Fax: 011 (33) 320538562
Tissue inhibitor of metalloproteinase 1 is an independent predictor of prognosis in patients with nonsmall cell lung carcinoma who undergo resection with curative intent
Article first published online: 7 MAR 2005
Copyright © 2005 American Cancer Society
Volume 103, Issue 8, pages 1676–1684, 15 April 2005
How to Cite
Gouyer, V., Conti, M., Devos, P., Zerimech, F., Copin, M.-C., Créme, E., Wurtz, A., Porte, H. and Huet, G. (2005), Tissue inhibitor of metalloproteinase 1 is an independent predictor of prognosis in patients with nonsmall cell lung carcinoma who undergo resection with curative intent. Cancer, 103: 1676–1684. doi: 10.1002/cncr.20965
- Issue published online: 4 APR 2005
- Article first published online: 7 MAR 2005
- Manuscript Accepted: 16 NOV 2004
- Manuscript Revised: 21 OCT 2004
- Manuscript Received: 7 JUL 2004
- matrix metalloproteinases;
- tissue inhibitors of matrix metalloproteinases;
- lung carcinoma;
Matrix metalloproteinases (MMPs) and tissue inhibitors of metalloproteinases (TIMPs) play a role in the processes of extracellular matrix degradation. Changes in their expression levels have been observed in various tumor types, including lung carcinoma. However, their clinical significance and their prognostic importance in the progression of nonsmall cell lung carcinoma (NSCLC) remain to be specified. In this study, mRNA expression levels of MMP-1, MMP-9, TIMP-1, and TIMP-2 were evaluated in patients with resected NSCLC, and their associations with disease progression and prognosis were determined.
Between June 1996 and December 1999, 116 patients underwent resection for NSCLC. Expression levels of MMPs and TIMPs were evaluated using Northern blot analysis in these NSCLC tissue samples and in 39 matched samples of normal lung tissue.
MMP-1, MMP-9, and TIMP-1 expression levels were increased in tumor samples compared with matched, corresponding normal tissues. In contrast, TIMP-2 expression was decreased in tumor samples. MMP-1 tumor expression was correlated significantly with the evolution of lymph node status and tumor-lymph node-metastasis (TNM) stage. In contrast, MMP-9 tumor expression was correlated significantly with increased T stage. TIMP-1 overexpression was an independent predictor of worse survival in patients with NSCLC that was not associated with other prognosis factors, such as TNM stage.
The overexpression of TIMP-1 was an independent prognostic marker in patients with NSCLC, and evaluating TIMP-1 may be important for identifying patients who are at greater risk of disease recurrence. Cancer 2005. © 2005 American Cancer Society.
Lung carcinoma is the leading cause of malignancy-related deaths in both men and women living in the Western world. Approximately 80% of lung carcinomas are classified histopathologically as nonsmall cell lung carcinoma (NSCLC) and include squamous cell carcinomas, adenocarcinomas (including the bronchoalveolar cell type), large cell carcinomas, and mixed types. Patient survival with NSCLC is correlated with the current international staging system.1
Surgery, either alone or associated with radiotherapy or chemotherapy, is the first-choice treatment for patients with localized tumors, including patients with Stage I and II NSCLC and some patients with Stage III NSCLC. Nevertheless, despite these multimodal treatments, patients with the same disease stage can show a wide range of 5-year survival rates, from 60% to 80% in patients with Stage I disease, from 39% to 55% in patients with Stage II disease, from 10% to 23% in patients with Stage IIIA disease, and < 10% in patients with Stage IIIB disease. Consequently, the development of a prognostic classification system based on molecular alterations is crucial to provide additional accurate and useful tools for selection of the most effective therapeutic options.2, 3
Degradation of the extracellular matrix (ECM) and penetration of the basement membrane play important roles in tumor invasion and metastasis development. Many proteases, including urokinase-plasminogen activator (u-PA), matrix metalloproteinases (MMPs), and cathepsins, have been associated with this ability.4 u-PA participates in the plasmin cascade that directly degrades components of the ECM.5 Plasmin also activates MMPs, allowing them to degrade the ECM.6 All members of the MMP family can be regulated by their endogenous inhibitors, the tissue inhibitor metalloproteinases (TIMPs), which bind MMP noncovalently in a 1:1 complex.7–9 The rate of ECM turnover is regulated by the balance between activated MMP levels and free TIMP levels.4 Four TIMPs have been identified (TIMP-1–TIMP-4).10 Although both TIMP-1 and TIMP-2 inhibit the activity of most MMPs, TIMP-1 preferentially forms a 1:1 complex with activated MMP-1, MMP-3, and MMP-2.8 Moreover, TIMP-1 forms a complex with pro-MMP-9 that blocks its activation by stromelysins.9 However, TIMPs exhibit a variety of other cellular activities. Accordingly, it has been shown that TIMP-1 has cell growth-promoting activity on human keratinocytes11 and on several other cell types.12
Among the proteinases, overexpression of u-PA was found in NSCLC,13, 14 and u-PA transcripts were found mainly within stromal cells but also within a smaller percentage of tumor cells.15 Both epithelial u-PA expression and stromal u-PA expression have been linked to tumor size, and stromal u-PA expression also have been linked to lymph node involvement.16 MMP-2 was reported as the most commonly expressed MMP in NSCLC, and its expression was located mainly in the stromal cells closest to tumor cells.17–19 MMP-2 overexpression was reported as an independent prognostic parameter for unfavorable outcome using log-rank and multivariate regression analyses in patients with NSCLC without lymph node involvement.20 MMP-9 was present in the tumor cells of squamous cell carcinoma21 and adenocarcinoma,18 and higher levels of MMP-9 have been found in patients with Stage III NSCLC compared with Stage I and II NSCLC.6 Increases in MMP-1 and MMP-11 expression levels were identified in patients with NSCLC compared with patients who had nonneoplastic lung disease,16–18, 21–23 and stromal MMP-11 expression was linked with tumor size and lymph node involvement in most lung carcinomas.16 Studies of TIMP expression have been controversial in NSCLC. Accordingly, it was found that TIMP expression decreased very early during tumor progression17 or was expressed strongly in tumor cells and increased progressively from the stage of epithelial atypia to invasive carcinoma.23–25
We previously analyzed the level of several proteases, including u-PA, MMP-2, MMP-11, cathepsin B, and cathepsin L, by Northern blot analysis in a series of 116 patients who underwent resection for NSCLC. We found that all studied proteases, except cathepsin L, were increased significantly in tumor. Moreover, the tumor:normal ratio of MMP-11 was linked significantly to lymph node involvement (P < 0.05).26 In the current study, we evaluated the expression of 2 other MMPs, MMP-9 and MMP-1, and 2 TIMPs, TIMP-1 and TIMP-2, in the same series of patients, comparing the expression of these MMPs and TIMPs with clinical and pathologic parameters and looking for a potential correlation with survival after long-term follow-up.
MATERIALS AND METHODS
From June 1996 to December 1999, 116 patients (95.7% men; mean age, 59 years; range, 31–79 years) who underwent surgery for NSCLC in the Division of Thoracic Surgery of University Hospital of Lille (Lille, France) were included in this study. Only patients who underwent complete resection with potentially curative intent were enrolled. The clinicopathologic features of the specimens were assessed according to the World Health Organization classification system27 and the tumor-lymph node-metastasis (TNM) staging system.1 Patients with unique metastatic sites who underwent surgery with curative intent (i.e., brain or adrenals) after or before lung resection and patients who received induction chemotherapy or mediastinal adjuvant radiotherapy also were enrolled.
Exclusion criteria were typical carcinoid tumors, patients with positive surgical margins, multiple metastatic spread, a history of extrapulmonary carcinoma, synchronous lung lesion histologically different from the resected NSCLC, and patients who died from surgical complications during their postoperative course. Clinicopathologic data from the patients are shown in Table 1. The time of last follow-up was July 31, 2002; the median follow-up duration was 31 months (range, 2–71 months). Fifty-three percent of patients died from recurrence of their primary lung carcinoma. The actuarial 5-year survival rate was 30.6%, and the median survival was 39.5 months. Conventional histopathologic variables known to influence prognosis were evaluated in addition to the biologic parameters studied, such as the size and subtype of the primary tumor, the degree of differentiation, or evidence of vascular invasion.
|Clinicopathologic parameters||All patients (n = 116 patients)||Patients with matched tumor and normal tissue samples (n = 39 patients)|
|Squamous cell carcinoma||58||14|
|Large cell carcinoma||2||0|
|Lymph node status|
|With capsular rupture||29||9|
Tumor specimens of NSCLC were taken from 116 patients. The tissues were frozen immediately in liquid nitrogen and stored at − 80 °C. Thirty-nine samples were matched with corresponding adjacent normal tissue, which was taken at a minimal distance of 5 cm from the tumor margins. The tumor was examined carefully for measurement in its greatest dimension in centimeters and for the presence of visceral pleura involvement, lymph node invasion, and vascular invasion. Morphologic control was done on 4-mm sections of each frozen specimen.
Northern Blot Analysis
Total RNA was isolated after homogenization of lung tissue in guanidinium isothiocyanate and centrifugation through cesium chloride gradients. Fifteen micrograms of total RNA were electrophoresed on 0.9% agarose-2.2 M formaldehyde gels and subsequently transferred onto nylon membranes (Hybond N+; Amersham Pharmacia Biotech, Rainham, U.K.). The membranes were hybridized overnight with 32P-labeled probes (random primed DNA labeling kit; Roche Molecular Biochemicals, Meylan, France). The MMP-2, MMP-11, cathepsin B, and cathepsin L probes were described in our previous study.26 MMP-1 was obtained from the American Type Culture Collection (Manassas, VA). The MMP-9 probe spanned nucleotides 757–2123 (British Biotech, Oxford, U.K.). The TIMP-1 and TIMP-2 probes spanned the complete coding sequence.28, 29
After hybridization, the blots were exposed to a Phosphoimager screen (Amersham Biosciences, Saclay, France). For normalization, the membrane was hybridized with the human actin complementary DNA (cDNA) probe. The results are expressed as the ratio between the labeling obtained after hybridization with the proteinase cDNA probe and the labeling obtained after hybridization with the actin cDNA probe.
Continuous variables are described as the mean ± standard deviation, and categorical variables are described as frequencies and percentages. The normality assumption of the various parameters was assessed with the Shapiro–Wilk test. Because several parameters did not appear to be distributed normally, we used nonparametric tests.
Comparison of means were performed using the Wilcoxon paired test, the Mann–Whitney test, or the Kruskall–Wallis test. Comparisons of frequencies were performed using the chi-square test or the Fisher exact test. Correlations were evaluated using the Spearman rank test.
The date of initial treatment was considered the starting day of observation; patients who died of other causes without evidence of recurrent disease or who were unavailable for follow-up were censored either at the time of death or at the last follow-up. The overall survival curves were calculated according to the Kaplan–Meier method. Analyses of survival according subgroups were performed using the log-rank test. Numeric parameters were separated into two categories by the determination of a cut-off value. The optimal cut-off value was determined by maximizing the hazard ratio30, 31 and was confirmed graphically using a Kernel density estimation, which is a nonparametric method for estimating and displaying the distribution of a numeric parameter.
A multivariate analysis was performed next using a stepwise Cox model. Differences were defined as statistically significant at P < 0.05. Statistical analyses were performed using SAS software (version 8.2; SAS Institute, Cary, NC).
Tumor specimens of NSCLC were taken from 116 patients. For 39 of 116 patients, corresponding adjacent normal tissue could be sampled. Expression levels of MMPs and TIMPs were analyzed in the 116 tumor specimens and in the 39 corresponding normal tissue specimens.
Expression of MMP-1, MMP-9, TIMP-1, and TIMP-2
Transcripts were analyzed by Northern blot analysis (an illustration is provided in Fig. 1). The levels of the different transcripts in lung carcinoma specimens and in control lung tissues were quantified by direct measurement of radioactivity on the membranes and then normalized to actin expression. The variation in RNA level of each gene in matched tumor and control tissues (n = 39 samples) is illustrated in Figure 2. The expression levels of MMP-1, MMP-9, and TIMP-1 were significantly greater in tumor samples compared with the corresponding control lung tissues (P < 0.01, P < 0.001, and P < 0.001, respectively) (Table 2). The mean fold increase in tumor tissue was 2.8-fold for MMP-1, 5.5-fold for MMP-9, and 7.3-fold for TIMP-1. In contrast, the expression of TIMP-2 was lower in tumor samples compared with the corresponding control tissues (P < 0.001).
|Expression||Normal tissue||Tumor tissue||P value|
|Paired tissue samples (n = 39)|
|MMP-1 (n = 26)||0.0047 ± 0.0033||0.0134 ± 0.0153||< 0.01b|
|MMP-9 (n = 39)||0.0076 ± 0.0110||0.0422 ± 0.0422||< 0.001b|
|TIMP-1 (n = 39)||0.0825 ± 0.0885||0.6033 ± 0.05321||< 0.001b|
|TIMP-2 (n = 39)||0.0442 ± 0.1091||0.0104 ± 0.0074||<0.001b|
|Normal tissue (n = 39) vs. all tumors (n = 116)|
|MMP-1||0.0047 ± 0.0033 (n = 26)||0.0270 ± 0.0380 (n = 104)||< 0.001c|
|MMP-9||0.0076 ± 0.0110 (n = 39)||0.0313 ± 0.0363 (n = 115)||< 0.001c|
|TIMP-1||0.0825 ± 0.0885 (n = 39)||0.6902 ± 1.0333 (n = 116)||< 0.001c|
|TIMP-2||0.0442 ± 0.1091 (n = 39)||0.0147 ± 0.0154 (n = 116)||0.2416c|
The expression of these genes was then analyzed in the whole collection of tumor samples (n = 116 samples), and the mean values in tumor tissues were compared with the mean values in control tissues (n = 39 samples) (Table 2). For MMP-1, MMP-9, and TIMP-1, the mean expression in tumor tissues was significantly higher than the mean expression in control tissues (P < 0.001). In contrast, no significant difference was observed for TIMP-2.
Correlations with Clinicopathologic Parameters
Correlations between tumor expression of MMP-1, MMP-9, TIMP-1, and TIMP-2 and conventional clinicopathologic parameters with prognostic significance in patients with NSCLC were analyzed (n = 116 patients). Comparison of MMP-1 tumor expression with TNM stage showed a progressive increase of tumor MMP-1 expression from Stage IA to Stage IIB, and statistical significance was achieved between Stage IA and Stage IIB (P = 0.016) (Fig. 3). No correlation was found between MMP-1 expression and T classification.
In contrast, MMP-9 overexpression did not show any relation with lymph node involvement or tumor stage but increased with T classification (Fig. 4). A statistical difference was found for T2 tumors versus T1 tumors (P = 0.017) and for T4 tumors versus T1 tumors (P = 0.003). For TIMP-1 and TIMP-2, no significant correlation was observed between expression levels and clinicopathologic parameters.
Correlations between Expression Levels of MMPs and TIMPs
The existence of eventual correlations between different transcripts was examined (Table 3). Correlations were found for MMP-9 and TIMP-1 (n = 115 patients; r = 0.423; P < 0.001), for MMP-9 and TIMP-2 (n = 115 patients; r = 0.412; P < 0.001), for MMP-1 and TIMP-2 (n = 104 patients; r = 0.360; P < 0.001), and for TIMP-1 and TIMP-2 (n = 116 patients; r = 0.344; P < 0.001).
|Parameter in tumor tissue||MMP-1||MMP-9||TIMP-1|
|No. of patients||103|
|P value||0.113||< 0.001|
|No. of patients||104||115|
|P value||< 0.001||< 0.001||< 0.001|
|No. of patients||104||115||116|
Analysis of Survival
We analyzed the correlations between expression levels of MMPs and TIMPs and survival. In addition, correlations of other previously analyzed proteases (u-PA, MMP-2, MMP-11, cathepsin B, and cathepsin L) were studied in the same cohort of patients. The cut-off values were used first to determine subsequent groups, and the survival difference was studied by log-rank test. Univariate analysis revealed that tumor stage, pathologic lymph node classification (TNM-N), and TIMP-1 expression were significant prognostic factors (P = 0.001, P < 0.001, and P = 0.0123, respectively). For TIMP-1 expression, 96 patients (82.8%) had tumor TIMP-1 expression < 1.2, and 20 patients (17.2%) had tumor TIMP-1 expression ≥ 1.2. The median survival for patients who had tumor TIMP-1 expression < 1.2 was 57.8 months, whereas the median survival for patients who had tumor TIMP-1 expression ≥ 1.2 was 21 months (P = 0.0123; log-rank test). The respective Kaplan–Meier curves are presented in Figure 5 and show 5-year survival rates of 48.6% for patients with tumor TIMP-1 expression < 1.2 and 22.6% for patients with tumor TIMP-1 expression ≥ 1.2 (P = 0.042; chi-square test). In contrast, no survival difference was found for pathologic tumor classification (TNM-T), MMP-1, MMP-2, MMP-9, u-PA, cathepsin B, cathepsin L, or TIMP-2.
A multivariate Cox regression analysis was used to evaluate whether the relation between TIMP-1 overexpression and shortened disease-related survival resulted from an association of TIMP-1 with other prognostic factors or whether TIMP-1 maintained its own prognostic value. TNM classification (P = 0.004) and TIMP-1 overexpression (P = 0.016) were identified as independent prognostic predictors (Table 4).
|Prognostic factor||Adjusted HR||95% CL||P value|
|Stage (I and II vs. III and IV)||2.764||1.388–5.503||0.004|
|TIMP-1 > 1.2||2.207||1.159–4.203||0.016|
Tumor invasion, metastasis, and angiogenesis require controlled degradation of ECM, and increased expression of proteases is associated with these processes in malignant tumors. Our previous study of proteases (u-PA, MMP-2, MMP-11, cathepsin B, and cathepsin L) by Northern blot analysis showed that the expression of MMP-11 was linked to lymph node involvement.26 In the current work, the study of proteases was extended to look for independent prognostic markers.
Increased expression of MMP-1 and MMP-9 was found in tumor tissue compared with corresponding normal tissue. This finding is in accordance with several studies mainly based on Northern blot, immunohistochemistry, in situ hybridization, or radioimmunoassay (RIA) analysis.16, 18, 21, 23, 32–34 TIMP-1 and TIMP-2, respectively, showed an increase and a decrease in tumor tissues compared with corresponding normal tissue. Although it has been reported that TIMP-1 acts as a metastasis suppressor gene, some studies using Northern blot analysis, immunohistochemistry, and RIA have shown increased expression of TIMP-1 in tumor cells, in accordance with the current results.23, 35, 36 In contrast, although TIMP-2 was expressed by tumor cells, tumor expression was generally at a similar level or was decreased compared with control cells.17, 18
In the current study, MMP-1 expression increased progressively with tumor stage up to Stage IIB, with a significant difference between Stage IA disease and Stage IIB disease. In contrast to MMP-1, MMP-9 expression was correlated with an increase in T classification, suggesting that MMP-9 expression may be involved in the local steps of tumor proliferation. High MMP-9 expression was associated previously with Stage III NSCLC.6 In our previous, study we found a correlation between MMP-11 overexpression and lymph node involvement in the same series of patients.26 Together, these data suggest a specific role for each MMP at different steps in the progression of NSCLC. In contrast, no correlation between TIMP-1 or TIMP-2 and the different clinical parameters was found, in accordance with the majority of the studies.17, 35 Only one study using immunohistochemistry showed an association between TIMP-1 overexpression and more advanced stage disease.24
To identify independent prognostic markers, expression values of MMPs and TIMPs were compared with the overall survival of patients using log-rank and multivariate regression analyses. No prognostic markers of survival were observed for MMP-1, MMP-9, or TIMP-2 in either analyses or for the different markers that were evaluated in our previous study (i.e., u-PA, MMP-2, MMP-11, cathepsin B, and cathepsin L).26 Herbst et al. also found no prognostic value for MMP-2 and MMP-9 mRNA expression in NSCLC,37 but the results for E-cadherin expression (i.e., the mean ratio of the expression of MMP-2 and MMP-9 to E-cadherin expression) indicated significant prognostic value. However, a prognostic value was reported for MMP-9 evaluated by immunohistochemistry in NSCLC.32, 38 In contrast, our study showed that TIMP-1 mRNA expression had a prognostic impact in our series of patients with NSCLC. Accordingly, in a univariate analysis, patients who had TIMP-1 expression (ratio to actin expression) ≥ 1.2 had a shorter survival compared with patients who had TIMP-1 expression < 1.2 (P = 0.0123). Furthermore, multivariate regression analysis showed that TIMP-1 expression was a significant, independent prognostic predictor for shortened survival in patients with NSCLC (P < 0.01) that was not linked to other prognostic factors, such as TNM stage. TIMP-1 expression was analyzed in connection with overall survival in a previous study of 45 patients with NSCLC by Kaplan–Meier analysis and also was associated with a poor outcome.35 Our current results confirm the association of TIMP-1 expression with overall survival, and the multivariate regression analysis further shows that TIMP-1 is an independent prognostic factor for survival in NSCLC. A recent immunohistochemical study also showed that overexpression of TIMP-1 protein was associated with an adverse outcome.39
An interesting question arising from these observations concerns the possible mechanisms that may account for a putative MMP inhibitor that potentially is associated with tumor progression. Indeed, it is possible that TIMPs may be protective against carcinogenesis due to their capacity for inhibiting the degradation of basement membrane and ECM. However, in addition to their MMP-inhibitory activities, TIMPs may exert many other biologic functions.40, 41 It has been shown that TIMP-1 has erythroid-potentiating activity,28, 42 promotes cell growth,11, 12 and inhibits apoptosis.43, 44 A particular proangiogenic activity was shown for TIMP-1 through the enhancement of vascular endothelial growth factor expression in mammary carcinoma.45 In a mouse model of epithelial carcinogenesis, it was shown recently that TIMP-1 contributes functionally to neoplastic development.46
With regard to the role of MMPs in matrix degradation in tumor invasion, MMP inhibitors were studied as potential therapeutic agents, and both synthetic MMP inhibitors and natural MMP inhibitors, such as TIMPs, currently are being investigated. In the latter investigations, the knowledge of the role of a TIMP associated with poor prognosis reinforces the concept of the design of selective synthetic inhibitors or the use of individual TIMPs.41
The results of the current study highlight the important role of MMPs and TIMPs in the neoplastic progression of NSCLC, and the finding that TIMP-1 is an independent prognostic marker may be relevant clinically. TIMP-1 expression may be helpful in identifying patients with NSCLC who have a greater risk of recurrence after surgery who may benefit from adjuvant chemotherapy.
- 27Histological typing of lung and pleural tumors. International histological classification of tumors. Geneva: World Health Organization, 1999., , , et al.
- 31Cutpoint determination methods in survival analysis using SAS. In: SAS Institute Inc., editor. SAS user group, Volume 261–28. Cary, NC: SAS Institute Inc., 1999., , .