Matrix metalloproteinase-9 (MMP-9) expression is known to enhance the invasion and metastasis of tumor cells. In previous work based on a proteomic screen, we identified the serpin protease nexin-1 (PN-1) as a potential target of MMP-9. Here, we show that PN-1 is a substrate for MMP-9 and establish a link between PN-1 degradation by MMP-9 and regulation of invasion. PN-1 levels increased in prostate carcinoma cells after downregulation of MMP-9 and in tissues of MMP-9–deficient mice, consistent with PN-1 degradation by MMP-9. We identified three MMP-9 cleavage sites in PN-1 and showed that mutations in those sites made PN-1 more resistant to MMP-9. Urokinase plasminogen activator (uPA) is inhibited by PN-1. MMP-9 augmented uPA activity in the medium of PC3-ML cells by degrading PN-1. Prostate cancer cells, overexpressing PN-1 or treated with MMP-9 shRNA, had reduced cell invasion in Matrigel. PN-1 siRNA restored uPA activity and the invasive capacity. PN-1 mutated in the serpin inhibitory domain, the reactive center loop, failed to inhibit uPA and to reduce Matrigel invasion. This study shows a novel molecular pathway in which MMP-9 regulates uPA activity and tumor cell invasion through cleavage of PN-1. Cancer Res; 70(17); 6988–98. ©2010 AACR.
Matrix metalloproteinase-9 (MMP-9) has been long recognized as a key enzyme for the proteolytic degradation of extracellular matrix (ECM) during tumor invasion and metastasis (1). Its expanding roles include regulating cancer progression, activating angiogenesis, and recruiting macrophages or other bone marrow–derived myeloid cells to the preexisting metastatic niche (1, 2). These varied functions of MMP-9 have made it an extremely promising target for preventing metastasis in cancer patients (3, 4). However, in the last decade, clinical trials of MMP inhibitors have failed to produce breakthroughs (3). This may be attributed to the lack of specificity of the inhibitors used, with more global MMP inhibition resulting in unacceptable side effects. If the particular proteolytic substrates of this enzyme could be identified, then potentially more precise inhibition profiles could be targeted.
Besides cleaving ECM components such as collagens and fibronectin, MMP-9 can degrade many noncollagenous substrates (1). MMP-9 cleavage alters the biological activity of chemokines, and its activity can result in the shedding of cell surface receptors (5). These molecules influence many biological and pathologic functions involved in cell adhesion, proliferation, angiogenesis, cell invasion, and metastasis (5, 6). MMP-9 has long been known to enhance cancer cell invasion; however, the underlying molecular mechanisms of how MMP-9 regulates tumor cell invasion and metastasis remain poorly understood (1, 6). To identify MMP-9 targets and potentially unveil new molecular mechanisms, we previously performed a label-free quantitative proteomics analysis to identify MMP-9 substrates in cancer cells (7). A number of novel MMP-9 targets were revealed, including the ECM protein protease nexin-1 (PN-1; ref. 7).
PN-1, also called serpin E2 or glial-derived nexin, belongs to the serpin family of regulatory proteins (8). It is a serine protease inhibitor known to potently and irreversibly inhibit several proteases, including thrombin, urokinase plasminogen activator (uPA), tissue plasminogen activator (tPA), and trypsin (9, 10). Many of these proteins are involved in tissue remodeling and tumor invasion (11). Although many serpins are found in plasma, PN-1 is found predominantly in tissues and platelets (12, 13). PN-1 is a 43-kDa secreted protein and can be produced by a multitude of cell types, including endothelial cells, fibroblasts, tumor cells, smooth muscle cells, and astrocytes (14–16). PN-1 is present in the extracellular space where it can bind to glycosaminoglycans (17) and collagen IV (18).
Notably, PN-1 contains a reactive center loop (RCL) region at its COOH terminus, which is the critical structural feature shared by most serpins and is necessary for inhibitory activity (19, 20). Serpins are usually present in a metastable state with the RCL region exposed. Upon contact with the target protease, the RCL is cleaved, leading to a covalent linkage between a COOH-terminal portion of the cleaved serpin and the target protease. The protease-serpin complex then reverts to a more stable and energetically favorable state, retaining the covalent, inhibitory linkage to target protease (20). This dramatic conformational change is the structural basis of the inhibitory effect of serpins against most proteases (19, 20).
In mammals, extracellular serpin-protease complexes are rapidly cleared from circulation through low-density lipoprotein receptor-related protein (LRP)–mediated endocytosis (21). Serpin-protease complexes bind to the LRP and are internalized, thus triggering subsequent signaling events and finally resulting in transport to the lysosomes (22). For example, PN-1-thrombin and PN-1–uPA complexes are internalized through the LRP (23).
PN-1 mRNA overexpression has been identified in head and neck squamous cell cancers (24) and in colon carcinoma (25). No functional role was described for this overexpression. Confirmation of altered protein levels was also not provided (24, 25). The regulation of PN-1 itself remains to be characterized. In our previous study, we found that recombinant MMP-9 cleaved PN-1 in a dose- and time-dependent manner. We also determined the possible MMP-9–dependent cleavage sites using a proteomics approach (7). Our current study confirmed that PN-1 is a direct target of MMP-9; thus, its inhibition of proteases, such as uPA, is under the control of MMP-9. Although PN-1 does not have a clearly defined role in cancer biology, our results suggest that it may have a significant impact on tumor invasion.
Materials and Methods
All animal experiments were performed in accordance with UK Home Office regulations. MMP-9–deficient C57/B6 mouse strains were kindly provided by Dr. Ghislain Oppendernakker (University of Leuven, Leuven, Belgium; ref. 26). After sacrifice, organs were harvested, rinsed in PBS, and stored at −80°C. The frozen organs were individually weighed, smashed, homogenized, and lysed in RIPA lysis buffer (Pierce) containing protease inhibitor cocktail (Roche), 4 mmol/L dithiothreitol, and 0.2 mmol/L phenylmethylsulfonyl fluoride at 4°C for 1 hour. The total protein concentration was determined by BCA assay (Pierce).
Plasmids and mutagenesis
pcDNA3-PN-1 plasmid was a kind gift from Dr. Peter Andreasen's laboratory (Aarhus, Denmark). pcDNA3-His-PN-1 was generated by incorporating 10× His at the COOH terminus of PN-1. Point mutations were generated based on pcDNA3-PN-1 through a site-directed mutagenesis kit (Invitrogen), and each was confirmed by DNA sequencing. The sequence pairs of primers are summarized in Supplementary Table S1.
Cell culture and treatment
A PC-3ML cell line (prostate cancer cell derived from PC-3) was obtained from Dr. Mark Stearns (Drexel University, Philadelphia, PA; ref. 27) and maintained as previously described (7, 27). A MMP-9 stable knockdown PC-3ML cell line was established by us as previously described (7). Panc1 (CRL1469), PC-3 (CRL1435), and TRAMP-C2 (CRL2731) cell lines were from the American Type Culture Collection. HT1080 cells were maintained as previously described (28). All cell lines were regularly tested to ensure the absence of Mycoplasma contamination (MycoAlert, Lonza). A new stock vial of each cell line was thawed every 3 to 4 months, and cell morphology was regularly checked to ensure that there was no cross-contamination of cell lines. Where indicated, PN-1 plasmids were transfected into cells using FuGENE 6 transfection reagent (Roche). Cell conditioned medium was collected and concentrated as previously described (7).
For siRNA experiments, predesigned siRNA against uPA, PN-1, MMP-9, and siRNA negative control oligos were from Ambion. Each siRNA, at 10 or 20 nmol/L, was transfected into cells using the siPORT NeoFX transfection agent (Ambion). After 48 hours, cells were washed and the medium was replaced with serum-free medium. Conditioned medium was collected 24 hours later. Bone marrow–derived cells (BMDC) were flushed from femurs of C57/B6 or MMP-9 KO mice and cultured as previously described (29).
Immunoblotting and uPA activity assay
Whole cell lysates were extracted as described before (7). Conditioned medium was concentrated and normalized according to cell numbers and intracellular proteins. Immunoblots were performed as previously described (7). The following antibodies were used: anti-human PN-1, anti-mouse PN-1 (R&D Systems), anti-uPA antibody (American Diagnostica), anti–MMP-9 (Abcam, R&D Systems), and anti–β-actin (Santa Cruz, Abcam). uPA activity was measured by a uPA activity assay kit (Chemicon). In brief, concentrated cell conditioned medium was mixed with assay buffer and incubated with a chromogenic substrate in 96-well plates at 37°C for 3 to 6 hours. The absorbance was read at OD405, and the activity (units) was extrapolated from a standard curve.
Matrigel invasion assay
PC-3ML or MMP-9 KD PC-3ML cells (1 × 105), subjected to siRNA or transfection treatment for 24 hours, were seeded on BD Biocoat Growth Factor Reduced Invasion Inserts (BD 354483). After 48 hours, invaded cells were stained with Hoescht, and nine fields from triplicate experiments were counted.
All statistical analyses and plots were performed with Prism GraphPad 5.0 software.
PN-1 is cleaved by MMP-9 in tumor cells
In our previous work, we identified PN-1 as a putative target for MMP-9 proteolysis (7). We then showed that incubation of purified MMP-9 with recombinant PN-1 protein in solution resulted in PN-1 degradation with demonstrable cleavage intermediates (ref. 7; again confirmed in Supplementary Fig. S1). Now, additional experiments were performed to determine whether PN-1 is a natural target of MMP-9 in mammalian cells and animals. PN-1 is secreted by prostate carcinoma PC-3 cells and their derivative, PC-3ML (7). After stable knockdown of MMP-9 by shRNA in PC-3ML cells (7), PN-1 levels increased in conditioned medium (Fig. 1A), consistent with the hypothesis that PN-1 is a substrate for MMP-9. Moreover, increased PN-1 expression was found in additional cell lines after MMP-9 downregulation (Supplementary Fig. S2). PN-1 expression induced by transfection of a PN-1 expression vector was greater after transfection into cells with downregulated MMP-9 (MMP-9 KD) compared with control cells (Fig. 1A).
MMP-9 does not recognize substrates through recognition of an exact linear sequence; instead, substrates must fit into a groove adjacent to the catalytic site (30). Previously, we identified numerous cleavage sites of PN-1 by MMP-9 through mass spectrometry after extensive digestion (7). More limited digestions allowed the identification of preferred cleavage sites (7). Based on the size of these cleavage fragments as determined by tandem mass spectrometry, four putative cleavage sites were identified (Fig. 1B, arrowheads). Only one of these sites at position 58/59 contained a sequence that corresponded to the major MMP-9–specific cleavage consensus sequence P-X-X-|Hy-S/T (Hy: hydrophobic amino acid) as previously described (31). Mutations of the proline at position 55 (P55A) or position 58 from a T to an I (I58T) were generated in PN-1 expression vectors. Transfection of P55A into PC-3ML cells and MMP-9 KD PC-3ML cells did not lead to accumulation of PN-1, suggesting that this mutation had destabilized the protein. However, transfection of the I58T mutation led to increased amounts of PN-1 in the conditioned medium of wild-type (WT) cells compared with the more moderate accumulation of WT PN-1 (Fig. 1C). This effect was dependent on MMP-9 because PN-1 levels after transfection into MMP-9 KD PC-3ML cells were equivalent for both (Fig. 1C). These data are consistent with position 58 of PN-1, affecting susceptibility to MMP-9 cleavage. Similar results were obtained with the PN-1 mutation I107T (Fig. 1C). Other mutations at positions 58, 107, and 368 all led to accumulation of PN-1 in PC-3ML cells (Fig. 1C). Thus, these three sites seem to be dominant cleavage sites in PN-1 for MMP-9. Collectively, of 21 mutations we generated and tested, mutations including I58T, I58T+I107T, I58T+P368H, and I58T+P368Y seemed to be substantially more resistant to MMP-9 cleavage than the WT (Fig. 1C).
PN-1 is targeted by MMP-9 in vivo
The hypothesis that PN-1 levels are controlled through degradation by MMP-9 leads to the prediction that PN-1 levels should be higher in MMP-9–deficient mice. The expression of PN-1 was screened in homogenates made from different organs of WT or MMP-9 KO mice (Fig. 2A). PN-1 was most abundantly expressed in seminal vesicles, consistent with previous reports (32). We also readily detected PN-1 in the lung, liver, spleen, and prostate. The levels of PN-1 were greatly increased in the lung, prostate, and pancreas and slightly increased in the seminal vesicles of the MMP-9 KO mice compared with the WT (Fig. 2A). With longer exposure, a PN-1 degradation fragment was apparent in the lysates from the seminal vesicles and the prostate (Fig. 2A, iii), consistent in size with cleavage at amino acids 107 or 368. In contrast, in the liver and spleen, PN-1 was equally expressed in both WT and MMP-9 KO mice, suggesting that other enzymes might be more important for PN-1 degradation in these organs (Fig. 2A). Thus, in the prostate, lung, and pancreas, MMP-9 would seem to play a major role in degradation of PN-1. There was more PN-1 in conditioned medium of BMDCs from MMP-9 KO than that from WT mice (Fig. 2B).
We examined the expression of PN-1 in situ by immunohistochemistry. The images taken from seminal vesicles, prostate, and pancreas showed the accumulation of PN-1 in MMP-9 KO mice (Fig. 2C). In particular, staining for PN-1 was mainly in the ECM surrounding the glands of the seminal vesicles, pancreas, and prostate (Fig. 2C). MMP-9 staining is shown in Supplementary Fig. S3. Thus, PN-1 is localized to the ECM and its amount is regulated by MMP-9. Taken together, these results indicate that PN-1 seems to be degraded by MMP-9 in cell culture and in mice in the lung, pancreas, and prostate.
MMP-9 regulates uPA activity through cleavage of PN-1
uPA is an inhibitory target of PN-1 (9). Increased expression of PN-1 due to transfection with an expression vector led to decreased uPA activity (Fig. 3A). Downregulation of MMP-9 led to increased PN-1 amounts and to inhibition of uPA activity (Fig. 3A). Both manipulations together led to even lower uPA activity (Fig. 3A). To determine whether MMP-9 has a direct impact on uPA level and activity, we incubated these two proteins together. uPA remained intact and active after incubation with MMP-9 (Fig. 3B and C). Thus, the effect of MMP-9 on increased uPA activity must be indirect and could be mediated by MMP-9 degradation of PN-1.
PN-1 is known to form a SDS-stable covalent complex with its target proteases such as uPA (23). Here, we showed that uPA was present in the PN-1 complex isolated using the His-tag engineered into PN-1 (Fig. 3D). In addition, complexes associated with PN-1 also contained MMP-9, supporting the hypothetical regulation axis of MMP-9/PN-1/uPA.
MMP-9 and PN-1 regulate tumor cell invasion by targeting uPA
As uPA is widely recognized to be integral to migration, metastasis, and tumor dormancy (11), we examined the effect of MMP-9 and PN-1 on uPA-dependent tumor cell invasion. The MMP-9 KD PC-3ML cell line had a marked reduction in Matrigel invasion compared with WT PC-3ML cells (Fig. 4A), consistent with the established role of MMP-9 in promoting cell invasion. Importantly, PN-1 overexpression in PC-3ML cells significantly reduced invasiveness. When PN-1 was overexpressed in MMP-9 KD PC-3ML cells, invasion was even more substantially inhibited (Fig. 4A). Therefore, MMP-9 and PN-1 can oppose each other in the regulation of tumor cell invasion.
To understand whether MMP-9 and PN-1 induce this effect through targeting uPA, we knocked down the expression of uPA using siRNA (Fig. 4B, ii and iii). uPA downregulation greatly reduced invasion (Fig. 4B, iv). Ectopic expression of PN-1 also inhibited uPA activity and invasion. With the effector molecule uPA depleted by siRNA, overexpressed PN-1 did not further reduce invasion (Fig. 4B, iv). This result suggests that PN-1 inhibits invasion through inhibition of uPA and that MMP-9 can enhance invasion through degradation of PN-1.
We tested whether MMP-9–resistant PN-1 mutants retained their inhibitory effect against uPA. By measuring uPA activity in cell conditioned medium, we found that each of the mutant forms tested was capable of uPA inhibition (Fig. 5A). The expression levels of uPA in the conditioned medium remained unchanged after transfection with the expression vectors for PN-1 and its mutant forms (Fig. 5A). Because the inhibitory function of serpins depends on conformational changes at many regions of the molecules, mutagenesis was based on structural data to attempt to best preserve the overall structure of PN-1 (33). Based on homology to PAI-1, the initial cleavage of PN-1 triggered by the proteases would be anticipated to be between R365 and S366 (34). R365P and S366P mutations seemed to prevent inhibition of uPA activity (Fig. 5B).
The effects on invasion of various PN-1 mutations were then tested. MMP-9–resistant mutations I58T+I107T and I58T+P368H displayed similar inhibition of cell invasion as WT PN-1 (Fig. 5C). Unexpectedly, despite inhibiting uPA activity in the same fashion as WT PN-1, I58T mutant was even more effective in reducing the cell-invasive capacity (Fig. 5C). In contrast, R365P and S366P had greatly reduced ability to inhibit invasion, in accordance with their reduced uPA inhibitory activity (Fig. 5D). Mutations in H48A and D49A were reported to impair PN-1 binding to LRP (23, 35). These PN-1 mutants retained the ability to inhibit uPA (Fig. 5B, ii) and inhibited tumor cell invasion to the same extent as WT PN-1 (Fig. 5D). These data indicate that the inhibitory activity of PN-1 is crucial for its effect on tumor cell invasion.
PN-1 is essential for MMP-9–regulated tumor cell invasion
An important consideration is whether PN-1 is a key mediator that links MMP-9 and uPA. To further address this question, we used PN-1 siRNA to reduce the expression of PN-1 in cells [confirmed by quantitative reverse transcriptase-PCR (qRT-PCR)] and in conditioned medium (confirmed by immunoblot; Fig. 6A). Importantly, downregulation of PN-1 led to increased uPA activity (Fig. 6B). Downregulation of PN-1 led to similar increases in uPA activity in cells deficient in MMP-9 (Fig. 6B), suggesting that MMP-9 has little direct effect on uPA activity itself. PC3-ML cell invasion was enhanced by downregulation of PN-1 (Fig. 6C). Likewise, reducing PN-1 led to enhanced uPA activity and invasion in cells with downregulated MMP-9 (Fig. 6C). Taken together, the data indicate a novel molecular pathway in which MMP-9 regulates tumor cell invasion through the selective targeting of PN-1, thus affecting uPA activity (Fig. 6D).
In this study, we identified a novel pathway through which MMP-9 regulates tumor cell invasion through degradation of PN-1. Consistent with the identification of PN-1 as an enzymatic substrate of MMP-9, PN-1 dramatically accumulated in cell lines with downregulated MMP-9 and in many organs from mice genetically deficient in MMP-9 (Figs. 1 and 2). Mutation of any of the three predominant sites of MMP-9 cleavage in PN-1 led to resistance to degradation (Fig. 1C). Next, we showed that MMP-9 regulation of uPA activity is mediated by control of PN-1 levels by degradation (Fig. 3). Finally, we showed that negative regulation of tumor cell invasion by PN-1 was mediated by PN-1 protease inhibition because mutations in PN-1 that impaired its inhibitory function abrogated its ability to inhibit migration (Figs. 4–6). These results lay the foundation for a new concept arguing that MMP-9 controls the activity of uPA by cleavage of PN-1.
The invasive capacity of tumor cells is an important aspect of tumor progression and a major factor of cancer morbidity and mortality (36). Importantly, these data reveal a protective role of PN-1 against invasion by a prostatic carcinoma cell line. PN-1 overexpression significantly prevents the invasion of tumor cells, analogous to its inhibition of thrombin-dependent migration by vascular smooth muscle cells (37). The ECM in the prostate proved to be a site of high levels of endogenous PN-1 in the mouse (Fig. 2A and C). Other serpins have been well documented in cancer invasion and metastasis, including PAI-1, which has paradoxical roles in both stimulating and inhibiting tumor progression (38). Whether PN-1 has a similar proangiogenic role as PAI-1, in which inhibition of plasmin generation reduces fasL release and endothelial apoptosis (39), remains to be investigated. Maspin, a tumor suppressor, is homologous to serpins but does not have demonstrable protease inhibitor activity (40). PN-1 targets uPA and thrombin, both implicated in tumor progression (41).
uPA is an independent prognostic marker for poorer prognosis in breast, gastric, pulmonary, prostate, and ovarian cancer (42). The uPA-uPAR system has great impact on prostate cancer progression and metastasis (36). uPA is secreted as a proenzyme and efficiently converts inactive plasminogen into the active serine protease plasmin by binding its membrane receptor uPAR. In addition to triggering a cascade of enzymatic activity, binding to uPAR triggers a signaling cascade including a consequent activation of extracellular signal-regulated kinase (36). Downregulation of either uPA or uPAR has been shown to inhibit prostate cancer cell invasion and its tumorigenicity in vivo (43). Some uPAR effects are due to its interactions with integrins, as especially documented for α3β1 and α5β1 (44). PAI-1 or PN-1 binding to uPA reduces these interactions and alters integrin recycling (45). Thus, there are several mechanisms through which altered uPA activity might alter migration. Thrombin is also known to regulate tumor cell invasion and metastasis (41). However, prostate cancer cells and PC-3 in particular produce little thrombin endogenously (46). Thus, in this system, PN-1 inhibition of thrombin is less likely to play a significant role than uPA does. In this study, we did not find alterations in uPA levels after changes in PN-1 expression (Fig. 5A). In our previous research (7), we found significant changes of uPA and tPA levels after MMP-9 downregulation by proteomics, which may be due to the changes of PN-1 levels (>10-fold). Decreased uPA could result from LRP-mediated clearance of uPA–PN-1 complexes (Supplementary Table S2).
We have shown that MMP-9 can regulate uPA through cleavage of PN-1. This regulation may be more complex. In THP-1 monocytes, uPA was shown to activate MMP-9 expression at the transcriptional level (47). Moreover, a recent report suggests that PN-1 can induce MMP-9 expression in breast cancer cells when it is complexed with its target proteases such as thrombin and tPA (16). However, even after the addition of large concentrations of recombinant uPA and PN-1, the increase in MMP-9 levels was modest (∼2-fold; refs. 16, 47). Therefore, we also tested the effect of additional PN-1 on MMP-9 production in PC-3ML cells (Supplementary Fig. S4). A slight elevation of MMP-9 (<2-fold) occurred despite greatly increased PN-1 (Supplementary Fig. S4). It is possible that there is a regulatory feedback loop in which PN-1 might induce MMP-9 that would then lead to decreased PN-1, possibly resulting in relatively minor changes in the overall level (Fig. 6D). The regulation may be a cell type–specific response to PN-1. More importantly, in vivo, the situation is more complicated because the feedback loops are not only between the same cell type but also occurring with host stromal cells and PN-1 in the stroma itself.
Clearly, the interaction between tumor cells and the stromal environment is crucial during tumor progression and metastasis. This study has shown the regulatory axis along MMP-9, PN-1, and uPA in tumor cells. Because MMP-9, PN-1, and uPA are all secreted into extracellular spaces in tissues, they closely interact with ECM components. It will be crucial to assess the effects of PN-1 in the context of in vivo assays of tumorigenicity and metastasis to determine whether these targets represent novel entry points for pharmacologic intervention in the future.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
We thank Dr. Sally Hill and Spela Ferjancic for experimental assistance, Dr. Ghislain Opdenakker for providing MMP-9–deficient mice, Dr. Mark Stearns for PC-3ML cells, and Dr. Peter Andreasen for pcDNA3-PN-1 plasmid.
Grant Support: Cancer Research UK.
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.
- Received January 21, 2010.
- Revision received May 17, 2010.
- Accepted June 16, 2010.
- ©2010 American Association for Cancer Research.