How ADAM-9 and ADAM-11 Differentially From Estrogen Receptor Predict Response to Tamoxifen Treatment in Patients with Recurrent Breast Cancer: a Retrospective Study

The ADAMs, which stands for a disintegrin and metalloproteinase, also known as MDCs, are a newly discovered family of membrane proteins. All ADAMs possess some or all of the following domains: a signal peptide, a propeptide, a metalloprotease, a disintegrin, a cysteine-rich domain, an epidermal growth factor–like domain, a transmembrane sequence, and a cytoplasmic tail. The propeptide might be involved in latency with activation upon loss, the metalloprotease domain in proteolysis, the disintegrin domain in adhesion, the cysteine-rich domain in fusion and adhesion, the epidermal growth factor–like domain in growth factor activity, and the cytoplasmic tail in cell signaling. The possession of these multiple domains with their potential functions makes them likely candidates to play a role in cancer cell invasion and metastasis. Indeed, some of these ADAMs have already been linked to various diseases including cancer (1–5). Despite the above findings, a definite role for any ADAM in either cancer formation, progression, or response to therapy, remains to be shown.

In recurrent breast cancer, steroid hormone receptor status is one of the variables often used to determine the choice of endocrine therapy. Thus far, tamoxifen is the most extensively used hormonal treatment, although only 50% to 60% of the treated patients will benefit (6–8). Because proteases such as the urokinase-type plasminogen activator have been shown to be associated with failure of tamoxifen therapy in patients with recurrent breast cancer (9, 10), we hypothesized that specific ADAMs might also be associated with therapeutic failure.

Thus far, >30 different ADAMs have been described, of which 19 appear in humans. For our study, we selected four ADAMs for which no pseudogenes have been described and which have already been shown to be expressed to some extent in human breast cancer. Of these four, only ADAM-11, also named MDC, does not possess an active matrix metalloproteinase (MMP)–like domain. However, based on its location within a loss of heterozygosity region of chromosome 17q21 (11, 12), ADAM-11 has been proposed to be a candidate tumor suppressor gene for human breast cancer (13, 14) and was therefore included in our study. The other three members included in this study, ADAM-9 (MDC9, meltrin-γ), ADAM-10 (Kuz, SUP-17, MADM), and ADAM-12 (meltrin-α) all possess an active MMP-like domain, and in addition have all been reported to be increased in malignant compared with normal breast tissue (3, 15). We also included two well-known members of the MMP-family, MMP-2 and MMP-9, which have also been reported to be increased in malignant compared with normal breast tissue (16–20). Because most MMPs are localized to the tumor stroma (17, 21), we suspected that this might also be the case for the ADAMs. We therefore compared mRNA levels measured from human breast tissue sections containing predominantly (>70%) tumor cells with those measured in sections containing at least 30% stromal cells.

In this report of our retrospective study which includes 259 patients with estrogen receptor (ER)–positive primary breast tumors treated with tamoxifen for recurrent breast cancer, we show that, especially in stroma-enriched primary tumors, ADAM-9 and ADAM-11 are able to predict the efficacy of first-line tamoxifen therapy.

Patients. The Medical Ethical Committee of the Erasmus Medical Center Rotterdam, the Netherlands, approved our study design (MEC 02.953). This retrospective study included 259 female breast cancer patients for which the following inclusion criteria were used: all patients should have measurable disease that was treated with tamoxifen as first-line treatment for metastatic disease; all patients underwent primary surgery for breast cancer; diagnosis took place between 1979 and 1996; the primary tumor should be ER-positive and >100 mg tissue should be available. Exclusion criteria were: neo-adjuvant therapy or adjuvant hormonal treatment; if the follow-up period during tamoxifen treatment was only 6 months or less and patient was still alive but showed no response or therapy was stopped for other reasons than progression (e.g., subjective or objective toxicity) during these 6 months; if previous other cancers were experienced (except basal cell skin cancer or early-stage cervical cancer stage Ia/Ib). Following the above criteria, 340 tumors were available for analysis. Of the tissues, 15% were excluded from this study because the sections contained <30% tumor cell nuclei (see below). Another 9% were excluded because of poor RNA quality (see below). The remaining 259 eligible patients were treated either with breast-conserving surgery (36%) or with modified mastectomy (64%). An axillary dissection was done in 94% of the patients (n = 244). Twenty-five patients received cyclophosphamide, methotrexate, 5-fluorouracil, whereas 17 patients received anthracyclin-containing adjuvant chemotherapy. Relevant clinicopathologic characteristics of the patients and their primary tumor are given in Table 1. Follow-up scheduling of physical and instrumental exams, which, depending on the type of metastasis, included computerized tomography scan, bone scan, magnetic resonance imaging, X-rays as well as plasma tumor marker levels, were done as described recently in detail (22). The date of diagnosis of metastasis was defined as that at confirmation of metastasis after symptoms reported by the patient, detection of clinical signs, or at regular follow-up. Twenty-four patients presented with distant metastasis at diagnosis or developed distant metastasis (including supraclavicular lymph node metastasis) within 1 month after primary surgery (M1 patients). These 24 patients and the 235 patients who developed a recurrence during follow-up (24 patients with local-regional relapse, 211 patients with distant metastasis) were treated with first-line tamoxifen (40 mg daily). Of the 235 M0 patients, the median time between primary surgery and start of therapy was 27 months (range, 4-164 months). At the time of surgical removal of the primary tumor, the median age of the patients was 58 years (range, 26-89 years), and at the start of tamoxifen therapy for recurrent disease, the median age of the patients was 61 years (range, 29-90 years). Response to tamoxifen therapy was defined by standard Unio Internationale Contra Cancrum criteria (23). Objective response was observed in 53 patients (12 complete remission and 41 partial remission), and 87 patients had an increase in tumor size of 25% or more, or showed new tumor lesions within 3 months (progressive disease). The 119 patients with no evident tumor reduction of 50% or more (complete remission and partial remission) or a tumor-progression (progressive disease), were considered as patients with no change. These patients with no change were divided into 103 patients who had no change at >6 months (defined as stable disease) and 16 patients with no change at ≤6 months. The median progression-free survival ratios were: complete remission, 37 months; partial remission, 16 months; stable disease, 14 months; no change at ≤6 months, 5 months; and for progressive disease, 3 months. Because the patients with stable disease had a progression-free survival similar to patients with partial remission, we classified these patients as responders to tamoxifen as advised by the European Organization for Research and Treatment of Cancer (24). Therefore, as has been done before (25–27), disease control was defined in our study as complete remission + partial remission + stable disease. For 156 patients (60%), disease was controlled by tamoxifen therapy. The median follow-up of patients alive after surgery was 90 months (range, 10-190 months) and 37 months (range, 4-131 months) after start of tamoxifen therapy. At the end of the follow-up period, 238 (92%) patients had developed tumor progression and 202 (78%) patients had died.

Tissue processing. After primary surgery, a representative part of the tumor was selected by the pathologist, frozen in liquid nitrogen, and sent to our laboratory for routine determination of ER and progesterone receptor (PgR) by ligand binding assay or enzyme immunoassay (28). Tumor cytosols were prepared and processed as recommended by the European Organization for Research and Treatment of Cancer (29). The cut-point used to classify tumors as ER- or PgR-positive was 10 fmol/mg cytosolic protein. The remainder of the tumor tissue was stored in our liquid nitrogen tumor bank at the Erasmus MC. For RNA isolation, 20 to 60 cryostat sections of 30 μm, corresponding to 30 to 100 mg, were cut from these tissues. Before, during, and after cutting the sections for RNA isolation, 5-μm sections were cut for H&E staining to assess the amount of tumor cells relative to the amount of surrounding stromal cells. The amount of nuclei evidently of epithelial tumor cell origin relative to the amount of surrounding stromal cells was estimated with a 100-fold magnification in 10 different areas covering the area of each of the three H&E sections. The fraction of tumor cells over stromal cells throughout the sections did not change greatly between the first and last section (mean coefficient of variation, 6%). Only specimens with at least 30% of the nuclei evidently of epithelial tumor cell origin and distributed uniformly over at least 70% of the section area were included.

RNA isolation and cDNA synthesis. Total RNA was extracted with RNABee (Campro, Veenendaal, the Netherlands) according to the manufacturer and stored aliquoted in RNase/DNase-free water at −80°C. Five micrograms of total RNA sample aliquots were reverse-transcribed with oligo(dT)_12-18 and random hexamer primers in a final volume of 40 μL using the Superscript II RNase H- kit from Invitrogen (Breda, the Netherlands) and used according to the manufacturer's instructions. Prior to PCR, the resulting cDNA samples were treated for 30 minutes at 37°C with four units of RNase H- (Ambion, Huntingdon, United Kingdom). The quantity and quality of the isolated RNA was established by UV spectroscopy, by examination of rRNA bands after agarose gel electrophoresis, and by the ability of the sample to be linearly amplified in a serial dilution with our housekeeping gene set (see next section for further details). Samples of total RNA not showing both the 18S and 28S bands (6%) or at 15 ng reverse-transcribed total RNA not amplifiable within 26 cycles at our fixed threshold value of 0.02 (see below) with our housekeeping set, which was the case for 3% of our samples, were excluded from this study.

Quantification of specific mRNA species. Real-time quantitative PCR was done in an ABI Prism 7700 Sequence Detection System (Applied Biosystems, Nieuwerkerk a/d IJssel, the Netherlands) using both the Assay-on-Demand kits from Applied Biosystems and the intron-spanning forward and reverse primer combinations shown in Table 2. PCR reactions were done in a final volume of 25 μL containing cDNA synthesized from 5 to 15 ng of total RNA, 330 nmol/L forward and reverse primer and 12.5 μL SYBR-green PCR master mixture (Applied Biosystems) or Brilliant SYBR Green Master Mix (Stratagene, Amsterdam, the Netherlands). For the Assay-on-Demand kits, the protocol with 40 rounds of amplification recommended by the manufacturer was used. For the SYBR-based assays, the following protocol was used. After 10 minutes of denaturation and activation of the Taq-DNA polymerase, PCR products were amplified in 35 cycles with 15 seconds of denaturing at 95°C, 30 seconds of annealing at 62°C, 10 seconds of ramping to 72°C, 20 seconds of extension at 72°C, 10 seconds of ramping to 79°C, and 20 seconds at 79°C. To avoid possible detection of primer-dimers, which usually melt at lower temperatures, SYBR green fluorescent signals of the products were acquired after each cycle at 79°C for PCR products with melting temperatures >80°C and only at 72°C for those with melting temperatures <80°C. A reference dye, ROX, was included in all assays to normalize data for non–PCR-related signal variation. Initial PCRs followed by product-melting curve analyses and gel electrophoresis experiments were done to ensure that with the PCR conditions and the different primer sets used, only one product of the expected size was amplified, and that for each gene, an additional cycle resulted in a doubling of PCR product, i.e., that all genes were amplified with an efficiency of at least 95%. In addition, the PCR efficiency of each gene-specific real-time PCR session was validated with a standard curve constructed from a simultaneously run serially diluted cDNA pool of human breast fibroblasts and cell-lines. Negative controls included samples without reverse transcriptase and samples where total RNA and cDNA was replaced with genomic DNA. Quantitative values were obtained from the threshold cycle (Ct) at which the increase in SYBR green or TaqMan probe fluorescent signal associated with an exponential increase of PCR products reached the fixed threshold value of 0.02, which was in all cases, at least 10-fold the standard deviation of the background signal. To enable comparison of the levels of specific mRNAs in different samples, they were evaluated relative to the average expression levels of three housekeeping genes: the low abundance housekeeping gene porphobilinogen deaminase (PBGD), the medium abundance housekeeping gene hypoxanthine-guanine phosphoribosyltransferase (HPRT), and the high abundance housekeeping gene β-2-microglobulin (β2M). With this set of housekeeping genes, the potential influence of sample-specific fluctuations in one of the housekeeping genes will be minimized. Levels of the target genes expressed relative to this housekeeping set were quantified as follows: mRNA target = 2^{(mean Ct housekeeping genes − mean Ct target)}.

Immunohistochemistry. To assess the source of the relevant mRNA species for this study, formalin-fixed, paraffin-embedded breast tumor tissues were analyzed by immunohistochemistry. Formalin-fixed, paraffin-embedded tumors were sectioned at 5 μm, mounted on StarFrost slides, dried, deparaffinized in xylene and rehydrated in graded solutions of ethanol and distilled water. Prior to immunostaining, specimens were pretreated with 1 mmol/L EDTA (pH 8.0) for 10 minutes at 121°C in an autoclave, cooled to room temperature, rinsed in PBS followed by a 15-minute peroxidase (0.3%) and a 30-minute bovine serum albumin (5%) block. The following primary antibodies were used: anti-ADAM-9 goat polyclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA; clone C-15; dilution 1:200); anti-ADAM-11 goat polyclonal antibody (Santa Cruz Biotechnology, clone H-19; dilution 1:200); anti-PR mouse monoclonal antibody (Dako Diagnostica GmbH, Hamburg, Germany; clone 1A6; dilution 1:320); anti-ER-α mouse monoclonal antibody (Dako Diagnostica; clone 1D5; dilution 1:320). After the primary antibody, ADAM-9 and ADAM-11 immunoreactions were visualized by a standard streptavidin-biotin-peroxidase complex (Strept ABC) method (DAKO, Diagnostica GmbH, Hamburg, Germany) followed by 3,3′-diaminobenzidine enzymatic development. ER-α and PgR were visualized using the DAKO EnVision+System-HRP mouse kit (DAKO). Sections were counterstained with hematoxylin. The specificity of immunostaining was controlled using normal goat and mouse IgG and by omitting the primary antibodies.

To assess the correlation between ER-α and PgR mRNA and protein levels, ER-α and PgR immunoreactivity was also assessed in 108 randomly selected frozen sections matching the frozen sections used for RNA isolation. These sections were cut, fixed in 4% paraformaldehyde, and analyzed for ER-α and PgR immunoreactivity as described above for the paraffin-embedded samples, except for the deparaffinization and pretreatment steps and with antibodies diluted 1:320. The percentage of tumor cells with positive nuclei was estimated with a 100-fold magnification in 10 different areas covering the section and scored in five categories as follows: 0% (1), <10% (2), 10% to 25% (3), 25% to 50% (4), >50% (5).

Statistics. Differences in levels were assessed with the Mann-Whitney U test or Kruskal-Wallis test, including a Wilcoxon-type test for trend, when appropriate. In these tests, patient and tumor characteristics were used as grouping variables. The strengths of the associations between continuous variables were tested with the Spearman rank correlation (r_s). For the analysis of treatment benefit, transformations of the variables were explored with fractional polynomials. The gain in χ² values was not substantial when using transformations other than log-transformations. Fractional polynomials did not result in statistically significantly better fit. The relation with disease control-to-therapy was examined with logistic regression analysis. Odds ratios (OR) were calculated and are presented with their 95% confidence interval (CI). The likelihood ratio test in logistic regression models was used to test for differences. The Cox proportional hazard model was used to calculate the hazard ratio and 95% CI in the analysis of progression-free survival. Progression-free survival was the time that the patients were treated with tamoxifen as first-line systemic treatment for recurrent disease. The start of tamoxifen therapy was set at zero and the end point at the stop-date of tamoxifen therapy or last date of follow-up. The proportionality assumption was investigated using a test based on the Schoenfeld residuals (30). Three equal thirds were used to categorize the variable to low, intermediate, and high. Survival curves were generated using the method of Kaplan and Meier (1958) and the log-rank test was used to test for differences. All P values are two-sided and P < 0.05 was considered statistically significant. Computations were done with the use of STATA statistical package, release 8.2 (STATA Corp., College Station, TX).

Correlations between biological factors. To verify that the sections used for RNA isolation were representative of the whole tumor with respect to ER and PgR levels, all samples were analyzed for ER and PgR mRNA expression. In agreement with the selection of ER protein–positive samples, none of the RNA samples tested negative for ER mRNA. In addition, ER and PgR mRNA levels correlated significantly with the amount of ER or PgR protein as measured in the cytosols (Spearman rank correlation, r_s = 0.62; P < 0.001 for ER, n = 259; and r_s = 0.63; P < 0.001 for PgR; n = 255) and by immunohistochemistry (Kruskal-Wallis test: χ² = 31.09; df = 4; P < 0.001 for ER, n = 108; and χ² = 55.95; df = 4; P < 0.001 for PgR, n = 108). Spearman rank correlation further revealed meaningful (defined as P < 0.001 for n = 250 to 259) correlations between ADAM-9 and ADAM-10 (r_s = 0.28), ADAM-12 (r_s = 0.28), MMP-2 (r_s = 0.36), and MMP-9 (r_s = 0.27). In addition, ADAM-10 correlated with ADAM-12 (r_s = 0.41), MMP-2 (r_s = 0.34), and MMP-9 (r_s = 0.24), ADAM-12 with MMP-2 (r_s = 0.69) and MMP-9 (r_s = 0.34), and MMP-2 with MMP-9 (r_s = 0.34). ER-α mRNA only correlated with PgR mRNA (r_s = 0.25), and ADAM-11 showed no correlation (P < 0.001) with any of the biological factors studied.

Associations of the expression levels with clinicopathologic factors. The associations of clinicopathologic factors with the biological factors at the median mRNA level are depicted in Table 1. None of the mRNA levels correlated with the dominant site of relapse or disease-free interval. ER-α mRNA levels were inversely related with grade and were higher in tumors from postmenopausal patients compared with premenopausal patients. MMP-2 and MMP-9 mRNA expression levels were inversely related with tumor size, and ADAM-10 expression levels varied between histologic subtypes. The association with nodal status is less straightforward. Although PgR mRNA levels in these ER-positive tumors were significantly lower in node-negative patients, ER-α, ADAM-9, ADAM-11, and ADAM-12 mRNA levels were negatively related with the number of positive lymph nodes.

Univariate and multivariate analysis for disease control. In our analysis of the predictive value of the ADAMs and the MMPs, the main clinical end point was the measurable effect of tamoxifen therapy on tumor size (disease control) from the start of therapy. In univariate analysis using log-transformed continuous variables, increasing levels of ER-α, PgR, ADAM-9, and ADAM-11, and decreasing levels of MMP-9 predicted a favorable disease control (Table 3). In contrast, no significant associations with treatment benefit were observed for ADAM-10, ADAM-12, and MMP-2 (Table 3). The predictive value of the factors for disease control was studied with multivariate logistic regression analysis (Table 3). For this multivariate analysis, we used the same base multivariate model including the traditional predictive factors as described previously for a larger group of 691 patients treated with first-line tamoxifen for recurrent disease (10). This base multivariate model includes the traditional predictive factors menopausal status, dominant site of relapse, disease-free interval, and ER and PgR tumor levels. The contributions of the biological factors that were shown to be significantly related with benefit of tamoxifen treatment in the univariate analysis were separately included as log-transformed continuous variables (Table 3). The analyses showed that only ADAM-9 (OR, 1.41; P = 0.015) and MMP-9 (OR, 0.81; P = 0.035) provided additional predictive information over the traditional predictive factors of the base model.

Effect of tumor cell percentage. Because most MMPs are localized to the tumor stroma, we hypothesized that this might be the case for the related ADAMs as well. Therefore, we also checked for possible correlations between the fraction of tumor cell nuclei (range, 30-90%; median, 70%) and stromal-derived cell nuclei in the primary tumor and the predictive power of the various biological factors with respect to tamoxifen benefit. For this, we split our samples at the median level of 70% tumor cell nuclei. The respective groups consisted of n = 106 patients with >70% tumor cell nuclei (>70% TC) and a group of n = 153 patients containing ≤70% tumor cell nuclei (≤70% TC) in their primary tumor. In the group of patients with >70% TC tumors (Table 4), only mRNA levels of ER-α (OR, 2.10; P = 0.004) and PgR (OR, 1.24; P = 0.025) showed significant correlations with treatment outcome. On the other hand, whereas the traditional predictive factors ER-α and PgR lost their predictive power in the group of tumors containing at least 30% stromal cells, ADAM-9, ADAM-11, and MMP-9 mRNA levels gained predictive power with respect to benefit of tamoxifen treatment (OR, 1.59; P = 0.007 for ADAM-9; OR, 1.65; P = 0.001 for ADAM-11; and OR, 0.78; P = 0.045, for MMP-9, respectively; Table 4).

Having established that the predictive power of some of the biological factors, as log-transformed continuous variables, depended on the cell type composition of the primary tumor, we extended our analysis for these factors. For this, we explored the significance of these factors with respect to disease control and progression-free survival after categorizing the mRNA levels of the biological variables in the specific tumor cell subpopulations in three equal thirds (low, intermediate, high; Table 5). Figure 1 shows progression-free survival as a function of the categorized ER-α tumor levels in all 259 patients (Fig. 1A) against the 106 patients with >70% TC tumors (Fig. 1B) and likewise for ADAM-9 in all patients (Fig. 1C) against the 153 patients with ≤70% TC tumors (Fig. 1D). Because the proportional hazards assumptions for ER-α, ADAM-9, and ADAM-11 were violated for the total follow-up time of 130 months (P < 0.005), we analyzed the relationships of these factors with progression-free survival during the first 9 months of follow-up, the time that half of the patients treated for advanced disease showed disease progression on tamoxifen (10, 31). In this short-term analysis (Fig. 1), with 130 failures in all 259 patients, 53 in the group of 106 patients with >70% TC tumors, and 77 failures in the 153 patients with ≤70% TC tumors, the proportional hazards assumption was no longer violated (P > 0.1) for any of the factors further evaluated. We therefore restricted our exploration of the relationship of the factors with progression-free survival to failures during the first 9 months of follow-up (Table 5). The most notable findings in these univariate analyses were: (a) that MMP-9 mRNA levels analyzed as a categorized variable lost its predictive value and (b) that ER-α mRNA levels when measured in >70% TC tumors and ADAM-9 and ADAM-11 mRNA levels when measured in ≤70% TC tumors were strong predictors for disease control by first-line tamoxifen therapy and for the length of progression-free survival after the start of treatment.

Localization of ER-α, ADAM-9, and ADAM-11 protein in human breast tissues. Finally, we employed immunohistochemistry to determine the location of ADAM-9 and ADAM-11 protein in our human breast tumor tissues and compared this with the location of ER-α protein. Representative results are shown in Fig. 2 for staining of preexistent mammary gland tissue (Fig. 2A-D), carcinoma in situ components (Fig. 2E-H), and lobular breast carcinomas (Fig. 2I-L). Whereas ER-α staining is mainly localized to the nuclei of tumor cells, ADAM-9 and ADAM-11 are most commonly found in the cytoplasm and less commonly at the cell membrane. Immunohistochemical staining of ADAM-9 and ADAM-11 protein in human breast carcinomas yielded heterogeneous results with both proteins found in tumor cells (Fig. 2J and K), adipocytes, smooth muscle cells of vessel walls, and the myoepithelial and luminal layers of nonneoplastic epithelium of the mammary gland (Fig. 2B and C).

Endocrine therapy is the most common treatment in breast cancer patients with tumors that express ER-α and/or PgR. Even though the ER-α is the prime target for endocrine therapy, the failure or success of this therapy is poorly understood. Systemic endocrine therapy in patients with recurrent disease at distant sites is merely palliative and accomplishes a disease control in about 50% to 60% of the patients. However, progression is inevitable in these patients because of the occurrence of acquired therapy resistance. From a biological point of view, first-line single-agent endocrine therapy in patients with recurrent breast cancer is an excellent setting to study response to therapy because it is less subject to prognostic influences unavoidably present when a similar study would be done in the adjuvant setting. In the present study, the effect of endocrine therapy on size of the metastatic or the occurrence of new lesions were used as the main clinical end point. We defined the type of response strictly beforehand, and when there was any doubt, patients were not included in this study. The size of the metastases or the occurrence of new lesions is an objective measure of treatment effect. However, because of the retrospective nature of our study, the differentiation between partial remission and no change was difficult to assess, especially in patients with bone metastasis (60%). In our study, the progression-free survival of patients with stable disease (no change >6 months) was comparable with the progression-free survival of patients with partial remission and could therefore be considered as responders. This is in agreement with a previously published prospective study which also reported that objective benefit was not always easy to assess and in which prolonged stable disease was categorized as response (6).

In this study, ER and PgR were determined in cytosols by biochemical methods and the cutoff used to classify tumors as ER- or PgR-positive was 10 fmol/mg cytosolic protein. These data correlated significantly with ER and PgR mRNA expression levels. However, although these quantitative procedures are the most accurate methods, it is not currently the most widely used method to evaluate hormonal receptor status in breast cancer. In fact, immunohistochemistry is nowadays more commonly used for routine ER and PgR measurements. Because this study shows a possible application for current clinical practice, we compared the biochemical and immunohistochemical methods in a randomly selected subgroup of patients. In agreement with a previously published study in which ligand binding assay and immunohistochemistry were compared in predicting response to tamoxifen in 205 patients with ER-positive metastatic breast cancer (32), ER and PgR levels also showed comparable differences in response rates in our study, whether defined by mRNA, by biochemical methods, or by immunohistochemistry.

The main findings of our study are that ADAM-9 and ADAM-11 differentially from ER predict the type of response to tamoxifen treatment in patients with recurrent breast cancer and that the fraction of tumor cells and stromal elements are important in this respect. The actual ER level in the ER-positive tumors (>10 fmol/mg cytosol protein) containing >30% stromal elements did not further contribute to the rate of response. This finding supports the results of a previous report that showed an association between ER level and the volume fraction of actual cancer cells present in the tumors (33). Therefore, it was advised that, when quantitative ER levels are used to predict the response of tumors to hormonal therapy, the cellularity of tumors should be taken into consideration. We followed this approach in our study by discriminating between tumors with >70% tumor cells and tumors with 30% to 70% tumor cells. Our results show that for tumors with a relatively low percentage of epithelial tumor cells, a marker set including ADAM-9 and ADAM-11 may have potential to assess the efficacy of tamoxifen therapy.

Of the ADAMs and MMPs studied, all, except ADAM-11, were readily detected by real-time PCR in all samples. The absence of detectable ADAM-11 mRNA levels in 29% of our primary breast tumors is most likely a reflection of the loss of heterozygosity on chromosome 17q21, where ADAM-11 is located (13), as described to be the case for 30% of the tumors (11). Patterns of copy number gains and losses define breast tumors with distinct clinicopathologic features and patient prognosis (34). Several studies have already shown that ERBB2 amplification is associated with a shorter disease-free and overall survival in the subgroup of patients receiving adjuvant tamoxifen therapy when compared with the untreated group (35–37). However, whereas ERBB2 is located on cytoband 17q12, a region of copy gain, ADAM-11 is located on cytoband 17q21, a region of copy loss. Our finding that low tumor ADAM-11 mRNA levels are associated with poor efficacy of tamoxifen treatment supports the hypothesis that ADAM-11 is a candidate tumor suppressor gene for human breast cancer (13, 14), and extends its role as a candidate tumor suppressor gene to a candidate tamoxifen susceptibility gene.

Our study shows that ADAM-9 and ADAM-11 mRNA levels are especially informative with respect to tamoxifen treatment outcome in tumors containing a relatively large proportion of stromal cells. In agreement with a previously published study describing the expression of ADAM-9, ADAM-10, ADAM-12, ADAM-15, and ADAM-17 in breast cancer specimens (38), immunohistochemical staining of ADAM-9 and ADAM-11 protein in human breast carcinomas yielded heterogeneous results with both proteins found in tumor cells, adipocytes, nonneoplastic epithelium of the mammary gland, and smooth muscle cells of vessel walls. The question of how ADAM-11 and ADAM-9, either stromal or tumor cell–derived, might prevent the development of tamoxifen resistance remains to be solved. Because proteases such as urokinase-type plasminogen activator (9, 10) and MMP-2 (39) have been shown to be related to tamoxifen resistance, we hypothesized that specific ADAMs might also be related to tamoxifen resistance. We found that high levels of ADAM-9 and ADAM-11 mRNA were related to a better response rate. This is in contrast with the findings for urokinase-type plasminogen activator (10), showing high levels to be associated with poor benefit of tamoxifen treatment in recurrent breast cancer, and for MMP-2 (39), showing that high levels predicted failure to adjuvant antiestrogen therapy. This suggests that ADAM-9 and ADAM-11 function differently from urokinase-type plasminogen activator and MMP-2, and that it is therefore perhaps not the protease function of the ADAMs that is important in the prevention or delay of tamoxifen resistance. Increasing evidence indicates that abnormalities occurring in growth factor signaling pathways, as currently well-documented for epidermal growth factor receptor (ERBB1) and ERBB2 (HER2/neu), could dramatically influence steroid hormone action and may be critical for anti–hormonal-resistant breast cancer cell growth (7, 8, 36, 40–43). From this point of view, one might expect factors that target growth factor signaling pathways are potentially able to prevent the development of tamoxifen resistance. Many intercellular signaling molecules are membrane-anchored proteins, which are proteolytically processed after becoming membrane-bound, to liberate their extracellular domains (ectodomain shedding). Genetic and biochemical studies have shown that some ADAMs participate in these events (3). Therefore, it is perhaps the ectodomain shedding function of the ADAMs that plays a role in the prevention of tamoxifen resistance. Furthermore, the disintegrin domain of ADAM-9 can function as an adhesion molecule by interacting with an α(v)β(5) integrin (44), thus limiting the metastatic potential of the cell.

In summary, our study shows that patients with primary tumors exhibiting a high percentage of tumor cell nuclei over stromal cells combined with high levels of ER-α have a good chance to benefit from tamoxifen therapy. For patients with tumors displaying ≥30% stromal components intermingled with epithelial tumor cells, the additional assessment of tumor mRNA levels of ADAM-9 and ADAM-11 could be helpful to refine treatment strategies for these patients. However, taking into account that only patients with ER-positive primary tumors entered this study, this may only apply to patients with ER-positive primary tumors. Further studies are required to verify whether the results of our study can be adapted to fit all patients, irrespective of the ER status of the primary tumor. Based on recent advances in breast cancer management, endocrine therapy with aromatase inhibitors may become the treatment of choice for postmenopausal women (45). Because both aromatase inhibitors and tamoxifen aim to deprive the ER from estrogens, it would be interesting to learn whether ADAM-9 and ADAM-11 could also be linked to disease control of aromatase inhibitors. In addition, as the majority of patients receive adjuvant treatment today, it will be important to learn whether ADAM-9 and ADAM-11 could also be informative for determining the outcome of breast cancer patients treated with adjuvant endocrine therapy.

We thank Maxime Look for her expert support with clinical data analysis and Iris van Staveren and Maaike Kiel for helping with the RNA isolation. We especially thank the surgeons, pathologists, and internists of the St. Clara Hospital, Ikazia Hospital, St. Fransiscus Gasthuis, Erasmus MC at Rotterdam, and Ruwaard van Putten Hospital at Spijkenisse for their assistance in collecting the tumor tissues and patient's clinical follow-up data.