https://orcid.org/0000-0002-9131-9089
Abstract
The toroidal nucleus is a novel chromosomal instability (CIN) biomarker which complements the micronucleus. Understanding the specific biological stresses leading to the formation of each CIN-associated phenotype requires the evaluation of large panels of biological images collected from different genetic backgrounds and environmental conditions. However, the quantification of toroidal nuclei is currently a manual process which is unviable on a large scale.
Here, we present QATS (QuAntification of Toroidal nuclei in biological imageS), a tool that automates the identification of toroidal nuclei, minimizing false positives while highly agreeing with the manual quantifications. Additionally, QATS identifies micronuclei for a convenient comparison of both CIN biomarkers. QATS is an open-source ImageJ plugin with a user-friendly interface that enables a wide scientific community to easily assess the frequency of CIN biomarkers for the determination of CIN levels in cellular models.
QATS is an ImageJ plugin freely available at http://www.toroidalnucleus.org/qats. The user manual and the images used for the evaluation of QATS are included in the website. Supplementary data are available at Bioinformatics online.
1 Introduction
Cancer cells often exhibit large chromosomal alterations, as found in 80% of solid tumors (Hoevenaar et al. 2020). Chromosomal instability (CIN) is a hallmark of cancer, and understanding the mechanisms of chromosome missegregation is a key focus of cancer research. However, analysis of CIN in mitotic cells is challenging since mitosis is short and dividing cells make only 10% of the cell population. Therefore, biomarkers detectable in interphase cells are convenient for the study of CIN. Although chromosome missegregation can result in different aberrant nuclear morphologies (Gisselsson et al. 2001), the presence of micronuclei is usually the only phenotype considered to score for CIN level (Fig. 1A, Krupina et al. 2021). Thus, complementary biomarkers are needed to obtain an exhaustive analysis of CIN status in cells. We recently identified the toroidal nucleus as a novel CIN biomarker (Almacellas et al. 2021), which is characterized by a doughnut-like shape with a void encompassing cytosolic components (Fig. 1A). Toroidal nuclei have been so far detected in 80% of the screened cell lines (Almacellas et al. 2021), and also in S. cerevisiae (Garcia et al. 2022), highlighting its potential value as readout across eukaryotes.
(A) Cell nucleus with a micronucleus (top) and a toroidal nucleus (bottom). (B) Representative output image generated by QATS in which two toroidal nuclei (in magenta) and four micronuclei (in green) were identified. (C) Scatter plot showing the number of toroidal nuclei quantified manually (by visual inspection) and by QATS in 100 images, and their Pearson’s correlation coefficient. Dot size is proportional to the number of images.
Even if most cancer cell lines exhibit an increase in toroidal nuclei and micronuclei (Almacellas et al. 2021, Krupina et al. 2021), the specific stimuli leading to the formation of these phenotypes are mostly unknown and might differ (Almacellas et al. 2021). Characterization of these stimuli requires extensive comparisons across cell types and conditions. Also, the prevalence of the toroidal nucleus is not fully understood, and the existing collections of morphological images, both in human, like the Broad Institute image collection (Ljosa et al. 2012), and yeast (Mattiazzi Usaj et al. 2020, Ohnuki et al. 2022), represent an invaluable resource to scan for its presence in very heterogeneous genetic backgrounds and conditions. Thus, further characterization of the toroidal nucleus demands the processing of large sets of images. However, quantification of toroidal nuclei currently relies on the manual analysis of fluorescent biological images (Pons et al. 2022), which is work-intensive and a limiting factor for large-scale analyses.
Here, we present QATS, an ImageJ plugin, open-source and freely available, for the efficient and reliable quantification of toroidal nuclei in large sets of fluorescent images.
2 Materials and methods
U2OS and HeLa cells were grown on sterile coverslips and fixed with 4% PFA for 10 min before counterstaining with DAPI for DNA detection, as explained previously (Pons et al. 2022). We acquired 100 images with a fluorescent microscope equipped with a 405 nm laser and a 40× air objective, containing nuclei of heterogeneous size and shape, with different levels of intensity saturation depending on their cell cycle phase when fixed. We visually inspected the images and quantified the number of toroidal nuclei and micronuclei.
We automated the quantitative analysis of toroidal nuclei by implementing QATS, an ImageJ (Schneider et al. 2012) plugin to process biological images containing DAPI-stained nuclei. QATS identifies nuclei by applying a dynamic intensity threshold, which accounts for the specific saturation level in each biological image. To identify toroidal nuclei, which should manifest as a void with an intensity similar to the background, QATS implements a single cell-based approach. After performing nuclei segmentation, it applies an intensity threshold to each nucleus proportional to the background and local intensities. The cell-based analysis allows the detection of brighter toroidal nuclei in saturated cells, and penalizes unexposed DAPI-stained nuclei. For the identification of micronuclei, QATS locates small circular particles and calculates the distance to their closest cell nucleus centroid. Particles far from any nucleus, such as isolated debris, are thus discarded. Also, multiple micronuclei close to the same nucleus, which probably originate from the same CIN event, are considered only once. Size and circularity ranges for nucleus, toroidal nuclei, and micronuclei can be automatically calculated or manually set by the user. It should be noted that QATS will be used for the screening of large collections of images in order to identify cell lines and conditions favoring the formation of toroidal nuclei. Thus, our approach and choice of parameters is inherently conservative, maximizing precision in order to minimize wrong calls that could compromise downstream analyses and conclusions.
QATS runs on single images and also on sets of images within a selected folder. For each run, QATS generates an output image highlighting all the identified relevant particles colored by their phenotype (Fig. 1B), including a header with the corresponding quantifications which are also reported in a separate text file. Importantly, QATS provides a spreadsheet per input image with topological descriptors of each identified nucleus, and whether they present any of the biomarkers, enabling future studies on morphological features of toroidal nuclei and micronuclei. If run on multiple images, QATS generates an additional summary spreadsheet with the corresponding quantifications.
3 Results
We ran QATS on a diverse set of 100 biological images, each consisting of 20 to 100 cells, from different experiments in U2OS and HeLa cells. We manually quantified the images and found the number of toroidal nuclei and micronuclei to be highly correlated with QATS quantifications (Pearson’s correlation = 0.91 and 0.84, respectively; see Fig. 1C). We individually verified each of the 209 toroidal nuclei identified by QATS and found 194 to be correct (precision = 0.93) with a recall of 0.77. Most errors could be explained by: (i) condensed DNA (mitotic or apoptotic cells) which induces a brighter signal; (ii) significantly small voids; and (iii) holes placed next to the nuclear envelope (see manual for illustrations). Precision and recall for micronuclei identification were 0.99 and 0.65, respectively. Most false negatives were linked to small micronuclei or micronuclei located in close proximity to the main nucleus. The average running time per image was 0.9 seconds on a laptop with a i7-1165G7 processor. We successfully tested QATS with ImageJ version 1.54f in Ubuntu 22.04 and Windows 10 using the set of 100 images mentioned above.
4 Conclusions
QATS is a computational image processing tool which enables the unbiased and reproducible identification of toroidal nuclei, representing a substantial improvement compared to the current time-consuming manual quantifications prone to human bias and error. Being the first automated approach for the identification of toroidal nuclei, QATS offers a unique opportunity for further understanding the mechanisms involved in chromosome missegregation and CIN, both specifically relevant in cancer research. In all, QATS enables the optimal detection of nuclear biomarkers to assess CIN in eukaryotic cells and provides an effective tool for genotoxicity screens.
Acknowledgements
We would like to thank the microscopy facility at CCiTUB campus clinic for their constructive suggestions.
Conflict of interest
None declared.
Funding
C.P. was supported by a Ramon y Cajal fellowship (RYC-2017-22959). C.M. was supported by the Ministerio de Ciencia e Innovación (PID2020-118768RJ-I00) and the AECC (LABAE222994MAUV).
