Detecting Phenotypical Subtypes of Breast Cancer with Multiplexed Immunohistochemistry
C Hoyt, J Rheinhardt, N Torsten, H Gardner. Cambridge Research & Instrumentation, Inc., Woburn, MA; Novartis Institutes for Biomedical Research, Cambridge, MA; British Columbia Cancer Agency, Vancouver, BC, Canada
Background: Treatment for breast cancer has benefited significantly from advances in molecular biology. IHC tests for protein receptors ER, PR, and Her2 have lead to a new patient classification system – Luminal A (ER+, PR+, and Her2-), Luminal B (reduced ER or PR percent positivity), Her2-positive, and Basel-like (ER-, PR-, Her2-). Traditional approaches to assessing multiple proteins use serial sections, staining for one protein per serial section. If multiple proteins are assessed in the same tissue section, co-expression can be detected on a per-cell basis. Percent double and triple positivity can be assessed, possibly revealing significant subtypes and leading to more targeted and more effective treatments and therapies.
Design: Having previously developed: a) multi-label immunohistochemical staining methods; b) an automated multispectral slide analysis system; and c) image analysis algorithms to differentiate relevant tissue regions (e.g., tumor vs stroma and necrosis, etc.) and to segment cellular compartments, the goal is now to perform pilot studies using archived clinical material, to determine whether per-cell co-expression subtypes can be uncovered.
Results: Multispectral imaging was performed on two sets of a 712-core TMA (356 patients represented in duplicate). The first set was stained for ER and ki67 plus counterstain, and the second for ER, PR and Her2, plus counterstain. IHC signals were spectrally 'unmixed' from each other and counterstain. Machine-learning-based automated image analysis was performed to locate cancer cells, segment subcellular compartments, and extract IHC signals on a per-cell basis. Per-cell co-expression subtypes were detected using flow-cytometry data analysis software. Each TMA core was imaged and processed at a rate of about three cores per minute. Image analysis algorithms were trained in under 2 hours, and then used to process core images at approximately 10 seconds per core. Percent double and triple positivity were determined, revealing subtypes. Correlation between subtypes and clinical outcomes will be the topic of future publications.
Conclusions: Together, the innovative multispectral platform and image analysis software, coupled with flow-cytometry analysis tools, sometimes termed 'tissue cytometry', can be used to reveal molecular subtypes, which may lead to new targeted strategies for breast cancer research and clinical care.
Category: Pan-genomic/Pan-proteomic Approaches to Diseases
Monday, March 22, 2010 1:00 PM
Poster Session II # 234, Monday Afternoon