Kallikrein-related peptidase 7 (KLK7) is upregulated in epithelial ovarian carcinoma (EOC) with high levels correlated with poor prognosis. However, the mechanisms underlying this relationship and the role of KLK7 in EOC progression are unknown. We report that two different KLK7 transcripts, KLK7-253 and KLK7-181, are simultaneously expressed in high-grade serous EOC. Multicellular aggregates (MCA), which promote cell survival and chemoresistance, were observed in SKOV-3 cells stably overexpressing KLK7-253 in particular. Importantly, these MCAs invade into a monolayer of mesothelial cells and form cancer cell foci. Blocking MCA using antibodies against KLK7 and α5β1 and β1 integrins confirmed the involvement of KLK7 and integrin-regulated cell adhesion. Increased levels of α5/β1 integrins and enhanced attachment to fibronectin and vitronectin, which was blocked with an anti–β1 integrin antibody, were also observed. Finally, Western blot and immunohistochemistry showed higher KLK7 and α5/β1 integrin levels in serous EOC cells from ascites and tumor samples from chemotherapy nonresponders with short postsurvival times. Additionally, both KLK7-253 and KLK7-181 clones were more resistant to paclitaxel treatment in vitro. These findings suggest a mechanism for the association of high KLK7 levels with chemoresistance and poor prognosis for serous EOC patients by promotion of peritoneal dissemination and reinvasion via increased MCA and α5β1 integrin–dependent cell adhesion. Cancer Res; 70(7); 2624–33
Ovarian cancer is the leading cause of death from gynecological malignancies, with serous epithelial ovarian carcinoma (EOC) being the predominant subtype. Late diagnosis and chemoresistance result in a ∼30% 5-year survival rate (1). Unlike other solid tumors, hematogenous metastasis rarely occurs in ovarian cancer; instead, EOC cells exfoliate from the primary site, form multicellular aggregates (MCA)/spheroids to aid survival in the peritoneal fluid, adhere to mesothelial cells on the surface of the peritoneum, invade into the underlying extracellular matrix (ECM), and grow secondary tumors (2, 3).
Dynamic modulation of cell adhesion molecules, such as integrins, plays a critical role in remodeling the malignant ovarian epithelium during peritoneal metastasis (4). Clinically, high levels of αv (5) and α5 (6) integrins are associated with poor outcome. Integrin signaling has been related to MCA/spheroid formation (7) and disaggregation (8) in EOC, and spheroids formed in three-dimensional suspension are more resistant to chemotherapy than two-dimensional monolayer cultured cells (9). Integrins can promote EOC cell dissemination and implantation in the peritoneal cavity via E-cadherin ectodomain shedding (10), loss of E-cadherin (6), and ECM protein degradation (11) by matrix metalloproteinases (MMP). However, the role of other proteases in EOC remains largely unknown.
The kallikrein-related peptidase (KLK) family comprises 15 serine proteases, with several members upregulated in EOC (12–14), including KLK7 (15–20), which is associated with poor prognosis in EOC patients (17, 18). KLKs degrade ECM molecules such as fibronectin and cell adhesion proteins contributing to skin desquamation (21) as well as activate growth factors and other proteases (12–14). Two KLK7 transcripts (KLK7-253 and KLK7-181) have been detected in cancerous but not normal ovarian epithelial (NOE) cells, implying cancer-specific expression (15), although little is known about their biological function in EOC. We now report that stable expression of KLK7-253 and KLK7-181 in SKOV-3 cells promotes paclitaxel resistance and enhances expression of α5/β1 integrins. The KLK7-253 clones, in particular, also form large compact spheroids, which were blocked by functional antibodies against KLK7 and α5/β1 integrins and which were capable of invasion into mesothelial cell monolayers. These data provide a mechanism for our clinically observed high KLK7 levels that were associated with chemoresistance and poor outcome.
Materials and Methods
Antibodies were purchased from Chemicon (integrins), R&D Systems (KLK7), and Invitrogen (anti-V5 and anti-mouse Alexa Fluor 488/568–conjugated secondary antibodies). All other chemicals were purchased from Sigma except where noted below.
Cell lines and ovarian tissue RNA
Cell lines were purchased from the American Type Culture Collection (serous EOC: SKOV-3, OVCAR3, OV-90, and Caov-3; fibroblast: NFF1) or Coriell Cell Repositories (peritoneal mesothelial: LP9) with PEO1, PEO4, PEO14, OAW42, JAM, and CI-80-13S as described previously (15). Other cell lines were generous gifts from Dr. Nelly Auersperg (University of British Columbia, Vancouver, British Columbia, Canada; IOSE144, IOSE379, and IOSE7576) and Dr. Samuel Mok (Harvard University, Cambridge, MA; HOSE17.1, OVCA420, and OVCA432). Four normal ovarian, 31 serous EOC tissue (22), and 7 EOC RNA (15) samples were described previously. Clinical and pathologic information of EOC patients was obtained from the Royal Brisbane Women's Hospital (Supplementary Table S1). Ethical approval was obtained from institutional ethics committees, and informed consent was obtained from all patients.
RNA extraction, reverse transcription-PCR, and reverse transcription-quantitative PCR
RNA was isolated using the RNeasy kit (Qiagen), treated with DNaseI (Invitrogen), and 2 μg RNA reverse transcribed using SuperScript III (Invitrogen). All KLK7-specific primers are shown schematically in Fig. 1A, and sequences of all primers, KLK7 microarray probe, and PCR conditions are listed in Supplementary Table S2. K75'S and K7Ex3AS were used to amplify both KLK7-253 (573 bp) and KLK7-181 (442 bp; Fig. 1A). PCR products were sequenced (Australian Genome Research Facility) to confirm transcript-specific amplicons (tBLASTN).
Quantitative PCR (qPCR) was performed on an ABI 7300 thermal cycler (Applied Biosystems) using SYBR Green following the manufacturer's instructions. KLK7 expression was normalized to 18S using the standard curve method. K7Ex1qS and K7Ex2qAS primers were used to amplify KLK7-253, whereas both KLK7-253 and KLK7-181 transcripts (total KLK7) were amplified with K7Ex4qS and K7Ex5qAS primers (Fig. 1A). The copy number of KLK7-181 was obtained by subtraction of KLK7-253 from total amplified KLK7. Microarray analysis was performed as previously described (22).
Generation of stable cell lines
KLK7-253 and KLK7-181 constructs are shown in Supplementary Fig. S1A. KLK7-253 and KLK7-181 (Fig. 1A) were amplified from OVCAR3 cDNA, subcloned (pcDNA3.1/V5-His, Invitrogen; COOH-terminal V5 tag), and transfected into SKOV-3 cells using Lipofectamine 2000 (Invitrogen). Monoclonal cells expressing KLK7 or vector control were selected using G418 (Invitrogen). Reverse transcription-qPCR (RT-qPCR) was performed using K7Ex4qS and K7Ex5qAS primers, and products were sequenced confirming transcript-specific expression.
Cells were grown on coverslips to ∼80% confluence, fixed [4% (w/v) paraformaldehyde], blocked [1% (w/v) bovine serum albumin/PBS], and incubated with anti-V5 antibody (1:500). Alexa Fluor 568 phalloidin (Invitrogen) was included with the secondary antibody (Alexa Fluor 488 anti-mouse IgG). Images were taken using a Leica-TCS SP5 confocal microscope (63× 1.4 oil immersion objective) and associated software.
In vitro functional assays
Cells (2 × 103 per well) were seeded in phenol red–free RPMI 1640/10% (v/v) FCS in 96-well plates (Nunc) and incubated at 37°C for 4, 24, 48, 72, and 96 h. WST-1 assays were performed per the manufacturer's instructions (Roche). The 4-h time point was used as “no-proliferation” control. The assay was performed thrice with eight replicates per clone per time point.
Cell survival after cisplatin/paclitaxel treatment.
Twenty-four hours after seeding, cells were treated with cisplatin (0, 1, 5, 10, 50, and 100 μmol/L) or paclitaxel (0, 0.01, 1, 10, and 100 nmol/L). WST-1 assays were performed 72 h after treatment as above. Cell survival was calculated as percentage of absorbance in nontreated cells.
MCA/spheroid formation and inhibition.
Generation of spheroids and MCA inhibition was based on the liquid overlay technique (7). Cells (5 × 103 per well) were layered on top of agarose-coated plates [60 μL of 0.5% (w/v) agarose/serum-free medium] and incubated at 37°C. For MCA inhibition, cells were resuspended in serum-free culture medium, and 5 × 103 cells per well were plated in agarose-coated 96-well plates. Blocking antibodies (anti-α5β1/β1 integrins and anti-KLK7) or control mouse IgG was added (10 μg/mL). Cells were cultured for 2 d and images were taken using a Nikon Eclipse TE2000-U digital camera (4× objective) and V++ software. The experiment was repeated thrice in quadruplicate.
Cell survival of MCA/spheroids after paclitaxel treatment.
Spheroid formation was conducted as described above. After 4 d of suspension, cultures were treated with paclitaxel (0, 0.01, 1, 10, and 100 nmol/L), and after 72 h, alamarBlue (Invitrogen) was added [10% (v/v) final concentration] for 5 h per the manufacturer's instruction. Fluorescence was detected using a BMG PolarStar plate reader. The assay was performed thrice with triplicates per clone per concentration. Cell survival was calculated as percentage of fluorescence in nontreated cells.
Monolayer mesothelial cell invasion.
LP9 mesothelial cells (5 × 103) were seeded in 96-well plates and grown to ∼80% confluence. Spheroids were washed in PBS, incubated in CellTracker Green CMFDA (4 μmol/L; Invitrogen), added on top of mesothelial monolayers (approximately four to six spheroids/200 μL), and cultured at 37°C. At 4 h and 1, 2, and 3 d from the initial spheroid plating, images were taken as above. The experiment was performed thrice in quadruplicate.
Attachment assays were performed using a published protocol (23) with 2 × 104 cells per well on plates coated (10 μg/mL) with fibronectin, vitronectin (BD Biosciences), and collagen I or IV (Calbiochem). For blocking assays, cells were seeded in serum-free medium with functional blocking β1 integrin antibody or mouse IgG (10 μg/mL). The assay was repeated thrice with four replicates per ECM protein.
Cells (2 × 104) were incubated with primary integrin antibodies for 1.5 h with anti-mouse IgG as control, followed by Alexa-conjugated secondary antibodies, washed, and resuspended in 0.4 mL PBS before flow cytometry (Beckman Coulter) and data analysis using CXP software (Beckman Coulter). The experiment was repeated thrice per integrin.
Whole-cell lysates (20 μg) were separated by SDS-PAGE (15) and blotted with primary antibodies α5 and β1 (1:1,000) integrins, KLK7 (1:500), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH; 1:10,000) and secondary antibody IRDye 680/800–conjugated mouse/rabbit IgG (LI-COR Biosciences). Images were obtained using the Odyssey (LI-COR Biosciences) system, and densitometric analyses were performed using Odyssey software version 2.0.
Immunohistochemistry was performed as previously described (15) on paraffin-embedded sections of well-differentiated and poorly differentiated tumors from seven serous EOC patients who had responded or not to chemotherapy (ID nos. 7 and 11: nonresponders, poorly differentiated; ID nos. 28, 31, and 34: responders, poorly differentiated; ID nos. 23 and 37: responders, well differentiated; Supplementary Table S1). Following antigen retrieval and incubation with antibodies against KLK7 (1:200), α5 or β1 integrin (1:400), or rabbit IgG (negative control; Dako) overnight at 4°C, the EnVision peroxidase system (Dako) was applied. Sections were counterstained with Mayer's hematoxylin, visualized using a digital camera (Olympus BX41), and photographed (QCapture Pro 6.0 software). Images were processed using Adobe Photoshop CS3 and displayed using CorelDraw.
KLK7-253 and KLK7-181 are highly expressed in serous EOC cells and tissues
The relative expression levels and cancer specificity (15) of KLK7-253 and KLK7-181 transcripts were examined using a cohort of normal ovarian tissues (NOT), NOE, and serous EOC cells. In high-grade (G2 and G3) primary serous EOC cells, PEO1, PEO4, and OVCAR3 cell lines, two amplicons were observed, with KLK7-253 bands (573 bp) generally more intense than those of KLK7-181 (442 bp; Fig. 1B). In contrast, no amplicons were detected in NOT, NOE, low-grade (G1), or SKOV-3 cancer cells (Fig. 1B), with β2-microglobulin (β2M) as a loading control. KLK7 protein was also detected in lysates from three of four high-grade EOCs, PEO1 (faint bands), and PEO4 cell lines, but not NOE, low-grade cancer, or SKOV-3 cells (Fig. 1C). The two bands at approximately 32 to 35 kDa likely represent glycosylated and unglycosylated KLK7-253 (24). No bands were detected at the suspected ∼20 kDa for the KLK7-181 isoform. RT-qPCR was performed for KLK7-253 and KLK7-181 transcripts using 4 NOE, 11 serous EOC lines, 3 NOT, and 16 serous EOC tissues. Both KLK7 transcripts are significantly upregulated in EOC cells and tissues compared with NOE and NOT, respectively (P < 0.001; Fig. 1D, left). These data were supported by microarray analysis, which showed significantly higher total KLK7 levels in cancerous than normal and benign ovarian tissues (P < 0.001; Fig. 1D, right).
KLK7-253 cells form MCAs that invade into mesothelial cell monolayers
MCA formation is thought to be a mechanism for EOC cell survival in ascites before peritoneal invasion (25, 26). To determine if KLK7 played a role in MCA formation, we generated stable transfectants of KLK7-253 and KLK7-181 in SKOV-3 cells (Supplementary Fig. S1B and C; Fig. 2A). Confocal microscopy confirmed protein over-expression in each of the KLK7-253 and KLK7-181 (at a lower level) clones, but negligible expression in vector control clones (Fig. 2A; Supplementary Fig. S1C). We then compared the ability of KLK7-253 and KLK7-181 clones to form spheroids (Fig. 2B). By 2, 4, and 7 days after seeding, vector controls and native SKOV-3 cells formed small spheroids compared with the single large compact MCA generated by all three KLK7-253 clones and less compact MCA formed by one (5D11) KLK7-181 clone (Fig. 2B). The other two KLK7-181 clones had a similar aggregation pattern to vector controls possibly due to low levels of KLK7-181 expression (Fig. 2A; Supplementary Fig. S1B and C).
We then examined the invasiveness of two representative MCAs (3E4 and 5D11) when added onto peritoneal mesothelial cell monolayers. As shown by bright-field and fluorescent imaging, both the compact KLK7-253 (3E4) and the less compact KLK7-181 (5D11) MCAs adhered on top of live mesothelial cell monolayers on the day of seeding and gradually formed an extensive front of invading cancer cells within 2 to 3 days (Fig. 2C), with the larger 3E4 MCA still spreading for up to 7 days (data not shown). However, similar migratory and invasive capacity of both KLK7-253 and KLK7-181 clones was apparent in two-dimensional Transwell migration and Matrigel invasion assays compared with vector controls (Supplementary Fig. S3A). Indeed, the attachment of KLK7-transfected clones to mesothelial cells did not differ from that of native SKOV-3 cells or vector controls (Supplementary Fig. S3B and C), suggesting that the only phenotypic difference induced by KLK7 overexpression was the ability to form MCAs.
KLK7-181 and KLK7-253 cell attachment to fibronectin/vitronectin is inhibited by a β1 integrin blocking antibody
To determine how disseminating KLK7-expressing cells invade into the peritoneum, given no difference in their adhesiveness to mesothelial cells (shown above), we examined their adhesion to ECM proteins from the matrix underlying mesothelial cell layers. Both KLK7-253 and KLK7-181 clones displayed increased adhesion to fibronectin and vitronectin (KLK7-181 only) compared with vector control cells (Fig. 3A and B), with greater adhesion exhibited by KLK7-181 clones (P < 0.01). There was no difference for collagen I/IV attachment (data not shown). The increased adhesion to fibronectin/vitronectin was reduced to basal levels by a β1 integrin blocking antibody for KLK7-181 clones and less for KLK7-253 clones, especially on vitronectin [Fig. 3A and B; KLK7-181, P < 0.01; KLK7-253, P < 0.05 (fibronectin only)]. As both α5 and β1 integrins regulate MCA formation (7), we used these functional blocking antibodies in MCA assays. Blocking α5β1, β1, or KLK7 reduced aggregation of both KLK7-253 (3E4) and KLK7-181 (5D11) clones, whereas cells remained aggregated with an IgG control (Fig. 3C).
High α5 integrin levels are associated with poor outcome for EOC patients (6), and increased α5/β1 integrins are responsible for serous EOC cell aggregation (7, 27), disaggregation, and invasion (8, 10). Of further interest, correlative analysis of our microarray data revealed a trend between KLK7 and α5 integrin expression (P = 0.09). Thus, we investigated whether integrin levels were elevated in KLK7-expressing clones. Flow cytometric analysis showed significant upregulation of cell surface α5 integrin expression in all KLK7-253 clones and β1 and αvβ3 integrin expression in KLK7-253 4B12, 3F11, and KLK7-181 5D11 clones compared with vector control cells (Fig. 4A). Although there was not a strict correlation for all clones between α5/β1/KLK7 levels and MCA formation, KLK7-253 clones in particular expressed the highest KLK7, formed the most compact MCAs, and had increased α5 (three of three) and β1 (two of three) integrin levels.
KLK7 and α5/β1 integrins are coexpressed in serous primary EOC cells and tissues
To confirm the relationship between KLK7 and integrin expression clinically, we performed Western blot and densitometric analysis on paired samples of serous EOC cells from primary tumors and ascites from six women (Fig. 4B and C). Compared with primary EOC cells, three of six ascites samples (ID nos. 1, 3, and 5) showed more intense KLK7 bands, four of six ascites samples (ID nos. 1, 2, 3, and 5) showed more intense α5 integrin bands (although this was not as apparent on densitometry), and two of six ascites samples (ID nos. 1 and 2) showed elevated β1 integrin expression. Interestingly, levels of all three proteins (KLK7 and α5 and β1 integrins) were greater in ascites compared with primary tumor–derived cells from a patient who only survived 6 months (Fig. 4B and C; ID no. 1). Conversely, samples (ID nos. 4 and 6) from patients who survived 23 and 18 months, respectively, following surgery had little KLK7 and β1 integrin but similarly intense α5 integrin bands in their primary tumor–derived and ascites-derived cells.
Representative immunohistochemistry on serial sections of well-differentiated and poorly differentiated tumors from seven serous EOC patients who had responded or not to chemotherapy (Supplementary Table S1; Materials and Methods) is shown in Fig. 4D. Weak KLK7 and α5/β1 integrin immunostaining on the apical membrane was observed for both well-differentiated tumors that responded to chemotherapy (Fig. 4D, a, d, and g; ID no. 23). For poorly differentiated serous EOC tissues, two different immunostaining patterns were observed. Staining intensities for the responders were similar (KLK7), if not weaker (α5/β1), compared with well-differentiated tumors (Fig. 4D, b, e, and h; ID no. 34). Strikingly, however, the staining was more intense for all three antigens on tissues from the two patients who did not respond to chemotherapy (Fig. 4D, c, f, and i; ID no. 11). No staining was detected in the negative control (rabbit IgG; data not shown).
KLK7-expressing cells grow faster and are more resistant to paclitaxel
MCA formation of EOC cells is associated with increased resistance to certain chemotherapy regimens (9, 28, 29). Our clinical data (Supplementary Fig. S4; Supplementary Tables S1 and S3) and a previous report showed that patients who had shorter progression-free and overall survival time after surgery had increased KLK7 expression (18). We then investigated if KLK7-expressing cells proliferate faster and how they respond to chemoreagents. Clones overexpressing NH2-terminal truncated KLK7-181 grow faster than vector controls at 72 and 96 hours (P < 0.05), whereas KLK7-253 clones seemed to have increased proliferation without reaching significance (Fig. 5A). There was no significant difference in survival after cisplatin treatment (Fig. 5B), but both KLK7-253 and KLK7-181 cells showed less paclitaxel-induced cell death over the dose range of 1 to 100 nmol/L than vector controls (P < 0.05; Fig. 5C). Paclitaxel treatment of MCA/spheroids revealed that MCA-forming KLK7-253 (3E4) and KLK7-181 (5D11) clones are similarly more resistant toward paclitaxel (P < 0.001, 3E4; P < 0.01, 5D11) than the non–MCA-forming KLK7-181 (4A12) or vector cells (Fig. 5D). Although the non–MCA-forming 4A12 clone was more resistant to paclitaxel at 10 nmol/L (P < 0.05), resistance of both 3E4 and 5D11 MCAs is much higher. These data indicate that KLK7-induced chemoresistance to paclitaxel in two-dimensional monolayers is maintained in the MCAs formed in three-dimensional monolayers, suggesting one possible mechanism by which chemoresistance to paclitaxel is induced in vivo, leading to poorer patient outcome in KLK7-expressing EOCs.
Approximately 75% of EOC patients are diagnosed with intra-abdominal dissemination (30). These patients are given platinum-paclitaxel chemotherapy after surgery, but most will develop resistance. Our work (Supplementary Fig. S4; Supplementary Table S3) confirmed that increased KLK7 expression is a hallmark (15) of EOC, particularly the serous subtype and an association between high KLK7 levels with high tumor grade, chemoresistance, and shorter survival time (16–18). We showed that KLK7 directs MCA formation and chemoresistance likely by upregulating α5β1 integrin pathways. These observations indicate that KLK7 has potential not only as a prognostic marker for paclitaxel resistance but also as a therapeutic target in a subset of serous EOC patients who are paclitaxel nonresponders and have elevated KLK7 levels.
Expression of cell adhesion proteins, along with proteases, alters in EOC development (25, 26), although interaction between these molecules is more important during peritoneal invasion. The initiating step of EOC dissemination is that cells exfoliate from the primary site into the peritoneal cavity. KLK7 contributes to skin desquamation via degradation of cell adhesion glycoproteins (21), and thus, KLK7 may also help initiate EOC cell shedding by a similar mechanism. In pancreatic carcinoma, KLK7 cleaves E-cadherin to generate an ectodomain that promotes cell proliferation and invasion (31). Loss of E-cadherin and gain of α5 integrin, a subunit of the major fibronectin receptor, have been related to an aggressive EOC phenotype (6). Interaction between α4β1 integrin and vascular cell adhesion molecule-1 promotes EOC peritoneal metastasis (32); however, we did not see increased α4 integrin in KLK7-expressing clones (data not shown). Collagen-binding integrins, such as α2β1 and α3β1, promote MMP9-dependent shedding of E-cadherin ectodomains (10). We did not see any evidence of E-cadherin processing in our KLK7-overexpressing clones nor changes in E-cadherin, N-cadherin, or cadherin 11/13 mRNA levels (data not shown). However, KLK7-253–expressing cells showed increased levels of cell surface α5/β1 integrins (Fig. 4A) from which we propose that α5β1 integrins may promote the observed cell aggregation in these cells as previously reported (6, 7). Although integrins are classically thought to mediate cell-ECM adhesion, cell-cell adhesion can also be promoted (33). This is evidently the mechanism seen here as no MCA was observed when α5/β1 integrins were inhibited (Fig. 5C). Compact MCAs are associated with an invasive EOC phenotype (34), and multicellular aggregation and disaggregation are regulated by MMP14 (35). Our data imply that other proteases, such as KLK7, are also involved by promoting EOC cell-cell adhesion in suspension via increased levels of α5/β1 integrins. Knocking down α5/β1 integrins reduced peritoneal seeding of ovarian cancer cells (36), and thus, in conjunction with our data, KLK7 blocking might be an alternative therapeutic target for inhibiting serous EOC peritoneal invasion.
Inside the abdominal cavity, single cancer cells and MCAs/spheroids attach to the peritoneum composed of a single mesothelial cell layer over the ECM substrata. Collagen I/IV, vitronectin, and fibronectin are predominant components of the peritoneal ECM targeted by EOC cells (11, 37). KLK7-253 and, in particular, KLK7-181 clones displayed increased attachment to fibronectin/vitronectin (Fig. 3A and B). Interestingly, there is no increased adhesion of KLK7 cells to mesothelial cells grown in two-dimensional culture compared with vector cells (Supplementary Fig. S3B and C); nevertheless, MCAs formed by KLK7 cells attached to mesothelial cell layers and invaded to form cancer cell foci (Fig. 2C). Although mechanisms underlying interactions between MCAs and the peritoneal membrane remain unknown, in both in vitro omentum three-dimensional coculture and in vivo mouse models, EOC cells express and secrete more MMP2 that degrades peritoneal fibronectin and presumably increases tumor invasion (11), which may be one of the mechanisms observed here given the MMP2 activity seen on zymography (Supplementary Fig. S2B). Yet, KLK7 may be another protease that acts similarly; KLK7 degrades fibronectin to contribute to cell shedding (38) but reduces cell adhesion to vitronectin and enhances urokinase-type plasminogen activator receptor shedding in pancreatic carcinoma (39). This may be an additional mechanism in our study, as vitronectin adhesion was not increased to the same degree in KLK7-253 compared with KLK7-181 clones (Fig. 3B). Cytokines, including tumor necrosis factor and interleukin-1β, a KLK7 substrate (40), disrupt the peritoneal integrity by altering the morphology of mesothelial cells (41). This increases the exposure of the ECM (41) and promotes the invasion of attached malignant cells. Collectively, these findings may be reflected in more aggressive clinical phenotypes of patients with high KLK7-expressing and α5/β1 integrin–expressing tumors (Fig. 4B–D) observed here.
MCA/spheroid formation helps single cells overcome anoikis and to survive in an ascites suspension environment (25, 26). Paclitaxel resistance is also related to spheroid formation in three-dimensional suspension cultures (9, 29). Paclitaxel changes integrin expression, including α5β1 integrin (42). Thus, the increased cell survival of KLK7-expressing cells with increased α5/β1 integrins following paclitaxel treatment may reflect the role of these ECM receptors in cellular responses to treatment (43) perhaps by the bcl-2 pathway, as both integrins suppress apoptosis via this mechanism (44). EOC cells exposed to paclitaxel accumulate at G2-M and a sub-G1 apoptotic phase when cultured as monolayers but not as spheroids. This relative sensitivity of two-dimensional monolayer cultures was associated with decreased bcl-X(L) levels after paclitaxel exposure but not in three-dimensional spheroids (9). MCAs are more resistant to chemotherapy than monolayers in vitro (9). However, in our hands, both KLK7-253 and KLK7-181 cells were less responsive to paclitaxel in either two-dimensional monolayers or three-dimensional MCAs (Fig. 5C and D), underscoring the robustness of KLK7-induced paclitaxel chemoresistance and perhaps reflecting increased KLK7/α5/β1 levels seen in patients who were poor responders to chemotherapy (Fig. 4D; Supplementary Table S1).
Northern blot analyses showed two KLK7 mRNA transcripts, KLK7-253 and KLK7-181, in skin (45) and serous EOC cells (15). KLK7-253 encodes the known active serine protease, whereas the exon 1–deleted KLK7-181 variant (15) encodes an NH2-terminal truncated KLK7-181 isoform that is not secreted or a protease. Its concurrently high expression with KLK7-253 in EOC implies that this truncated KLK7-181 isoform has an adjunct intracellular function, although it is not yet detected at the protein level in clinical samples. We also could not determine the relative contribution (whether synergistic or dominant negative) of the KLK7-181 isoform from coculture experiments (Supplementary Fig. S2C), an attempt to mimic the simultaneous presence of both KLK7 variants in EOC. Nevertheless, although all KLK7-253 clones formed large MCAs, suggesting that proteolysis plays a role, one KLK7-181 clone (5D11) also formed spheroids (Fig. 4A), suggesting that proteolysis alone is not sufficient for MCA formation. That the functional KLK7 blocking antibody abolished MCA formation not only of KLK7-253 (4B12) clones but also of KLK7-181 (5D11) clones (Fig. 5C) further supports this notion. KLK7-181 cells showed increased proliferation (Fig. 5A) and increased adhesion to fibronectin/vitronectin (Fig. 3A and B), whereas KLK7-253 clones only showed increased attachment to fibronectin, further underscoring functional differences between both isoforms, which remain to be clarified. Native SKOV-3 cells treated with activated KLK7 enzyme also aggregated faster than those treated with the proteolytically inactive pro-KLK7 (Supplementary Fig. S2A), so presumably there are both proteolytic and nonproteolytic aspects of the functional roles of KLK7 isoforms. Notably, proteolytic-dependent and proteolytic-independent effects induced by prostasin, an extracellular serine protease, have been documented in prostate cancer cells (46); thus, there is precedence for this paradox.
In summary, we have confirmed that high KLK7 levels are indicative of poor outcome for EOC patients and, to our knowledge, for the first time have shown a link with paclitaxel chemoresistance. We have addressed a potential mechanism of action of KLK7 in EOC progression. Our data support the notion that the serine protease KLK7-253 and the NH2-terminal truncated nonproteolytic KLK7-181 isoform promote MCAs, which are paclitaxel resistant. Increased levels of α5/β1 integrins in KLK7-expressing cells and intense staining of both integrins in EOC tissue sections with high KLK7 expression imply their association in promoting EOC cell dissemination and chemoresistance. Collectively, these findings suggest that KLK7 enhances the expression and function of integrin adhesion receptors and that both KLK7-253 and KLK7-181 isoforms function in EOC peritoneal invasion by mediating MCA and paclitaxel chemoresistance.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
We thank Dr. Maria Brattsund (Umeå University, Umeå, Sweden) for the gift of KLK7 antibody and Dr. Michael McGuckin (Mater Medical Research Institute, Brisbane, Queensland, Australia) for access to the primary ovarian cells.
Grant Support: National Health and Medical Research Council (NHMRC) of Australia grants 242220, 390123, and 550523. J.A. Clements is a NHMRC Principal Research Fellow.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.