Abstract LB-108: Automated immunohistochemistry-based identification of molecular subtypes in colorectal cancer

BACKGROUND Recent molecular profiling of colorectal cancer has identified three distinct subtypes: two good-prognosis subtypes characterised by chromosomal instability (Subtype 1) and microsatellite instability (Subtype 2), and a poor prognosis subtype defined by epithelial-mesenchymal transition and stemness (Subtype 3). A key challenge lies in the subsequent implementation of this classification system in the clinic, where the requirement for sufficient bulk tumour and high cost hampers the widespread use of genomic profiling. In this proof-of-concept study, we have adapted our molecular signatures to an automated immunohistochemistry (IHC)-based classifier which can be used a rapid screening tool, and validated the feasibility of this approach in a multi-centre study. PATIENTS AND METHODS A total of 1080 patients from four different centres were used in this study. Tumour microarrays (TMAs) from each patient were stained using immunohistochemistry for a panel of five stains in conjunction with microsatellite instability status (MSS). An automated image analysis pipeline was developed to quantitate and normalize all staining. A classifier to distinguish colorectal cancer subtypes was trained on 74 patients from a single centre and applied to the three remaining cohorts. RESULTS We have developed an automated IHC-based classification system which demonstrated 84% correlation with our genomic profiling system. The prognostic value of colorectal cancer subtyping was validated in three independent cohorts, taking into account age, sex and stage (Hazards Ratios with 95% Confidence Interval: 1.8 (1.2 - 2.6), 1.4 (1.1 - 1.7) and 1.2 (1.0 - 1.6)). We evaluated the predictive value of the subtyping system in a retrospective analysis of a late-stage clinical trial and demonstrated the benefit of adjuvant cetuximab in KRAS/BRAF wild type Subtype 1 patients (Hazards Ratio 0.6 (0.4-0.9)) but not in Subtype 3 patients. CONCLUSION In this proof-of-concept study we have demonstrated the effectiveness of an automated IHC-TMA classifier as a surrogate for genomic profiling. Using this approach, we have validated the prognostic value of colorectal subtyping in a multi-centre study, and shown the predictive value of subtyping for adjuvant cetuximab therapy. This approach has demonstrated high clinical potential as a rapid screening tool, in particular in retrospective examination of patient cohorts where only formalin-fixed paraffin-embedded (FFPE) tissue is available. Citation Format: Anne Trinh, Filipe De Sousa E Melo, Xin Wang, Joan H de Jong, Evelyn Fessler, Marnix Jansen, Gerrit KJ Hooijer, Jan Paul Medema, Florian Markowetz, Louis Vermeulen. Automated immunohistochemistry-based identification of molecular subtypes in colorectal cancer. [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 LB-108. doi:10.1158/1538-7445.AM2015-LB-108