Lung cancer remains the most common cause of cancer-related death in Europe and the United States.1, 2, 3 Non-small cell lung cancer (NSCLC) affects approximately 80% of all lung cancer patients.3, 4 The standard treatment for early stage NSCLC is surgery resulting in a 5-year survival rate of only 50–60% in Stage I and II.5, 6 A better understanding of the molecular mechanisms of lung cancer progression might help to identify patients at risk for an unfavorable outcome with potential consequences for adjuvant therapy.
Penetration of basement membranes and degradation of extracellular matrix are essential steps in the dissemination of tumor cells from primary tumors to distant organs.7 One group of proteolytic enzymes that has been associated with these abilities is the matrix metalloproteinase family (MMP). Matrix metalloproteinase 9 (MMP-9) (92 kDa Type IV collagenase or gelatinase B) has the potency to degrade gelatin and Type IV collagen,8 which is a major component of basement membranes. The clinical relevance of MMP 9 in NSCLC remains in discussion. First clinical trials investigating the efficacy of MMP-inhibitors in advanced cancer were rather disappointing.9 Furthermore, previous studies examining the prognostic impact of immunohistochemical detection of MMP-9 in NSCLC10, 11 are discussed controversially because they did not demonstrate an independent prognostic impact of MMP-9 considering all standard prognostic factors included in the international system for staging lung cancer.6 To clarify the prognostic impact of MMP 9 in operable NSCLC, we applied a new immunohistochemical evaluation system, focusing on homogeneous expression of MMP-9. The prognostic value was analyzed in relation to established prognostic factors of the TNM-system. Our study shows that the prognostic impact of homogeneous MMP-9 expression remains independent from possible prognostic joint effects of pT-status, pN-status, tumor histology and tumor grading. Because of this independence, MMP-9 may be considered as an interesting target for adjuvant anticancer therapy in operable NSCLC using selective MMP-inhibitors with high specificity for MMP-9.
MATERIAL AND METHODS
After approval by the ethical committee of the University of Munich and after written informed consent, tissue specimens of 147 consecutive patients with completely resected NSCLC were collected. The tumors were classified according to the international union against cancer (UICC) TNM-classification.6 The preoperative staging of all patients had resulted in the diagnosis of resectable tumors (T1–T4) without evident distant metastasis (M1) or contralateral or supraclavicular lymph node involvement (N3). In general, a lobectomy or pneumonectomy with systematic mediastinal lymphadenectomy was carried out in our Department of Thoracic Surgery. Only patients with confirmed postoperative R0 stage were selected for further evaluations. The final study population comprised 33 female and 110 male patients. The median age at the time of surgery was 60 years with a range of 37–76 years. Patients whose primary tumors were classified as pT3 or pT4 received adjuvant percutaneous radiotherapy of the tumor bed and patients with mediastinal lymph node involvement (pN2) received percutaneous radiotherapy of the entire mediastinum.
The median follow-up duration was 72 months (range = 12–144 months). Follow-up studies included physical examination, chest X-ray and blood tests in 3-month intervals and an additional thoracic CT scan, abdominal ultrasound and bronchoscopy in 6-month intervals. Close follow-up was documented by contacting family practitioners with questionnaires concerning local relapse, distant metastasis and death. If possible, a relapse was confirmed at our institution and the patient was admitted for subsequent therapy.
Immunohistochemical staining of primary tumors
The presence of MMP-9 in primary tumors was analyzed by immunohistochemical staining using the labeled streptavidin-biotin (LSAB) method. Briefly, paraffin sections were dewaxed, rehydrated and subsequently cooked for 10 min in 10 mM citric acid buffer at pH 6.0 using a microwave oven. Endogenous peroxidase activity was blocked by treating the specimens with 30% hydrogen peroxide for 10 min. Non-specific antibody binding was prevented by a commercial blocking agent (LSAB-kit, Dako Corp., Carpinteria, CA). Excess blocking agent was drained and the sections were incubated overnight at 4°C with an anti-MMP-9 primary antibody that was kindly provided by R. Lichtinghagen. The polyclonal affinity-purified rabbit antibody recognizes both the latent proenzyme and the active form of MMP-9 specifically.12, 13 The antibody solution (1 mg/ml) was used in 1:2,000 dilution with background reducing antibody diluent (Dako Corp.). For the negative controls the primary antibody was replaced with rabbit non-immune IgG (Southern Biotechnology Associates Inc, Birmingham, AL).
Immunohistochemical staining was continued by incubating the slides with biotinylated antimouse/antirabbit secondary antibody solution (LSAB-kit, Dako Corp.) for 30 min at 25°C. Peroxidase was introduced using a streptavidin conjugate (LSAB-kit, Dako Corp.). Between each step of the procedure the specimens were thoroughly rinsed with 0.1 M Tris-HCl buffer pH 8.2. Peroxidase reactivity was visualized using aminoethyl carbazole (AEC) (Sigma, St. Louis, MO) dissolved in dimethylformamide and 0.1 M acetate buffer pH 5.2 creating a red-brown staining with 30% hydrogen peroxide. Finally, the sections were counterstained with hematoxylin and mounted in Kaiser's glycerol gelatin.
Evaluation of immunohistochemical staining
Immunohistochemical staining of tumor sections was examined independently by 2 observers who were unaware of the clinical data. The slides with discrepant evaluations were re-evaluated and a consensus was reached (n = 10). The slides were examined under light-microscopes using objectives with 10× and 40× magnification.
The staining intensity for MMP-9 in tumor cells was assessed in comparison to normal ciliated bronchial or bronchiolar epithelium. Bronchiolar epithelium is known to express MMP-9 and is therefore suitable to serve as an internal positive control and reference concerning the level of MMP-9 expression.14 One-hundred tumors cells in the center of each tumor were counted and the percentage of tumor cells that showed staining at least as intensive as bronchiolar epithelium was calculated. MMP-9 expression was categorized into: negative, ≤5% tumor cells stained; heterogeneous, >5% and <95% tumor cells stained; and homogeneous, ≥95% tumor cells stained. Staining specificity was assessed by evaluation of the negative controls in which the primary antibody had been replaced with normal rabbit non-immune IgG.
Statistical analysis was carried out using the SPSS software package, version 11.0 (SPSS, Inc, Chicago, IL). The 2-tailed Pearson's χ2 test was used to analyze the association between clinicopathological variables. All variables were dichotomized. For analysis of follow-up data, life table curves were calculated using the Kaplan-Meier method and survival distributions were compared by log-rank statistics. The primary end point was cancer-related survival, as measured from the date of surgery to the time of the last follow-up or cancer-related death. Data of patients who were still alive and without evidence of disease at the end of the study were censored. The joint effects with already recognized prognostically relevant variables were examined via Cox proportional hazards analysis. pT-status, pN-status, tumor histology and tumor grading were entered stepwise forward into the model to test these covariables for possible prognostic joint effects with homogeneous MMP-9 expression. The threshold for statistical significance was chosen at p = 0.05.
Immunohistochemical evaluation of primary tumors
Because MMP-9 is expressed not only in tumor cells but also in stromal cells, tissue specimens were separately evaluated with respect to these different localization of MMP-9 expression. The specimens were divided into 3 groups according to the immunoreactivity of cancer cells in the center of the tumor to distinguish between negative expression, heterogeneous expression and homogeneous expression of MMP-9. The respective cut-points based on a preceding evaluation of our immunostaining system that showed that approximately 30% of all NSCLC-derived specimens exhibited positive staining in at least 95% of tumor cells and that further 30% of all specimens showed no staining or staining of <5% of tumor cells. Therefore, MMP-9 expression was assessed as negative if ≤5% tumor cells, as heterogeneous if >5% and <95% tumor cells and as homogeneous if ≥95% tumor cells were stained at least similar to normal bronchiolar epithelium, which served as an internal positive control and reference concerning the level of MMP-9 expression. All inflammatory cells exhibited immunohistochemical staining. Their staining, however, was less intense than staining of normal bronchiolar epithelium and was not considered as reference concerning the level of MMP-9 expression of tumor cells.
As a result of the exclusion of 4 unspecifically stained specimens, tumors of 143 patients were available for analysis of MMP-9 expression. Tissue sections of 13 (21.7%) patients showed no MMP-9 expression. Specimens of 31 (51.7%) patients exhibited heterogeneous MMP-9 expression and tumors of 26 (18.2%) patients exhibited homogeneous MMP-9 expression (Fig. 1, Table I). No significant correlation became apparent between homogeneous MMP-9 expression and pT-status, pN-status, tumor histology, grading or gender (Table I). Negative and heterogeneous MMP-9 expression did not correlate with these clinicopathological parameters either (Table I).
Table I. Patients' Characteristics and Tumor Variables According to MMP-9 Expression in NSCLC
Assessment of tumor stroma either showed no immunoreactivity or weak immunostaining of maximally 50% of fibroblasts. In all specimens staining of tumor stroma was less intense than staining of normal bronchiolar epithelium that served as an internal positive control. There was no significant correlation between MMP-9 expression by tumor stroma and clinicopathological parameters such as pT-status, pN-status, tumor histology, grading or sex (data not shown).
Exclusion of 13 patients with cancer-unrelated death or incomplete follow-up resulted in 130 eligible patients for survival analyses. The median follow-up duration was 72 months (range = 12–144 months). Within the observation period, distant metastases developed in 49 patients (38%), and local recurrence occurred in 31 (24%) of the 130 eligible patients. Relapse was defined as diagnosis of distant metastasis or local recurrence. In total, relapse was observed in 78 patients among the 130 eligible patients (60%) and led to cancer-related death in all of the 78 patients within the observation period.
Survival analysis of cancer-related death, comparing patients with homogeneous MMP-9 expression by tumor cells (n = 24) to the group with negative and heterogeneous MMP-9 expression (n = 106) showed a significant correlation between homogeneous MMP-9 expression and unfavorable outcome (Fig. 2). The log-rank test p-value was 0.016. In contrast, cancer-related survival distributions of patients with homogeneous MMP-9 expression and with heterogeneous MMP-9 expression by tumor cells (n = 69) compared to MMP-9-negative patients (n = 61) were almost equal (Fig. 3; p = 0.82; log-rank test). Furthermore, MMP-9 expression by stromal fibroblasts was not of prognostic impact (p = 0.82; log-rank test). Additional Kaplan-Meier analyses of local recurrence and occurrence of distant metastases demonstrated that homogeneous MMP-9 expression was correlated significantly with development of local recurrence (p = 0.012; log-rank test) and distant metastases (p = 0.032; log-rank test).
A multivariate analysis was conducted to evaluate whether the correlation between homogeneous MMP-9 expression in tumor cells and shortened cancer-related survival resulted from an association of MMP-9 expression with other prognostically relevant factors or whether it could maintain its own prognostic value (Table II). Homogeneous expression of MMP-9, pT-status, pN-status, tumor histology and tumor grading were tested for possible prognostic joint effects. This Cox regression analysis demonstrated that homogeneous MMP-9 expression was a significant and independent prognostic parameter for shortened cancer-related survival in patients with operable NSCLC (p = 0.045). The relative risk for cancer-related death was 1.8-fold increased in case of homogeneous MMP-9 expression compared to patients with negative or heterogeneous MMP-9 expression (95% CI = 1.0–3.3). Comparison of the relative risks of homogeneous MMP-9 expression, pT-status, pN-status, tumor histology and tumor grading showed that homogeneous MMP-9 expression (relative risk 1.8) was the second strongest prognostic parameter after pN-status (relative risk 2.5).
Table II. Univariate and Multivariate Analysis of Cancer-Related Survival1
Penetration of basement membranes and degradation of extracellular matrix are essential steps in the dissemination of tumor cells from primary tumors to distant organs. MMP-9 has been associated with invasion and metastasis because of its potency to degrade gelatin and type IV collagen8 as major components of basement membranes. To assess the clinical impact of MMP-9 expression on long-term survival in patients with operable NSCLC, an immunohistochemical study of MMP-9 expression by primary tumors was carried out. We used a non-commercial antibody for immunohistochemical detection of MMP-9, because several tested commercial antibodies resulted in unsatisfactory immunohistochemical staining of paraffin sections (e.g., N40402 Novus Molecular, Inc., CA; Ab-MMP-9 Amersham Biosciences Corp., NJ; Ab-3, Oncogene/Calbiochem, Bad Soden, Germany). Our non-commercial antibody has been proven to be highly specific.12, 13 The antibody has been already successfully used by other research groups.15, 16 It is produced continuously and will be provide for free to the scientific community upon request to Dr. Lichtinghagen. Thus, the antibody is well established and will be available to the research community in the future.
Our present study shows that detection of homogeneous MMP-9 expression in cancer cells is of independent prognostic impact in operable NSCLC (Fig. 2, Table II). To our best knowledge, only 2 previous studies investigated the prognostic impact of immunohistochemically detected MMP-9 in NSCLC.10, 11 These studies did not demonstrate an independent prognostic impact of MMP-9 considering all standard prognostic factors included in the international system for staging lung cancer.6 Furthermore, these 2 previous studies observed MMP-9 expression in 65 or 52% of the patients, respectively. In view of a recent histopathological study of differential expression of MMPs in NSCLC,14 which reported MMP-9 expression in 36.5% and MMP-9 homogeneous expression in 2.6% of tissue sections, immunohistochemical finding of MMP-9 in 52–65% of specimens might rather be assessed as detection of MMP-9 expression, but may not be considered as detection of homogeneous expression of MMP-9. The high rates of positive immunostaining could result from the cut point of 20% stained cancer cells in both studies10, 11 mentioned above. In contrast, we applied a new immunohistochemical evaluation system that focused exclusively on homogeneous expression of MMP-9. In our present study, expression of MMP-9 was categorized into negative, heterogeneous and homogeneous. To focus on homogeneous expression, a cut-point of 95% stained cancer cells was chosen as threshold between heterogeneous expression and homogeneous expression of MMP-9. This cut point was clearly reproducible when 2 independent observers evaluated the stained tumor sections. Discrepant results were obtained in only 10 specimens (7%) and a consensus decision could be reached in all respective cases. The cut-point of 95% stained cancer cells based on a preceding evaluation of our immunostaining system that showed that 30% of all NSCLC-derived specimens exhibited positive staining in at least 95% of tumor cells. This cut-point resulted in comparable ratios of positive section in the present study: 26 (18%) tumors with MMP-9 homogeneous expression and in 51 (36%) specimens with heterogeneous MMP-9 expression. This cut point appears to be unusual at the first glance. There are many studies in the literature, however, in which “negative staining” is defined as ≤5% stained cells, a classification that also relies on the reproducible identification of a minor subpopulation.
Cox et al.10 found MMP-9 expression in 88 (52%) of 169 specimens and Shou et al.11 reported MMP-9 expression in 78 (65%) of 119 tumors. To verify these 2 previous studies with immunohistochemical detection in 52–65% of specimens, we calculated a survival curve for a combined group of patients with heterogeneous and homogeneous expression that exhibited a similar rate of immunohistochemical MMP-9 detection amounting to 69 (53%) of 130 eligible specimens. Interestingly, in this analysis, the survival of patients with homogeneous and heterogeneous MMP-9 expression was almost equal to the survival of MMP-9-negative patients (Fig. 3) and contrasted with the data of Cox et al.10 and Shou et al.11 Focusing on homogeneous expression, we demonstrated that exclusively homogeneous MMP-9 expression is of prognostic impact in operable NSCLC (Fig. 2).
Although Shou et al.11 did not focus on homogeneous MMP-9 expression, they reported an association between MMP-9 expression and unfavorable outcome in univariate log-rank analysis. In their study, MMP-9 expression correlated with the pT-status and the pN-status, suggesting that the observed prognostic value of MMP-9 could result from possible joint effects with the pT-status or pN-status. A multivariate regression analysis that could examine this possible prognostic interaction was not shown. Cox et al.10, 11 carried out a multivariate analysis, which only considered the pN-status of the standard prognostic factors included in the international system for staging lung cancer.6 In contrast, our study analyzed the prognostic impact of MMP-9 in relation to all established prognostic factors of the TNM-system, showing for the first time that the prognostic impact of homogeneous MMP-9 expression remains independent from possible prognostic joint effects of pT-status, pN-status, tumor histology and tumor grading.
Yet, it has to be mentioned that Immunohistochemistry can not distinguish between the proenzyme and the active form of MMP-9 because of the structural similarities of both forms.8, 13 To date, activity of MMPs in tissue sections can only be measured using in situ zymography.17, 18 For this method however, fresh cryosections of primary tumors are needed. Due to logistical reasons, such specimens were not available in the present study examining paraffin sections of a well characterized population with a follow-up of 12 years. Nevertheless, because homogeneous expression of MMP-9 is associated with unfavorable outcome in our patients, it may be assumed that active MMP-9 contributes to the effects of homogeneously expressed MMP-9. Verification of the prognostic impacts of different MMPs might be helpful in choosing selective MMP inhibitors as adjuvant therapeutic agents. Recent Phase III trials evaluating the efficacy of broad spectrum MMP inhibitors in small cell lung cancer19, 20 and Stage IV NSCLC.21 were disappointing. Selective MMP-inhibitors might be of more benefit than the tested broad spectrum MMP inhibitors, because side effects that could promote tumor progression are supposed to be reduced in the former ones.22
Moreover, subgroup analyses of Phase III trials evaluating the efficacy of MMP inhibitors in gastric and pancreatic cancer illustrated that improvement of survival by MMP inhibitors can be found in patients with low tumor burden or early stage disease.23 As patients with surgically removable cancers frequently have small tumor burden or early stage disease,24 MMP inhibitors might be of benefit in adjuvant therapeutic settings in operable NSCLC. Therefore, MMP-9 could be considered as a potential target for studies investigating the benefit of adjuvant MMP inhibitor therapy after surgery for NSCLC. For such studies it is recommended to measure the targeted MMP in tumor tissue to identify patients who are more likely to respond to MMP inhibitor therapy.9
Our data suggest that for immunohistochemical examination of MMP-9 expression in tumor specimens, an evaluation system that focuses on homogeneous expression is recommended. The present study may provide a basis for studies that examine the value of homogeneous MMP-9 expression for the preselection of patients to be included in trials investigating specific protease inhibitor therapy after surgical resection of NSCLC.
This work was supported by grants from the Sonderforschungsbereich (SFB) 469 of the Ludwig-Maximilians University, Munich, Germany. We would like to thank A. Wanger for her support in the manuscript preparation. We also gratefully acknowledge the technical assistance of C. Huber.