Cancer Cell Biology
EMMPRIN-induced MMP-2 activation cascade in human cervical squamous cell carcinoma
Version of Record online: 19 JAN 2006
Copyright © 2006 Wiley-Liss, Inc.
International Journal of Cancer
Volume 118, Issue 12, pages 2991–2998, 15 June 2006
How to Cite
Sier, C. F.M., Zuidwijk, K., Zijlmans, H. J.M.A.A., Hanemaaijer, R., Mulder-Stapel, A. A., Prins, F. A., Dreef, E. J., Kenter, G. G., Fleuren, G. J. and Gorter, A. (2006), EMMPRIN-induced MMP-2 activation cascade in human cervical squamous cell carcinoma. Int. J. Cancer, 118: 2991–2998. doi: 10.1002/ijc.21778
- Issue online: 27 MAR 2006
- Version of Record online: 19 JAN 2006
- Manuscript Accepted: 22 NOV 2005
- Manuscript Received: 12 AUG 2005
- matrix metalloproteinase;
- in situ zymography;
Tumor progression and recurrence of cervical cancer is associated with upregulation of matrix metalloproteinase 2 (MMP-2). We evaluated the location, origin and activity of MMP-2 in cervical squamous cell carcinomas in comparison with MT1-MMP (MMP-14), TIMP-2 and extracellular matrix metalloproteinase inducer (EMMPRIN). Positive immunostaining for MMP-2 in malignant cells was detected in 83% of the patients. Two patterns of tumor cell MMP-2 staining were observed: either homogenous in all tumor cells or confined to the cells neighboring the stroma (tumor-border staining pattern, TBS). Fluorescence in situ zymography showed active MMP-2 mainly around tumor nodules displaying TBS. The MMP-2 staining of TBS tumors correlated significantly with the presence of TIMP-2 and MT1-MMP, proteins involved in docking MMP-2 to the cell surface and essential for MMP-2 activation. In situ mRNA hybridization in TBS tumors demonstrated more abundant presence of MMP-2 mRNA in neighboring myofibroblasts than in the adjacent tumor cells. Moreover, the TBS MMP-2 pattern correlated with the presence of EMMPRIN (p = 0.023), suggesting that tumor cells induce MMP-2 production in nearby stromal cells. This pro-MMP-2 could subsequently be activated on tumor cells via the presence of MT1-MMP and TIMP-2. The biological relevance of this locally activated MMP-2 was underscored by the observation that only the TBS pattern of MMP-2 significantly correlated with decreased survival. In conclusion, the colocalization of EMMPRIN, MT1-MMP and TIMP-2 in human cervical carcinomas seems to be involved in a specific distribution pattern of tumor cell bound MMP-2, which is related with local proteolytic activity and therefore might be associated with worse prognosis of the patients. © 2006 Wiley-Liss, Inc.
Cervical cancer is the second most common cancer among women worldwide, primarily caused by human papillomavirus. About 80% of primary cervical cancers arise from preexisting squamous dysplasia.1 Upregulation of certain members of the matrix metalloproteinase family, i.e. MMP-1, MMP-2, MMP-9 and MT1-MMP (MMP-14) has been associated with malignancy in cervical squamous cell carcinoma (CSCC).2, 3, 4 The ability of MMPs to degrade various structural extracellular matrix (ECM) components, as shown in physiological processes such as embryonic development, angiogenesis and wound healing suggest a similar role for these enzymes in malignancies. The importance of MMPs in cancer was shown in different model systems5, 6 in which the invasion and metastasis of malignant cells was reduced by blocking/inhibition of MMPs. Although upregulation of several MMPs has been described in cervical cancer,4, 7, 8, 9, 10, 11 a clear correlation with survival, as has been found for other proteinases involved in invasion, has not been established yet.12 The mere presence of MMPs, however, does not imply proteolytic activity: MMP activity is regulated at multiple levels, including transcription, translation, posttranslational modification and the presence of inhibitors. Except for the membrane bound MMPs (MT-MMPs), MMPs are secreted into the pericellular space as inactive proenzymes. Although specific receptors for MMPs are not found yet (possible exception MMP-1313), several secreted MMPs have been found to dock to the cell surface via various membrane bound proteins. Association to the cell surface has been reported for MMP-2, linked via a complex with MT1-MMP and TIMP-2.14 More recently, docking of MMP-1, MMP-7 and MMP-9 has been described via CD147,15 CD44, TM4SF, RECK and LRP, respectively.16 Association to the cell surface has found to be important for MMP activity, because it stimulates activation of MMPs by other proteinases and offers protection against inactivation by inhibitors. The commonly used zymographic technique to estimate the activities of MMP-2 and MMP-9 in tumor homogenates provides an estimation of activity, but does not take into account the localization or the cell types involved. Immunohistochemical studies do not distinguish activated MMPs, nor are they able to clarify the origin of the MMPs. MMPs found on cervical malignant cells may be produced by neighboring stromal cells, e.g. myofibroblasts, as has been postulated for other tumor types.17, 18 Indeed, tumor cells have been found to be able to induce MMP production by stromal cells via soluble and membrane bound factors, followed by tumor cell associated activation.19 Extracellular matrix metalloproteinase inducer (EMMPRIN) is a 58 kDa protein found on the surface of tumor cells and has recently been shown to stimulate fibroblast and endothelial cell MMP production.20, 21
We hypothesized that the presence of MMP-2 activity in the interface zone between stromal and tumor cells would be most advantageous for the tumor and disadvantageous for the prognosis of the patient. Previous studies reported indeed that in some cases of CSCC invasive edges of tumor cell nodules stained more intense for MMP-17 and MMP-2.8, 22 In addition, a recent study showed that high levels of active MMP-2, as determined by gelatin zymography, were associated with the stage of the tumors, lymphatic spread and tumor recurrence.8 Therefore, we decided to determine the presence of MMP-2 in 30 cases of CSCC with special focus on the edges of the tumor nodules. The expression and localization was verified with in situ hybridization and immunofluorescence double staining, respectively. Second, we correlated the MMP-2 staining with the presence of MT1-MMP and TIMP-2, both implicated in the cell surface activation of MMP-2,23 in the same CSCC. Third, we determined the activity profile for MMP-2 in comparison with the immunohistochemical occurrence using fluorescence in situ zymography. Fourth, the presence of MMP-2 was correlated with tumor cell associated EMMPRIN, an inducer of fibroblast MMP. Finally, the immunohistochemical staining patterns of MMP-2 were analyzed with respect to the survival of the patients. The results of our study indicate that mere presence of MMP-2 alone does not correlate to the biological behavior of cervical cancer. The colocalization of MMP-2 with MT1-MMP and TIMP-2 and hence the activity state of MMP-2 at specific locations, i.e. at the interface zone between stromal and cervical tumor cells, is more relevant. Furthermore, cervical tumor cells could indeed be involved in the induction of MMP-2 in neighboring stroma cells via EMMPRIN.
Patients, material and methods
Formalin-fixed, paraffin-embedded tissue blocks from 30 patients with CSCC squamous cell carcinomas of the uterine cervix, who underwent a radical hysterectomy with lymphadenectomy between 1985 and 1995, were randomly chosen from the archives of the Department of Pathology, Leiden University Medical Center.24 None of the patients had received any therapy before surgery. For immunohistochemistry, paraffin blocks containing a representative part of the tumor were used. Slides of all tumors were reviewed using conventional histologic sections stained with hematoxylin and eosin. The tumors were staged as FIGO class Ib (n = 20), IIa (n = 8), IIb (n = 1) and IIIb (n = 1). From most tumors, frozen material was available for MMP-2 activity determination.
Immunohistochemistry was performed on 4 μm sections using aminopropyl-ethoxysilane-coated slides. Paraffin sections were deparaffinized and rehydrated, and endogenous peroxidase was quenched with 0.3% H2O2 in methanol for 20 min. The primary antibodies used are listed in Table I. Incubations were performed overnight at room temperature. Phosphate-buffered saline (PBS) containing 1% bovine serum albumin was used as a diluent for all antibodies. Washing between incubations was performed 3 times for 5 min each in PBS. Biotinylated rabbit anti-mouse or goat anti-rabbit immunoglobulins, followed by biotinylated horseradish peroxidase (HRP)-streptavidin complex (both Dako, Glostrup, Denmark) were applied for 30 min each. To visualize immune complexes, a 0.05% solution of diaminobenzidine (Sigma, St. Louis) containing 0.0018% H2O2 in a 0.05 M Tris-HCl buffer (pH 7.6) or alternatively Nova Red (Vector Laboratories, Burlingame, CA) was applied, resulting in brown, respectively, red staining. Citrate antigen retrieval was used wherever indicated by the manufacturers (Table I). Mayer's hematoxylin was used for counterstaining of the slides. Appropriate positive control sections were stained simultaneously and staining the similar sections omitting primary antibodies was used as negative control. The specificity of the staining patterns of the polyclonal rabbit antibodies (anti-MMP-9 and anti-MT1-MMP; both from TNO-PG) was confirmed by similar results on frozen sections from the same patients and in case of MMP-9 also by similar staining on frozen tissue using commercial available monoclonal antibody (clone GE-213, NeoMarkers, Fremont, CA).
|MMP-2||Mouse, mAb||(pro)MMP-2||Citrate||1/200||NeoMarkers, Fremont, CA|
|MMP-9||Rabbit, pcAb||(pro)MMP-9||Citrate||1/500||TNO-PG, Leiden, The Netherlands|
|MT1-MMP||Rabbit, pcAb||MT-1MMP||None||1/1000||TNO-PG, Leiden, The Netherlands|
|TIMP-2||Mouse, mAb||TIMP-2||Citrate||1/400||NeoMarkers, Fremont, CA|
|EMMPRIN||Rabbit, pcAb||EMMPRIN||Citrate||1/150||Zymed, San Francisco, CA|
|CD31||Mouse, mAB||Endothelial cells||Citrate||1/100||Alexis-Benelux, The Netherlands|
|CD68||Mouse, mAb||Macrophages||Citrate||1/100||Dako, Glostrup, Denmark|
|EpCAM||Mouse, mAb||Epithelial cells||None||1/100||Centocor, Leiden, The Netherlands|
|Desmin||Mouse, mAb||Muscle cells||Citrate||1/100||Sanbio, Uden, The Netherlands|
|Vimentin||Mouse, mAb||Mesenchymal cells||Citrate||1/100||Sanbio, Uden, The Netherlands|
|α-smooth muscle actin||Mouse, mAb||Myofibroblasts, smooth muscle cells||Citrate||1/1000||Progen, Heidelberg, Germany|
Staining for MMP-2, MT1-MMP, TIMP-2 and EMMPRIN in tumor cells was scored semiquantitatively, according to a system proposed by Ruiter et al.25 As final score, the mean result of 2 independent individuals (C.S. and K.Z.) was used. The percentage of tumor cells that stained positive were scored as follows: 0, absent; 1, 1–5% sporadic; 2, 6–25% local; 3, 26–50% occasional; 4, 51–75% majority and 5, 76–100% large majority. The intensity of tumor cell staining was scored as: 0, no staining; 1; weak staining; 2, moderate staining and 3, intense staining. A total score was calculated by adding the scores for percentage and intensity, resulting in values from 0 to 8: 0, negative; 2–4, weak; 5–6, moderate and 7–8 strong.
Immunofluorescence double staining
The sections were incubated with mAbs against MMP-2 (IgG1) and Ep-CAM (IgG2a) in appropriate dilutions in PBS with 1% BSA (PBS-BSA) for 1 hr in a moist chamber, washed in PBS-BSA and incubated with Alexa Fluor 488- and 546-conjugated goat anti-mouse IgG1 and IgG2a, respectively, (Molecular Probes, Leiden, The Netherlands) diluted in PBS-BSA. The sections were mounted in Vectashield (Vector Laboratories). Background immunofluorescence was assessed by using irrelevant isotype matched control antibodies or omitting the primary antibodies. A Zeiss LSM 510 confocal microscope equipped with argon (488 nm) and He/Ne (543 nm) lasers and a 25× objective were used to obtain the images.
RNA in situ hybridization
RNA in situ hybridization was performed, as described earlier.26 In brief, tissue paraffin sections (4 μm thick) were rehydrated, digested with proteinase K (5 μg/ml, 10 min, 37°C) and postfixated with 4% formaldehyde. The sections were hybridized at 55°C for 16h with 100 ng/ml digoxigenin labeled riboprobe (MMP2, kindly provided by G. Murphy and P. Birembaut27) in 50% deionized formamide, 250 μg/ml salmon sperm DNA, 25 μg/ml tRNA, 10 mM DTT, 1× Dernhardt's solution, 10% dextran sulphate and 2× SSC. Sections were washed twice in 2× SSC at 55°C, once with 20 mM β-mercaptoethanol in 0.1× SSC at 55°C and treated with 2 U/ml RNase T1 for 30 min at 37°C. The DIG-labeled hybrids were visualized with mouse anti-DIG (Sigma, St. Louis, MO), rabbit anti-mouse and APAAP (both DAKO, Glostrup, Denmark), respectively, in 1% blocking reagent (Roche, Basel, Switzerland) in 0.15 M NaCl and 0.1 M Tris pH 7.5. Transcripts were visualized with 5-bromo-4-chloro-3-indolylphosphate/nitro-blue tetrazolium (BCIP/NBT) (Roche), 1% levamisol (Vector Laboratories) and 10% polyvinylalcohol (PVA) at 50°C for 3 hr, resulting in blue staining of mRNA. Adjacent tumor slides, hybridized with MMP-2 sense probes, were included as negative controls and did not show staining. Colorectal cancer tissue served as a positive control.
In situ zymography
Gelatinolytic activity was demonstrated in unfixed cryostat sections from the same patients, using DQ-gelatin (EnzChek, Molecular Probes, Eugene, OR) as a substrate.28 Cryostat sections (5 μm) were air-dried for 10 min. DQ-gelatin was dissolved in water (1 mg/ml) and 1:10 diluted in 1% (w/v) low gelling agarose (Gibco-BRL) in PBS containing 0.5 μg/ml propidium iodide (PI) to counterstain the nuclei. The mixture was applied to the sections, followed by a coverslip. After gelling of the agarose at 4°C, the sections were incubated at RT for 1–3 hr. Placenta tissue with or without the addition of 20 mM EDTA in the agarose mixture was used as negative and positive control, respectively. Fluorescent gelatinolytic activity was detected using a confocal microscope (green) with excitation at 460–500 nm and emission at 512–542 nm. PI nuclear staining (red) was detected with excitation at 540–580 nm and emission at 608–682 nm.
Spearman correlations between parameters, survival curves, log rank analysis and 2D unsupervised hierarchical cluster analysis were performed using the SPSS 10.0 software package (SPSS, Chicago Illinois), considering p values ≤0.05 significant.
Figure 1 shows representative immunohistochemical stainings for MMP-2 in CSCC and adjacent normal cervix. Staining of MMP-2 in tumors was localized to the cytoplasm and observed in peritumoral stroma cells, endothelial cell and tumor cells. Normal cervical epithelial cells, present in most of the tissue slides, did not stain, but in some cases, stromal cells were positive (Fig. 1c). Table II shows an overview of MMP-2 staining intensities, the percentages of stained malignant cells and an overall score for MMP-2 in 30 CSCC patients. MMP-2 staining in tumor cells was found in the majority of tumors (25, 83%). A subgroup of tumors (16, 53%) showed particular staining of tumor cells restricted to the cells at the tumor–stroma interface (tumor border staining, TBS, Fig. 1a), whereas 47% of the tumors showed uniform staining of the majority of tumor cells (Fig. 1b) or no staining. The nature of the MMP-2 positive cells in TBS tumors was investigated with immunomarkers for CD68 (macrophages), desmin (muscle), CD31 (endothelium) and vimentin (other mesenchymal origin). The lack of staining of the cells at the tumor nodule borders with these markers (not shown) suggested the epithelial origin of the MMP-2 positive cells. The nature of the cells was further confirmed by fluorescence double staining for epithelial cell marker Ep-CAM and MMP-2. Figure 1d shows double staining mainly for cells at the border of the tumor nests in tumors with TBS, whereas all malignant epithelial cells stained for MMP-2 in the homogenous type (Fig. 1e). The cells around tumor nodules stained for α-smooth muscle actin (SMA), indicating that they were from myofibroblast origin (Fig. 1j).29 SMA was also found in walls of muscularized vessels, as expected.
|Staining intensity||Percentage positive cells||Total score||Border staining (TBS)|
The activity state of MMP-2 for the 2 different immunohistochemical staining patterns, i.e. TBS versus homogenous, was investigated using fluorescence in situ zymography on frozen material derived from the same tumors as the paraffin-embedded material. Tumors with concentrated MMP-2 staining at the tumor–stroma interface (tumor border staining, TBS) showed gelatinase activity at this specific location (Fig. 1f). Virtually no activity was found in the tumor nodules that stained homogenously for MMP-2 (Fig. 1h), whereas in the same sections abundant MMP activity was present around endothelial cells (Fig. 1h insert). The gelatinase activity was mainly caused by MMP-2, because of the immunohistochemically shown absence of the other known gelatinase (MMP-9) in this area (data not shown). Gelatinase activity completely disappeared after the addition of EDTA, indicating specific MMP activity.
To investigate the postulation that effective MMP-2 activation primarily takes place at the cell surface where MMP-2 is associated with MT1-MMP and TIMP-2 in a ternary complex, we stained successive sections of the same tumors for MT-1-MMP and TIMP-2. Staining for MT1-MMP was found in all tumors, primarily in the tumor cells (Table II, Fig. 1k). Endothelial cells were positive in almost all tumors, but the staining was less intense. Stromal cells were weakly positive for MT1-MMP in the majority of tumors (Fig. 1k), but not all tumors contained positive stromal cells. Staining of tumor nodules for MT1-MMP was in general homogeneous, but in some cases, aggregation was found at the tumor–stroma interface. Normal cervical epithelial cells did not stain (not shown). TIMP-2 was found in the majority of tumors (28, 93%; Table II, Fig. 1l). The percentage of positive malignant cells varied, but most tumors showed staining in the majority of malignant cells. Next to malignant cells, TIMP-2 was found in endothelial cells and less intense in infiltrating cells and fibroblasts. Aggregation of TIMP-2 staining at the border of the nodules was noticed in 12 (41%) of the tumors. The total score for every parameter from Table II was tested for correlation, according to Spearman's Rank test (Table III). Total MMP-2 staining did not correlate with TIMP-2, or MT1-MMP, but a significant correlation of the TBS pattern was found between MMP-2 and TIMP-2 (Table IV). TIMP-2 staining correlated well with MT1-MMP (rs = 0.533, p = 0.003).
The observation that MMP-2 was present on tumor cells does not automatically imply that MMP-2 was actually produced by these cells. In situ hybridization revealed that high levels of MMP-2 mRNA were present in malignant cells of the tumors that stained homogenously for MMP-2 (Fig. 1i). However, MMP-2 mRNA was considerably less present in the tumor cells with immunohistochemical MMP-2 border staining (Fig. 1g) and enhanced expression in these border cells was not observed. Instead, the as myofibroblast identified cells directly surrounding the tumor nodules demonstrated a more intense MMP-2 signal. Therefore, it is suggestive that the MMP-2 found in the interface zone is of stromal myofibroblast origin. To investigate the possibility that stromal MMP-2 production was induced by tumor cells, we stained successive sections of the same 30 tumors for EMMPRIN. EMMPRIN staining was present in the majority of the tumors, mainly in tumor cells with in some cases a more intense staining in the cells at the border of the nodules (see Table II, Fig. 1m). Normal squamous epithelium did not show EMMPRIN staining (data not shown). Total EMMPRIN staining did not correlate significantly with the total score of MMP-2 (Table III), but correlated significantly with the enhanced presence of MMP-2 at the tumor-stroma interface (TBS, rs = 0.413, p = 0.023, Table IV). The correlation between the immunohistochemically determined MMP-2, MT1-MMP, TIMP-2 and EMMPRIN was further characterized in a dendrogram constructed by 2D unsupervised hierarchical cluster analysis using the between-groups linkage method (Table V). Each parameter was divided in 2 groups: for MMP-2 and MT1-MMP TBS type (+) versus homogenous type staining (−), respectively, and for TIMP-2 and EMMPRIN TBS type OR total score >4 (+) versus homogenous type staining ≤4 (−), respectively. Grouping of the tumors according to similar MMP-2/MT1-MMP/TIMP/EMMPRIN profile indicated a significant correlation between the presence of the different combinations and the status of the patients (χ2 = 15.8; p = 0.04). The first order division (I) separated the patients with non-TBS MMP-2, non-TBS MT1-MMP, low EMMPRIN and any TIMP-2 (5 survivors from 5 patients) versus the other patients (7 survivors from 25, χ2 = 9.0; p = 0.03). The second order division (II) divided the patients based on TBS presence of MMP-2 (9 survivors from 14 patients versus 3 from 16, χ2 = 6.5; p = 0.01). To confirm this correlation between the immunohistological scores for MMP-2, TIMP-2 and EMMPRIN with survival of the patients, Kaplan-Meier curves were constructed using all possible low–high scores from Table II and TBS-scores of all single parameters. Subdivision of patients based on MMP-2 staining intensity, percentage of staining cells or total scores did not reach significance. Only the survival of the patients using MMP-2 TBS was significantly correlated with survival (log rank 4.10; p = 0.043, Fig. 2a), whereas MT1-MMP TBS showed a trend (log rank 3.63; p = 0.057, Fig. 2b).
Although the involvement of MMPs in cancer development is without doubts, because of the large number of existing MMPs, the different behavior of the various cancer types and the complex cascade of activation of the enzymes, the exact roles of MMPs are not defined yet. Typical for the situation is the disappointing results of clinical trials with synthetic MMP inhibitors,30 which worked well in in vitro systems and animal models, but were not successful in cancer patients. Compared to other cancer types relatively few studies are published about the enhanced presence of MMPs and TIMPs in cervical carcinomas. Investigating the presence of a broad range of MMPs, Sheu et al.31 showed a significant increase in particularly MMP-2 and MMP-9 comparing normal cervix to high-grade squamous intraepithelial lesions and CSCC. Positive correlations were observed between MMP-2 levels and the presence of lymphovascular invasion and nodal metastasis.8 These studies were supported by in vitro experiments, showing a correlation between MMP-2 expression and invasion in cervical cancer cell lines.9 However, a clear correlation between enhanced MMP-2 levels and worse survival of patients, which has been found for certain other cancer types has not been established yet for CSCC.10, 11
The traditionally scored immunostained tissue sections of patients with primarily stage Ib cervical carcinomas, in our study, did not show a correlation between MMP-2 levels and survival. However, when we used the observed particular TBS pattern of MMP-2, i.e. concentrated in the cells neighboring the stroma versus homogenous staining of all cells, we found a significant correlation with survival (p = 0.04), even in our relatively small group of patients. Recently, the same TBS pattern for MMP-2 staining has been described.22 In that study, the TBS pattern was associated with nonkeratinizing squamous cell carcinomas in a small group of patients. This correlation was not present in our group of patients, which could be due to the low number of cases in both studies and a possible difference in the used histologic grading sytems.32
The same type of TBS scoring showed a similar trend for MT1-MMP. MT1-MMP is a membrane type MMP that next to its proteolytic activity is involved in the binding and activation of MMP-2. Although not obvious from Tables III and IV, MMP-2 border staining was significantly correlated with the presence of MT1-MMP if staining in the large majority of tumor cells (Total score level 5) or MT1-MMP border staining was used (rs = 0.535, p = 0.002). In fact, if the same combination of scoring was applied for TIMP-2, 13 out of 30 cervical carcinoma patients coexpressed MMP-2, MT1-MMP and TIMP-2 on the border cells of tumor nodules; from which 10 patients died within 8 years. The relevance of coexpression of these components is emphasized in the dendrogram in Table V. From the group of patients with all 4 factors present in their tumor, 7 of 8 were deceased. Our results indicate the importance of coexpression of these molecules, which seems to be a prerequisite for MMP-2 activation. Our in situ zymography experiments confirmed that high MMP-2 levels as such did not automatically correlate with enhanced proteolytic capacity, but that localization at the periphery of the tumor nodules was required. Nagase et al. postulated that activation of proMMP-2 primarily takes place on the cell surface mediated by MT-MMPs.23 Later studies demonstrated that pro-MMP-2 is recruited to the cell surface by interacting with TIMP-2 bound to MT1-MMP, by forming a ternary complex. Another free MT1-MMP molecule, closely located to the ternary complex, subsequently activates proMMP-2, providing the cell with local proteolytic activity.33 Our results confirm these in vitro experiments and suggest that activation of myofibroblast-derived MMP-2 takes place at the surface of cervical cancer cells at the tumor–stroma interface (TBS). In addition, our results emphasize the recent finding that the E7 protein of high-risk types human papilloma virus lead to induction of MT1-MMP34 and hence to increased MMP-2 activation.
MMP-2 activity plays a role in invasion. Recent studies, however, indicate that MMP activity is not restricted to degradation of the ECM.35 In addition, MMPs are implicated in the cleavage or shedding of a whole range of substrates that are known to play a role in different stages of neoplastic growth: growth factors, cytokines and cytokine receptors.36 Local cleavage of these proteins by MMPs could induce (in)activation, which would have major implications for tumor progression. Pyke and coworkers suggested already in 1991 that tumor cells might actively induce the production of proteolytic enzymes in neighboring stromal cells.37 Indeed, a number of cytokines is found to induce MMP production in fibroblasts in vitro.38 A recent report, however, showed that the presence of EMMPRIN on tumor cells rather than humoral factors is involved in inducing pro-MMP-2 production in fibroblasts, followed by activation via MT1-MMP.39 In our study, EMMPRIN staining was mainly found on tumor cells and was significantly correlated with the presence of MMP-2 at the border of tumor nodules, where activity is expected. Our results are concordant with a role for EMMPRIN expressed by cervical cancer cells in inducing pro-MMP-2 production in fibroblasts in vivo. In conclusion, our results suggest that colocalization of EMMPRIN, MT1-MMP and TIMP-2 are involved in production and activation of MMP-2 in human CSCC. The presence of this presumed active MMP-2 correlated with the survival of the patients, indicating the possible relevance of this mechanism in cervical cancer. Our study also indicates that a delicate change in scoring of an immunostained parameter (e.g. localization or activity) could considerably change the outcome in survival analysis.
We thank Dr. H.A. Baelde and W.P. Bouwman for excellent technical assistance and Dr. P. Birembaut (Inserm, Reims, France) and Dr. G. Murphy (University of Cambridge, UK.) for providing the probes for MMP-2.
- 25Quality control of immunohistochemical evaluation of tumour-associated plasminogen activators and related components. European BIOMED-1 concerted action on clinical relevance of proteases in tumour invasion and metastasis. Eur J Cancer 1998; 34: 1334–40., , , , , , et al.