P2-01-17: L1-CAM Promotes Adhesion of Breast Cancer Cells to the Endothelium

Background: Overexpression of the adhesion molecule L1 -CAM (L1) in breast cancer is correlated with nodal involvement, high grading, and a shorter recurrence-free and overall survival, but the mechanism leading to this effect is poorly understood. L1 is also expressed in endothelial cells promoting the interaction between lymphatic cells and endothelia via homophilic L1-L1 interaction. This mechanism might also mediate adhesion of breast cancer cells to endothelia and thus promote metastasis. To examine this likely role of L1, the impact of L1 expression in breast cancer cells on their adherence to human pulmonary microvascular endothelial cells (HPMEC) was investigated and potential L1 ligands on these cells were identified. Methods: MDA-MB231-Fra2 breast cancer cells expressing high levels of L1 were stably transfected with shRNA vectors to generate two clones with strong L1 downregulation (L1low clones). Adherence of these clones to endothelial cells was studied in dynamic cell-flow adhesion assays in channels seeded with HPMEC simulating in vivo-conditions. Rolling or adherent cells per minute were counted using CapImage software. For the cell-flow assay HPMEC were activated with TNFα. In order to identify the L1 ligand, endothelial cells were partly pre-incubated with blocking antibodies directed to the potential binding partners L1, ALCAM, ICAM-1 and E-selectin before interaction with tumor cells. In additional experiments, flow chambers were coated with recombinant L1-CAM, ICAM1, ALCAM or E-selectin, and the adherence of MDA-MB231 cells with high or low L1 expression to these proteins was investigated.

Results: Adhesion of MDA-MB231 cells to activated HPMEC was significantly higher in L1high cells compared to L1low clones, where the number of adherent cells was only 40–50% of the L1high control (p=0.025; p=0.035). Blocking experiments showed that the adherence of L1high cells could be reduced by antibodies directed to ALCAM (p=0.0007), but not to ICAM1. Anti-L1 antibodies had a significant effect only in passages of endothelial cells which showed L1 expression, whereas E-selectin blocking strongly diminished adherence of breast cancer cells irrespective of their L1 expression. In addition, L1low clones showed significantly lower adhesion to recombinant L1, ALCAM and E-selectin proteins compared to the parental cells.

Conclusion: Our experiments indicate that L1 expression in breast cancer cells leads to an increased adherence to activated endothelial cells via homophilic (L1-L1) or heterophilic (L1-ALCAM) interactions. This mechanism is a possible explanation for the increased metastatic potential and poor prognosis in L1-positive carcinomas observed in vivo. Our results suggest that this adhesion molecule might be a suitable target for therapeutic interventions.

Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P2-01-17.

P3-05-05: Cyclin D1 Gene Amplification Is Rarely Heterogeneous in Breast Cancer

Background: Amplification of Cyclin D1 (CCND1) occurs in about 10–20% of breast cancers and has been suggested to predict resistance to anti-hormonal therapy. As the diagnostic accuracy of predictive biomarkers can be substantially limited by regional expression differences within tumors, heterogeneity of CCND1 amplification was assessed in this study. To assess heterogeneity, a novel tissue microarray based analysis platform was developed.

Material and Methods: To comprehensively asses the three-dimensional molecular composition of breast cancers, a “heterogeneity TMA” was constructed containing 8 different tissue cylinders from as many different cancer containing tumor blocks as possible (at least 4) from 147 primary breast cancers. Additional tissue samples were taken from 1–4 corresponding nodal metastases from 35 of these patients. Dual labeling fluorescence in situ hybridization (FISH) with probes for CCND1 and centromere 11 was applied.

Results: The analysis revealed amplification in 29 of 133 (21.8%) patients with interpretable FISH data. CCND1 amplification was more frequently seen in ductal (22 of 87; 25.29%) than in lobular type (5 of 32; 15.63%) (p=0.251). CCND1 amplification was also associated with high tumor grade with amplification rates of 1 of 18 (5.56%) in grade 1, 15 of 72 (20.83%) in grade 2 and 12 of 40 (30%) in grade 3 carcinoma (p=0.075). CCND1 amplification was more frequently seen in ER positive cases (27 of 110; 24.55%) than in ER negative cases (1 of 17; 5.88%) (p=0.052). No association could be found between CCND1 amplification and tumor stage (p=0.445) and CCND1 amplification and PR status (p=0.752). Heterogeneous amplification status was detected in 9 of 29 (31.0%) amplified tumors, i.e. in 6.8% of all informative cases. Heterogeneity was successfully validated on large sections in all 4 heterogeneous cases with high level amplification. In the remaining 5 “heterogeneous cases” discordant results were due to variable interpretation of borderline amplification results with CCND1/centromer 11 ratios between 1.7 and 2.3. There were no discrepancies seen between primary tumors and matched lymph node metastases.

Discussion: The high degree of homogeneity seen for CCND1 amplification suggests that this alteration represents an early event in tumor development/progression in a subset of breast cancers. CCND1 status determined in a small biopsy will be highly representative of the entire tumor and will thus be appropriate for predicting treatment outcome.

Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P3-05-05.

P1-06-26: The EndoPredict Score Is a Response Predictor for Neoadjuvant Chemotherapy in ER-Positive, HER2−Negative Breast Cancer

Background:

The EndoPredict (EP) score is a multigene classifier to predict the likelihood of distant recurrence in ER-positive, HER2−negative breast cancer patients treated with adjuvant endocrine therapy. Two large randomized phase III trials involving endocrine therapy only (n > 1700) demonstrated additional prognostic information of the EP score independent from clinicopathological parameters by classifying 49% as low risk. However, the predictive role of the EP is not clear. Therefore, we examined whether the EP Score also predicts sensitivity towards neoadjuvant chemotherapy in ER-positive, HER2−negative breast cancer patients.

Methods: Four publicly available gene expression data sets (Affymetrix HG-U133A) were retrieved from the gene expression omnibus (GEO) data repository. All analyzed breast cancer patients were treated with anthracycline or taxane/anthracycline-based neoadjuvant chemotherapy. Microarray cel files were MAS5 normalized with a global scaling procedure and a target intensity of 500. The analysis was restricted to ER-positive, HER2−negative breast cancer patients according to pre-specified cut-off levels for the respective ESR1/ERBB2 Affymetrix probesets. The EP score was calculated and patients were classified as having low or high risk according to the pre-specified validated cut-off value. Pathological complete response (pCR) — defined as no residual invasive cancer in the breast or lymph nodes — was used as the primary endpoint for the assessment of treatment response.

Results: The EP Score was examined in 221 ER-positive, HER2−negative breast cancer patients treated with neoadjuvant therapy. Among the 221 patients, 61 tumors (27.6%) were classified as EP-low-risk, whereas 160 tumors (72.4%) were EP-high-risk. Only one of the EP-low-risk tumors achieved a pCR after neoadjuvant therapy, whereas 24 of the 25 pCR events were classified as EP high risk. The sensitivity of the EP score was 96% and the negative predictive value 98% with an area under the receiver operating characteristic curve of 0.73.

Conclusions: The EP Score is a predictor of chemosensitivity in the neoadjuvant setting. The test correctly identified all but one of the patients achieving a pCR suggesting that the benefit of cytotoxic chemotherapy is limited to the EP high risk group.

Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P1-06-26.

Abstract P3-10-07: T5 Is a New Molecular Predictor of Distant Recurrence in Estrogen Receptor-Positive, HER2-Negative Breast Cancer

Background: Molecular tests predicting outcome of breast cancer patients may be useful for treatment decisions in addition to standard clinicopathologic features.

Methods: Using human genome HG-U133A array and qRT-PCR datasets, we developed and validated a gene-expression signature predicting the likelihood of distant recurrence in postmenopausal, early-stage breast cancer patients with estrogen receptor-positive, HER2-negative tumors treated with adjuvant endocrine therapy. RNA levels assessed by qRT-PCR in formalin-fixed paraffin-embedded tumor specimens were used to calculate a risk score (T5) and to determine a risk group (low or high) for each patient. The prospectively defined T5 risk score was then validated independently in patients from two large randomized phase III trials. Distant recurrence-free survival and overall survival were analyzed with Cox models adjusted for clinicopathological factors. The primary endpoint was time to distant recurrence.

Results: In a training set of 964 tumors, we identified a gene-expression signature consisting of three proliferation-related genes (BIRC5, UBE2C, DHCR7), five estrogen-regulated genes (RBBP8, IL6ST, AZGP1, MGP, STC2), and three reference genes (CALM2, OAZ1, RPL37A). For the validation, RNA analysis was possible in 1702 of 1725 (99%) tumors of both validation sets. Women were classified as having low risk (n=832; 49%) or high risk (n=870; 51%) by the T5 risk score. The T5 risk score provided prognostic information independent from clinicopathologic risk as estimated by Adjuvant!Online or Ki67 labeling index. Patients with a higher T5 risk score had a significantly shorter time to distant recurrence (adjusted hazard ratio, 1.24; 95% confidence interval [CI], 1.15 to 1.33; P<0.001) and overall survival (adjusted hazard ratio, 1.13; 95% CI, 1.06 to 1.19; P<0.001) compared to patients with a lower T5 risk score. The addition of the risk characterized by the T5 risk score to the clinicopathological risk resulted in 10-year distant recurrence-free survival rates of 95% in combined low risk patients and 82% in combined high risk patients (P<0.001).

Conclusions: Using formalin-fixed paraffin-embedded tumor specimens, the multigene T5 risk score provides prognostic information independent of Adjuvant!Online or Ki67 labeling index. By combining the T5 risk score with clinicopathological risk, we were able to accurately identify breast cancer patients with low risk or high risk for distant recurrence. Using this new easy-to-use multigene tool in clinical practice will assist in optimizing adjuvant therapy by reducing both undertreatment and overtreatment and thus improves outcome and quality of life of patients with early-stage breast cancer.

Citation Information: Cancer Res 2010;70(24 Suppl):Abstract nr P3-10-07.

Comparison of Prognostic Signatures in Node-Negative Tamoxifen-Treated Breast Cancer Patients

Background: A number of molecular signatures have been published to aid breast cancer prognosis and therapy response prediction. A 76-gene signature has been developed in a node-negative patient cohort that did not receive systemic therapy.1 Another prognostic 97-gene signature measures predominantly proliferation-associated genes.2 Finally, a 21-gene signature has been developed for node-negative and estrogen receptor–positive breast cancer patients treated with tamoxifen in the adjuvant setting.3 Here we compare the prognostic performance of all three published algorithms in a cohort of 189 node-negative breast cancer patients treated with tamoxifen.Materials and Methods: Fresh-frozen tumors from node-negative patients were profiled on HG-U133a arrays. In addition, HG-U133a datasets with clinical annotation were downloaded from GEO (http://www.ncbi.nlm.nih.gov/geo/). All patients received tamoxifen only as adjuvant treatment after surgery. We determined the molecular subclass on the basis of ESR1 and ERBB2 mRNA expression; only ESR1-positive and ERBB2-negative tumors were considered for further analysis, leaving 141 in-house and 48 public datasets. After mapping of the gene signatures to the HG-U133a platform, we performed Cox regression and ROC curve analysis with distant metastasis as endpoint.Results: Cox regression analysis yielded a significant outcome association for all three gene signatures (76-gene signature: P = 0.0018; 97-gene signature: P = 0.0294; 21-gene signature: P = 0.0025) in the whole ESR1+/ERBB2– cohort. However, the 97-gene signature did not yield a significant result by ROC analysis for distant metastasis at 5 years (76-gene signature: AUC = 0.645, CI = 0.505–0.785; 97-gene signature: AUC = 0.608, CI = 0.462–0.754; 21-gene signature: AUC = 0.744, CI = 0.633–0.854). In addition, when the analysis was restricted to grade 2 tumors (n = 114), only the 21-gene signature remained prognostic by Cox regression (76-gene signature: P = 0.2405; 97-gene signature: P = 0.9001; 21-gene signature: P = 0.047) as well as by ROC curve analysis at 5 years (76-gene signature: AUC = 0.575, CI = 0.398–0.751; 97-gene signature: AUC = 0.532, CI = 0.347–0.717; 21-gene signature: AUC = 0.699, CI = 0.549–0.849).Discussion: While all three gene signatures reveal a significant outcome association in our whole patient cohort by Cox regression analysis, only the 21-gene signature remains significant in a subset analysis of grade 2 tumors. Since this group is clinically particularly challenging, further comparison between prognostic gene signatures in homogeneously treated patient cohorts is needed.

Citation Information: Cancer Res 2009;69(24 Suppl):Abstract nr 104.

Prediction of chemotherapy response in breast cancer by using a four-gene panel including osteopontin, activating leukocyte cell adhesion molecule (ALCAM), HER2 and estrogen receptor (ER)

Abstract #2033

Background: Several prognostic factors such as TNM-stage, hormone receptor and HER2 status refer to therapeutical decisions towards adjuvant chemotherapy. Up to now it is still difficult to predict the response to chemotherapy for a single patient, which is a challenging problem in order to spare patients from cost-intensive ineffective treatment and adverse side effects. Therefore there is urgent need for additional predictive markers. Our previous findings implicated that tumor tissue expression of Activated Leucocyte Cell Adhesion Molecule (ALCAM/CD166) is associated with chemotherapy response. Osteopontin (OPN) has been described as prognostic marker in breast cancer and recently it was shown that high OPN levels in vitro mediate chemoresistance of breast cancer cells by inhibition of apoptosis. If these findings result in chemoresistance in vivo has not been investigated so far. Material and Methods: Primary breast cancer tissues from 100 patients treated with Taxane-free adjuvant standard chemotherapy regimen were collected at surgery prior to any additional therapy. The median follow-up time was 81 months. We analyzed ALCAM and OPN mRNA expression in the biological context of additional and predefined breast cancer markers, like estrogen receptor (ER), Her-2/neu (HER2), members of the urokinase plasminogen activator (uPA) system, vascular endothelial growth factor (VEGF) using oligonucleotide microarrays (Affymetrix HG-U133A). Hierarchical cluster analysis was performed to develop a gene algorithm predicting outcome after standard chemotherapy. Results: In contrast to ALCAM overexpression, which was associated with better outcome upon chemotherapy, OPN overexpression was associated with a significantly higher recurrence rate (p=0,027), but not with shorter overall survival. In addition, there were significant positive correlations of OPN with uPA, PAI-1, uPAR and VEGF-A mRNA expression. By cluster analysis based on the four markers ER, HER2, ALCAM and OPN, we identified a group of high-risk patients, which was characterized by low ER and HER2 expression, high OPN and low ALCAM levels, and with significantly shorter recurrence-free and overall survival (p<0,001). Discussion: In our cohort of patients the combination of ALCAM, OPN, ER and HER2 expression levels turned out as valuable predictive marker for response to adjuvant Taxane-free chemotherapy. The ratio of high OPN and low ALCAM levels was particularly associated with chemoresistance and a poor prognosis in tumors having low ER and HER2 expression levels. We hypothesize that this simple four-gene panel could help to reassign an optimized therapy to the single patients.

Citation Information: Cancer Res 2009;69(2 Suppl):Abstract nr 2033.