The Phosphorylated ATM Immunofluorescence Assay: A High-performance Radiosensitivity Assay to Predict Postradiation Therapy Overreactions

Abstract

Purpose: The ability to identify, before treatment, those patients who will overreact to radiation therapy would have sound positive clinical implications. By focusing on DNA double-strand breaks recognition and repair proteins after irradiation, we recently demonstrated that the maximal number of phosphorylated ATM (pATM) nuclear foci in the first hour (pATMmax) after ex vivo irradiation correlated with postradiation therapy toxicity severity. We performed additional analyses of our whole collection of fibroblast lines to refine the predictive performance of our assay.

Methods and materials: Immunofluorescence experiments were performed on 117 primary skin fibroblast lines irradiated at 2 Gy. The toxicity response was split into 2 binary classes: 0 if the toxicity grade was <2 and 1 otherwise. To assess the relationship between the quantity of pATMmax foci and toxicity grade, we applied a correlation and then a supervised classification analysis. Training data sets from 13 radiosensitive patients randomly drawn using a random undersampling technique were constituted. Receiver operating characteristic analyses were performed using a Monte-Carlo method to estimate the optimal threshold and discriminate the responses for each data set. The discrimination cutoff was estimated as the maximum value of the 10⁴ thresholds computed from each training subset.

Results: As expected, we confirmed a quasi-linear dependence between toxicity and pATMmax (Pearson correlation coefficient -0.85; P < 2.2e^-16). When taken as a binary predictive assay with the optimal cutoff value of 34.5 pATM foci/cell, our assay showed outstanding predictive performance (sensitivity, specificity, negative predictive value, positive predictive value, and area under the curve: 100%, 92%, 100%, 99%, and 0.987, respectively).

Conclusions: The results of these experiments allowed us to identify pATMmax as a high-performance predictive parameter of patients with postradiation therapy overreactions. Additional studies are in progress to confirm that this radiosensitivity assay reaches the same performance level in any condition to adapt clinical practice.