Background: The number of DNA double stand breaks (DSBs) induced after ionizing radiation is strongly correlated with cell death, mutation and cancer. Although a wide variety of techniques has been used to quantify DNA damage, immunostaining of DNA repair proteins, like γ-H2AX or Rad 51, that localize to the site of DNA damage within seconds following IR exposure has become a method of choice due to its sensitivity. DNA damage foci are imaged using microscopy, and then counted either manually, or with some form of computer assistance. A number of open source or vendor provided focus counting programs are available as either stand-alone applications or as plugins to image processing utilities such as ImageJ. While counting efficiency is greatly enhanced through the use of these tools, they are limited in that the analysis is confined to a single plane, they are often not conducive to high throughput analysis, and finally, they suffer from user-variability. We therefore developed a fully automatic, MATLAB-based algorithm which can process and uniquely identify foci in a 3D image stack, with only limited user interaction.
Methods: We compared three different methods of quantifying foci for uniformly irradiated cell cultures: a) 2D counting on single plane with Nikon NIS elements software and b) 2D or c) 3D counting with the MATLAB-based algorithm. 3D image stacks of stained γ-H2AX and Rad51 foci, as well as DAPI stained nuclei were acquired at 100X magnification with 200 nm slice spacing. The automated image analysis process comprises three steps: (i) nuclei segmentation, (ii) dynamic background characterization for threshold determination and (iii) segmentation of separate DNA damage foci of the entire 3D stack to ensure the identification and counting of unique foci.
Results: The accuracy of the 3D counting algorithm was determined to be 99% using a numerical phantom containing a known number of randomly placed foci in a 3D volume. In cell culture, 2D counting on a single plane using NIS Elements identified a mean foci count 0.09 and 31.7 per cell for 0 Gy and 4 Gy respectively while our algorithm counted 0.24 and 26.7 foci/cell, demonstrating remarkable consensus. Counting in 3D yielded 0.39 (0 Gy) and 77.58 (4 Gy) per cell, demonstrating how single plane counting underestimates DNA damage. The algorithm yielded a six fold speedup in processing a single field of view over manual counting, with additional time savings realized by removing the need for user interaction.
Conclusions: We have successfully implemented a fully automated 3D foci counting algorithm which identifies both foci, as well as surrounding nuclei, with little to no human interaction. This algorithm allows for the fast processing of large quantities of immunofluorescence data, thereby providing greater confidence in our DNA damage experiments.
Citation Format: Alexander Katsis, Eric Abel, Raisa Pavlyuchkova, Shilpa Senapati, Swati Girdhani, Renate Parry. Implementation of a fully automated 3D foci counting algorithm to determine DNA damage in cells. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 3599.