21055 Background: We have shown with cDNA microarrays that inflammatory breast cancer (IBC) and non-IBC are distinct biological entities. The purpose of this study was to confirm our previous results using Affymetrix chips. Methods: RNA was extracted from 19 IBC samples and 42 non-stage matched non-IBC samples. RNA was hybridized onto Affymetrix HG U133 Plus 2.0 chips. Gene expression data were normalized using GCRMA and genes with a gene expression of at least 250 in 50% of the cases were filtered in. Hierarchical clustering and principle component analysis was executed. Identification of the different cell-of-origin subtypes in our expression data set was done using the intrinsic gene list. A NFkB signature, a MAPK signature and our own IBC signature were tested by clustering analysis. Results: Clustering using 11341 genes resulted in the identification of two clusters: one containing 14/19 IBC samples and a second containing 32/42 non-IBC (Pearson χ2; p<0.0001). Principle component analysis separated IBC from non-IBC samples along the first principle component. Interestingly, IBC samples more closely resemble T1 - T2 tumours than T3 - T4 tumours. Application of the intrinsic gene set to our IBC/non-IBC data set resulted in the classification of 14/19 IBC samples as basal-like or ErbB2-overexpressing tumours compared to only 4/42 non-IBC tumours (Pearson χ2; p<0.0001). Our own IBC signature was confronted with the new data set and performed well in separating IBC specimens form non-IBC specimens. Clustering identified three clusters from which one cluster contained 18 samples, including 12 IBC specimens (p<0.0001). Using the NFkB and MAPK signatures, similar results were obtained. Conclusions: These results confirm our findings that IBC is a distinct biologic phenotype, characterized by activation of NFkB, possibly through activation of MAPK's. IBC tumours more often demonstrate characteristics from basal-like and ErbB2-overexpressing breast tumours. The fact that IBC tumours are rapidly developing tumours instead of longstanding tumourigenic processes might explain the close resemblance of the IBC gene expression profile to the gene expression profile of T1 and T2 tumours. No significant financial relationships to disclose.
Effects of type 2 diabetes mellitus on gene expression profile of meibomian glands in mice