Quantitative evaluation of TP53 immunohistochemistry to predict gene mutations: lessons learnt from a series of colorectal carcinomas

Abstract

This study addressed if TP53 immunohistochemistry as a surrogate method for gene sequencing could be applied to colorectal carcinomas as successfully as recently reported for ovarian cancers. Sanger sequencing of the coding exons 2-11 of 87 tumors yielded a total of 65 mutations in 61 of the tumors. Immunohistochemistry was done with the Do-7 antibody. By a pattern recognition evaluation of immunohistochemistry, 44 cases were classified as "overexpressors" and 20 as having "wild-type" immunostaining; complete absence of or cytoplasmic immunostaining was seen in 9 and 4 cases, respectively. However, for 10 tumors, a confident distinction between overexpression and wild-type immunostaining was not possible ("indeterminates"). Quantitative analysis on digital images (i) using QuPath to determine the percentage of immunopositive cells and (ii) WEKA segmentation to obtain an index that quantified the intensities of tumor cells' nuclear immunostaining showed a continuous distribution of the data, explaining failure of assessment by pattern recognition in some cases. Quantitative data were then used to define cutoffs by receiver operator curve analysis, which allowed for predicting the mutational status of the TP53 gene with sensitivities of 0.89 and 0.95 for the 2 methods, respectively, and specificities of 0.81 for both. In conclusion, by a dedicated approach, TP53 immunohistochemistry works well as a surrogate method for molecular studies. Considering the potential predictive role of TP53 gene mutations in chemotherapy decisions, TP53 immunohistochemistry may be of value alongside with molecular gene studies, possibly even across different cancers.