Introduction. Breast cancer can be divided into several subgroups characterized by unique patterns of pathway activation. Platelet-derived growth factor receptor (PDGFR) signalling has not yet been included in this classification scheme, although it has been reported to be a potential target for therapy. In this study, we have constructed a PDGFR-activation signature and investigated its relevance in breast cancer. Materials and Methods. Sixteen PDGFR-modulated genes were identified by intersecting two published PDGFR-modulated gene lists. The resulting gene signature was applied onto a publicly available gene expression data set of GIST (GSE17743) using principle component analysis. The segregation of PDGFR- and KIT-mutated GIST samples was investigated using permutation analysis and classification sensitivity and specificity were assessed. Using the regression coefficients from the first principal component, a PDGFR-activation score was constructed and applied onto a second data set in order to validate the score (GSE1923). Finally, the score was applied onto a gene expression data set of 389 breast cancer ***samples, including 137 samples from patients with IBC. Results. Sixteen PDGFR-modulated genes (NR4A1, EGR3, JUNB, IER3, TIEG, JUN, BCL3, MYC, NR4A3, PLAU, MCL1, DUSP1, DUSP5, DUSP6, SGK, GADD45A) were able to discriminate PDGFR-mutated GIST samples from KIT-mutated GIST samples with a sensitivity of 75% and a specificity of 85%. Application of the PDGFR-activation score onto a data set of control and PDGF-treated glioblastoma cells showed a significant increase in the PDGFR-activation score in the treated condition (P=0.0302). Application of the PDGFR-signature onto our series of IBC and nIBC samples demonstrated a significant and molecular subtype-independent increase in PDGFR-activation in IBC (P=0.0015; FDR=3%). In addition, in our series of nIBC samples only, PDGFR-activation was associated with decreased DMFS and RFS (P=0.0038 and P=0.0137 respectively). In fact, PDGFR-activation was an independent prognosticator in a multivariate model incorporating the molecular subtypes. Discussion. We identified a gene signature composed of 16 genes able to predict PDGFR-activation in tissue samples by gene expression analysis. PDGFR-activation is significantly increased in samples from patients with IBC, an aggressive form of locally advanced breast cancer. In addition, in nIBC, PDGFR-activation is associated with DMFS and RFS, independently of the molecular subtypes suggesting that PDGFR-activation might add another level of clinically relevant heterogeneity in breast cancer. Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P5-01-01.

Introduction: The fibrotic focus (FF) is a practical, easily assessable and reproducible integrative histological prognostic parameter in breast cancer. Its prognostic value has been shown before (Van den Eynden et al. Histopathology 2007). In this study we investigated whether the assessment of the FF adds prognostic information to the relapse score based on gene expression analysis of 76 genes as previously described (Wang et al. Lancet 2005). Materials and Methods: All patients of 2 previous prognostic breast cancer gene expression studies for whom FFPE slides of the tumor were available (Wang et al. Lancet 2005 and Yu et al. BMC Cancer 2007) were selected, leading to a study population of 176 lymph node negative breast cancer patients. The presence and size (<1/3 or >1/3 of tumor area) of a FF were assessed on standard HE slides. These data were compared to the 76-gene relapse score, to standard clinicopathological variables and to metastasis-free survival. Results: A small and large FF were found in 31 (17.6%) and 20 (11.4%) of patients, respectively. In 120 (68.2%) patients there was no FF and in 5 patients the presence of a FF could not be assessed due to insufficient FFPE material. 64 (36.4%) and 112 (63.6%) patients had respectively a good and poor prognostic 76-gene relapse score. There was a significant correlation between the presence of a FF and a poor 76-gene relapse score, 18 of 20 patients with a large FF had a poor relapse score (p = 0.03). Patients with a tumor with a FF and especially with a large FF had a significantly reduced metastasis free survival (Log rank p<0.001). The same was true for patients with a poor 76-gene relapse score (Log rank p<0.001). When only patients with a poor relapse score were taken into account, patients with a tumor with a large FF had a significantly decreased metastasis-free survival compared to patients without a FF or with a small FF (Log rank p=0.005). In patients with a good relapse score, the number of patients with a FF was too small for a separate analysis. In a multivariate Cox regression model for metastasis-free survival including age, ER and PR status, T stage, the FF and the 76-gene relapse score status, the FF (OR 1.5, p=0.02) and the relapse score status (OR 3.4, p = 0.001) were significant independent prognostic factors. Comparable results were found if the presence of a FF was dichotomized in large FF versus no or a small FF. Conclusion: The assessment of the presence and size of a FF adds independent significant prognostic information to the prognostic 76-gene expression signature, especially in selecting a subgroup of patients with a very poor prognosis. Since the assessment of the FF is practical, easy, reproducible and cheap it should be considered to become part of the standard pathological examination of breast cancer resection specimens. Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P5-14-14.