Translation of breast cancer gene expression profiles into an immunohistochemistry-based classifier. Poster at the American Association for Cancer Research (AACR) Annual Meeting 2007

Hege G. Russnes ¹, Therese Sørlie ¹, Bjørn Naume ¹, Elin Borgen ¹, Anne-Lise Børresen-Dale ¹, Rolf Kåresen², Robert S. Seitz³, Rodney A Beck ³, Brian Z. Ring ³, William J. Shasteen ³, Douglas T. Ross ⁴.
¹Rikshospitalet-Radiumhospitalet Medical Center, Oslo, Norway, ² Ullevål University Hospital, Oslo, Norway, ³ Applied Genomics Inc., Huntsville, AL, ⁴Applied Genomics Inc., Burlingame, CA.

Abstract

Gene expression profiling of breast cancer has identified distinct subtypes: Luminal A, Luminal B, ERBB2+, Basal-like and Normal-like with defined molecular properties and clear clinical implications. However, the requirement for specialized technology has slowed the validation and translation of gene expression-based classifiers into clinically robust tests. For routine clinical diagnostics, fluorescence in situ hybridization (FISH) and immunohistochemistry (IHC) technologies are robust, efficient, widely distributed and therefore attractive alternatives. We previously demonstrated that IHC reagents targeted to cognate proteins of subtype signature genes, identified biologic properties and stratified patient samples as predicted by the gene expression signatures. Here we directly compare classification using a model based upon IHC reagents targeting ESR1, HER2, SLC7A5, NAT1 and TRIM29 and their classification by gene expression signature. Fresh tumor samples from 123 patients were prospectively collected at primary surgery and gene expression profiles examined using whole-genome cDNA microarrays. Tissue microarrays were constructed from formalin-fixed paraffin-embedded tissues from the same patients and IHC stains performed with the five antibodies and FISH for the HER2 oncogene.
IHC reagents targeting NAT1 (LumA), SLC7A5 (LumB), TRIM29 (Basal-like), HER2 (ERRB2+), and ESR1 (Luminal A/B) were significantly associated via logistic regression analysis with their cognate patient subtypes as assigned by gene expression signature across the 123 samples. An IHC/FISH-based taxonomy constructed to recapitulate gene expression-based classification correlated with the gene expression-based subclasses in 63% of the informative cases. Nineteen percent failed to show concordant classification while 19% were unclassified.
These data demonstrate that classification of breast carcinomas by targeted IHC reagents can recapitulate the main traits of the gene expression-based taxonomy. Further comparison of the clinical associations of IHC-based to gene-based classification will reveal the relative merits of each approach. The further exploration of the IHC-based classifiers will allow rapid exploration of patient classification in larger and otherwise inaccessible cohorts and identification of a clinically useful IHC-based stratifier of breast cancer.

	(5.3 MB PDF)