SimplySmart_v1 accurately recognizes DNA damage and processes a large batch of imagesTreatment of cells with etoposide induces DNA damage, which can be detected by staining for γH2AX, a marker of DSBs. The signal for γH2AX forms characteristic foci at the site of DNA breaks. First, the application was tested on a small set of images, including 3 control images and 4 images of etoposide-treated mouse primary neurons. After induction of DNA damage with etoposide (ETOP), mouse cortical neurons were stained for γH2AX to detect DNA damage and stained with DAPI to visualize the nucleus. The analysis was performed using neurons with a nuclear area in the range of 1000–1300 pixels. Nuclei on the edges of images or clumped nuclei were excluded from the statistical analysis to improve accuracy and reduce potential errors. In total, 13 control DMSO-treated and 8 etoposide-treated neurons were included in the analysis.DNA damage was greater in the ETOP group than in the DMSO group. Although this small probe did not significantly affect the number of foci (average number of foci: DMSO = 1.92, ETOP = 3.37, p > 0.05, t test, Fig. 2), the area of DNA damage foci was significantly greater in the ETOP group than in the DMSO group (average area of foci: DMSO = 103.6, ETOP = 367.6, *p < 0.05, t test, Fig. 2).Fig. 2SimplySmart_v1 revealed the increase of DNA damage in mouse cortical neurons after etoposide treatment. A Representative images of mouse cortical neurons stained for the DNA damage marker γH2AX (red). Nuclei were visualized with DAPI (cyan). Upper row, DMSO-treated neurons; lower row, etoposide (ETOP)-treated neurons. B Graphs showing the average number and area of damaged DNA per neuron obtained from a small set of images (left graphs) and the average number and area of damaged DNA foci per neuron obtained from multiple copies of small set images (right graphs). *p < 0.05, ***p < 0.001, t testTo test the application further, the measurements were performed on a large batch of images obtained by multiplication of the same images (16 × copies). This resulted in 48 single images for the DMSO group (for each channel) and 64 images for the ETOP group (for each channel). Despite the large amount of data available for processing, the application quantified DNA damage within a few seconds and generated data with the same quality as that obtained on the original small set. After visual inspection of the images, the incorrectly recognized single cells were removed as previously described. As expected, the results reflected an increase from the single small set with 208 cells analysed for the DMSO group, with 16-fold more results than the single set and the same average value = 1.92 foci per cell, and 128 cells analysed for the ETOP group, corresponding to 16-fold more results than the single set and the same average value = 3.37 foci per cell. Compared with that in the DMSO control group, the number of DNA damage foci in the ETOP group was significantly greater (***p < 0.001, t test, Fig. 2). Similarly, the area of DNA damage foci was significantly greater in the ETOP group (average area = 367.6) than in the DMSO control group (average area = 103.6) (***p < 0.001, t test, Fig. 2).Thus, the application accurately detected an increase in DNA damage in ETOP-treated neurons compared with that in DMSO-treated neurons and was capable of analysing large batches of images.SimplySmart_v1 accurately recognizes DNA damage foci in NSC-34 cellsNext, we tested the performance of SimplySmart_v1 on an immortalized cell line, NSC-34, which was treated with etoposide. Interestingly, we observed less heterogeneity in the appearance of DNA damage foci in NSC-34 cells than in primary mouse neurons, even after treatment with the same dose of etoposide (13.5 µM). Examples of more structured foci and pannuclear damage in neurons are shown in Fig. 3A. In both cases, SimplySmart_v1 was capable of identifying damaged regions (Fig. 3A). However, we noted that, in contrast to NSC-34 cells, the analysis performed on primary neurons may require a few more adjustments of thresholding.Fig. 3Representative images of DNA damage recognized in mouse primary neurons (A) and NSC-34 cells (B) by SimplySmart_v1. A Substantial heterogeneity in the appearance of DNA damage was observed across mouse primary neurons. B Representative image showing DNA damage foci in NSC-34 cells recognized by SimplySmart_v1Similar to primary neurons, SimplySmart_v1 recognized DNA damage foci in NSC-34 cells with high accuracy (Fig. 3B). These findings imply that although SimplySmart_v1 was optimized in mouse primary neurons, it can be successfully used in the analysis of DNA damage in other cell types.SimplySmart_v1 quantifies DNA damage in the desired fraction of cells on the basis of the size of the nucleusAmong neurons, other cell types, such as astrocytes, oligodendrocytes or microglia, are often present in primary neuronal cultures. The staining of a specific type of cell helps to identify the desired cell type. However, such staining is not always possible due to experimental requirements or the limited number of dyes. Moreover, the size of the nucleus may impact the magnitude of DNA damage [6]. Therefore, the ability of SimplySmart_v1 to analyse DNA damage in neurons, which were selected on the basis exclusively of nucleus size, was assessed. First, we investigated whether the size of the nucleus impacts the magnitude of DNA damage. Thus, we divided the nucleus size into three groups, 1000–1500, 1500–2000, and 2000–2500 pixels, on the basis of the histogram generated by SimplySmart_v1. We then compared the number of foci and total area of foci measured in DMSO-treated or ETOP-treated cells with SimplySmart_v1 (Fig. 4). In all the groups, the ETOP-treated cells presented more DNA damage than did the DMSO-treated cells for both parameters, as expected. However, these differences were more prominent in the last group (2000–2500 pixels) than in the other groups for the number of foci (****p < 0.0001 vs. **p < 0.01 and **p < 0.01, t test) and in the first group (1000–1500 pixels) than in the other groups for the foci area (****p < 0.0001 vs. *p < 0.05 and **p < 0.01, t test). This implies that the nucleus size impacts the magnitude of the change between the control and experimental conditions (Fig. 4A). Interestingly, the nucleus size affected the magnitude of DNA damage in the DMSO control group itself, as the area of DNA damage in the DMSO-treated group (1500–2000 pixels) was significantly larger than that in the other DMSO-treated groups (1000–1500 and 2000–2500 pixels, ***p < 0.001 and *p < 0.05, respectively; one-way ANOVA, Tukey’s post hoc test) (Fig. 4B). Hence, the influence of nucleus size on the results may be particularly important in the case of less prominent changes in DNA damage, e.g., in experimental settings without pharmacological induction of DNA damage.Fig. 4The nucleus size impacts the magnitude of DNA damage. A Quantification of foci number and foci area within different ranges of nucleus size revealed the variability in the magnitude of the increase in DNA damage after etoposide treatment compared with that under DMSO control conditions (foci number, **p < 0.01, **p < 0.01 or ****p < 0.0001 was reached depending on the nucleus size, 1000–1500, 1500–2000, 2000–2500 pixels, respectively; foci area, ****p < 0.0001, *p < 0.05 or **p < 0.01 was reached depending on the nucleus size, 1000–1500, 1500–2000, 2000–2500 pixels, respectively). B The size of the nucleus affected the magnitude of DNA damage in the DMSO-treated group. The area of DNA damage in cells within the nucleus size range of 1500–2000 pixels was significantly larger than that in the other groups, ***p < 0.001 vs. the 1000–1500 pixel group and *p < 0.05 vs. the 2000–2500 pixel group; one-way ANOVA with Tukey’s post hoc test was usedTo further test the utility of the option of nucleus size selection provided by SimplySmart_v1, we determined whether it is possible to filter neurons on the basis of only the nucleus size. The size of the neuronal nucleus was determined via 8-bit images displaying DAPI staining and the neuronal marker NeuN with FijiJ software. The values are expressed in pixels. The quantification of 11 nuclei from two random images revealed that the area of the neuronal nucleus ranged between 1273 and 2534 pixels. Thus, this range was used in the analysis using SimplySmart_v1. Only single and whole nuclei were included in the statistical analysis. The application recognized 74 nuclei within the DMSO group and 69 nuclei within the ETOP group; one nucleus was excluded from further analysis because it was a clump of two nuclei.The accuracy of neuron selection on the basis of nucleus size was subsequently verified with manual counting. Thus, the size of the nuclei in the images was measured with respect to the expression of the neuronal marker NeuN using Fiji J software. In the control DMSO group, 75 cells with clear expression of the marker NeuN and a nucleus size ranging from 1167 to 2504 pixels were obtained. Only 3 neurons were outside of the previously identified size range between 1273 and 2534, accounting for 4% of the NeuN-stained cells.In addition, 12 cells with weak NeuN expression were obtained within both sizes ranging from 1167 to 2504 and from 1273 to 2534, and 5 cells with a lack of NeuN expression and a nucleus size below these ranges were observed. Moreover, 4 cells with a lack of NeuN expression and a nucleus size above these size ranges were noted, and one likely apoptotic cell with remnants of NeuN expression and a nucleus size below this size range was identified.In etoposide-treated neurons, 83 cells with NeuN expression and nuclear sizes ranging between 1020 and 2546 pixels were observed. Of these, 10 neurons were outside of the previously defined size range of 1273–2534 pixels, for a total of 12.1%. Five nuclei within the size range with weak expression of NeuN, 7 nuclei below the range without NeuN expression, and 2 nuclei above the range without NeuN expression were observed (Fig. 5).Fig. 5The nucleus size of the NeuN-positive cells was within an initially defined range, but that of the NeuN-negative cells was outside this range. The graph shows the ranges of nucleus sizes of NeuN-positive cells determined by counting 11 random cells with NeuN expression vs. determining the nucleus size of all cells with respect to NeuN expression. Overall, these calculations revealed that the number of NeuN-positive cells fell within the initially defined nucleus size range, whereas the nucleus size of NeuN-negative cells was outside of this range. The numbers in oval shapes indicate the number of NeuN-negative cells with nucleus sizes outside the rangeThus, initial determination of nucleus size allowed the selection of NeuN-positive cells with high precision and the simultaneous removal of cells with a clear lack of NeuN expression using SimplySmart_v1.The results obtained with SimplySmart_v1 and manual counting are consistentNext, the accuracy of the DNA damage measurements performed with SimplySmart_v1 was compared with that of manual counting for the above data. SimplySmart_v1 revealed that the average number of foci per neuron was 3.92 for the control group and 8.02 for the etoposide group, with a statistically significant difference of ****p < 0.0001 according to the t test. The average total area of DNA damage per cell was 229.1 for the control group and 467.6 for the etoposide group, with a statistically significant difference of ***p < 0.001 according to the t test (Fig. 6 A and B, the foci area is not shown as it cannot be compared with manual counting results).Fig. 6Manual validation of the data revealed the accuracy of the data obtained with SimplySmart_v1. A Representative images of mouse cortical neurons stained for the neuronal marker NeuN (green), the DNA damage marker γH2AX (red) and nuclei (cyan). B Graphs showing data obtained with the SimplySmart_v1 application (left) and manual counting (right). Compared with the DMSO control group, the ETOP group presented an increase in the average number of DNA damage foci (****p < 0.0001, t test). C Comparison of the data obtained with SimplySmart_v1 and manual counting. Left panel, DMSO group (p > 0.05); right panel, ETOP group (*p < 0.05, t test)The number of DNA damage foci for NeuN-positive cells obtained by manual counting was 2.45 foci/cell in DMSO-treated neurons with clear NeuN expression and 10.59 in ETOP-treated neurons, with a statistically significant difference of ****p < 0.0001 according to the t test (Fig. 6A and B). Thus, manual validation of the data confirmed that SimplySmart_v1 accurately assessed increased DNA damage in neurons treated with etoposide.In addition, no differences were noted in the values obtained for the DMSO groups using manual counting and SimplySmart_v1 (p > 0.05, t test). However, the ETOP groups differed with *p < 0.05 (p = 0.03; t test) (Fig. 6C). Because visual DNA damage foci counting is more vulnerable to bias and human error, it is difficult to determine the actual reason for this discrepancy. Nevertheless, both manual counting and SimplySmart_v1 revealed an increase in the number of DNA damage foci after etoposide treatment, which was consistent with our findings.Comparison of SimplySmart_v1 with Fiji, CellProfiler and the focinatorWe tested the utility of SimplySmart_v1 in comparison with other open-source tools, such as Fiji (Fiji-win32) [9], CellProfiler (4.2.5) [10] and a focinator [11]. The tests were performed by a person without background in image analysis to determine the simplicity of these tools. In total, 6 images of DMSO-treated neurons and 6 images of etoposide-treated neurons were analysed. The evaluation was performed on the basis of the number of γH2AX foci. The use of CellProfiler needed a significant amount of time for initial learning and understanding of the software. The configuration and development of the pipeline was a complex process, and an accurate understanding of the software was needed. This software was found to be a better option for experienced programmers than for novice users. Compared with CellProfiler, Fiji required less time for initial understanding but required more time than did SimplySmart_v1 (Fig. 7A).Fig. 7The open-source software comparison advantages SimplySmart_v1 over other tools. A Compared with Fiji and CellProfiler, SimplySmart_v1 requires the shortest time to comprehend the usage of the software (left) and to process the images (right). B Average number of γH2AX foci per nucleus obtained with different tools on the same set of imagesAnalysis with Fiji required at least 4.5 min to analyse the DMSO group and 6.5 min to analyse the ETOP group. Similarly, approximately 2 min per group was required for CellProfiler as a first-time user of the software. The analysis performed with SimplySmart_v1 was exceptionally rapid and took approximately 1 min without any specific training prior to the analysis. Although all tools performed the analysis relatively quickly, evaluating the time spent on training as well as the actual analysis advantages SimplySmart_v1 (Fig. 7A).We also undertook an attempt to test the focinator, however, the installation turned out to be too challenging. Notably, this requires a specific version of Fiji (Fiji 1.51 bundled with Java, the Bio-Format importer, and the R-3.5.1 language), a basic understanding of the bioformats package Jar and R language. An individual without a related background found it too difficult and complex. Therefore, we focused on comparing SimplySmart_v1 with Fiji and CellProfiler.Both Fiji and CellProfiler performed well; however, the accuracy of the foci count with Fiji might be affected when the images have weak contrast between the background and objects of interest. Fiji enabled the acquisition of information such as total area, average size of the foci, % area and mean value.The effectiveness of CellProfiler is heavily dependent on pipeline creation. In accordance with the instructions in the CellProfiler manual, an example pipeline for speckle counts was used. However, the configuration problem of the inability of the pipeline to identify image sets was encountered several times as a first-time user of the software. Additionally, building a pipeline from scratch was a challenging procedure. Despite having several challenges, more information, such as the execution time to identify the targeted objects or object intensity, is provided by CellProfiler.In contrast, the analysis with SimplySmart_v1 was straightforward and effective and was supported by a very user-friendly interface. While Fiji is also user-friendly, the procedure is arduous and time-consuming. The ‘user-friendly’ aspect was found to be the greatest drawback of CellProfiler and focinator.All tools were capable of detecting increased amounts of DNA damage in the tested ETOP group compared with the DMSO control group. Figure 7B shows the analysis of the average number of foci per nucleus quantified with all three software programs. The analysis revealed that the average quantification result deviated between the DMSO control and ETOP treated groups, with the variation being lower in the ETOP treated groups among all three tools. Because of the variation, highly consistent results were obtained with SimplySmart, whereas the susceptibility to changes in quantification was observed with respect to Fiji and CellProfiler. This confirms the reliability of the quantification performed with SimplySmart_v1.Utilization of the full capabilities of SimplySmart_v1 showed induction of DNA damage in neurons expressing mutant TDP-43 (A315T) compared to wild type TDP-43To further test the ability of SimplySmart_v1, the induction of DNA damage was quantified in mouse primary neurons expressing wild-type TDP-43 (WT) or its amyotrophic lateral sclerosis-associated mutant, A315T. Primary neurons were transduced with wild-type or A315T TDP-43 lentiviruses with puromycin as a selective marker. Treatment of neurons with puromycin resulted in a number of apoptotic cells, making the identification of a single nucleus on the basis of DAPI challenging. However, staining for TDP-43 allowed us to identify surviving, transduced neurons, and this staining was used as a marker in the analysis with SimplySmart_v1 (Fig. 8 A). DNA damage was detected by immunostaining for γH2AX as previously described. Only DNA damage foci larger than 2 pixels were counted. After visual inspection of the output images, clumps of cells or cells located at the edges of the images were excluded from the statistical analysis.Fig. 8Compared with WT TDP-43, TDP-43 mutant A315T induced more DNA damage, as calculated with a one-sided t test. A Representative images of mouse cortical neurons stained with TDP-43 and DAPI. The selection of cells treated with puromycin resulted in many apoptotic nuclei; however, staining for TDP-43 revealed surviving transduced neurons. B The upper panel shows DNA damage areas identified using SimplySmart_v1 in neuronal nuclei expressing A315T (color-mapped regions). The graphs show that the average number of DNA damage foci in the WT TDP-43 group was significantly lower than that in the A315T group (*p < 0.05, one-sided t test), whereas the foci area was unchanged. The images were analysed with SimplySmart_v1 with a threshold value = 140, a marker area range of 1000–5000 pixels, and a minimal size > 2 pixels. Forty-nine nuclei from the WT group and 38 from the A315T group were analysed from 3 technical replicatesIn our previous study, we reported that neurons expressing mutant TDP-43 A315T are characterized by increased DNA damage compared with those expressing wild-type TDP-43 after pharmacological induction of DNA damage with etoposide [8]. Here, we used SimplySmart_v1 to investigate whether neurons expressing TDP-43 A315T exhibit increased DNA damage compared with those expressing wild-type TDP-43 without any exogenous induction of DNA damage. Consistent with our previous findings, we did not observe a statistically significant difference in the magnitude of DNA damage between TDP-43 A315T neurons and wild-type TDP-43 neurons according to two-sided t test (p > 0.05). However, neurons expressing mutant TDP-43 A315T were characterized by a significantly greater number of DNA damage foci than were those expressing wild-type TDP-43 when a one-sided t test was applied (*p < 0.05, t test). No statistically significant differences in the area of DNA damage foci were noted between the investigated groups, regardless of the type of test used (two- or one-sided) (p > 0.05, t test) (Fig. 8 B). Although the data were obtained with different quantification techniques and in different experimental settings, they are consistent with our previous findings [8]. They also showed that the flexibility of SimplySmart_v1 makes it a powerful tool that is applicable for DNA damage analysis in real experimental settings.
