell anchorage status, results in rapid stimulation of seprase production.

Figure 1.

Comparison of seprase RNA expression in cells treated with TIC. RT-PCR was used to assess seprase expression compared to levels in SB247 cells cultured in the absence of any ECM proteins. Significance was calculated by Student’s T-test and error bars are the SEM. All values were normalized to β-actin. (a), A comparison of substrates. TIC (1 mg/ml) diluted in DMEM or Growth Factor Reduced Matrigel™ was coated onto tissue culture dishes, allowed to set and then dried overnight. Prior to use each of the dishes was washed twice with PBS and 5 × 10⁵ SB247 cells were suspended in CCC, plated on the substrates, and incubated for 24 hours. *** indicates a significant difference of P< 0.001. (b) Comparison of TIC in solution added to adherent cells. SB247 cells were plated and allowed to adhere to plastic tissue culture dishes in CCC over night. Cells were then exposed to the indicated concentrations of TIC diluted in DMEM for l hour and 4 hours. Fold change is compared to cells in culture media containing no TIC at each time point. (c) Comparison of TIC dilutions under non-adherent cell culture conditions. Cells were suspended in solution of TIC dilutions ≤ 200 μg/ml and incubated in bacterial dishes for 2 hrs to simulate the ascitic fluid environment.

Seprase knockdown cell lines have a diminished capacity for TICg disk contraction and invasion

Three subclones were isolated and were plated on TICg to evaluate the expression of seprase RNA. The control subclone containing an RNAi hairpin with no target (NT) was compared to subclone 1384 #1 and a significant (P=0.036), 44.1% reduction in seprase expression was detected (Fig. 2a). Further, RNAi against seprase is most effective in subclone 1384 #4 in that it nearly prevents seprase expression and significantly (P=0.0009) reduces it to 0.3% of the NT control (Fig. 2a). Neither proliferation in vitro as shown in Fig 2b, nor the ability of these cells to form colonies in soft agar (data not shown) appears to be dependent on seprase RNA expression. However, a reduction in seprase expression in the 1384 #1 and 1384 #4 cell lines corresponded with a significant (P=0.0037 and P<0.0001, respectively) inhibition of these cells to invade through TICg in vitro (Fig. 2c).

Figure 2.

Characterization of monoclonal cells stably expressing seprase shRNA. (a) RT-PCR comparing seprase expression of SB247 monoclonal cells containing the RNAi vector with a no target hairpin (NT), or a hairpin targeting seprase (1384 #1, 1384 #4). Cells were plated overnight on dried 1 mg/ml TIC. (b) Proliferation of monoclonal cells in culture. Cells (10⁵) from each cell line were cultured in 10 cm cell culture dish and incubated at 37°C under 5% CO₂. Cells were counted every day under a microscopy for 6 days. (c) Transwell migration of monoclonal cells to cross a TIC coated membrane. Values presented are the means ± SEM of 9 filters and corrected from control experiments with uncoated membranes. Significance was determined by a Student’s t-test and is indicated as follows: *=P < 0.05, **= P < 0.01 and ***= P < 0.001. (d-g) Contraction of TICg by monoclonal cells. Gel contractility assay was performed to examine the ability of cells to reorganize and contract TICg. TIC diluted in DMEM to 220 μg/ml (d), 360 μg/ml (e) and 570 μg/ml (f) was added to the three SB274 subclones. The diameters of the major and minor axes were taken daily and averaged. Each condition was evaluated in triplicate for three experiments and the vertical lines represent ± the SEM of the cumulative experiments. (g) Images of the disks contracted at Day 5. The inset numbers reflect the percent of average diameter reduction at the end of the five day incubation.

The in vitro gel contractility assay was useful to assess the tumor cells’ capacity to adhere to the peritoneal membrane and form secondary lesions.²⁴ When applied to a gel contractility assay, over the five day time course and across all three TICg concentrations (220 μg/ml, 360 μg/ml and 570 μg/ml) the NT cell line retained the greatest ability to contract the TICg disks, whereas the increasing inhibition of seprase induction in the 1384 #1 and 1384 #4 cell lines was coincident with a progressive decline in disk contraction (Fig. 2d-2g). In fact, statistical analysis revealed a significant difference between the NT cell line and the 1384 #4 cell lines at all three concentrations; 220 μg/ml (P=0.0105), 360 μg/ml (P=0.0005) and 570 μg/ml (P=0.0034). In addition, all three cell lines were significantly different from one another at the 360 μg/ml concentration (P<0.05). Therefore, by diminishing seprase production, we have reduced invasion and TICg contraction, which are characteristics of invasive tumor cells.

Seprase expression is associated with greater secondary lesion formation in vivo

Apparent by fluorescence microscopy and then confirmed by flow cytometry, we found that the NT subclone expressed the lowest levels of GFP as indicated by mean fluorescence index, 1384 #1 showed 1.86-fold higher levels, and 1384 #4 exhibited the highest GFP signal at 2.52 times greater than the NT cell line. For example, if a cluster of 1000 1384 #4 cells were required to elicit a strong enough GFP signal to be visualized using the in vivo imaging equipment, an equivalent GFP signal would only be produced when 2,520 NT cells were clustered. Even though heightened levels of GFP in cells that exhibited the least seprase expression may have enabled visualization of smaller lesions in the two knockdown cell lines, quantification by visual inspection showed that the number of lesions was still greatest in mice inoculated with the NT cell line that has the greatest seprase expression, but weakest GFP signal (Fig. 3a, 3b). The GFP positive cell clusters were first visualized on the abdominal surface using the in vivo imaging equipment, but localization and adherence to the peritoneal membrane was only verified upon necropsy. Dissection of the peritoneal membrane, followed by a PBS rinse, allowed enumeration of individual lesions that were GFP positive and remained tightly adhered to the peritoneal membrane. Therefore only attached lesions were counted and free-floating tumor cell clusters in the ascitic fluid were not quantified. Further, this method allowed us to distinguish lesions formed on the ventral surface of the peritoneal membrane from tumors localized to the spleen and pancreas regions (Fig. 3a *). Overall, a significant reduction in tumor formation between the control NT and knockdown 1384 #4 subclones (P =0.034) was revealed (Fig. 3a, 3b).

Figure 3.

Seprase suppression corresponds to inhibition of peritoneal membrane tumor lesion. Ovarian tumor lesion was examined in vivo using i.p. inoculation of the SB247 cell line expressing high (NT), medium (1384 #1) and low (1384 #4) levels of seprase in nude mice. (a) Visualization of tumor cells created by GFP labeled subclones expressing different levels of seprase. A total of 15 Balb/c ^nu/nu mice (5 mice/ subclone) were injected with 10⁶ cells and their ventral sides were monitored via Maestro fluorescence imager. Images were captured at Day 24. The green areas indicate regions of tumor lesion and the light pink regions are visualization of the intestines containing rodent food that fluoresces red. The bottom two mice in the NT and 1384 #4 groups as well as the bottom mouse in the 1384 #1 group all appear with white patches, which indicate a full bladder. The open white arrows point to the tumor that developed at the site of injection, which are different sizes due to difference in GFP intensity across the three cell lines. The small white arrows indicate representative i.p. tumor lesions. Higher magnification above the boxed images was shown to indicate small tumor lesions. The asterisks (*) indicate the lesions developed in the pancreas and spleen region, but are not attached to the ventral surface of the i.p. membrane as determined by fluorescence imaging during necropsy. (b) Enumeration of peritoneal tumor lesions by visual inspection. During necropsy the peritoneal membrane was removed, rinsed with PBS and viewed under fluorescence to quantify only adherent tumor clusters. (c) Quantification of Ki67 positive cells in IHC of peritoneal tumor lesions. Positive cells were counted as a percent of the total cells in four random fields. Values presented are the means ± SEM and significance was determined by a Student’s t-test and is indicated as follows: *=P < 0.05 and ***= P < 0.001.

The in vivo imaging system used during the longitudinal study enabled us to see that tumors formed initially at the site of injection merely four days post injection and persisted over the course of the experiment, which confirmed delivery of tumor cells at the outset of the experiment (Fig. 3a, open arrows). The difference in GFP levels across the cell lines accounts for the apparent size differences in the site of injection lesion as detected by the fluorescence imager (Fig. 3a, open arrow). Necropsy revealed that tumor cell clusters were also found within the folds of the intestinal mesentery, an extension of the peritoneum, and on the surface of other internal organs in addition to the ventral peritoneal membrane surface (Fig. 3a, white arrows), which was the region used for comparison. Labeled GFP cells were found to cluster near the spleen and pancreas (Fig. 3a, *), but individual lesions were difficult to discern in this region. Moreover, cell proliferation assay as determined by IHC utilizing the polyclonal antibody against Ki67, revealed that 1384 #1 cells within the lesions proliferated at a significantly greater rate than both NT cells (P=0.0006) and 1348 #4 cells (P=0.001) and proliferation detected in NT generated lesions were significantly greater (P=0.024) than 1384 #4 (Fig. 3c). Despite a weaker GFP signal and differences in proliferation, the NT subclone, with the greatest expression of seprase, yielded a greater number of peritoneal membrane tumor lesions than either knockdown subclone.

β1 integrin functions in collagen gel contraction, seprase induction and tumor progression

Previously, the C27 mAb directed against β1 integrin was described and characterized for its functional blocking of cellular adhesion to and migration on collagenous substrata. ^21,22,23 Using a gel contractility assay, the contraction capability of the SB247 cells is inhibited when they are exposed to 200 μg/ml of mAb C27 against β1 integrin before or 1-hour after cell-gel formation, but the control mAb C37, directed against cell surface glycoprotein gp90, does not exhibit the same effect (Fig. 4a). The NT subclone reduced the average diameter of the TICg disk to 26.3% of its original size when exposed to the control mAb C37, but contraction was inhibited when mAb C27 was applied and only a 47.8% reduction of the original dimensions was seen. The effect of mAb C27 on the NT subclone was similar to the effect of the 1384 #1 partial knockdown of seprase in the presence of the control antibody reflected by the 47.2% reduction. In the same way, when mAb C27 was applied to the 1384 #1 it diminished TICg disk contraction to 66.3% of original dimensions, which is comparable to the 1384 #4 subclone and the control antibody at 66.3% overall reduction. Finally, the 1384 #4 subclone, in the presence of mAb C27 was still capable of disk contraction, but this was significantly lessened to an 81.1% reduction.

Figure 4.

Inhibition of gel contraction and induction of seprase expression in cells by anti-β1 integrin antibody. (a) Gel contractility by cells expressing high (NT), medium (1384 #1) and low (1384 #4) levels of seprase in the presence of either mAb C27 against β1 integrin (open symbols) or mAb C37 against glycoprotein-90 (solid symbols). (b-d) Seprase RNA expression of cells exposed to the mAb C27 under distinct adherence conditions. All measurements were normalized to β-actin. (b) Comparison of mAb dilutions added to adherent cells. Cells were plated and allowed to adhere to the dish overnight. Dilutions of mAb C27 were added to the culture media and incubated for 2hrs. (c) Comparison of cells added to pre-bound mAb C27 dilutions. Dilutions of mAb C27 were applied to CellBIND® (Corning, Lowell, MA) high-affinity binding plates and incubated overnight at 37°C. SB247 cells were then plated and incubated for 2hrs. (d) Comparison of mAb C27 dilutions under non-adherent conditions. Cells were mixed with dilutions of mAb C27 and incubated for 2hrs on bacterial plates to maintain the cells in suspension. To determine if mAb C27 caused the crosslinking of β1 integrin subunits and clustering induced seprase expression, mAb C27 was digested with papain and the Fab portions were purified using Fab Preparation Kit (Pierce, Rockford, IL) according to manufacturers instruction and were applied to SB247 cells in suspension.

To examine the impact of β1 integrin engagement on seprase expression of adherent cells, the SB247 cells cultured overnight were exposed to a series of mAb C27 dilutions. For cells growing on plastic that were exposed to media containing 300 μg/ml of mAb C27, seprase expression was induced 1.86-fold (Fig. 4b). In contrast, when the antibody was first immobilized on high-affinity binding plates, 600-800 μg/ml of mAb C27 was necessary to induce 3.31-fold elevation in seprase expression (Fig. 4c). In addition, when incubated in non-adherent conditions similar to that of an ascitic fluid environment, SB247 cells expressed elevated seprase levels that were 6.8-fold greater in response to the 300 μg/ml concentration of mAb C27 (Fig. 4d), suggesting the involvement of β1 integrin available for mAb C27 binding in seprase expression. To determine if mAb C27 cross-linking of β1 integrin is involved in the induction, mAb C27 was first papain digested into Fab fragments and then applied to the cells in non-adherent conditions, and no such response was detected (Fig. 4d). Together, our data strongly suggest that seprase expression is mediated through β1 integrin clustering on the tumor cell surface.

To investigate if β1 integrin signaling of seprase expression occurs during ovarian tumor metastasis, the capability of GFP-SB247 cells to metastasize in SCID mice were examined in the presence of mAb C27 (directed against β1 integrin) and control IgG matched mAb C37 (directed against gp90). Approximately 5 × 10⁵ GFP-SB247A cells that stably express GFP were inoculated i.p. with mAb C27 or control IgG into a SCID mouse. Peritoneal metastases in aggregates of greater than 3 × 10³ cells in both control and experimental animals were first recorded 4-6 weeks after inoculation using Maestro™ in vivo imaging system (Fig. 5a). Based on total GFP fluorescence, mAb C27 significantly blocked formation of peritoneal aggregates (n=4, p<0.01), in comparison with the control IgG group (Fig. 5a). Similarly, based on GFP colony numbers per mouse, the number of regional tumor colonies in the control group was significantly higher than these of the experimental group (Fig. 5b, p<0.0002), suggesting that antibody inhibitors could block formation of peritoneal metastases in ovarian cancer. After six weeks, the GFP signal in lung and liver tissues was examined by fluorescence microscopy and Western immunoblotting to indicate metastases in the lung and liver (Fig. 5c, 5d). In comparison with the control IgG group (Fig. 5c, 5d, left bars), mAb C27 (Fig. 5c, n=4, p<0.0002) (Fig. 5d, n=4, p<0.02) significantly blocked formation of lung and liver metastases. These data support the notion that induction of seprase expression in SB247 tumor cells promotes development of peritoneal and distant metastases, and antibodies directed against β1 integrin can help suppress metastatic progression.

Figure 5.

Inhibition of ovarian tumor metastasis by mAb C27 against β1 integrin. (a) Peritoneal metastases examined based on total GFP signal. Inhibition of peritoneal tumor growth is significant (*) for mAb C27 in comparison with the control IgG (P < 0.01). (b) Peritoneal metastases examined based on number of GFP colonies per mouse. The number of peritoneal tumor colonies in the control group was significantly higher than these of the experimental group (*, p<0.0002). (c) Inhibition of lung and liver metastases by mAb C27. Fluorescence microscopy was used to detect the number of GFP colonies per mm² of lung or liver tissues. Inhibition of lung and liver metastases is significant for mAb C27 (P < 0.0002; n=4). (d) Inhibition of lung and liver metastases by mAb C27. Western immunoblotting was used to measure GFP per μg of lung or liver tissues as an indicator of lung and liver metastases. Inhibition of lung and liver metastases is significant for mAb C27 (P < 0.02; n=4).

Discussion

Seprase is present on tumor cells invading the ECM, endothelial cells involved in angiogenesis, and fibroblasts involved in wound closure, but is not expressed in inactive cells of the same type.^21,22,23 This transient expression suggests that seprase is induced by some component to which only active cells are exposed. Indeed, we found rapid seprase induction in the SB247 ovarian carcinoma cell line when exposed to low TICg concentrations regardless of cell adhesion status (Fig. 1). While Matrigel™ contains other matrix proteins, it does not contain TIC. The elevation of seprase RNA expression in the SB247 cells through plating on rehydrated TICg as compared to cells plated on plastic or growth factor reduced Matrigel™ indicates a defined induction that requires not merely any substrate to which cells can adhere, but rather TICg in particular. This is further supported by the dramatic increase in seprase expression of cells when exposed to 50 μg/mL TICg under adherent conditions (Fig. 1b) and 30-40 μg/mL TICg under non-adherent conditions. There appears to be a threshold where individual gelatin strands are at a maximal concentration just prior to forming secondary or tertiary structures, where seprase expression is maximal. At greater TIC concentrations, higher order gelatin/collagen structures may form that reduces the accessibility of seprase to its substrate. Our experiments suggest that, in the absence of a suitable substrate, seprase induction wanes.

The effect of seprase on the ECM is also reflected in formation of peritoneal tumor lesions in an ovarian cancer peritoneal metastasis model in that mice inoculated with seprase knockdown cell lines formed fewer lesions (Fig. 3a-b). Tumors were found to form at the site of injection regardless cell lines and these tumors were used to confirm successful delivery of tumor cells. It is reasonable to find a tumor developing within the peritoneal cavity at the injection site due to mesothelial cells activation for wound healing purposes secreting matrix molecules that would either attract or catch tumor cells irrespective of seprase expression.

There is much support for the involvement of integrins in gelatinase activity. Integrin α3β1 has been implicated in promoting the invasion of hepatocellular carcinoma, melanoma and breast carcinoma cells through collagenous matrices.²⁵ Antibodies directed against β1 integrin have been shown to interrupt ascitic fluid spheroids of ovarian cancer patients from binding to a TIC substrate.¹ In addition, the TIC receptor β1 integrin, co-localizes with seprase on the surface of actively invading tumor cells in response to TIC exposure and endothelial cells during angiogenesis.^19,26 In this study, we found that mAb C27 against β1 integrin inhibited gelatin contraction and induced seprase expression when the SB247 cell lines interacted with TIC (Fig. 4). These findings suggest that the production of seprase alone is not sufficient for collagen gel contraction, but rather the TICg substrate must be bound to β1 integrin, an interaction which is disrupted by mAb C27.

Figure 4c shows the induction of seprase expression only at the high concentration of mAb C27 anchored to the high affinity binding plate, suggesting that antibody induced cross-linking of β1 integrin subunits is responsible for the elevation of seprase expression. Consistently, Figure 4d shows a specific and intense spike in seprase induction in response to the intact mAb C27 when the cells are in suspension, but not to Fab fragments alone, which lack the Fc portion necessary for cross-linking. When the cells are in suspension it is likely that the β1 integrin subunits are equally dispersed over the cell surface. The spike reflects the concentration of mAb C27 at which β1 integrin subunits can be crosslinked, but at higher concentrations of antibody no further cross-linking can occur because the subunits have been saturated. This integrin clustering effect appears to be a common mechanism for gelatinase production in tumor cells because not only has integrin interaction with the ECM been shown to promote the production and activation of both soluble and membrane bound MMPs, but MMP-2, MMP-9, and MMP-14 have all been induced through exposure to antibodies or cross-linkers targeting β1 integrin.²⁷

From these data it appears that it is not only the clustering of β1 integrin and resulting seprase expression alone that allows seprase to exert its gelatinase activity, but rather the proximity of seprase to TICg, which is held by β1 integrin. Because β1 integrin is bound to the substrate of seprase it may act as a beacon towards which seprase can be directionally shuttled to the surface, colocalize with β1 integrin, thereby moving to the vicinity of the bound substrate. However, in the absence of the TICg substrate, when seprase is directionally shuttled to β1 integrin, there is no substrate upon which to act and therefore seprase is unable to act as a gelatinase. Furthermore, it is likely that when intact TIC substrate binds to β1 integrin it is first processed into TICg by members of the MMP family such as MMP-2 ²⁸ and MMP-9 ²⁹, both of which are known to cleave fibrillar TIC, and is subsequently acted on by gelatinases such as seprase. When the TICg has been digested by seprase, those fragments are released, and β1 integrin is prompted to interact with a new TIC strand, thereby advancing the protruding invadopodia into the substrate. As shown by seprase production via β1 integrin stimulation with mAb C27 and the concurrent inability to contract TIC, seprase does not appear to interact with β1 integrin in the same way when lacking the TIC substrate. We have shown that mAb C27 stimulates the production of seprase, but seprase is then either unable to properly colocalize with TICg, or is not properly activated by other proteases that are reliant on the integrin-substrate specific interaction.

Although small-molecule inhibitors are known to effectively inhibit the proteolytic activity of different classes of proteases, the disruption of interactions between β1 integrin, seprase and ECM components may prove to be an additional tool against tumor progression.^30,31 The simultaneous inhibition of seprase proteolytic activity and prevention of β1 integrin interaction with the ECM could be useful to prevent formation of peritoneal tumor lesions in progressing cases of ovarian carcinoma. In this way tumor cell interaction with TIC would be hindered and may prove a useful strategy in combating tumor progression in late stage ovarian cancer. It is clear that seprase plays a role in the ascites environment and likely is a key player in metastatic lesion formation in epithelial ovarian carcinoma. Further, it is the transient and inducible nature of seprase that suggests seprase to be a specific target ideally suited for inhibition.

Abbreviations

TICg

type I collagen gel

mAb

monoclonal antibody

RNAi

RNA interference

ECM

extracellular matrix

TIC

type I collagen

a.a.

amino acid

MMP

matrix metalloprotease

DPP4

dipeptidyl peptidase IV

DMEM

Dulbecco’s modified Eagle’s medium

PBS

phosphate buffered saline

CCC medium

cancer cell culture medium

FACS

fluorescence activated cell sorting

GFP

green fluorescent protein

RT-PCR

reverse transcriptase- polymerase chain reaction

i.p.

intraperitoneal

SDS-PAGE

sodium dodecyl sulfate polyacrylamide gel electrophoresis

shRNA

short hairpin RNA

BSA

bovine serum albumin

TT

Trizma-Tween-20

GFRM

growth factor reduced matrigel

TIC_FRAG

type I collagen fragment

TICg

type I collagen gel

IHC

immunohistochemistry