Using Immunohistochemistry for Practical Molecular Subtyping of Bladder Cancer

In recent years, the molecular subtyping of muscle-invasive bladder cancer using gene expression patterns has been a major area of research. This has allowed the categorization of cells into luminal or basal major subtypes and several classification schemes. However, several factors limit the widespread use of RNA expression-based subtyping in the clinical setting. In a recent study, Hardy et al. examined the use of immunohistochemistry as a more practical alternative for molecular subtyping. As this technique is routinely used in pathology laboratories and clinical settings, it can be feasible to perform clinical subtyping effectively and rapidly.

Using the Lund bladder cancer datasets with mRNA expression data from muscle-invasive bladder cancer samples and immunohistochemistry, the authors developed a panel of 24 immunohistochemical features that could be used to identify five tumor subtypes of muscle-invasive bladder cancer: genomically unstable, urothelial-like, basal/squamous cell carcinoma-like, mesenchymal-like, and small-cell neuroendocrine-like. Luminal (urothelial-like and genomically unstable) tumors exhibited high expression of GATA3, EPCAM, and FOXA1, while basal tumors showed increased expression of KRT5 and KRT14. The authors then simplified the classifier by focusing on the core protein markers KRT14, KRT5, CDH3, FOXA1, GATA3, PPARG, RB1, CCND1, CDKN2A (p16), FGFR3, and TP63. Finally, they investigated the ability of their classifier to distinguish basal versus luminal tumors using only GATA3 and KRT5 and reported a subtyping accuracy of 91%. Upon adding p16 to the classification model, the accuracy was 78% for urothelial-like, genomically unstable, and basal subtypes and 91% for luminal versus basal subtypes. An alternate model using expression of KRT14 followed by RB1 exhibited a higher accuracy of up to 86% for distinguishing between basal, urothelial-like, and genomically unstable subtypes.

To test whether evaluating the expression of a fuller set of protein markers would increase the model's accuracy, a model based on eight proteins was used and achieved an accuracy of approximately 82% on the validation set. This indicates that additional protein markers do not necessarily lead to higher subtyping accuracy. There was no significant correlation with overall survival based on subtyping. Hardy et al. prioritized markers that showed distinct patterns of expression that the screening pathologists could efficiently utilize. GATA3, KRT14, KRT5, and RB1 performed well in this regard as they exhibited bimodal expression patterns with a clear distinction between basal versus luminal subtypes in contrast to p16 and CCND1, which showed variability across subtypes and intermediate staining.

Hardy et al.'s findings are valuable for establishing parsimonious protein-based assays that can eliminate redundancy and enhance ease and practicality in the clinical setting. Establishing a model that identifies genomically unstable subtypes is important as subtyping typically only distinguishes between luminal versus basal subtypes. Further prospective studies validating these models on a large number and wide range of samples will be needed to consolidate these findings.

Written by: Bishoy M. Faltas, MD, Director of Bladder Cancer Research, Englander Institute for Precision Medicine, Weill Cornell Medicine, New York City, New York

References:

Hardy, C., Ghaedi, H., Slotman, A., Sjödahl, G., Gooding, R. J., Berman, D. M., & Jackson, C. L. (2022). Immunohistochemical Assays for Bladder Cancer Molecular Subtyping: Optimizing Parsimony and Performance of Lund Taxonomy Classifiers. The journal of histochemistry and cytochemistry : official journal of the Histochemistry Society, 70(5), 357–375. https://doi.org/10.1369/00221554221095530

Read the Abstract

Login