Receptor tyrosine kinases play an important role in malignant transformation of epithelial cells by activating signal transduction pathways important for proliferation, invasion and metastasis. In a pilot study (n = 40), we evaluated expression of the c-Met and Her2/neu receptor tyrosine kinases and the c-Met ligand hepatocyte growth factor/scatter factor (HGF/SF) in primary breast cancers and their lymph node metastases using both conventional immunohistochemistry and confocal immunofluorescence. Neither c-Met and HGF/SF nor Her2/neu expression correlated with established prognostic factors such as age, lymph node involvement, estrogen receptor (ER), progesterone receptor (PR), tumor size, or grade. Both staining methods confirmed a significant correlation between c-Met overexpression and a high risk of disease progression. Furthermore, among tumors with c-Met overexpression, only 50% also overexpress Her2/neu, thus identifying a subset of patients with aggressive disease in addition to Her2/neu. Median disease-free survival in patients with c-Met overexpressing tumors was 8 months compared to 53 months when c-Met expression was low (p = 0.037; RR = 3.0). This significant impact of c-Met on tumor aggressiveness independent of Her2/neu was also confirmed by multivariate analysis. In conclusion, the role of c-Met expression as a prognostic variable and consequently as an interesting target for novel therapeutic approaches deserves further analysis in a larger cohort of patients.
The Her2/neu gene encodes an Mr 185,000 transmembrane glycoprotein receptor that has partial homology with the other 3 members of the EGF receptor tyrosine kinase family. Overexpression of Her2/neu is associated with increased cell proliferation, antiapoptotic signals and invasion through activation of tumor-associated proteases, therefore contributing to cell transformation. Numerous clinical studies have demonstrated that abnormal Her2/neu expression in invasive breast cancer is correlated with aggressive disease. This led to the development of a humanized monoclonal antibody, trastuzumab, directed against the extracellular domain of Her2/neu. Clinical trials have shown that trastuzumab administered alone or together with chemotherapy provides clinical benefit in patients with metastatic breast cancer in which Her2/neu is highly overexpressed or amplified.1, 2 However, amplification of the Her2/neu gene or overexpression of its protein product is observed in at most 20–30% of human breast cancers and therefore most patients are upfront ineligible for treatment.
Given that the concept of inhibiting a receptor tyrosine kinase (RTK) such as Her2/neu has translated in efficient clinical treatment of breast cancer, we were interested in analyzing other RTKs that activate similar signaling pathways and compare their expression to Her2/neu. One interesting candidate is the RTK c-Met, the product of the protooncogene c-Met, which is overexpressed in many tumor types, including breast cancers, renal cell carcinomas, malignant gliomas and endometrial cancer.3 c-Met is an RTK that mediates the diverse activities of its ligand, hepatocyte growth factor/scatter factor (HGF/SF), which is a multifunctional heterodimeric growth and mitogenic factor. The concept that HGF/Met activation drives cells to be transformed, grow, invade and metastasize has been confirmed in cell culture and animal model systems. c-Met overexpression in breast tumors is associated with breast cancer progression and high HGF/SF correlates with poor survival in ductal breast carcinomas4, 5, 6, 7, 8, 9 (for a comprehensive literature review of the prognostic role of HGF/SF and c-Met in cancer, see http://www.vai.org/vari/metandcancer/index.aspx). In view of the important role of c-Met in tumor invasion and metastasis and the fact that Her2/neu is only positive in at most 20–30% of breast cancers, we investigated if c-Met and Her2/neu expression is linked in breast tumors. Using 2 independent assays, we show in node-positive breast cancer that c-Met is higher expressed in corresponding lymph node metastases than in the primary tumor and that its overexpression is associated with an unfavorable clinical outcome. c-Met overexpression identifies a different subgroup of patients at risk for recurrence than Her2/neu, suggesting c-Met as an interesting therapeutic target in addition to Her2/neu.
Material and methods
Patients and tumor samples
Clinical relevance of c-Met and Her2/neu expression was retrospectively evaluated in 40 primary breast cancer patients having 3 or more positive axillary lymph nodes. Patients received their primary therapy at the Department of Obstetrics and Gynecology of the Technical University of Munich, Germany, between 1989 and 1997. Informed consent for analysis of tumor biologic factors was obtained before surgery. Treatment decisions were based solely on consensus recommendations at the time. After surgery, 2 patients did not receive any adjuvant systemic therapy, 21 patients received adjuvant chemotherapy, 13 patients received endocrine therapy and 4 patients received combined chemoendocrine therapy. Median patient age was 54 years (range, 28–80 years).
Expressions of c-Met, HGF/SF and Her2/neu were determined by immunohistochemistry (IHC) in formalin-fixed, paraffin-embedded breast samples using the avidin-biotin peroxidase technique and the following primary antibodies: c-Met (C28; Santa Cruz Biotechnology, Santa Cruz, CA), Her2/neu (A0485; Dako, Glostrup, Denmark) and HGF/SF. The HGF/SF antibody was prepared in the Laboratory of Monoclonal Antibody Production at the Van Andel Institute and has been characterized and validated as described.10, 11 Paraffin sections were dewaxed and digested with 0.1% proteinase K (Sigma, St. Louis, MO). Endogenous peroxidase activity was blocked by 3% hydrogen peroxide in methanol for 30 min at RT followed by 4% milk/1% nonimmunized goat serum in PBS. The c-Met (1:300), HGF/SF (3 mg/ml) and Her2/neu (1:500) primary antibodies were incubated for 1 hr at RT. As positive controls, an NIH 3T3 cell line stably transfected with human Met and HGF/SF as well as Her2/neu overexpressing breast cancer tissue was used.12 Negative controls were performed by omission of the respective primary antibody and incubation with nonspecific mouse IgG2b. Immunohistochemical reactivity was scored independently without knowledge of clinical outcomes by 2 experienced pathologists (D.P., T.R.) according to the immunoreactivity score (IRS) first described by Remmele and Stegner13 and commonly used for hormone receptor analysis. For cytoplasmic c-Met (c-Metcyt) and HGF/SF staining, the percentage of stained tumor cells was grouped into 4 categories: < 10% = 1; < 40% = 2; < 70% = 3; > 70% = 4. A scale from 0 (no staining) to 3 (intense staining) was assigned for staining intensity and an immunoreactivity score from 0 to 12 was calculated by multiplying positivity with staining intensity. For c-Met membrane staining (c-Metmem), a score from 0 (no membrane staining) to 2 (strong membrane staining) was applied. Optimized cutoffs determined by log-rank statistics were used for c-Metcyt (score 6), c-Metmem (score 1) and HGF/SF (score 3) to discriminate between high and low immunostaining. c-Metcyt and c-Metmem were combined in order to assess total c-Met expression, c-Metmem + cyt. Her2/neu immunostaining was performed and scored (0 to 3+) according to the clinical routine protocol and 3+ was considered Her2/neu overexpression.
c-Met was stained by immunofluorescence (IF) using the same polyclonal antibody against c-Met (C28) and a rhodamine-labeled secondary antibody against rabbit immunoglobulin. The Her2/neu antigen was identified using a monoclonal antibody (Oncogene Science, Cambridge, MA). The indirect staining for Her2/neu was done using an FITC-labeled secondary antibody against mouse immunoglobulin. We used secondary-only stained controls as well as positive control cells from a cell line stably transformed with human c-Met that was prepared into a paraffin block. The staining was imaged using a Zeiss 310 laser scanning confocal microscope as described in detail previously.6, 14 Briefly, tumors were evaluated using the Nomarski optics to identify histopathology and tumor areas were then imaged using the appropriate laser line. Superimposing the fluorescent image onto the Nomarski image created overlay images. Individuals were evaluated using the Media Cybernetics image analysis program, Image Pro (Media Cybernetics, Silver Spring, MA). A region/mask was defined, the fluorescence intensity of that area measured and average staining intensity ranked against adjacent normal epithelium and the control samples. For statistical analysis, the computer-calculated expression data were grouped as follows: for c-Met and for Her2/neu, increased expression in tumor compared to surrounding normal tissue was considered overexpression. Lack of normal tissue in some sections did not allow determination of an IF ratio score.
Correlations between continuous variables were analyzed using the Spearman rank test. Associations between continuous and/or categorical variables were analyzed using the Mann-Whitney U-test, the Pearson chi-square test, the κ-test, or the McNemar test as appropriate.
Optimized cutoffs determined by log-rank statistics were used for Metcyt (score 6), Metmem (score 1) and HGF/SF (score 3) to discriminate between high and low immunostaining in the primary tumor tissue. For univariate analysis of disease-free survival (DFS), Kaplan-Meier curves were plotted and compared using log-rank statistics. Multivariate analysis was performed in a stepwise forward fashion by applying the Cox proportional hazards model using the SPSS software package.15
Immunostaining for c-Met, HGF/SF and Her2/neu
All 40 patients had lymph node-positive breast cancer, 29 of which had metastases to mobile ipsilateral axillary lymph nodes (pN1) and the remaining 11 had nodal metastases fixed to one another or to other structures (pN2). Median follow-up in patients still alive at the time of analysis was 70 months (range, 12–122 months); 21 patients (53%) experienced disease recurrence, the majority (n = 18) at distant sites. Median survival for all patients was 25 months. Her2/neu, c-Met and HGF/SF protein expression were analyzed by IHC and IF in primary tumors and corresponding lymph node metastasis. Two distinct IHC staining patterns were observed for c-Met: cytoplasmic (c-Metcyt) and membrane staining (c-Metmem; Fig. 1). c-Met membrane staining was observed only in breast cancer cells. c-Met cytoplasmic staining in tumor cells exhibited considerable variation between different tumors but was substantially stronger than in surrounding normal tissue. In normal breast tissue, only weak cytoplasmic c-Met staining was seen, with higher expression observed in nonneoplastic alterations such as apocrine metaplasia, adenosis and epithelial hyperplasia. In normal axillary lymph node tissue, no c-Met staining was seen and specific staining was found only in infiltrating tumor cells (Fig. 1). Determination of c-Met overexpression both by conventional IHC and confocal IF rendered concordant results (r = 0.41; p = 0.01; Table I, Fig. 1). Four out of 6 tumors with high total c-Metmem + cyt expression by IHC were classified as having strong c-Met expression compared to adjacent uninvolved breast epithelial cells by IF.
Table I. Distribution of c-Met, HGF/SF and Her2/neu in Primary Breast Carcinomas and Axillary Lymph Node Metastases as Assessed by IHC and IF
cyt, cytoplasmic staining; mem, membrane staining; mem + cyt, membrane and cytoplasmic staining. Small deviations from the total sample number (n = 40) are caused by insufficient tumor tissue on some sections for immunocytochemical evaluation or lack of sufficient adjacent benign tissue for immunofluorescence quantification.
IHC cutoffs optimized using log-rank statistics; IF expression is considered high if expression in tumor tissue is higher than in adjacent benign tissue.
Total c-Metmem + cyt (IHC)
Total c-Met (IF)
The expression pattern of c-Met differed significantly between primary tumors and lymph nodes (p = 0.027): 9 of 11 cases with high c-Met membrane expression (c-Metmem) showed overexpression of c-Metmem in the corresponding lymph node, while only 12 of the 28 cases with low expression in the primary tumor showed overexpression in the lymph node. As for total c-Metcyt + mem, 3 tumors showed high expression only in the primary tumor, 3 in both primary tumor and lymph node, 7 only in the lymph node metastasis and the remaining 26 had low expression in both tissues. For HGF/SF, only 7 of 24 cases with overexpression in the primary tumor also showed overexpression in the corresponding lymph node metastasis and an additional 2 cases overexpressed HGF/SF only in the lymph node. Her2/neu overexpression in the primary tumor corresponded to overexpression in the lymph node in 8 of 9 cases. Six additional cases overexpressed Her2/neu in the lymph node but not the primary tumor.
In primary tumor tissue, Her2/neu expression levels were significantly, though weakly, correlated with c-Metcyt (r = 0.39; p = 0.013), c-Metmem (r = 0.39; p = 0.013) and HGF/SF (r = 0.36; p = 0.028). c-Metmem, c-Metcyt, HGF/SF and Her2/neu expression levels in the primary tumor were not significantly correlated to established prognostic factors (age, number of involved lymph nodes, tumor size, grade, estrogen receptor (ER), progesterone receptor (PR).
Out of 40 node-positive breast cancer tissues, 9 showed Her2/neu overexpression, 12 overexpression of c-Metcyt, 11 of c- Metmem and 6 of total c-Metmem + cyt (Table I). Only 4 of the 9 patients with Her2/neu-positive staining overexpressed c-Metmem, while 7 additional patients overexpressed c-Metmem but not Her2/neu. Three of the 6 tumors with high total c-Metmem + cyt by IHC overexpressed Her2/neu, or the other way around: 6 of 9 cases had overexpression of Her2/neu but not of total c-Metmem + cyt. These results were confirmed using confocal IF analysis, where we found that in the 7 patients with c-Met overexpression, 3 overexpressed Her2/neu.
Patients with high c-Metmem by IHC had a median DFS time of 14 months, in contrast to 56 months for the patients with low c-Metmem expression; this difference did not reach statistical significance. However, when total c-Metmem + cyt was measured, the survival disadvantage associated with high c-Metmem + cyt overexpression as determined by IHC was significant (p = 0.021): A median DFS time of 8 months was seen in the high c-Metmem + cyt expressing breast tumors compared to 53 months in the low expressing tumors (Fig. 2). All relapses in the high-risk group occurred within the first 14 months after primary therapy. Interestingly, all 3 patients with c-Metmem + cyt as well as Her2/neu overexpression relapsed within 14 months of primary therapy. c-Metmem + cyt overexpression was associated with a 3-fold increased risk of relapse (p = 0.037; RR = 3.0; 95% CI = 1.1–8.3) in univariate Cox analysis. These results were confirmed for the breast tumor samples analyzed by IF (data not shown). Breast tumors with c-Met overexpression by IF had a significantly worse DFS (5-year DFS, 30%) than those with low c-Met (5-year DFS, 54%; p = 0.045; RR = 2.5; 95% CI = 1.02–6.01).
With regard to overall survival (OS), only c-Met overexpression by IF reached borderline significance (p = 0.067); tumors overexpressing c-Met by IHC (c-Metmem, c-Metmem + cyt, c-Metcyt) showed a nonsignificant trend toward shorter OS. Her2/neu overexpression had no significant prognostic impact in our patient collective. Even though there was a trend toward a better prognosis for patients overexpressing HGF, HGF as a single factor did not reach statistical significance with regard to DFS or OS. However, when stratifying the patients according to the combination of c-Metmem + cyt and HGF/SF, the subgroup of patients whose tumors overexpressed HGF/SF, but not c-Metmem + cyt, had a favorable prognosis with regard to DFS (Fig. 3) as well as OS (p = 0.0313). In contrast, overexpression of both HGF/SF and c-Metmem + cyt identified a subgroup of patients with adverse outcome (p = 0.009). Of the 6 patients with high HGF/SF and high c-Metmem + cyt, 5 patients had disease recurrence within 12 months of primary therapy and 4 of them died within the first 3.5 years.
Due to the rather small patient collective, multivariate Cox analysis was performed using only a limited number of variables: tumor size and hormone receptor status as established prognostic factors as well as the 2 strongest tumor biologic variables, total c-Metmem + cyt and the combination of c-Metmem + cyt and HGF/SF. For DFS, total c-Met overexpression (p = 0.007; RR = 17.6; 95% CI = 2.2–141.6) and the combination of c-Met and HGF were significant (p = 0.044; RR = 0.33; 95% CI = 0.1–0.97). For OS, none of the factors tested turned out to be significant.
In this study, we show that c-Met is overexpressed in a subgroup of patients with lymph node-positive breast cancer. c-Met expression is higher and more frequent in lymph node metastasis compared to the primary tumor. Moreover, overexpression of c-Met identifies a subset of patients with an adverse outcome independent of Her2/neu, thus making c-Met an interesting molecular target in addition to Her2/neu.
Ghoussoub et al.5 analyzed breast tissue from 91 patients with breast cancer for c-Met expression and showed that it is of prognostic significance in lymph node-positive and -negative patients. In the subgroup of lymph node-positive breast cancers (which corresponds to our group), 23% showed overexpression of c-Met as determined by immunofluorescence. A different study found higher expression of c-Met in breast cancer compared to adjacent normal tissue in 20% of the patients.6 In our study, 26% of all patients had c-Met overexpression by immunofluorescence, confirming these results. When we assessed c-Met expression by immunohistochemistry, 32% of tumors had high cytoplasmic c-Met expression, 27% had high membrane c-Met expression and 16% had high expression of c-Met in the cytoplasm as well as on the membrane. The present study is the first study comparing 2 different methods for c-Met detection and it shows that both immunofluorescence and IHC render similar results for c-Met overexpression and its clinical relevance. In one study, c-Met overexpression was found in 73% of all breast cancers. However, in this study, expression was defined as any staining in more than 30% of tumor cells without accounting for staining intensity, therefore including many cases that we and others would have rated as negative.8 We conclude from our study as well as others5, 6, 7 that 20–30% of breast cancers overexpress c-Met irrespective of nodal involvement. The simplicity of the immunohistochemical technique for c-Met as well as the available antibodies would make it easy to test for c-Met in patients with breast cancer.
We observed 2 distinct staining patterns for c-Met, cytoplasmic and membrane staining. Even though membrane staining seemed to be associated more strongly with tumor aggressiveness, the clinically most informative results were obtained by focusing on total c-Met expression (cytoplasmic + membrane staining) using either IF or conventional IHC. The fact that most relapses in patients with high c-Met expression occurred within the first 14 months hints at a failure of conventional adjuvant systemic therapy in these patients. c-Met overexpression was found both in primary tumor tissue and in corresponding lymph node metastasis, suggesting an association with tumor progression and metastatic potential. Additional high-risk patients were identified by high c-Met expression in the corresponding lymph node metastasis. Overexpression of both HGF/SF and c-Met in the primary tumor was associated with rather poor clinical outcome. Since our patient collective is small and treatment after relapse tends to be heterogeneous, prognostic information on DFS is likely to be more conclusive than information on OS.
Twenty to 30% of all breast cancer patients show Her2/neu overexpression; of these patients, only about 20% will eventually have a complete or partial response when treated with trastuzumab alone,1 thus leaving the majority of patients without the option to be treated by this approach.1, 16, 17 Therefore, it is highly desirable to identify and characterize other biologic targets in the group of patients who are not appropriate or do not respond to trastuzumab therapy. In view of the vast amount of in vivo and in vitro data suggesting a role of c-Met and its ligand HGF/SF in tumor progression and metastases, both are potential candidates as targets for breast cancer therapy. In a pilot study, we tested the hypothesis that c-Met, HGF/SF and Her2/neu identify different subsets of breast cancer patients. We conclude from our results using confocal IF and IHC that only half of the tumors that express c-Met concomitantly overexpress Her2/neu. In node-negative breast cancer patients, Ocal et al.7 have also found no correlation between Her2/neu, EGFR and c-Met expression as determined by immunohistochemistry. Thus, c-Met identifies a subset of patients with an adverse outcome independent of Her2/neu both in node-positive and -negative breast cancer. Her2/neu and c-Met expression within the same group of patients were also examined in gastric cancer. Similar to the findings in breast cancer, c-Met identified gastric cancer patients with adverse clinical outcome independent of HER2/neu.18
These clinical findings suggest that c-Met and Her2/neu exert different in vivo biologic effects in breast cancer cells. We know that at least in tissue culture, the phosphorylated c-Met receptor activates the PI3-kinase/AKT signaling pathway, the STAT3 cytokine pathway and through Gab1 the MAPK pathway, depending on which tyrosine in its cytoplasmic domain is phosphorylated.12, 19 In contrast, Her2/neu activates mostly the Ras/MAPK and protein kinase C pathway. In an elegant study culturing untransformed mouse mammary cells in 3-dimensional spheroids, Niemann et al.20 show that stimulation with HGF/SF induces branched tubule formation and proliferation of mammary cells, while neuregulin as an Her2/neu ligand induces fast-growing alveolar-like aggregates. Several therapeutic approaches for the inhibition of HGF/Met signaling have been employed in the past, targeting HGF/SF as well as c-Met. Binding of wild-type HGF/SF to the receptor was inhibited by HGF/SF antibodies; c-Met was inhibited by dominant negative constructs and by peptides. All these approaches show that HGF/Met signaling with its tumor-promoting effects can be inhibited at least in cell culture and in mouse models.10, 21 The next step would be development of a drug for clinical trials in order to evaluate if the concept of HGF/Met inhibition as an active cancer treatment is feasible. Given the findings of other investigators as well as the present study, patients with c-Met overexpressing breast cancer should be part of such studies.