Multispectral Imaging to define morpho-molecular classes of human HCC

Question:

Hepatocellular carcinoma (HCC) represents a molecularly and morphologically heterogeneous tumor entity. Molecular profiling allowed for molecular HCC classification, finally resulting in the definition of morpho-molecular subtypes by the new WHO classification. Although there is some evidence that immunohistology may allow for the identification of molecular HCC classes, a decisive protein-based HCC classification has not been performed yet and the complex interplay between pathways commonly altered in HCC as well as the proteomic heterogeneity of human HCC remain elusive.

Methods:

Formalin-fixed, paraffin-embedded (FFPE) tissues of 59 human HCCs were analyzed using multispectral Imaging (MSI, Vectra® 3.0 Automated Quantitative Pathology Imaging System), which enables the parallel detection of six biomarkers, while preserving the morphologic (image) information. Antibody panels were designed to interrogate molecular HCC subclasses (BerEP4, p-S6K, CRP, CTNNB1, GS, ARID1A) as well as important HCC pathways (p-AKT, TP53, MDM 2, p-ATF2). As conventional image analysis software do not allow for the evaluation of cross-talks between signaling pathways, T-distributed Stochastic Neighbor Embedding (t-SNE) and machine learning based approaches were applied to analyze the multi-dimensional MSI data sets.

Results:

MSI-based classification of human HCC is feasible, but may be limited by background fluorescence. Importantly, both tumor and stromal cells contribute to molecular HCC signatures and MSI allows for the visualization of tumor heterogeneity as well as the interrogation of mechanism contributing to therapy resistance.

Conclusion:

Multiparameter immunohistology and advanced data analyses allow for morpho-molecular typing of HCC in routine FFPE biopsies, thereby demonstrating its potential power to identify predictive biomarkers highlighting the importance of HCC biopsy in clinical trials.