Abstract 2175: A diagnostic gene expression assay for the classification of pheochromocytoma

Abstract Background/Aims: Pheochromocytoma (PCC) is a neoplastic growth of adrenal or extra-adrenal chromaffin cells. PCC is one of the most heritable cancers with up to 35% of cases attributed to a germline mutation in one of 13 or more genes. Previous microarray gene-expression studies have observed that PCC clusters into two or more subtypes corresponding to the underlying mutations that cause either dysregulation of hypoxia inducible factor (HIF) (e.g. VHL, SDHx) or constitutive activation of receptor tyrosine kinase (RTK) signalling (e.g. RET, NF1). The aim of this study was to conduct an unbiased assessment of the number of gene-expression subtypes of PCC using a compendium of published gene-expression data. Furthermore, we aimed to design a gene-expression based diagnostic assay to accurately classify tumours as an adjunct to clinical mutation analysis and for research purposes. Methods: Consensus clustering was applied to a gene-expression compendium of 253 samples to assess the most stable number of subtypes. Differential gene-expression analysis was performed to identify genes specific to each class. The expression of these genes in the microarray data was used to train a cross-platform k-NN classifier. The classifier was tested on an independent RNA-seq dataset to assess accuracy. Forty-seven genes (plus five control genes) were selected for Nanostring gene-expression analysis. The Nanostring assay was performed on RNA isolated from FFPE blocks corresponding to samples in the RNA-seq dataset to assess concordance and accuracy of the cross platform classifier. Results: Consensus clustering identified six robust PCC subtypes. Based on clinical annotation from the published data, four classes belong to the RTK signalling group, three of which (annotated RTK1 to 3) contain a mixture of NF1, RET, RAS, and TMEM127 gene mutations, and a fourth (annotated MAX-like) containing samples with mutations in the MAX gene.The remaining two classes belong to the HIF signalling group, one representing samples with a VHL mutation and the other mutations in one of the SDHx subunits. Applying the classifier we were able to classify RNA-seq data for samples with RET (6), NF1 (6), RAS (2), and TMEM127 (1) as belonging to the RTK 1-3 groups. Accurate classification of nine of eleven VHL samples and two SDHx samples was achieved. Classification of corresponding Nanostring data provided concordant classification for eleven of twelve samples. Conclusion: PCC can be divided into six biological classes indicative of the underlying driver mutation. The ability to accurately classify samples from FFPE material would provide a valuable adjunct to current practices in genetic testing laboratories, providing corroboration for known mutations and guidance on samples with mutations of unknown significance in PCC genes. Furthermore, the classifier can be used for research purposes to subtype tumours with unknown driver mutations for the interpretation of candidate driver genes. Citation Format: Aidan Flynn, Diana Benn, Roderick Clifton-Bligh, Joshy George, Anthony J. Gill, Rodney J. Hicks, Richard W. Tothill. A diagnostic gene expression assay for the classification of pheochromocytoma. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 2175. doi:10.1158/1538-7445.AM2015-2175