Int. J. Cancer: 118, 2991–2998 (2006) ' 2006 Wiley-Liss, Inc.
EMMPRIN-induced MMP-2 activation cascade in human cervical squamous cell carcinoma Cornelis F.M. Sier1,2*, Kim Zuidwijk1, Henry J.M.A.A. Zijlmans1,3, Roeland Hanemaaijer4, Adri A. Mulder-Stapel4, Frans A. Prins1, Enno J. Dreef1, Gemma G. Kenter3, Gert Jan Fleuren1 and Arko Gorter1 1 Department of Pathology, Leiden University Medical Center, Leiden, The Netherlands 2 Department of Gastroenterology, Leiden University Medical Center, Leiden, The Netherlands 3 Department of Gynaecology, Leiden University Medical Center, Leiden, The Netherlands 4 Gaubius Laboratory TNO-PG, Leiden, The Netherlands 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 (MMP14), 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 (tumorborder 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 5 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. Key words: cervix cancer; matrix metalloproteinase; immunohistochemistry; in situ zymography; MT1-MMP; TIMP-2; EMMPRIN
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).2–4 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,4,7–11 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 Publication of the International Union Against Cancer
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 *Correspondence to: Department of Gastroenterology/Pathology, Leiden University Medical Center Building 1, C4, P.O. Box 9600, 2300 RC Leiden, The Netherlands. Fax: 131-71-5248158. E-mail:
[email protected] Received 12 August 2005; Accepted 22 November 2005 DOI 10.1002/ijc.21778 Published online 19 January 2006 in Wiley InterScience (www.interscience. wiley.com).
2992
SIER ET AL. TABLE I – ANTIBODIES USED IN THIS STUDY Antigen
Source
Detects
Pretreatment
Dilution
Supplier
MMP-2 MMP-9 MT1-MMP TIMP-2 EMMPRIN CD31 CD68 EpCAM Desmin Vimentin a-smooth muscle actin
Mouse, mAb Rabbit, pcAb Rabbit, pcAb Mouse, mAb Rabbit, pcAb Mouse, mAB Mouse, mAb Mouse, mAb Mouse, mAb Mouse, mAb Mouse, mAb
(pro)MMP-2 (pro)MMP-9 MT-1MMP TIMP-2 EMMPRIN Endothelial cells Macrophages Epithelial cells Muscle cells Mesenchymal cells Myofibroblasts, smooth muscle cells
Citrate Citrate None Citrate Citrate Citrate Citrate None Citrate Citrate Citrate
1/200 1/500 1/1000 1/400 1/150 1/100 1/100 1/100 1/100 1/100 1/1000
NeoMarkers, Fremont, CA TNO-PG, Leiden, The Netherlands TNO-PG, Leiden, The Netherlands NeoMarkers, Fremont, CA Zymed, San Francisco, CA Alexis-Benelux, The Netherlands Dako, Glostrup, Denmark Centocor, Leiden, The Netherlands Sanbio, Uden, The Netherlands Sanbio, Uden, The Netherlands Progen, Heidelberg, Germany
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 Tissue samples 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 5 20), IIa (n 5 8), IIb (n 5 1) and IIIb (n 5 1). From most tumors, frozen material was available for MMP-2 activity determination. Immunohistochemistry Immunohistochemistry was performed on 4 lm 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 (antiMMP-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). Immunohistochemical evaluation 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 in-
dependent 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 PBSBSA 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 253 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 lm thick) were rehydrated, digested with proteinase K (5 lg/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 lg/ml salmon sperm DNA, 25 lg/ ml tRNA, 10 mM DTT, 13 Dernhardt’s solution, 10% dextran sulphate and 23 SSC. Sections were washed twice in 23 SSC at 55°C, once with 20 mM b-mercaptoethanol in 0.13 SSC at 55°C and treated with 2 U/ml RNase T1 for 30 min at 37°C. The DIGlabeled 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 lm) were air-dried for 10 min. DQ-gelatin was dissolved in
MMP-2 ACTIVATION IN CERVICAL CARCINOMA
2993
FIGURE 1 – Representative stainings of MMP-2 in human squamous cell cervical carcinomas. (a) Immunohistochemical staining for MMP-2 mainly present in cells at the border of tumor nodules (TBS, arrow) and in stromal cells (3100). (b) Homogenous MMP-2 staining in tumor nodule and stroma (arrow, 3100). (c) Normal epithelial cells do not stain for MMP-2 (arrow), whereas stromal cells in some cases do. (d) Fluorescence double staining in TBS type cervical carcinoma. Red MMP-2 staining is mainly present in cells at the border of tumor nodules (arrow) and stroma. Green Ep-CAM staining is only found in malignant epithelial cells (3250). (e) Idem for tumor homogenously staining for MMP-2. All epithelial cells (Ep-CAM, green) are also staining red (MMP-2). (f) In situ zymography for MMP-2 activity (green) in TBS type cervical carcinoma (250). Activity is mainly present around tumor nodules (arrow). Red staining is propidium iodide nuclear staining. (g) In situ hybridization for MMP-2 mRNA in TBS-type cervical carcinoma (3100). Blue indicates MMP-2 mRNA, present mainly in fibroblast-like cells. (h) In situ zymography for MMP-2 activity in cervical carcinoma with homogenous MMP-2 immunohistochemical staining (3250). Although MMP-2 is abundantly present, no activity (green) is detected in tumor cells; however, some endothelium in the same cryosection is staining green (insert). (i) In situ hybridization for MMP-2 mRNA in cervical carcinoma with homogenous MMP-2 immunohistochemical staining (3100). mRNA for MMP-2 is present in all malignant cells. (j) a-SMA staining in tumor nodule surrounding myofibroblasts (arrow). (k) Strong MT1MMP staining in all tumor cells and less intense in stromal cells in TBS type cervical carcinoma (3100). (l) TIMP-2 staining is present in tumor nodules and stroma cells in TBS type cervical carcinoma (3100). (m) EMMPRIN staining is only present in tumor cells (3100). All stainings are performed on serial sections from the same tumor (TBS-type) except for (b and e), which were both from a tumor with uniform MMP-2 staining type.
water (1 mg/ml) and 1:10 diluted in 1% (w/v) low gelling agarose (Gibco-BRL) in PBS containing 0.5 lg/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.
Statistical analysis 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.
Results 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).
2994
SIER ET AL. TABLE II – SEMIQUANTITATIVE ANALYSIS OF IMMUNOHISTOCHEMICAL PRESENCE OF MMP-2, MT1-MMP, TIMP-2 AND EMMPRIN IN MALIGNANT CELLS OF CERVICAL SQUAMOUS CELL CARCINOMAS FROM 30 PATIENTS Staining intensity 0
MMP-2 MT1-MMP TIMP-2 EMMPRIN
1
2
3
Percentage positive cells 0
1
5
Total score
2
3
4
0
2
5 5 10 10 5 9 5 0 7 15 8 0 0 0 2 20 8 0 2 5 3 3 13 11 3 3 1 6
4 3 1 7
1 6 5 0 9 18 0 0 2 17 2 5 7 6 3 1
3
4
5
6
7
8
6 10 2 1 2 0 0 3 11 13 3 0 3 10 7 2 4 9 10 1
4 3 0 0
Border staining (TBS)
16 11 12 8
Intensity: 0, no staining; 1, weak; 2, moderate; 3, intense; percentage: 0, no staining; 1, 1–5%; 2, 6– 25%; 3, 26–50%; 4, 51–75% and 5, 76–100%; total score: sum of staining intensity (0–3) and percentage of positive cells (0–5); TBS: (enhanced) staining of malignant cells at the border of the nodules. TABLE III – CORRELATIONS BETWEEN THE PRESENCE/INTENSITY OF MMPS AND RELATED PARAMETERS (TOTAL SCORE FROM TABLE II) IN TUMOR CELLS FROM 30 CERVICAL SQUAMOUS CELL CARCINOMAS ACCORDING TO SPEARMAN’S RANK TEST
MMP-2 MT1-MMP TIMP-2
MT1-MMP
TIMP-2
EMMPRIN
0.239 ns
0.194 ns 0.533 0.003
0.081 ns 0.301 ns 0.313 ns
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 a-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. 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 cor-
TABLE IV – CORRELATIONS BETWEEN THE BORDER-STAINING PATTERN OF MMPS AND RELATED PARAMETERS AT THE BORDER OF TUMOR NODULES IN 30 CERVICAL SQUAMOUS CELL CARCINOMAS ACCORDING TO SPEARMAN’S RANK TEST
MMP-2 MT1-MMP TIMP-2
MT1-MMP
TIMP-2
EMMPRIN
0.296 ns
0.391 0.036 0.274 ns
0.413 0.023 0.323 ns 0.344 0.068
relation, 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 5 0.533, p 5 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 5 0.413, p 5 0.023, Table IV). The correlation between the immunohistochemically determined MMP-2, MT1MMP, 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 (1) versus homogenous type staining (2), respectively, and for TIMP-2 and EMMPRIN TBS type OR total score >4 (1) versus homogenous type staining 4 (2), 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 (v2 5 15.8; p 5 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, v2 5 9.0; p 5 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, v2 5
TABLE V – TWO-DIMENSIONAL UNSUPERVISED HIERARCHICAL CLUSTER ANALYSIS OF THE MMP-2 INDUCTION ACTIVATION PATHWAY IN CERVICAL CANCER
The analysis was done using the between-groups linkage method. Rows: individual cases. Columns: MMP-2, MT1-MMP TIMP-2, EMMPRIN and status (0, alive; 1, dead). I and II are first and second order division, respectively.
2996
SIER ET AL.
FIGURE 2 – Overall survival curves for 30 patients with cervical squamous carcinoma subdivided for presence of TBS, border-pattern staining of tumor cells (MMP-2 and MT1-MMP). Patients with TBS are indicated with normal curves, whereas for patients with homogenously staining tumors striped curves are used. (a) MMP-2, log rank 4.10, p 5 0,043, n 5 14 (5 events, – –) versus n 5 16 (13 events, —) and (b) MT1-MMP, log rank 3.63, p 5 0.056, n 5 19 (8 events, – –) versus n 5 11 (10 events, —).
6.5; p 5 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 5 0.043, Fig. 2a), whereas MT1-MMP TBS showed a trend (log rank 3.63; p 5 0.057, Fig. 2b).
Discussion 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 5 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 MT1MMP. MT1-MMP is a membrane type MMP that next to its proteolytic activity is involved in the binding and activation of MMP2. Although not obvious from Tables III and IV, MMP-2 border staining was significantly correlated with the presence of MT1MMP if staining in the large majority of tumor cells (Total score level 5) or MT1-MMP border staining was used (rs 5 0.535, p 5 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 MT1MMP 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, cyto-
2997
MMP-2 ACTIVATION IN CERVICAL CARCINOMA 36
kines and cytokine receptors. 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 conclu-
sion, our results suggest that colocalization of EMMPRIN, MT1MMP 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. Acknowledgements 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.
References 1. 2.
3.
4. 5. 6.
7.
8.
9. 10.
11.
12. 13.
14. 15. 16. 17. 18.
Waggoner SE. Cervical cancer. Lancet 2003;361:2217–25. Garzetti GG, Ciavattini A, Lucarini G, Goteri G, De Nictolis M, Biagini G. Microinvasive cervical carcinoma and cervical intraepithelial neoplasia: biologic significance and clinical implications of 72-kDa metalloproteinase immunostaining. Gynecol Oncol 1996;61:197–203. Nuovo GJ, MacConnell PB, Simsir A, Valea F, French DL. Correlation of the in situ detection of polymerase chain reaction-amplified metalloproteinase complementary DNAs and their inhibitors with prognosis in cervical carcinoma. Cancer Res 1995;55:267–75. Gilles C, Polette M, Piette J, Munaut C, Thompson EW, Birembaut P, Foidart JM. High level of MT-MMP expression is associated with invasiveness of cervical cancer cells. Int J Cancer 1996;65:209–13. Itoh T, Tanioka M, Matsuda H, Nishimoto H, Yoshioka T, Suzuki R, Uehira M. Experimental metastasis is suppressed in MMP-9-deficient mice. Clin Exp Metastasis 1999;17:177–81. Iwasaki M, Nishikawa A, Fujimoto T, Akutagawa N, Manase K, Endo T, Yoshida K, Maekawa R, Yoshioka T, Kudo R. Anti-invasive effect of MMI-166, a new selective matrix metalloproteinase inhibitor, in cervical carcinoma cell lines. Gynecol Oncol 2002;85:103–7. Brummer O, Bohmer G, Hollwitz B, Flemming P, Petry KU, Kuhnle H. MMP-1 and MMP-2 in the cervix uteri in different steps of malignant transformation––an immunohistochemical study. Gynecol Oncol 2002;84:222–7. Sheu BC, Lien HC, Ho HN, Lin HH, Chow SN, Huang SC, Hsu SM. Increased expression and activation of gelatinolytic matrix metalloproteinases is associated with the progression and recurrence of human cervical cancer. Cancer Res 2003;63:6537–42. Kato Y, Yamashita T, Ishikawa M. Relationship between expression of matrix metalloproteinase-2 and matrix metalloproteinase-9 and invasion ability of cervical cancer cells. Oncol Rep 2002;9:565–9. Davidson B, Goldberg I, Kopolovic J, Lerner-Geva L, Gotlieb WH, Ben Baruch G, Reich R. MMP-2 and TIMP-2 expression correlates with poor prognosis in cervical carcinoma––a clinicopathologic study using immunohistochemistry and mRNA in situ hybridization. Gynecol Oncol 1999;73:372–82. Moser PL, Kieback DG, Hefler L, Tempfer C, Neunteufel W, Gitsch G. Immunohistochemical detection of matrix metalloproteinases (MMP) 1 and 2, and tissue inhibitor of metalloproteinase 2 (TIMP 2) in stage IB cervical cancer. Anticancer Res 1999;19:4391–3. Sidenius N, Blasi F. The urokinase plasminogen activator system in cancer: recent advances and implication for prognosis and therapy. Cancer Metastasis Rev 2003;22:205–22. Barmina OY, Walling HW, Fiacco GJ, Freije JM, Lopez-Otin C, Jeffrey JJ, Partridge NC. Collagenase-3 binds to a specific receptor and requires the low density lipoprotein receptor-related protein for internalization. J Biol Chem 1999;274:30087–93. Seiki M, Koshikawa N, Yana I. Role of pericellular proteolysis by membrane-type 1 matrix metalloproteinase in cancer invasion and angiogenesis. Cancer Metastasis Rev 2003;22:129–43. Guo H, Li R, Zucker S, Toole BP. EMMPRIN (CD147), an inducer of matrix metalloproteinase synthesis, also binds interstitial collagenase to the tumor cell surface. Cancer Res 2000;60:888–91. Fridman R, Toth M, Chvyrkova I, Meroueh SO, Mobashery S. Cell surface association of matrix metalloproteinase-9 (gelatinase B). Cancer Metastasis Rev 2003;22:153–66. Heppner KJ, Matrisian LM, Jensen RA, Rodgers WH. Expression of most matrix metalloproteinase family members in breast cancer represents a tumor-induced host response. Am J Pathol 1996;149:273–82. Nielsen BS, Timshel S, Kjeldsen L, Sehested M, Pyke C, Borregaard N, Danø K. 92 kDa type IV collagenase (MMP-9) is expressed in neu-
19. 20.
21. 22.
23. 24. 25.
26.
27.
28.
29.
30. 31. 32.
33.
34.
trophils and macrophages but not in malignant epithelial cells in human colon cancer. Int J Cancer 1996;65:57–62. Zucker S, Hymowitz M, Rollo EE, Mann R, Conner CE, Cao J, et al. Tumorigenic potential of extracellular matrix metalloproteinase inducer. Am J Pathol 2001;158:1921–8. Guo H, Zucker S, Gordon MK, Toole BP, Biswas C. Stimulation of matrix metalloproteinase production by recombinant extracellular matrix metalloproteinase inducer from transfected Chinese hamster ovary cells. J Biol Chem 1997;272:24–7. Caudroy S, Polette M, Nawrocki-Raby B, Cao J, Toole BP, Zucker S, Birembaut P. EMMPRIN-mediated MMP regulation in tumor and endothelial cells. Clin Exp Metastasis 2002;19:697–702. Malina R, Motoyama S, Hamana S, Maruo T. Laminin-5 g2 chain and matrix metalloproteinase-2 expression in the neoplastic changes of uterine cervical squamous epithelium. Kobe J Med Sci 2004;50: 123–30. Nagase H. Cell surface activation of progelatinase A (proMMP-2) and cell migration. Cell Res 1998;8:179–86. Hazelbag S, Gorter A, Kenter GG, Van Den Broek L, Fleuren G. Transforming growth factor-b1 induces tumor stroma and reduces tumor infiltrate in cervical cancer. Hum Pathol 2002;33:1193–9. Ruiter DJ, Ferrier CM, van Muijen GN, Henzen-Logmans SC, Kennedy S, Kramer MD, et al. Quality 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. de Boer WI, van Schadewijk A, Sont JK, Sharma HS, Stolk J, Hiemstra PS, van Krieken JH. Transforming growth factor b1 and recruitment of macrophages and mast cells in airways in chronic obstructive pulmonary disease. Am J Respir Crit Care Med 1998;158: 1951–7. Bisson C, Blacher S, Polette M, Blanc JF, Kebers F, Desreux J, et al. Restricted expression of membrane type 1-matrix metalloproteinase by myofibroblasts adjacent to human breast cancer cells. Int J Cancer 2003;105:7–13. Mook OR, Van Overbeek C, Ackema EG, Van Maldegem F, Frederiks WM. In situ localization of gelatinolytic activity in the extracellular matrix of metastases of colon cancer in rat liver using quenched fluorogenic DQ-gelatin. J Histochem Cytochem 2003;51:821–9. Barth PJ, Ramaswamy A, Moll R. CD34(1) fibrocytes in normal cervical stroma, cervical intraepithelial neoplasia III, and invasive squamous cell carcinoma of the cervix uteri. Virchows Arch 2002;441: 564–8. Rudek MA, Venitz J, Figg WD. Matrix metalloproteinase inhibitors: do they have a place in anticancer therapy? Pharmacotherapy 2002; 22:705–20. Sheu BC, Hsu SM, Ho HN, Lien HC, Huang SC, Lin RH. A novel role of metalloproteinase in cancer-mediated immunosuppression. Cancer Res 2001;61:237–42. Stock RJ, Zaino R, Bundy BN, Askin FB, Woodward J, Fetter B, et al. Evaluation and comparison of histopathologic grading systems of epithelial carcinoma of the uterine cervix: Gynecologic Oncology Group studies. Int J Gynecol Pathol 1994;13:99–108. Chattopadhyay N, Mitra A, Frei E, Chatterjee A. Human cervical tumor cell (SiHa) surface avb3 integrin receptor has associated matrix metalloproteinase (MMP-2) activity. J Cancer Res Clin Oncol 2001; 127:653–8. Smola-Hess S, Pahne J, Mauch C, Zigrino P, Smola H, Pfister HJ. Expression of membrane type 1 matrix metalloproteinase in papillo-
2998
SIER ET AL.
mavirus-positive cells: role of the human papillomavirus (HPV) 16 and HPV8 E7 gene products. J Gen Virol 2005;86:1291–6. 35. Egeblad M, Werb Z. New functions for the matrix metalloproteinases in cancer progression. Nat Rev Cancer 2002;2:161–74. 36. Overall CM, McQuibban GA, Clark-Lewis I. Discovery of chemokine substrates for matrix metalloproteinases by exosite scanning: a new tool for degradomics. Biol Chem 2002;383:1059–66. 37. Pyke C, Kristensen P, Ralfkiaer E, Grøndahl-Hansen J, Eriksen J, Blasi F, Danø K. Urokinase-type plasminogen activator is expressed in stromal cells and its receptor in cancer cells at invasive
foci in human colon adenocarcinomas. Am J Pathol 1991;138: 1059–67. 38. Hanemaaijer R, Koolwijk P, le Clercq L, de Vree WJ, van Hinsbergh VW Regulation of matrix metalloproteinase expression in human vein and microvascular endothelial cells. Effects of tumour necrosis factor a, interleukin 1 and phorbol ester. Biochem J 1993;296(Pt 3):803–9. 39. Suzuki S, Sato M, Senoo H, Ishikawa K. Direct cell-cell interaction enhances pro-MMP-2 production and activation in co-culture of laryngeal cancer cells and fibroblasts: involvement of EMMPRIN and MT1-MMP. Exp Cell Res 2004;293:259–66.
