Identification of cell-of-origin breast tumor subtypes in inflammatory breast cancer by gene expression profiling

• Published in: Breast cancer research and treatment 2006-02, Vol.95 (3), p.243-255
• Main Authors: VAN LAERE, Steven J, VAN DEN EYNDEN, Gert G, VAN DER AUWERA, Ilse, VANDENBERGHE, Melanie, VAN DAM, Peter, VAN MARCK, Eric A, VAN GOLEN, Kenneth L, VERMEULEN, Peter B, DIRIX, Luc Y
• Format: Article
• Language: English
• Subjects: Adenocarcinoma - classification; Adenocarcinoma - genetics; Adenocarcinoma - metabolism; Biological and medical sciences; Biomarkers, Tumor - genetics; Biomarkers, Tumor - metabolism; Breast cancer; Breast Neoplasms - classification; Breast Neoplasms - genetics; Breast Neoplasms - metabolism; Cancer research; Cancer therapies; Carcinoma, Ductal, Breast - genetics; Carcinoma, Ductal, Breast - metabolism; Classification; Female; Gene expression; Gene Expression Profiling; Gynecology. Andrology. Obstetrics; Humans; Immunoenzyme Techniques; Inflammatory diseases; Mammary gland diseases; Medical sciences; Oligonucleotide Array Sequence Analysis; Tumors
• ISSN: 0167-6806; 1573-7217
• DOI: 10.1007/s10549-005-9015-9

Inflammatory breast cancer (IBC) is an aggressive form of locally advanced breast cancer with high metastatic potential. Most patients have lymph node involvement at the time of diagnosis and 1/3 of the patients have distant metastases. In a previous study, we demonstrated that IBC is a distinct form of breast cancer in comparison with non-IBC. The aim of this study was to investigate the presence of the different molecular subtypes in our data set of 16 IBC and 18 non-IBC specimen. Therefore, we selected an 'intrinsic gene set' of 144 genes, present on our cDNA chips and common to the 'intrinsic gene set' described by Sorlie et al. [PNAS, 2003]. This set of genes was tested for performance in the Norway/Stanford data set by unsupervised hierarchical clustering. Expression centroids were then calculated for the core members of each of the five subclasses in the Norway/Stanford data set and used to classify our own specimens by calculating Spearman correlations between each sample and each centroid. We identified the same cell-of-origin subtypes in IBC as those already described in non-IBC. The classification was in good agreement with immunohistochemical data for estrogen receptor protein expression and cytokeratin 5/6 protein expression. Confirmation was done by an alternative unsupervised hierarchical clustering method. The robustness of this classification was assessed by an unsupervised hierarchical clustering with an alternative gene set of 141 genes related to the cell-of-origin subtypes, selected using a discriminating score and iterative random permutation testing. The contribution of the different cell-of-origin subtypes to the IBC phenotype was investigated by principal component analysis. Generally, the combined ErbB2-overexpressing and basal-like cluster was more expressed in IBC compared to non-IBC, whereas the combined luminal A, luminal B and normal-like cluster was more pronounced in non-IBC compared to IBC. The presence of the same molecular cell-of-origin subtypes in IBC as in non-IBC does not exclude the specific molecular nature of IBC, since gene lists that characterize IBC and non-IBC are entirely different from gene lists that define the different cell-of-origin subtypes, as evidenced by principal component analysis.