Blood Collection and Irradiation

Blood was collected by venipuncture from healthy adult donors (between the ages of 27–48 years) with informed consent and with approval by Columbia University's Research Ethics Board (approved IRB protocol IRBAAAE-2671) in 5-ml lithium-heparinized Vacutainer® tubes (BD Vacutainer™, Franklin Lakes, NJ). All donors were non-smokers in relatively good health at the time of donation with no obvious illnesses such as colds, flu or infections and no known exposures to medical ionizing radiation within the last 12 months. Aliquots of blood samples (1 ml) were dispensed into 15-ml conical bottom tubes (Santa Cruz Biotechnology® Inc., Dallas, TX) and transported to a Gammacell® 40 ¹³⁷cesium (¹³⁷Cs) irradiator (Atomic Energy of Canada Ltd., Chalk River, Canada). The blood sample tubes were placed on their side in the middle of the chamber and exposed to 0 (control), 1.0, 2.0, 3.0 or 4.0 Gy of γ rays at a dose rate of 0.73 Gy/min. The ¹³⁷Cs irradiator is calibrated annually with TLDs and homogeneity of exposure across the sample volume was verified using EBT3 Gafchromic™ film with less than 2% variation within the sample (Ashland Advanced Materials, Bridgewater, NJ).

Small Volume CBMN Assay Protocol

Aliquots of 100 µl from the irradiated blood samples were added to 1.4 ml 2D Matrix microtubes (Thermo Scientific™, Waltham, MA) containing 900 µl PB-MAX karyotyping media (Life Technologies, Grand Island, NY). Each microtube was mixed five times using a 1.2-ml multichannel electronic pipet (Eppendorf Inc., Westbury, NY or Thermo Scientific). The rack containing microtubes was placed into an incubator at 37°C, 5% CO₂. After 24 h of incubation, cytochalasin-B (Cyto-B; Sigma-Aldrich®, St. Louis, MO) was added to cultures at a final concentration of 6 µg/ml to block cytokinesis of proliferating lymphocytes. After an additional 44 h of incubation, the samples were harvested and the microtube plates were spun down before the supernatant was removed.

Samples were then resuspended in 900 µl FACS™ Lysing Solution for 2 min (FLS; BD Biosciences, Franklin Lakes, NJ) and spun down. The cells were resuspended in 950 µl FLS and incubated for 13 min, and then spun down again. Each sample was then washed twice with 950 µl 1× phosphate buffered saline (PBS; Gibco®, Gaithersburg, MD) and resuspended in a final volume of 25 µl PBS. All steps in the procedure were performed at room temperature and microtubes in racks were spun at 250g for 3 min. Finally, DRAQ5 (eBioscience, San Diego, CA) was added to the sample at a final concentration of 20 µM and incubated at room temperature for a minimum of 5 min.

Data Acquisition and Analysis on the ISX and IDEAS

All samples in suspension were run on the ISX MKII system (MilliporeSigma, Seattle, WA) at 40× magnification with the 488-nm laser set to the maximum 200 mW. BF images were collected in channel 1 while DRAQ5 images were collected in channel 5. All other channels were disabled during data collection. Data were collected using the ISX INSPIRE® software with only the Area feature applied. Events with area less than 100 pixels (25 µm²) were gated out to minimize the collection of small debris. A total of up to 200,000 events were collected per sample and each data file typically took approximately 13 min to acquire, on average.

All data were analyzed based on the gating strategy developed in the IDEAS software package described elsewhere (24). Briefly, the analysis identifies BNCs that possess nuclei of high circular morphology, similar areas and DRAQ5 intensities, and with sufficient spatial separation (i.e., not overlapping). Once the final BNC population was obtained, MN within the BNCs were automatically identified and enumerated. The analysis template was applied to all data files and automatically batch processed, which enabled the number of BNCs and MN per file to be extracted and exported to Office 365® Microsoft Excel for analysis.

Dose-Response Calibration Curve

All calibration curves were generated using GraphPad Prism version 6 (La Jolla, CA), and fitted with a non-linear regression quadratic equation of the form:

where Y is the number of MN per BNC, D is the dose, c is a constant, and α and β are constants corresponding to the linear and quadratic components of the dose, respectively. Goodness of fit (R²) of the calibration curve was tested. Figure 1 shows the average dose-response calibration curve obtained from 100-µl whole blood samples from four healthy donors. Coefficients for all dose-response calibration curves are shown in Table 1. It can be seen that MN frequency increases according to a linear-quadratic dependence with dose as expected from previously reported results (23, 24). The average number of BNCs from all samples was 509 at 0 Gy and 250 at 4 Gy, sufficient for triage biodosimetry (27, 28).

FIG. 1.

Dose-response calibration curve representing the average rate of micronuclei (MN) per binucleated cell (BNC) in lymphocytes from 100-µl whole blood samples, which were collected from four healthy donors and ex vivo exposed to 0 to 4 Gy. Error bars represent the standard error of the mean.

TABLE 1

Coefficient Values Calculated for each Linear-Quadratic Fit and Goodness of Fit (R²) from each Calibration Curve for Figs 1–4

Blood-to-Media Ratio

To ensure that the maximum number of BNCs would be generated within the limited culture volume of 250 µl, we tested blood-to-media ratios. Figure 2A shows a comparison of dose-response curves obtained from 100 µl of blood cultured in 900 µl of PB-MAX media (blood-to-media ratio 1: 9) with 200 µl of blood cultured in 800 µl of PB-MAX media (blood-to-media ratio 2: 8) using the 96-well 1.4 ml Matrix tube format described above (n ¼ 4 donors). Coefficients for the dose-response calibration curves are shown in Table 1. The results show that the mean MN frequency for both the 1: 9 and 2: 8 blood-to-media conditions was similar, indicating that increasing the blood-to-media ratio did not adversely affect MN frequency. Furthermore, due to the increased blood volume, the total number of BNCs is generally higher from the 2: 8 ratio than the 1: 9 ratio in samples with the same total culture volume across the 0–4 Gy dose range. On average, 544 BNCs were obtained from the 2: 8 ratio while only 300 BNCs were captured from the 1: 9 ratio, an improvement of 81%. Statistical significance was tested using the Wilcoxon signed rank test, with differences between the two data points at 0, 1 and 2 Gy being statistically significant (P < 0.05). As such, it was decided that a blood-to-media ratio of 2: 8 would be used for all subsequent experiments.

FIG. 2.

Optimization of blood-to-media ratio. Panel A: Dose-response calibration curves indicating a similar increase in MN frequency for both the 1: 9 and 2: 8 blood-to-media ratios. Panel B: The average number of BNCs scored for the 1: 9 and 2: 8 blood-to-media ratios. All data points represent the average of four donors and error bars represent the standard error of the mean. Statistical significance of BNCs between the two groups was examined by the Wilcoxon signed rank test (*P < 0.05).

Addition of Hypotonic Solution

The effect of increasing the blood-to-media ratio resulted in an increase in partially lysed RBCs and apoptotic/ necrotic cell debris (data not shown), which hindered the acquisition of isolated BNC images and impacted the accuracy of MN quantification. The most effective strategy to improve RBC lysis, while maintaining a simple processing method, is through the addition of a hypotonic solution step (75 mM KCl, 800 µl) to whole blood prior to FLS (2, 29). Figure 3A shows that in addition to improving RBC lysis and eliminating the apoptotic/necrotic cell debris, 75 mM KCl caused the lymphocytes to swell slightly prior to fixation. This improved the circular morphology of, and separation between, the main nuclei. In addition, the DRAQ5 staining was much more uniform in the main nuclei and the MN. The MN were also better separated from the main nuclei, and therefore easier to mask in IDEAS. Figure 3B shows the MN dose-response calibration curves generated from samples processed in the presence and absence of KCl. The inclusion of KCl reduced nonirradiated, baseline MN levels by 51% (0.055 ± 0.012 to 0.028 ± 0.014) and at the lower doses, removed many of the dim artifacts resulting from non-uniform DRAQ5 staining, which were close to the main nuclei that were incorrectly scored as MN (28, 30). Furthermore, R² values improved from 0.70 to 0.87 (Table 1).

FIG. 3.

Improvements to the BNC images and dose-response calibration curves through the addition of hypotonic solution (75 mM KCl). Panel A: More spatially separated nuclei and MN, with more uniform DRAQ5 staining intensity was observed with the use of KCl. Panel B: The dose-response calibration curves representing the average of four donors showing the rate of MN/BNC using sample processing protocols with or without the use of KCl. Error bars represent the standard error of the mean.

DRAQ5 Concentration

High-quality staining of the BNCs and MN is critical for their capture using the ISX and accurate identification in IDEAS. DRAQ5 was determined to be the best DNA dye for IFC-based CBMN assay with 488-nm laser (21), however, the optimal concentration of the dye has not been examined. If the staining is too dim, low-intensity fluorescent artifacts can be incorrectly identified as MN. Conversely, if the staining is too bright, the main nuclei can appear blurred or saturated and the number of legitimate BNCs will be underestimated. This also increases the chances that very bright artifacts can be incorrectly identified as MN. Thus, it was essential to determine the optimal concentration of DRAQ5.

Figure 4A demonstrates that samples stained with only 10 µM DRAQ5 showed relatively low fluorescent intensity in the two main nuclei and were not being correctly identified in IDEAS. Furthermore, a number of dim artifacts were incorrectly identified as MN, as evidenced by overestimation of the 0 Gy MN frequencies and high standard errors (Fig. 4B). In contrast, samples stained with 50 µM DRAQ5 exhibited overstaining and saturation of the BNCs and MN, resulting in a number of bright artifacts at the edges of the nuclei being incorrectly identified as MN, most noticeably at 0 Gy (Fig. 4B). Overall, the data confirmed that at 20 µM DRAQ5, the main nuclei and MN were more uniformly stained, leading to a better signal-to-noise ratio and identification of MN, resulting in an improved linear-quadratic dose-response calibration curve (Table 1).

FIG. 4.

Improvement in the MN and BNC images through the optimization of DRAQ5 concentration. Panel A: At 20 µM of DRAQ5, the BNC images demonstrate optimal signal-to-noise ratio. Panel B: Average dose-response calibration curves for four different donors using either 10, 20 or 50 µM as the final concentration of DRAQ5. Error bars represent the standard error of the mean.

Optimization of Data Acquisition on ISX

To further optimize the assay for small volume and increase the throughput of the assay, we modified the acquisition template to improve the efficiency of data collection on the ISX. A template was created that permitted rejection of a number of undesired events such as doublets, large aggregates, unfocused nuclear images and very dimly stained cells. This ensured that all data files contained mostly sharply focused, single cell images. Figure 5 shows the sequential application of each region that was applied during acquisition along with representative imagery. This optimization of the data acquisition template reduced the total number of collected events from 200,000 to approximately 30,000 while maintaining the same number of single cells collected. Previously, by only gating out very small debris, data file sizes reached up to 2 GB. By using the gating strategy described here during acquisition, data file sizes were reduced to approximately 300 MB, on average, which in turn, reduced the average analysis time per file in IDEAS by 78%, from 9 min to 2 min.

FIG. 5.

Data acquisition template used to collect only single, highly-focused, DNA-positive cells. Panel A: Bivariate plot of BF aspect ratio versus BF area, allowing for the selection of single cells and the removal of doublets and small and large debris. Panel B: Bivariate plot of DRAQ5 gradient RMS versus BF gradient RMS enabling the selection of sharply-focused cells while eliminating blurry cells. Panel C: Histogram of DRAQ5 intensity allows for the selection of DNA positive cells and the elimination of very dimly stained cells.

Optimization of the Analysis Template in IDEAS

To ensure that the IDEAS analysis template was performing correctly with the improvements made to the sample processing methodology, all images in the final BNC and MN populations were visually inspected. It was observed that some artifacts residing outside of the cytoplasm and between the two nuclei, as well as some dim and elongated artifacts, were being incorrectly scored as MN. Therefore, we added two additional panels in the IDEAS template (Fig. 6) to eliminate these false-positive MNs.

FIG. 6.

Elimination of false positive artifacts. Panel A: Use of the cytoplasm mask to identify false-positive MN residing outside of the cytoplasm (top image) and the skeleton mask to identify artifacts residing between the two nuclei that are incorrectly identified as MN (middle image). Panel B: Use of the aspect ratio of the MN mask to identify elongated false-positive MN and the ratio of the median pixel intensity of the MN and BNC images to eliminate dim artifacts.

In IDEAS, by using masks to highlight specific pixels in regions of interest and then using features to perform calculations based on those masks, certain attributes of the image can be quantified, such as area, aspect ratio, among others. Figure 6A shows how artifacts residing outside the cytoplasm or between the two nuclei are eliminated. A mask was created on the BF images to identify the cytoplasm of the cell (Fig. 6A, top image) and the distance between this mask and the MN mask was calculated using the Spot Distance feature. Distances greater than zero indicated that the MN resided outside of the cytoplasm. Next, to remove artifacts between the two nuclei, the skeleton mask, which highlights pixels along the center of the semi-major axis of the image, was created on the BNC image (Fig. 6A, middle image). If the distance between this mask and the MN mask was equal to zero, this indicated that the spot identified by the MN mask was actually an artifact between the two nuclei. Figure 6B shows how dim and elongated artifacts that are incorrectly identified as true MN were eliminated. The aspect ratio provides a measure of circularity, and objects with an aspect ratio of less than 0.45 were observed to be elongated artifacts (Fig. 6B, bottom image). The ratio of the median pixel values of both the BNC and MN masks was calculated, and potential MN that had a median pixel intensity less than one fourth of the median pixel intensity of the BNC mask were determined to be dim artifacts (Fig. 6B, top image). A region was then created that allowed for the exclusion of both of these false-positive events from MN (Fig. 6B, middle image). Finally, visual inspection of the BNC population containing MN before and after addition of the new panels in the template revealed that the false-positive rate of MN was reduced by 20% based on previous observations, from 15% (24) to 12%.