Test article formulation

Master stock solutions for each chemical were made fresh on the day of the experiment in DMSO.

Methyl methanesulphonate (MMS), CAS no. 66-27-3, carbendazim, each supplied from Sigma-Aldrich. The working concentrations for MMS (0.00, 0.31, 0.63, 1.25, 2.50, 5.00 μg/mL) and carbendazim (0.00, 0.40, 0.60, 0.80, 1.20 and 1.60 μg/mL) were selected based on the data produced by Verma et al. (2017). The vehicle control was dimethylsulphoxide (DMSO) (CAS no. 67-68-5).

Cell culture and growth media

Human, p53 competent, lymphoblastoid TK6 cells (Cat. No. 95111735, alternate collection no. ATCC CRL 8015) were used in this study and obtained from European Collection of Cell Cultures (ECACC) Salisbury (Branda et al. 2001). RPMI 1640 (Gibco) culture media supplemented with 1% penicillin streptomycin (pen strep) and 10% heat inactivated horse serum (Gibco) was used for TK6 cell culture. Cells were incubated at 37 °C in a humidified atmosphere of 5% (v/v) CO2. TK6 cells doubled every 16–18 h and once cells reached confluence sub-cultures were established. Each subculture did not exceed a confluence value of 1 × 106 cells/mL as per ECACC/ATCC recommendations.

Treatment of cell cultures

2 × 105 TK6 cells/mL were placed in a series of sterile vented tissue culture flasks and treated with either MMS or Carbendazim for a 1.5 cell cycle period with no recovery. Dose volume to cell culture did not exceed 1% i.e., 100 μL of dose was added to 9.9 mL of cell suspension. Any precipitation or colour change was noted upon chemical addition to cell culture flask. All incubation steps occurred at 37 °C, 5% (v/v) CO2 ± 0.5% in air. Each replicate when performed on the same day were generated from cells of a different passage. After the treatment period, cell counts were taken for each culture using a Beckman coulter counter. Cell cultures were then transferred to 15 mL centrifuge tubes and were centrifuged at 200g for 8 min, supernatant was discarded, and the pellet re-suspended in 5 mL pre-warmed RPMI HIHS culture media. Subsequently, the RPMI media was removed via centrifugation at 200g, the pellet was re-suspended and wash step repeated with 5 mL Phosphate-Buffered Saline (PBS). The highest concentration tested was one that allowed the maximum exposure up to 2000 μg/mL or 10 mM for freely soluble test articles, or the limit of solubility or toxicity, whichever is lower (OECD 2023). Where toxicity was a limiting factor, the maximum treatment concentration selected for analysis was that with relative cell growth at 30%.

Cell counts and cytotoxicity

Cells were counted using a Beckman Coulter Counter and Relative Cell Growth (RCG) was used to estimate cytotoxicity in treated samples. RCG was calculated as follows:

$$\text= \frac.\,\text/\text}.\,\text/\text} \times 100$$

Cell fixation and staining

Following cell treatment period wash steps, the cell pellet was resuspended in residual PBS and BD FACS Lyse was used to fix and permeabilise the cells. BD FACSLyse was diluted in a 1:10 ratio with distilled water (dH2O) (1 mL FACS Lyse: 9 mL dH2O). Avoiding further agitation of samples, 2 mL of FACS Lyse solution was added to each sample. Samples were incubated at room temperature for 12 min. Following incubation samples were centrifuged at 200g for 5 min. The supernatant was discarded and the cell pellet gently resuspended in the remaining solution, 5 mL PBS was added and samples returned to centrifuge for further 5 min, this wash step was repeated twice. Samples at this stage may be placed in the fridge and stained at a later date or stained immediately. Samples were stained with 300 μL of antibody (AB) master mix for a minimum of 60 min under agitation at room temperature. The master mix consisted of BV421 anti-ɣH2AX AB (Cat. No. 564720, BD Biosciences), AF488 anti-pH3 AB (Cat. No. 641003, BioLegend) and PE anti-p53 AB (Cat.No. 645805, BioLegend) in the ratio 3 μL of pH 3: 5 μL of ɣH2AX: 6 μL of p53: 286 μL of PBS. DRAQ5™ DNA stain (Cat. No. 564902, BD Biosciences) was used to label nuclei and MN. 2 μL DRAQ5™:98 μL PBS was mixed, 100 μL of the 1:49 ratio stain solution added to each 300 μL cell sample antibody solution making final staining ratio of 1:199. Samples were counterstained with DRAQ5™ for a minimum of 20 min. After the staining period ended samples were centrifuged and washed in 5 mL PBS.

ImageStream X Mark II® data acquisition

Samples were analysed on a Cytek® Amnis® ImageStream®X Mk IIimaging flow cytometer using Cytek® Amnis® INSPIRE® software version 6.2 (Merck Millipore, Nottingham UK). Prior to experimental analysis, following the manufacturer’s instructions for appropriate ImageStream X Mark II® set up and quality control procedures, all system calibrations performed and passed using ImageStream X Mark II® SpeedBead calibration reagents (Cat. No. 400041).

100 µL aliquots of cell suspension at a concentration of ~ 7 × 105 and no less than ~ 4 × 105 cells/mL were prepared in 1.5 mL Eppendorfs. Replicate 1 and 2 for both MMS and Carbendazim treated cell samples were analysed by placing the Eppendorfs on the sample port and a minimum of 50µl sample loaded. For replicate 3 of both chemicals, cell samples were transferred to a 96-well plate. Prior to data collection laser intensities were balanced to limit saturation events and fluorescence channel overspill. For data acquisition an INSPIRE® template was set up for in focus and single cellular event gating, using Aspect Ratio and Root Mean Square (RMS) features, ensuring collected cells were sufficiently circular and in focus. Once established, data were acquired at a low velocity 66.0 (mm/s), resulting in 15,000–30,000 single cellular events being collected in approximately 45 s at a magnification of × 40, and subsequently automatically saved for each experimental replicate. Acquisition of images for ɣH2AX pH3, p53 and DRAQ5™ assessment occurred in: Channel 1 and 9, bright field; channel 2, 3, 7 and 11 fluorescence; Channel 6 side scatter. Whilst these channels were of main interest data was acquired for all 12 channels. All samples were collected with no compensation applied.

Compensation sample files were acquired at the same time as sample analysis using the INSPIRE® acquisition compensation wizard or manually using Compensation beads (Cat.No.01-2222-41) from Thermofisher Scientific. Acquisition of compensation samples was performed without the presence of brightfield or side scatter. 488, 405 and 642 nm lasers were utilised at the same intensity values used during the experimental setup. Acquired files formed compensation matrices in Cytek® Amnis® IDEAS® 6.2 software.

IDEAS® 6.2 data analysis

Once data were collected for all samples, the acquired raw image files (.RIF) data files were analysed using IDEAS® v6.2 software (Merck Millipore, Nottingham, UK). Compensation matrices and template analyses were applied enabling generation of subsequent.CIF (compensated image file) and .DAF (data analysis file) files.

Use of a standardised analysis template allowed for batch processing of data files and extraction of the following metrics: Cell cycle, MN, ɣH2AX, pH3 and p53. Due to the slight shifting of the DNA cell cycle histogram the position of the cell cycle gates (G1, S, G2/M) were adjusted accordingly per sample. The finalised template used the following generated masks: Brightfield 1A, Brightfield 1B, Cytoplasm, Nuclear Mask 1, Nuclear Mask 2, Nuclear Mask 3, gH2AX mask, Nuclear gH2AX mask, pH3 Mask, Nuclear pH3 Mask, P53 Mask, P53 Cytoplasmic Mask, P53 Nuclear mask and Complete Final MN Mask (CFM). The functions in various combinations used to generate these masks were morphology, adaptive erode, threshold, range, dilate, intensity, spot, watershed and level set. Features in the IDEAS® system are mathematical expressions that assess, within the image, quantitative and spatial information. These allow for the generation of histograms and scatter graphs in the analysis area for cell population responses to be assessed. The features used were: Aspect ratio, Area, contrast, Gradient Root Mean Square (RMS), Intensity, Similarity, and spot. The mask development and features used were based on the recommendations within the IDEAS® 6.2 user manual 2015, the publication Imaging Flow Cytometry: Methods and Protocols 2016 and literature (Filby et al. 2016; Patterson et al. 2015; Rodrigues et al. 2018; Verma et al. 2018). Details of the various masks, features and functions used can be found in the example template file provided for download at the BioStudies database under accession number S-BSST1351. For additional clarity, Supplementary Table 3 (ST3) demonstrates the image function combinations used to generate the three masks for the CFM.

Cell populations used for analysis

Cells that are circular, single and in focus were determined by meeting the criteria of having an aspect ratio of >  ~ 0.6, brightfield area of >  ~ 70 but <  ~ 450 and gradient root mean square value > 50, respectively (Filby et al. 2016; Rodrigues et al. 2018; Verma et al. 2018). This single and in focus cell population was then assessed for cell health. The healthy cell population was determined by combining the nuclear area feature of nuclear mask 1 with the contrast feature applied to the cytoplasm mask. Healthy cells have a high nuclear area with pixels that meet the 50% threshold intensity and low cytoplasmic contrast (Filby et al. 2016, 2011; Rodrigues 2018; Rodrigues et al. 2018). A healthy cell population with DNA content showing > 1 × 105 nuclear intensity was used for pH3 and p53 assessment. The cell population used for ɣH2AX, and MN assessment was also required to be negative for pH3 staining. To determine the mononucleated cell population for MN assessment nuclear mask 3 was used in combination with the spot count feature. This generated a histogram from which healthy mononucleated cells could be extracted.

Biomarker and MN metric extraction

Gating parameters for ɣH2AX pH3 and p53 were determined based on the cell populations of unstained vehicle and stained vehicle samples compared to unstained and stained genotoxicant dosed samples on intensity histograms. Signal separation on scatter graphs and specific signal masking based on nuclear/cytoplasmic localization further refined the gating strategy. Development of the gating strategy is described in the results section. A minimum of ~ 13,000 single in focus healthy cells were assessed per dose per replicate to obtain the dose response for each biomarker.

The mononucleated healthy cell population was used to extract the MN data through applying the spot count feature to the complete final micronucleus mask. This automatically extracted the cell population containing MN. Development of the complete final micronucleus mask is described in the results section and applies the MN criteria for slide based scoring (Fenech et al. 2003). A minimum of ~ 10,000 mononucleated cells were assessed per dose, per replicate to obtain the MN dose response.

Dose–response comparison

Covariate BMD analysis using PROAST 71.1 in the R programming environment (version 4.2.1) was performed for carbendazim collected either by the ISMN-mb approach or scored manually from slides (manual data obtained from Verma et al. 2018). The results of this analysis are presented in Figure S1.

Micronucleus mask accuracy

To understand the effectiveness of the MN mask generated, two metrics were assessed based on the MN populations of four tool compounds at the highest analysable concentration. The first, ‘percentage accuracy’ (%Accuracy), is described as; of the cell population Identified by the MN mask as containing MN what percentage were true, this was confirmed by inspecting each saved cell image by eye. %Accuracy was calculated as follows:

$$\%\text= \frac}} \times 100$$

The second, ‘percentage Miss Rate’ (%Miss Rate) required a defined cell population where images could be assessed both by eye and using the MN mask. This population was selected by plotting the healthy mononucleated cell population as a histogram using the Gradient RMS feature on the DNA content. A gate termed DNA Focus was drawn to the right of the central point of the in-focus nuclear content histogram. This population taken forward provided cells with clear crisp nuclei staining, giving the MN mask the best chance of identifying all MN, whilst randomly selecting cells to help minimize technician population bias and still being representative of the whole cell population. The total number of cells assessed were 1000–1500 per data file. %Miss Rate was calculated as follows:

$$\%\text= \frac}} \times 100$$

Data evaluation criteria

All tables and graphs were generated using Excel Microsoft. Unless otherwise stated data displayed in tables and plotted on graphs for biomarker and MN metrics are presented in raw response fold change that has been square-rooted to normalise. Cell cycle data are presented as raw response cell percentages. Fold change cut offs were also applied to graphical data in line with industry standards. It is important to note, the fold changes have been taken from the literature for expediency and specific cut off values for the ImageStream platform have not yet been defined. For pH3 this was 1.3 fold increase and 0.7 fold decrease when compared to vehicle controls (Khoury et al. 2016). For ɣH2AX and p53 fold change increase of 1.5 when compared to vehicle controls was used (Dertinger et al. 2019; Smart et al. 2011). These fold change values were square-rooted giving values of: 1.2 for ɣH2AX and p53. 1.1 for pH3 for increase in signal and 0.8 for decrease in signal. The MN response was considered positive based on a statistically significant response (p < 0.05) compared to that of vehicle control (Johnson et al. 2014). Where statistical significance was not achieved the MN response was considered positive based on a greater than twofold increase compared to control (Shi et al. 2010; Takeiri et al. 2019), this was then square-rooted giving a fold change value threshold of 1.4. A genotoxic response is considered positive when the mean response relative to control exceeds the fold change cut off or is statistically significant. Where both MN and biomarker response exceed fold change cut offs and/or are statistically significant indication of MoA may be inferred (Bryce et al. 2017; Dertinger et al. 2019).

Statistical analysis

Dose–response data were tested for variance homogeneity and normality using Bartlett and Shapiro–Wilk tests, respectively. Where datasets ‘passed’ these tests (i.e., p > 0.05), a one-way ANOVA was run followed by pairwise testing versus control to establish response significance (p ≤ 0.05) carried out by Dunnett’s T-test method. In the event the data failed the distribution tests the non-parametric Dunn’s test was performed on the raw response data. All statistics were calculated using DRSMOOTH package in the R programming environment using methods described in Johnson et al. (2014).