Following a mass-casualty nuclear/radiological event, there will be an important need for rapid and accurate estimation of absorbed dose for biological triage. The cytokinesis-block micronucleus (CBMN) assay is an established and validated cytogenetic biomarker used to assess DNA damage in irradiated peripheral blood lymphocytes. Here, we describe an intercomparison experiment between two biodosimetry laboratories, located at Columbia University (CU) and Health Canada (HC) that performed different variants of the human blood CBMN assay to reconstruct dose in human blood, with CU performing the assay on isolated lymphocytes and using semi-automated scoring whereas HC used the more conventional whole blood assay. Although the micronucleus yields varied significantly between the two assays, the predicted doses closely matched up to 4 Gy – the range from which the HC calibration curve was previously established. These results highlight the importance of a robust calibration curve(s) across a wide age range of donors that match the exposure scenario as closely as possible and that will account for differences in methodology between laboratories. We have seen that at low doses, variability in the results may be attributed to variation in the processing while at higher doses the variation is dominated by inter-individual variation in cell proliferation. This interlaboratory collaboration further highlights the usefulness of the CBMN endpoint to accurately reconstruct absorbed dose in human blood after ionizing radiation exposure.

High throughput biodosimetry has been the focus of many studies over the last several years [Sproull and Camphausen, 2016; Sproull et al., 2017; Satyamitra et al., 2021], which has been driven by the requirements to rapidly assess radiation exposure in large populations following radiologic accidents such as Fukushima [Lee et al., 2012] or potential radiological terrorism [Homeland Security Council, 2006]. Traditional, microscope-based methods for biodosimetry, such as the dicentric chromosome assay, can be time consuming and labor intensive as it requires manual identification of chromosome damage [International Atomic Energy Agency 2001]. Further, these microscope-based methods are not suitable for mass casualty situations where laboratory capacity would quickly become overwhelmed.

There has been much effort focused on adapting cytogenetic biodosimetry assays to high throughput processing. The cytokinesis-block micronucleus (CBMN) assay [Fenech, 2007] is one method that lends itself particularly well to automation. The CBMN assay quantifies radiation-induced chromosome damage expressed as post-mitotic micronuclei (Mni). Lymphocytes are stimulated to divide in culture, but they are blocked at cytokinesis, preventing separation of the two new cells. Healthy lymphocytes form binucleated cells (BNCs), while those with chromosome damage can form additional Mni containing chromosomal fragments, with the yield of Mni per BNC increasing monotonically in a linear quadratic form with dose up to about 4 Gy [Vral et al., 2011]. Automation of the CBMN has been demonstrated through the use of automated slide scanning and image capture [Schunck et al., 2004], automated image analysis [Thierens et al., 2014], flow cytometry [Avlasevich et al., 2006], and imaging flow cytometry [Rodrigues et al., 2016] techniques and through the use of custom [Garty et al., 2010, 2011; Repin et al., 2014] or commercial robotics [Repin et al., 2017]. A key advantage of the CBMN assay is that the signal is fairly stable for many months post-exposure, removing the need for early acquisition of blood samples [Vral et al., 2011]. A disadvantage is that in some conditions, Mni yields were seen to drop at very high doses, though this can be overcome by incorporating extra information such as mitotic index [Garty et al., 2023] or dicentric yields [Shuryak et al., 2022], or by adjusting the assay chemistry [Pujol-Canadell et al., 2020].

Traditionally, the CBMN assay is performed on whole blood with manual processing of individual 1 mL blood samples [International Atomic Energy Agency, 2001; Fenech, 2007]. At Columbia University (CU), a method has been developed that isolates lymphocytes prior to culturing. This assay was developed to utilize robotics that would get clogged by the high concentration of red blood cells present in whole blood. The use of multiwell plates allows parallelization of sample handling by loading multiple donor samples into one multiwell plate [Garty et al., 2016; Repin et al., 2017]. In order to validate this method, a comparison was made with the traditional large volume, whole blood CBMN assay method with manually selected and manually scored Mni/BNCs conducted at Health Canada (HC) [McNamee et al., 2009]. HC is the lead laboratory of the Canadian Biological Dosimetry Network [Miller et al., 2007; Wilkins et al., 2015], which currently conducts biologically based dose estimates of radiation exposure using both the dicentric chromosome assay and the CBMN assay [Miller et al., 2007; McNamee et al., 2009; Wilkins et al., 2015]. This intercomparison exercise was used to evaluate Mni yields across donors from a tight demographic and Mni yields from the same donors analyzed several weeks apart.

This work demonstrated the importance of conducting interlaboratory comparisons to validate modifications on assay methodologies. It also confirms the importance of using calibration curves that are well matched with the exposure scenario and those that are current such that the scorers who generated the calibration curves are the ones performing the dose estimates.

Donor Recruitment

To minimize possible confounders due to inter-individual variability, all blood donors were recruited, and blood samples were collected, at CU, by the nurses in the Department of Radiation Oncology. Twenty healthy male nonsmoking volunteers (aged 20–29 years) with no known exposures to ionizing radiation in the previous year were recruited and informed consent was obtained according to the CU Institutional Review Board, protocol AAAE9551. Two volunteers were recruited each week and processed as a batch. Due to the logistics of sample shipping and culturing, blood collection was always performed on a Monday morning.

The first ten donors were used only at CU for the purpose of generating an in-house calibration curve for this study. While the standard practice is to use donors of both sexes and a wide age range, we intentionally used a tight demographic, to minimize contribution of any sex or age effects, if they exist. The second set of ten donors were used for the actual interlaboratory comparison study (references to donor numbers below refer only to this second set of donors). For these donors, the irradiated aliquots from the same donor were analyzed at CU and HC, with the samples randomized with 3-digit numbers so that neither site knew the donor ID or dose for each aliquot. Additionally, three donors from the second set were sampled three times each, on separate weeks of the study (with a gap of between 2 and 4 weeks between consecutive draws) providing an indication of the variability due to sample handling in the assay. To facilitate analysis, the repeat samplings were given unique donor numbers; thus, donors 4 and 12 were repeats of donor 1, donors 11 and 15 were repeats of donor 3, and donors 9 and 16 were repeats of donor 5. All other donors were unique.

Irradiation and Dosimetry

For each donor, at least 16 mL of peripheral blood was drawn by venipuncture into spray-coated sodium heparin (158 USP) vacutainer tubes (Becton Dickinson and Company, Franklin Lakes, NJ, USA). The blood was transported at room temperature (RT) from the Department of Radiation Oncology to the laboratory (both in the same building at Columbia University’s Health Sciences campus) and each donor sample was split into eight 2 mL aliquots and irradiated to a dose of 0, 1, 2, 3, 4, 5, 7.5, or 10 Gy with γ rays at a dose rate of 0.73 Gy/min using a Gammacell 40 cesium irradiator (Atomic Energy of Canada, Ltd., Chalk River, ON, Canada). Irradiations were performed as soon as practically possible, typically within an hour of the blood draw. Samples were irradiated at RT in air and then stored in an incubator prior to blinding. Calibration and dose rate monitoring were performed routinely by the chief medical physicist at Columbia University Irving Medical Center’s Department of Radiation Oncology using thermoluminescent dosimeters.

Shipping

Following irradiation, each blood aliquot was further split into two 1 mL aliquots and assigned a random three-digit identifier. One set of the paired aliquots was shipped to HC. Shipping was performed using temperature-controlled shipping boxes (Credo 22-248, Minnesota Thermal Science, Plymouth, MN, USA). The Credo 22-248 guarantees a temperature range of 15–25°C for up to 48 h. The gel inserts for the Credo system were pre-conditioned at 30°C over the weekend prior to shipping and the package included a temperature logger (Lascar Electronic Limited, Whiteparish, UK) to verify that the temperature remained between 15 and 25°C. Boxes were shipped by FedEx overnight and arrived at HC within 24 h of blood collection. Samples in shipments exceeding the 24 h ship time or the predefined temperature range would have been discarded had they occurred. To mimic shipping conditions, the CU samples were left on the bench under ambient conditions until the HC samples had arrived and processing on both samples was performed simultaneously.

HC CBMN Assay Protocol

The CBMN assay was performed, at Health Canada, according to a modified procedure of Fenech et al. [Fenech and Morley, 1985]. Briefly, 1 mL of whole blood per sample was diluted 1:9 with RPMI-1640 culture medium (Gibco) containing 10% inactivated fetal bovine serum (Sigma), 2 mM L-glutamine, 100 U/mL penicillin, and 0.1 mg/mL streptomycin (Sigma Aldrich, Oakville, ON, Canada) to achieve 10 mL cultures in 25 cm2 vented flasks. The cultures were mitogen stimulated by the addition of 100 μL phytohemagglutinin (PHA; Invitrogen, Waltham, MA, USA). After stimulation with 1% PHA, blood cultures were incubated (37°C, 5% CO2) for 44 h before the addition of cytochalasin B (final concentration: 4 μg/mL; Sigma, St. Louis, MO, USA). After an additional 28 h incubation period, cell suspensions were transferred to 15 mL centrifuge tubes and the cells were centrifuged for 8 min at 200 g, at RT. The supernatant was removed and the cell pellets were then resuspended in 10 mL of RT 75 mm KCl and incubated for 5 min at RT; 2 mL of fixative (5:1 methanol:glacial acetic acid) was then added to the hypotonic cell suspension and gently mixed. The samples were allowed to stand for a further 10 min and were then centrifuged for 8 min at 200 g, at RT. The supernatant was removed, and the cell pellets were resuspended in 10 mL of fixative and allowed to stand for 10 min. This fixation step was repeated twice more to remove cellular debris. Finally, the fixed cell suspension was incubated for 24 h at 4°C. To improve cell membrane fixation, 250 μL of 37% (v/v) formaldehyde (Fisher Scientific, Ottawa, ON, Canada) was added and samples centrifuged for 8 min at 200 g, at RT, and resuspended in a small volume (∼200 μL) of fixative to achieve the desired cell concentration for slide preparation. 15 μL of cell suspensions was dropped onto tilted pre-cleaned Fisherfinest glass slides (Fisher Scientific) maintained in a pre-warmed/humidified Hanabi Metaphase Spreader (Adstec Inc., Japan), then dried overnight on a 37°C slide warmer.

Scoring

Slides were stained with 50 μg/mL final concentration acridine orange (Sigma) on the day of analysis, visualized at ×400, and scored at ×600 magnification under fluorescence (BX51, Olympus, Center Valley, PA, USA) and Mni were scored according to criteria similar to that of Fenech et al. [Fenech et al., 2003]. Mni were quantified in 500 BNCs by each the of two scorers for a total of 1,000 BNCs for each sample. For samples with very few BNCs, either as many BNCs per slide as possible were scored or scoring was stopped after a minimum of 200 BNCs. Cells containing neoplasmic bridges were scored as if they were regular BNCs.

CU CBMN Assay Protocol

Unless otherwise noted, all reagents and plasticware at CU were purchased through Fisher Scientific Inc. (Waltham, MA, USA). The processing at CU was begun 24 h after irradiation, concurrent with the start of processing at HC.

For each sample, 0.5 mL of irradiated blood was diluted in 1.5 mL of pre-warmed (37°C) medium (RPMI 1640 + 10% FBS + 1% Pen/Strep; Invitrogen, Carlsbad, CA, USA) and carefully layered over 2 mL of Histopaque-1077 (Invitrogen) in a 15 mL conical centrifuge tube. The conical tubes were centrifuged for 45 min at 300 g at RT. Following centrifugation, the buffy layer was transferred to a well in a 6-well plate, preloaded with 5 mL of pre-warmed (37°C) medium containing phytohemagglutinin (2% PHA; Invitrogen). Following lymphocyte incubation (44 h at 37°C, 5% CO2, 95% humidity), 500 μL of medium containing cytochalasin B (6 μg/mL in RPMI 1640 + 10% FBS + 1% Pen/Strep + 2% PHA) was added to the culture.

Following an additional 28 h of incubation (37°C, 5% CO2, 95% humidity), the contents of each well were transferred to a 15 mL conical centrifuge tube. Each well was further rinsed with 3 mL PBS, which was added to the tube. The tubes were centrifuged for 10 min at 200 g to pellet the lymphocytes. The supernatant was removed and discarded leaving 1.5 mL. The pellet was dispersed by bubbling using a transfer pipette. 1 mL of RT KCl (0.075 m, Sigma Aldrich, St. Louis, MO, USA) was added to each tube. After further mixing by air bubbling, KCl solution was added to bring the total volume to 8 mL. Air bubbling was used as it is strong enough to disaggregate cell clumps but also gentle enough not to cause any structural damage. At 10 min, 1 mL of cold fixative (3:1 methanol:acetic acid, pre-chilled on ice prior to use) was added to block KCL action; then, the tubes were mixed by gentle inversion. Samples were then centrifuged (10 min, 200 g), supernatant (KCL solution) was aspirated, and 1-2 mL of cold fixative was added to the cell suspension. Pellet was disaggregated by air bubbling and cold fixative was added to bring the total volume to 12 mL. The tubes were kept at RT for 10 min and then centrifuged again, the supernatant discarded, and 5 mL of fixative was added. The fixed cells were then stored in an explosion-proof fridge overnight, then moved to a −20°C explosion-proof freezer for longer storage.

Prior to slide preparation, the lymphocytes were centrifuged and the supernatant discarded by inverting the tube for 3 s. The pellet was resuspended and 25 μL of lymphocyte suspension was dropped at the center of a microscope slide and spread by tilting the slides. The slides were allowed to air-dry for 10 min before staining through the application of 50 μL of DAPI Vectashield Mounting Medium (Vector Laboratories, Burlingame, CA, USA) and a cover slip. Slides were pressed for 10–20 min, sealed with clear nail polish, and left overnight in the fridge prior to imaging.

Scoring

Slides were imaged using a ×40 objective lens on a Zeiss epifluorescent microscope (Axioplan2 imaging MOT, Carl Zeiss, Germany) driven by the Metafer MNScore (MetaSystems, Althaussen, Germany). The Metafer MNScore is software that automatically scans slides prepared for the CBMN assay. A minimum of 500 BNCs were scored per sample except for high doses where only 200 BNCs were scored. Mni were scored manually from all BNCs in the Metafer-generated galleries.

Calibration Curves and Dose Predictions

At HC, a previously generated calibration curve, based on 6 donors (3 male and 3 female) with one donor per gender at approximately 20, 40, and 60 years of age, was used [McNamee et al., 2009] and is shown in Figure 1:
Mni=0.038D2+0.045D+0.019
(1)
Fig. 1.

Data from the first set of ten donors, used to generate the calibration curve for CU (dotted line). The HC calibration curve (solid line), shown for reference, was generated based on data up to 4 Gy. The region between 4 and 5 Gy, shown as dash-dotted line, is extrapolated.

Fig. 1.

Data from the first set of ten donors, used to generate the calibration curve for CU (dotted line). The HC calibration curve (solid line), shown for reference, was generated based on data up to 4 Gy. The region between 4 and 5 Gy, shown as dash-dotted line, is extrapolated.

Close modal
This calibration curve was generated using 137Cs γ-rays at a dose rate of 0.82 ± 0.04 Gy/min. At CU, calibration curves were generated specifically for this study using the initial set of 10 donors, not shared with HC. The calibration curve was generated using an unweighted least-squared fit of the Mni yields to a linear quadratic curve, using the trendline feature in MS Excel, which implements polynomial regression based on QR decomposition. The resultant quadratic is (dotted line in Fig. 1) given by the following:
Mni=0.0198D2+0.0973D+0.0034
(2)

For both laboratories, these equations were used to generate dose predictions from the Mni yields prior to unblinding. Figure 1 shows the data obtained for these first ten donors at CU along with both calibration curves.

Dose Prediction

Figure 2 and online supplementary Tables 1 and 2 show the raw Mni yields for all test samples analyzed at both sites (this is the second set of 10 donors, of which three were irradiated and analyzed on three separate days and treated as 6 extra donors) (for all online suppl. material, see https://doi.org/10.1159/000533488). Mni yields monotonically increased from 0 to 5 Gy; however at higher doses, the yield reached a plateau or even decreased for some donors. Due to this plateau, only the 0–5 Gy data have been included in the intercomparison analysis. A discussion on possible methodology to address the high-dose range appears below.

Fig. 2.

Raw micronucleus yields for all samples analyzed in the intercomparison (10 donors, of which three were analyzed three times on separate days) (data in online supplementary Tables 1 and 2).

Fig. 2.

Raw micronucleus yields for all samples analyzed in the intercomparison (10 donors, of which three were analyzed three times on separate days) (data in online supplementary Tables 1 and 2).

Close modal

Figure 3 shows plots of the predicted dose for all donors, as a function of the physical dose. The average dose over all donors for each laboratory is tabulated in Table 1 (reconstructed doses for individual samples appear in online suppl. Tables 3 and 4). Figure 4 shows a comparison of the data scored at both sites. Although there is a large variation in Mni yields between the two sites, the dose predictions for each donor are better correlated.

Fig. 3.

Comparison of CU (a) and HC (b) dose prediction to physical dosimetry (based on the data in Fig. 2). The physical dose is indicated by the color/shape of the data point. The dotted line represents a predicted dose identical to the physical dose (see also online supplementary Tables 3 and 4).

Fig. 3.

Comparison of CU (a) and HC (b) dose prediction to physical dosimetry (based on the data in Fig. 2). The physical dose is indicated by the color/shape of the data point. The dotted line represents a predicted dose identical to the physical dose (see also online supplementary Tables 3 and 4).

Close modal
Table 1.

Average predicted dose for both sites

Physical doseDose prediction
HCCU
0 Gy 0.01±0.03 0.01±0.04 
1 Gy 0.7±0.2 0.8±0.2 
2 Gy 1.6±0. 3 1.7±0.3 
3 Gy 2.4±0.3 2.6±0.3 
4 Gy 3.1±0.5 3.6±0.6 
5 Gy 3.6±0.5 4.5±0.7 
Physical doseDose prediction
HCCU
0 Gy 0.01±0.03 0.01±0.04 
1 Gy 0.7±0.2 0.8±0.2 
2 Gy 1.6±0. 3 1.7±0.3 
3 Gy 2.4±0.3 2.6±0.3 
4 Gy 3.1±0.5 3.6±0.6 
5 Gy 3.6±0.5 4.5±0.7 
Fig. 4.

Comparison of HC and CU Mni yields (a) and predicted dose (b). Each data point corresponds to a single donor (or repeat) and dose, plotting the value obtained at Columbia versus the value obtained at Health Canada. The physical dose is indicated by the color/shape of the data point. The dotted line represents identical results obtained at the two laboratories.

Fig. 4.

Comparison of HC and CU Mni yields (a) and predicted dose (b). Each data point corresponds to a single donor (or repeat) and dose, plotting the value obtained at Columbia versus the value obtained at Health Canada. The physical dose is indicated by the color/shape of the data point. The dotted line represents identical results obtained at the two laboratories.

Close modal

This intercomparison was established to compare the dose estimate generated by two laboratories running slightly different assays, one based on isolated lymphocytes (CU) and the other based on whole blood (HC). This could influence the number of Mni scored due to differing sensitivity of isolated lymphocytes as compared to lymphocytes cultured in whole blood. Additionally, the CU protocol used twice as much PHA (2% vs. 1%), which can also affect yields. Although this effect was not seen in the 0–5 Gy dose range, it could be responsible for the fact that at higher doses the micronucleus yields from whole blood culturing plateau whereas the results of the lymphocyte assay decline.

At HC, cell staining was performed using acridine orange stain and cells were scored manually, from direct observation of the cells. This method allows a clearer visualization of the cell boundaries, as compared to DAPI-only imaging, which relies on proximity to identify binucleate cells, as done at CU, which could lead to misidentifying two closely located mononuclear cells as a single BNC. CU minimized this issue by ensuring the cell density on the slides was low. It was seen by us and others that fully automated scoring using the Metafer platform has a high false-positive rate and a lower rate of detection of CBMN at higher doses [Thierens et al., 2014]. For triage, this is less of an issue for high doses but can affect the lower limit of detection or falsely identify unexposed individuals as exposed. The recent RENEB (Running the European Network of Biological and retrospective physical Dosimetry) exercise showed that semi-automated scoring, where only the Mni-positive BNCs were confirmed, performs better in a blinded interlaboratory comparison [Vral et al., 2023]. CU used a slightly modified semi-automated scoring, with all BNCs automatically identified and imaged and the Mni manually scored in the captured digitized BNC images. It has been demonstrated that manual scoring with microscopy generally allows for more accurate identification of damage than in captured images [Wilkins and Beaton, 2023]. This can be due to the ability to adjust the plane of focus and view the slides at higher resolution.

RENEB intercomparisons [Endesfelder et al., 2023; Vral et al., 2023] have shown that micronucleus yields are systematically higher in manual scoring than when scoring is partially or fully automated. Despite these differences, in general, there was a fairly good agreement between the results of the assays implemented at the two sites, particularly at the lower doses (Fig. 4).

It has been suggested that the semi-automated scoring rejects cells with nucleoplasmic bridges, which are more likely to contain Mni [Cheong et al., 2013], and would therefore underestimate the overall micronucleus yield. In our study, visual inspection of the slides did not reveal a large yield of cells with nucleoplasmic bridges, such that this would likely be a small effect. In any case, since the calibration curve was established using the same scoring protocol (i.e., with the same underestimation), it would inherently correct for this.

Table 2 shows a stratification of the predicted dose into 1 Gy bins. At the low doses (0, 1, 2 Gy), both laboratories correctly classified the majority of the samples (within ±0.5 Gy). For doses above 2 Gy, the CU assay correctly classified most of the donors while the HC assay classified the donors in the next bin down, that is, slightly underestimated the actual dose.

Table 2.

Classification of donors to 1 Gy bins

Physical doseHCCU
Dose category0 Gy, %1 Gy, %2 Gy, %3 Gy, %4 Gy, %5 Gy, %0 Gy, %1 Gy, %2 Gy, %3 Gy, %4 Gy, %5 Gy, %
“0 Gy” (<0.5 Gy) 100 13 100 
“1 Gy” (0.5–1.5 Gy) 88 44 100 19 
“2 Gy” (1.5–2.5 Gy) 56 75 13 81 44 
“3 Gy” (2.5–3.5 Gy) 25 63 38 56 44 19 
“4 Gy” (3.5–4.5 Gy) 25 63 50 13 
“5 Gy” (4.5–5.5 Gy) 0 69 
Physical doseHCCU
Dose category0 Gy, %1 Gy, %2 Gy, %3 Gy, %4 Gy, %5 Gy, %0 Gy, %1 Gy, %2 Gy, %3 Gy, %4 Gy, %5 Gy, %
“0 Gy” (<0.5 Gy) 100 13 100 
“1 Gy” (0.5–1.5 Gy) 88 44 100 19 
“2 Gy” (1.5–2.5 Gy) 56 75 13 81 44 
“3 Gy” (2.5–3.5 Gy) 25 63 38 56 44 19 
“4 Gy” (3.5–4.5 Gy) 25 63 50 13 
“5 Gy” (4.5–5.5 Gy) 0 69 

The bold bins correspond to correct classification.

The CU assay resulted in a 14% average deviation from the real dose with 76% of the dose predictions being within 0.5 Gy of the real dose and 94% within 1 Gy. The HC assay resulted in a 24% average deviation from the real dose with 50% of the dose predictions being within 0.5 Gy of the real dose and 76% within 1 Gy. At HC, however, the calibration curve only covered the range of 0–4 Gy so that the curve had to be extrapolated to make the higher dose estimations. If only the test doses between 0 and 4 Gy are considered, 60% of the dose predictions were within 0.5 Gy of the real dose and 86% within 1 Gy.

Impact of Calibration Curve

As seen in numerous intercomparison studies [Wilkins et al., 2008; Willems et al., 2010; Romm et al., 2013], the interlaboratory agreement on predicted dose, using a laboratory-specific dose calibration curve (Fig. 4), is much better than the agreement in actual Mni yields (Fig. 3). This is to be expected as various systematic differences in the sample processing are accounted for in the calibration curve. Furthermore, the impact of using the appropriate calibration curve is highlighted here with the comparison of the dose estimates made at HC using an older calibration curve and those made using matched samples from this intercomparison to create a calibration curve. The dose estimates made with the calibration curve generated from the intercomparison samples were more accurate. That being said, in a realistic scenario requiring biodosimetry analysis, the laboratory can only use their pre-existing calibration curve and will not have the luxury of producing a calibration curve exactly matched to the exposure scenario and age and sex of the donors. This exercise was only completed in order to provide a more meaningful comparison between the manual and automated scoring procedures. It is important here that the dose estimates made by HC using their pre-existing calibration curve were accurate 86% of the time to within 1 Gy when analysis was limited to the range of the calibration curve and was able to clearly identify highly exposed samples from lower exposed samples.

HC used a pre-existing calibration curve based on 0–4 Gy data [McNamee et al., 2009]. Although this calibration curve matched the quality of the CU exposure system and used the same protocol for sample preparation, it was established 5 years prior to this intercomparison and was a compilation of data from six different scorers. Only two of these scorers contributed to the analysis of the Mni in this work. Conversely, the CU curve was generated immediately before the intercomparison with donors from the same age range and with the same scorer reviewing the image galleries and thus more closely matches the dataset against which it is being tested and we would expect a better dose estimate based on the “matched” calibration curve.

To test this hypothesis, a pair of calibration curves (HC and CU) based on a random selection of 4 out of the 16 blood collections (10 donors + 6 repeats) shared between the two sites were generated. Using these calibration curves, the doses for the remaining 12 samples were predicted. Figure 5a shows a typical set of results. This was repeated for all possible 164=1,820 combinations and the log of the probability of obtaining each pair of doses is plotted as a contour map in Figure 5b. Compared to Figure 4, there is a better correlation between the dose estimates derived at each laboratory as well as a better agreement between the HC data and the true dose (not shown).

Fig. 5.

a Typical correlation plot between the HC and CU predicted doses. For this graph, we randomly used donors 2, 4, 10, and 13 to generate the calibration curve and plotted the predicted dose (CU vs. HC) for the 12 other donors (or repeats). b Cumulative plot of (a) summed over all 1,820 possible combinations. The color corresponds to a log scale of the number of hits in each 0.1 Gy × 0.1 Gy bin (e.g., “HC predicted between 3 and 3.1 Gy and CU predicted between 2.5 and 2.6 Gy” 100 times). The solid, dashed, and dotted lines represent an exact match, match within ±0.5 Gy, and within ±1 Gy, respectively.

Fig. 5.

a Typical correlation plot between the HC and CU predicted doses. For this graph, we randomly used donors 2, 4, 10, and 13 to generate the calibration curve and plotted the predicted dose (CU vs. HC) for the 12 other donors (or repeats). b Cumulative plot of (a) summed over all 1,820 possible combinations. The color corresponds to a log scale of the number of hits in each 0.1 Gy × 0.1 Gy bin (e.g., “HC predicted between 3 and 3.1 Gy and CU predicted between 2.5 and 2.6 Gy” 100 times). The solid, dashed, and dotted lines represent an exact match, match within ±0.5 Gy, and within ±1 Gy, respectively.

Close modal

Inter- and Intra-Donor Variation

To examine the variation of the CBMN assay, three out of the ten donors shared between HC and CU were repeated an additional two times each with a gap of 2–4 weeks between consecutive draws. Online supplementary Figure 1 shows the results of these repeats with each panel corresponding to three repetitions of the same donor. Online supplementary Figure 2 shows the resulting variation in dose predictions. As a proxy for the inter-individual variability, the standard deviation of all 10 donors (without the repeat samplings) is also plotted (dashed line). Within each dose, the three repetitions, for each site (solid lines), varied by ±10–30%. This is comparable to the root mean-squared difference between the HC and CU dose predictions (dotted line). Both inter- and intra-donor variabilities tended to be similar up to about 4 Gy. At higher doses, the CU inter-donor variations increase, likely driven by (donor-dependent) reduced mitotic index, resulting in lower binucleate and Mni yields and hence larger error bars in the dose prediction. The HC inter-donor variations stay about the same throughout the dose range. Despite the increase in the variation in the CU data, the root mean-squared difference between HC and CU, averaged over all donors, remains similar to the low-dose inter- and intra-donor variability up to 7.5 Gy.

Higher Doses

The data analysis for this intercomparison was limited to the lower doses (0–5 Gy) due to the plateau and/or decrease in Mni frequency observed over 5 Gy where the Mni yields consistently depart from a linear-quadratic dose dependence. This departure is not surprising as it is well known that high doses of radiation can cause delays in the cell cycle, blocking cells either in G1 phase impairing highly damaged cells from entering S phase or in G2 phase and impairing cells with more damage from entering mitosis [Bernhard et al., 1995]. As only BNCs are scored, this cell cycle delay reduces the proportion of highly damaged cells analyzed, therefore lowering the yield of Mni. This decrease in Mni yield at high doses was more prominent with the lymphocyte cultures. This may be due to the difference between culturing in whole blood and lymphocyte cultures indicating the existence of pro-proliferation factors in whole blood that are not present in the medium used. Alternatively, it might be an artifact of the automated location of the BNCs. This large decrease in Mni yield after high doses was also observed by others using automated Mni analysis even when the cells were cultured as whole blood [Rodrigues et al., 2014]. It could be due to the ability of the human eye to better identify highly damaged BNCs or a higher efficiency of counting Mni directly through the microscope as opposed to using the computer screen-generated gallery.

The decrease/plateau of Mni at high doses was studied in detail by Muller and Rode (2002), Beinke et al. (2016) who have shown that it is possible to modify the micronucleus assay analysis to cover a dose range of 0–15 Gy. As the cell cycle slows down with increasing dose, the Mni/BNC drops such that the calibration curve becomes a downward parabola. This means that for each value of Mni/BNC, there could be two possible doses estimated. Adding additional parameters related to cell proliferation to the analysis, for example, the ratio of mononucleate cells to binucleate cells, could extend the dose of applicability. This can be done in several ways:

  • 1.

    Using a composite calibration curve, for example, using the ratio of mononucleate cells to binucleate cells at high doses (>5 Gy) and the standard Mni yields for the lower doses (<5 Gy)

  • 2.

    Using multiparameter analysis such that one set of parameters can classify the response as a high or low dose and then the yield of Mni can be used to estimate the dose, knowing the range in which the dose should fall. This type of approach has also been suggested with data from CBMN analyzed by imaging cytometry [Rodrigues et al., 2014] and has the potential to increase the dose range of applicability of the CBMN assay to up to 10 Gy. Extending beyond 10 Gy is limited by the plateau of proliferation index curve as very few cells would cycle through to mitosis

  • 3.

    As an alternative, biological approach, Pujol-Canadell and colleagues targeted the abrogation of G2/M and spindle checkpoints through the addition of caffeine and an Aurora kinase inhibitor, respectively, at specific times to the CBMN assay cultures and improved lymphocyte cell proliferation and the detection of Mni yields after high-dose photon (∼8 Gy) and neutron irradiations (∼4 Gy) [Pujol-Canadell et al., 2020].

Repeatability and Inter-Individual Variation

When the repeatability and inter-individual variation was examined by resampling the same donors three times, it was seen (online suppl. Fig. 2) that at low doses, the variation between donors (dashed lines) could be largely attributed to the day-to day assay performance variability (solid lines). This highlights the benefits of a fully automated assay where this variability is minimized. At higher doses, however, the variation could not be explained solely by day-to -day variability. The variation is likely stemming from large inter-individual variation in cell proliferation. Interestingly, the whole blood assay performed at HC had a lower variation than the isolated-lymphocyte-based assay, again indicating the existence of pro-proliferation factors in whole blood might be different to the isolated lymphocytes (where these factors would be absent) that are not present in the medium used. Additionally, the lower variation with the whole blood method may also be due to higher efficiency of detecting damage using the fully manual scoring directly through the microscope, as discussed above.

A laboratory intercomparison study was completed for the CBMN assay based on isolated lymphocyte cultures that can be developed for an automated microculture assay and those obtained using the standard CBMN assay in a functional biodosimetry laboratory using in-house calibration curves. Although the Mni yields obtained varied significantly between the two assays, the predicted doses matched rather well. These results highlight the importance of a robust calibration curve(s) across a wide age and sex range that matches the exposure scenario as closely as possible and that will account for differences in methodology between laboratories. The data also indicate that at low doses variability in the results may be attributed to variation in the processing while at higher doses the variation is dominated by inter-individual variation in cell proliferation. It has been shown in other studies [Garty et al., 2023; Royba et al., 2023] that the details of the irradiation (e.g., high dose rate vs. protracted and to some extent the type of radiation) also has implications on dose prediction, so the use of an appropriate laboratory-specific and exposure-specific calibration curve is very important.

We thank Lilian Oling and Annerys Rodriguez for the recruitment of the blood donor volunteers to this study and the nurses at the Department of Radiation Oncology, at Columbia University Medical Center, for assisting with the blood draws. We thank also Dr. John Ng for the randomization of the blood sample aliquots and Catherine Ferarrotto for assisting with scoring at Health Canada.

This study protocol was reviewed and approved by Columbia University’s Institutional Review Board, under protocol AAAE9551. Informed consent was obtained from all study participants.

The authors have no conflicts of interest to declare.

This project has been supported in whole or in part with federal funds from the Department of Health and Human Services; Administration for Strategic Preparedness and Response; and Biomedical Advanced Research and Development Authority (BARDA), under Contract No. HHSO100201000002C, and by the National Institutes of Health and National Institute of Allergy and Infectious Diseases (NIAID) (Grant No. U19-AI067773).

A.B. and H.C.T. conducted the experiments at Columbia University. R.C.W. and S.L. conducted the experiments at Health Canada. G.G. and D.J.B. designed the experiments and analyzed the data. All authors participated in manuscript preparation.

1.
Avlasevich SL, Bryce SM, Cairns SE, Dertinger SD. In vitro micronucleus scoring by flow cytometry: differential staining of micronuclei versus apoptotic and necrotic chromatin enhances assay reliability. Environ Mol Mutagen. 2006;47(1):56–66.
2.
Beinke C, Port M, Riecke A, Ruf CG, Abend M. Adaption of the cytokinesis-block micronucleus cytome assay for improved triage biodosimetry. Radiat Res. 2016;185(5):461–72.
3.
Bernhard EJ, Maity A, Muschel RJ, McKenna WG. Effects of ionizing radiation on cell cycle progression. A review. Radiat Environ Biophys. 1995;34(2):79–83.
4.
Cheong HS, Seth I, Joiner MC, Tucker JD. Relationships among micronuclei, nucleoplasmic bridges and nuclear buds within individual cells in the cytokinesis-block micronucleus assay. Mutagenesis. 2013;28(4):433–40.
5.
Endesfelder D, Kulka U, Bucher M, Giesen U, Garty G, Beinke C, et al. International comparison exercise for biological dosimetry after exposures with neutrons performed at two irradiation facilities as part of the BALANCE project. Cytogenet Genome Res. 2023. In press.
6.
Fenech M. Cytokinesis-block micronucleus cytome assay. Nat Protoc. 2007;2(5):1084–104.
7.
Fenech M, Chang WP, Kirsch-Volders M, Holland N, Bonassi S, Zeiger E, et al. HUMN project: detailed description of the scoring criteria for the cytokinesis-block micronucleus assay using isolated human lymphocyte cultures. Mutat Res. 2003;534(1–2):65–75.
8.
Fenech M, Morley AA. Measurement of micronuclei in lymphocytes. Mutat Res. 1985;147(1–2):29–36.
9.
Garty G, Chen Y, Salerno A, Turner H, Zhang J, Lyulko O, et al. The RABIT: a rapid automated biodosimetry tool for radiological triage. Health Phys. 2010;98(2):209–17.
10.
Garty G, Chen Y, Turner H, Zhang J, Lyulko OV, Bertucci A, et al. The RABiT: a rapid automated biodosimetry tool for radiological triage. II. Technological developments. Int J Radiat Biol. 2011;87(8):776–90.
11.
Garty G, Royba E, Repin M, Shuryak I, Deoli N, et al. Sex and dose rate effects in automated cytogenetics radiat prot dosimetry. 2023. In press.
12.
Garty G, Turner HC, Salerno A, Bertucci A, Zhang J, Chen Y, et al. The decade of the RABiT (2005–15). Radiat Prot Dosimetry. 2016;172(1–3):201–6.
13.
Homeland Security Council. National planning scenarios. Washington (DC); 2006. (final version 21.3).
14.
International Atomic Energy Agency. Cytogenetic analysis for radiation dose assessment: a manual. Vienna: IAEA; 2001. p. 127.
15.
Lee JK, Han E-A, Lee S-S, Ha W-H, Barquinero JF, Lee HR, et al. Cytogenetic biodosimetry for Fukushima travelers after the nuclear power plant accident: no evidence of enhanced yield of dicentrics. J Radiat Res. 2012;53(6):876–81.
16.
McNamee JP, Flegal FN, Greene HB, Marro L, Wilkins RC. Validation of the cytokinesis-block micronucleus (CBMN) assay for use as a triage biological dosimetry tool. Radiat Prot Dosimetry. 2009;135(4):232–42.
17.
Miller SM, Ferrarotto CL, Vlahovich S, Wilkins RC, Boreham DR, Dolling JA. Canadian Cytogenetic Emergency Network (CEN) for biological dosimetry following radiological/nuclear accidents. Int J Rad Biol. 2007;83(7):471–7.
18.
Muller WU, Rode A. The micronucleus assay in human lymphocytes after high radiation doses (5–15 Gy). Mutat Res. 2002;502(1–2):47–51.
19.
Pujol-Canadell M, Perrier JR, Cunha L, Shuryak I, Harken A, Garty G, et al. Cytogenetically-based biodosimetry after high doses of radiation. PLoS One. 2020;15(4):e0228350.
20.
Repin M, Pampou S, Karan C, Brenner DJ, Garty G. RABiT-II: implementation of a high-throughput micronucleus biodosimetry assay on commercial biotech robotic systems. Radiat Res. 2017;187(4):492–8.
21.
Repin M, Turner HC, Garty G, Brenner DJ. Next generation platforms for high-throughput biodosimetry. Radiat Prot Dosimetry. 2014;159(1–4):105–10.
22.
Rodrigues MA, Beaton-Green LA, Kutzner BC, Wilkins RC. Multi-parameter dose estimations in radiation biodosimetry using the automated cytokinesis-block micronucleus assay with imaging flow cytometry. Cytometry A. 2014;85(10):883–93.
23.
Rodrigues MA, Beaton-Green LA, Wilkins RC. Validation of the cytokinesis-block micronucleus assay using imaging flow cytometry for high throughput radiation biodosimetry. Health Phys. 2016;110(1):29–36.
24.
Romm H, Barnard S, Boulay-Greene H, De Amicis A, De Sanctis S, Franco M, et al. Laboratory intercomparison of the cytokinesis-block micronucleus assay. Radiat Res. 2013;180(2):120–8.
25.
Royba E, Repin M, Balajee AS, Shuryak I, Pampou S, Karan C, et al. Validation of a high-throughput dicentric chromosome assay using complex radiation exposures. Radiat Res. 2023;199(1):1–16.
26.
Satyamitra M, Reyes Turcu FE, Pantoja-Galicia N, Wathen L. Challenges and strategies in the development of radiation biodosimetry tests for patient management. Radiat Res. 2021;196(5):455–67.
27.
Schunck C, Johannes T, Varga D, Lörch T, Plesch A. New developments in automated cytogenetic imaging: unattended scoring of dicentric chromosomes, micronuclei, single cell gel electrophoresis, and fluorescence signals. Cytogenet Genome Res. 2004;104(1–4):383–9.
28.
Shuryak I, Royba E, Repin M, Turner HC, Garty G, Deoli N, et al. A machine learning method for improving the accuracy of radiation biodosimetry by combining data from the dicentric chromosomes and micronucleus assays. Sci Rep. 2022;12(1):21077.
29.
Sproull M, Camphausen K. State-of-the-art advances in radiation biodosimetry for mass casualty events involving radiation exposure. Radiat Res. 2016;186(5):423–35.
30.
Sproull MT, Camphausen KA, Koblentz GD. Biodosimetry: a future tool for medical management of radiological emergencies. Health Secur. 2017;15(6):599–610.
31.
Thierens H, Vral A, Vandevoorde C, Vandersickel V, de Gelder V, Romm H, et al. Is a semi-automated approach indicated in the application of the automated micronucleus assay for triage purposes?Radiat Prot Dosimetry. 2014;159(1–4):87–94.
32.
Vral A, Endesfelder D, Balázs J, Beinke C, Cuceu Petrenci C, Finot F, et al. RENEB inter-laboratory comparison 2021: the cytokinesis-block micronucleus assay. Radiat Res. 2023;199(6):571–82.
33.
Vral A, Fenech M, Thierens H. The micronucleus assay as a biological dosimeter of in vivo ionising radiation exposure. Mutagenesis. 2011;26(1):11–7.
34.
Wilkins RC, Beaton-Green LA, Lachapelle S, Kutzner BC, Ferrarotto C, Chauhan V, et al. Evaluation of the annual Canadian biodosimetry network intercomparisons. Int J Radiat Biol. 2015;91(5):443–51.
35.
Wilkins RC, Beaton LA. Development of high-throughput systems for biodosimetry. Radiat Prot Dosimetry. 2023. In press.
36.
Wilkins RC, Romm H, Kao T-C, Awa AA, Yoshida MA, Livingston GK, et al. Interlaboratory comparison of the dicentric chromosome assay for radiation biodosimetry in mass casualty events. Radiat Res. 2008;169(5):551–60.
37.
Willems P, August L, Slabbert J, Romm H, Oestreicher U, Thierens H, et al. Automated micronucleus (MN) scoring for population triage in case of large scale radiation events. Int J Rad Biol. 2010;86(1):2–11.