Multiparameter imaging flow cytometry-based cytokinesis-block micronucleus assay: Reduction of culture time and blood volume for improved efficiency

Abstract

In the event of a large-scale incident involving radiological or nuclear exposures, there is a potential for large numbers of individuals to have received doses of radiation sufficient to cause adverse health effects. It is imperative to quickly identify these individuals in order to provide information to the medical community to assist in making decisions about their treatment. The cytokinesis-block micronucleus assay is a well-established method for performing biodosimetry. This assay has previously been adapted to imaging flow cytometry and has been validated as a high-throughput option for providing dose estimates in the range of 0 to 10 Gy. The goal of this study was to test the ability to further optimize the assay by reducing the time of culture to 48 h from 68 h as well as reducing the volume of blood required for the analysis to 200 μL from 2 mL. These modifications would provide efficiencies in time and ease of processing impacting the ability to manage large numbers of samples and provide dose estimates in a timely manner. Results demonstrated that either the blood volume or the culture time could be reduced while maintaining dose estimates with sufficient accuracy for triage analysis. Reducing both the blood volume and culture time, however, resulted in poor dose estimates. In conclusion, depending on the needs of the scenario, either culture time or the blood volume could be reduced to improve the efficiency of analysis for mass casualty scenarios.

1. Introduction

In the event of a large-scale incident involving radiological or nuclear exposures, there is a potential for large numbers of individuals to have received doses of radiation sufficient to cause adverse health effects. It is imperative to quickly identify these individuals in order to provide information to the medical community to assist in making decisions about their treatment. Biological dosimetry is a tool that can be used post-exposure to determine the amount of radiation received by an individual based on biological indicators. The cytokinesis-block micronucleus (CBMN) assay is a well-established, standardized method in which the number of micronuclei (MN) in binucleated cells (BNC) indicates the amount of DNA damage induced from radiation [1–4] and has been validated as a biological dosimeter for in vivo ionizing radiation exposures [5]. The MN result from either whole or fragmented chromosomes that, rather than being incorporated into one of the two daughter nuclei after division, form distinct MN in the cytoplasm of the cell and can persist for up to one year following their formation[6]. These MN can be used to detect genetic damage from any type of genotoxin and has been applied as a biological dosimeter for ionizing radiation.

The CBMN assay has traditionally been done using manual microscopy, a method that can be time-consuming, requires slides to be made and also has a risk of scorer bias. Efforts to automate this assay include slide-based automated image analysis, laser scanning microscopy, automated sample processing, and flow cytometry and have been reviewed in Wilkins et al. [7]. These efforts have also assisted in the development of high-throughput CBMN analysis for mass casualty events.

There has also been much work conducted on developing the CBMN assay for the imaging flow cytometer (IFC) for high-throughput analysis of radiation exposure for mass casualty events [8–10]. Through this work, the methodology and data analysis has evolved to a point where it is now possible to measure doses up to 10 Gy (2 mL blood samples, 68 h culture incubation) through the use of multiparametric analysis [9]. The importance of extending the dose range of applicability is that individuals receiving doses between 5 and 10 Gy may be candidates for alternative therapies to assist them through the haematological and gastrointestinal effects of acute radiation syndrome. Determining the dose as accurately as possible would provide important information for medical interventions[11,12].

While these advances have increased the throughput of and the dose range of applicability of the CBMN assay, two major areas to further improve the efficiency are decreasing the initial blood volume and decreasing the sample culture incubation time. The traditional microscope-based assay requires 1 mL of blood and 72 h of culture time. It has been demonstrated that the assay can be reduced to 68 h [8,13], and preliminary studies further reduced this culture time in this study to 48 h [14,15]. The use of small-volume blood samples in combination with the automated CBMN assay has also been recently explored [15], and blood volumes as low as 30 μL have demonstrated success, maintaining a similar level of accuracy as larger volume cultures [16]. The focus of this work has been to use the multiparametric approach to examine the effects of changing the culture conditions (blood volume and incubation time) on the dose estimates in the range of 0–10 Gy using the optimal conditions of 2 mL culture and 68 h incubation as a benchmark on the IFC

2. Materials & methods

2.1. Blood draw and irradiations

Whole blood was collected from each volunteer donor (between the ages of 20–60 years) with informed consent and approval by Health Canada’s Research Ethics Board (protocol REB 2002–0012H). All donors selected were non-smokers in relatively good health at the time of donation. Donors had no known illnesses nor previous exposures to medical ionizing radiation within the last 12 months. Blood was drawn by venipuncture in 10 mL lithium-heparinized Vacutainer^® tubes (BD Vacutainer^™, Mississauga, ON, Canada) and kept at 37.0°C ±0.5°C. Blood tubes were kept warm, aliquoted and transported to an XRAD 320 X-ray cabinet biological irradiator (Precision X-Ray, N. Branford, CT). The blood sample tubes were placed in a water-equivalent phantom providing 4.5 cm of build-up material, 50 cm from the source and exposed to X-rays at a dose rate of 0.36 Gy/min. Irradiations were performed at 250 kVp and 15 mA with a filter composed of 0.75 mm Sn, 0.25 mm Cu and 1.5 mm Al (approximately 3.7 mm Cu half-value layer). All doses were measured using a PTW TW30010–10 ion chamber and a PTW UNIDOS T10002 electrometer (PTW Dosimetry, Lörracher Strasse 7, 79115 Freiburg) with N_k = 48.3 mGy nC⁻¹ at 250 kV, (calibrated by the National Research Council, Ottawa, Canada) assuming air kerma to be equal to dose. For DNA repair, all samples were incubated 2 hours at 37.0°C ±0.5°C as per IAEA requirements [1]. Two sets of exposures between 0 and 10 Gy were conducted: 1) for the generation of the dose response calibration models, samples from each of the 7 donors (4 male/3 female) were exposed at intervals of 1 Gy and 2) for validation of the method (“test data”), samples for 6 donors (4 male/2 female) were exposed to 1, 2, 4, 6, and 8 Gy and blinded.

2.2. Culture and Fixation

From each sample, four conditions were established, one for each culture time and blood volume combination (Condition 1: 68 h, 2 mL; Condition 2: 48 h, 2 mL; Condition 3: 68 h, 200 μL; Condition 4: 48 h, 200 μL). For Condition 1 and 2, two aliquots of 2 mL whole blood were transferred to T25 flasks (Thermofisher Scientific, Thermo Scientific^™ Nunc^™ EasYFlask^™) containing 8 mL complete media (86.5% RPMI 1640 media (Gibco), 10% heat-inactivated Fetal Bovine Serum (FBS) (Fisher Scientific-Cytiva-HyClone), 2.5% Phytohemagglutinin (PHA) (Gibco) to induce blastogenesis of lymphocyte white cells, and 1% L-Glutamine/ Penicillin-Streptomycin (Sigma-Aldrich)), and gently mixed. Similarly, for Condition 3 and 4, two aliquots of 200 μL whole blood were transferred to 1 mL Matrix tubes (Thermofisher Scientific, Thermo Scientific^™ Matrix^™) containing 800 μL complete media and gently mixed. For each culture condition, one control sample (CTRL) was made from a mixture of blood from each dose point to be used during setup of the IFC laser powers. The samples were incubated at 37.0°C ±0.5°C and 5% ±1% CO2, for 24 h. After 24 h, cytochalasin-B (Cyto-B; Sigma-Aldrich) was added to cultures at a final concentration of 6.0 μg/mL for the purpose of blocking cytokinesis of proliferating lymphocytes. Samples Condition 1 and 3 (68 h) were incubated for an additional 44 h before the fixation process, while Condition 2 and 4 (48 h) were incubated for an additional 24 h [8].

The fixation process has been previously described [9]. Briefly, samples were centrifuged, supernatant removed and pellet was resuspended in a hypotonic 75 mM KCl solution (Sigma-Aldrich) followed by a hypotonic soft fixation of cells, then centrifuged and resuspended in 1X FACS Lysing solution (Becton, Dickinson and Company (BD); Franklin Lakes, NJ) to lyse the red blood cells and fix the lymphocytes. Cell suspensions were washed one time with 1X FACS Lysing solution to complete the lymphocyte fixation then washed three times with phosphate buffer saline solution (PBS, pH 7.4; Teknova; Hollister, CA) before being resuspended in PBS to a total volume of 100 μL per cell suspension. Fixed samples were stored at 4°C for up to 2 weeks until being run on an IFC (ImageStream^®X Mark II (ISXMkII), Cytek Biosciences, Seattle, WA).

2.3. Staining and Acquisition

Two DRAQ5 (eBioscience, San Diego, CA) working solutions were prepared for the two different volume conditions. For Conditions 1 and 2 (2 mL), a working solution of 315 μM DRAQ5-in-PBS was prepared and for Conditions 3 and 4 (200 μL), a working solution of 45 μM DRAQ5-in-PBS was prepared. Immediately before analysis using the IFC, working solutions were added to the samples to obtain a final concentration of 35 μM for conditions 1 and 2 and 5 μM for Conditions 3 and 4 of fixed cell suspension and incubated in the dark at room temperature for 20 minutes.

Samples were introduced into the ISXMkII using a 96-well plate with approximately 45 μL in each well. All samples were acquired at 40× magnification with the 642-nm laser set between 15 and 70 mW. For each culture condition, the CTRL sample was loaded manually, and the laser was adjusted as necessary so that the fluorescence intensity of the desired population was between 1,000 and 3,000 raw max pixels in channel 5. The Brightfield (BF) LED was on and the side scatter (785 nm) laser was set to 2.5 mW. Images were collected on a single camera system with BF images collected in channel 1 and DRAQ5-stained images collected in channel 5. Data were collected using the ISX INSPIRE^® software (version 200.1.388.0) 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. Typically, 50,000 events were collected for the small-volume samples and 100,000 events collected for the large-volume samples. This yielded up to 800 and 10,000 BNCs collected in the small- and large-volume samples respectively. The number of BNCs was dependent on dose and tended to drop off at the higher doses.

2.4. Data processing

All data were processed using the IDEAS (version 6.2) software; BNC, mononucleated (MONO), trinucleated (TRI), quadranucleated (QUAD) and non-apoptotic (NAP) cells were identified along with BNCs containing 0, 1, 2, 3 or 4 MN (0MN, 1MN, 2MN, 3MN, 4MN) using the gating strategies outlined in Rodrigues et al [17] and Rodrigues [18].

2.5. Statistical analysis

To compare the calibration models generated with the different culture conditions, four different models were developed: Condition 1: 68 h, 2 mL; Condition 2: 48 h, 2 mL; Condition 3: 68 h, 200 μL; Condition 4: 48 h, 200 μL. The first objective was to test if the models differed with respect to culture time, holding blood volume constant (i.e., comparing Condition 1 vs Condition 2, and Condition 3 vs Condition 4). The second objective was to test if the models differed with respect to blood volume, holding culture time constant (i.e., compare Condition 1 vs Condition 3, and Condition 2 vs Condition 4). Data from these studies were correlated such that each person’s blood was exposed to each dose in the range of 0 to 10 Gy. As previously observed [9], the shape of the relationship followed a downward parabola for doses ranging from 0 to 10 Gy. A random coefficients model with Poisson errors was fit to the data. Simultaneous regression models were then used to compare if the fixed effects portion of the random coefficients models of the two models being compared were similar.

To compare the dose estimates made with each culture condition, dose estimates were generated using blind-irradiated samples and compared to dose-response calibration models derived under each of the four culture conditions. The multiparameter methodology used to establish the calibration models and generate the dose estimates has been previously described in detail [9] and is briefly described in the supplemental material. Prediction of the true dose under the different calibration models (i.e., one for each condition) was summarized using the following statistics: 1. the mean of the estimated doses at each true dose; 2. the percentage of estimated doses within 30% of the true dose; 3. relative bias (RBIAS); and 4. relative mean square error (RMSE). The RBIAS at each dose level was calculated as

$$RBIAS=\left(mean\left(d_e\right)-d_{True}\right)/d_{True}$$ (1)

except in the case of the control group where ${\mathrm{d}}_{True}=0$, the denominator was set to 1. In this case the RBIAS was simply a difference between the mean of the estimated doses and the true dose. The RMSE at each dose level was calculated as follows:

$$RMSE=MSE/d_{True}^2=\sum _{i=1}^s {\left(d_{e,i}-d_{True}\right)}^2/\left(s\mathrm{*}{d}_{True}^2\right)$$ (2)

Where s was the number of replicates in the dose level. Again, when the true dose was equal to 0 Gy the denominator was set to 1 and so in this case the RMSE was simply a mean square error.

All analysis was carried out in R version 4.1.0 (R Core Development Team).

3. Results

Calibration models were generated for each of the 4 culture time and blood volume conditions described. For each data set, the random coefficients model with Poisson errors was fit to the data (see supplemental Figure S1 for full data representation). Figure 1 presents the fixed portion of the random coefficients model for each of the conditions. Comparing calibration models for Condition 1 and Condition 2 that had the blood volume fixed at 2 mL and the culture time varying, the two models are statistically different (p<0.01) with respect to all parameter estimates. Similarly, when comparing calibration models for Condition 3 and Condition 4 with blood volume fixed at 200 μL and culture time varying, the two models are statistically different (p<0.01) with respect to all parameter estimates. In both cases, the higher response was measured after 68 h compared to 48 h. Similarly, when the blood volume was varied and the culture time was fixed, there was a statistically significant difference between the models of each blood volume, with the 2 mL volume having a higher MN/BNC response. Overall, the highest response was for conventional conditions of 68 h culture and 2 mL of blood volume (Condition 1), however, except for the small volume and short time culture (Condition 4), an increase in MN/BNC was detected for doses below 5 Gy.

Figure 1:

Random coefficients model with Poisson errors fit to the CBMN data under different culture time and blood volume conditions. The lines represent the fixed portion of the random coefficients model.

Using these calibration models, the conditions were compared on their ability to predict the true dose applied to blood samples. The first step was to determine whether the dose was above or below the dose where the calibration model reached its maximum value ${(D}_{max})$ using the logistic discriminant function (LDF) based on the data used to derive each calibration model. A sensitivity analysis was conducted to determine the probability with the greatest proportion of data being correctly classified. This probability was set to 60%. That is, if the calculated probability based on the observed proportions BNC/NAP, Mono/NAP, Tri/NAP and QUAD/NAP from the LDF was less than or equal to 60% then the observation was classified as having a dose less than or equal to the $D_{max}$. Conversely, if the probability was greater than 60% then the observation was classified as having a dose greater than $D_{max}$. Results of this classification are shown in table 1 where it can be seen that Condition 2 had the best success (100%) while Conditions 1 and 3 had more than 80% correct classification and Condition 4 only had 73.3% correct classification.

Table 1.

Culture conditions and classifications

Condition	Culture Time (h)	Blood volume	Dmax (Gy)	% Correctly classified
Condition 1	68	2 mL	6.85	83.3
Condition 2	48	2 mL	4.75	100
Condition 3	68	200 μL	5.89	85.7
Condition 4	48	200 μL	5.92	73.3

The next step was to use the classification from the LDF and the observed rate of TotalMN/BNC in the appropriate calibration model to estimate the dose. Figures 2 to 4 summarize how well the calibration models performed. In Figure 2 the percentage of samples for each dose in each condition that agreed within 30% of the actual dose is presented. From Figure 2 it was noted that the benchmark calibration model (Condition 1: culture time 68 h and blood volume 2 mL) reported the highest proportion of observations within 30% of the true dose (27 out of 30 samples) correctly estimating all the doses except at 6 Gy where 3 samples were overestimated. Calibration model Condition 2 also had a high number of dose estimates within 30% of the true dose (25/29 samples), however the 4 misidentified samples were in the low dose range and were still within 1 Gy of the true dose. For Condition 3 this value was 70%, again with 4 of the misidentified samples in the low dose range and still falling within 1 Gy of the true dose. One of these misidentified was due to and initial misclassification to the low dose region. Finally, dose estimates made using Condition 4 had the smallest proportion of dose estimates within 30% of the true dose (9 of 30 samples), performing poorly at all doses.

Figure 2:

Probability that the estimated dose is within 30% of the true dose.

Figure 4:

A) Relative BIAS (RBIAS) measured at each true dose group by calibration model based on the different conditions. B) Relative mean square error (RMSE) measured at each true dose group by calibration model based on the different conditions.

Figure 3 presents the mean of the estimated dose determined from the observed TotalMN/BNC versus the true dose using the calibration models based for each condition. Individual data points can be seen in supplemental figure S3. For all conditions except Condition 4, the mean dose estimates were close to or within 1 Gy of the true dose, with the estimates being mostly within 0.5 Gy of the true doses at 4 Gy and below. Also evident is that for the smaller volumes where fewer BNCs were collected, there was much more variability in the dose estimates as demonstrated by the larger error bars.

Figure 3:

Estimated dose from each observed TotalMN/BNC versus the true dose using the calibration models based on Panel A Condition 1: 68 h, 2 mL; Panel B Condition 2: 468 h, 2 mL; Panel C Condition 3: 68 h, 200 μL; and Panel D Condition 4: 48 h, 200 μL. In each panel, the estimated dose within ½ Gy of the true dose (light gray band) and within 1Gy of the true dose (dark gray band) can be observed. Error bars represent the standard error on the mean of the dose estimate.

Finally, Figures 4 a and b present the RBIAS and RMSE results as an alternative demonstration of the performance of each condition. The absolute value of RBIAS ranged between 0 and 1.8 with most values valing below ±0.5. The highest RBIAS was observed with a calibration model based on Condition 4 (culture time 48 h, blood volume 200 μL) at all dose levels. The lowest RBIAS was observed with standard culture conditions (culture time 68 h, blood volume 2 mL). The RMSE was considered low at all dose groups for calibration models for Conditions 1, 2 and 3 and high for Condition 4.

The results differ considerably between the 4 calibration models. As compared to the standard conditions (culture time 68 h and blood volume 2 mL) Conditions 2 and 3 performed adequately, but the calibration model based on Condition 4 (culture time 48 h and blood volume 200 μL) performed poorly.

4. Discussion

The CBMN assay developed for the IFC has previously been shown to be a rapid method for estimating the dose of radiation received by individuals [10]. It has also been demonstrated that, by incorporating multiple parameters from the image analysis from the IFC, the upper dose limit for applicability of this assay has been extended to 10 Gy [9]. This assay has great potential for use in largescale emergencies where hundreds to thousands of individuals without physical dosimetry may be exposed to radiation. This study set out to test the limits of the assay for reducing the time of culture as well as reducing the volume of blood required for the analysis. Both of these modifications could impact the ability to manage large numbers of samples and provide dose estimates in a timely manner.

In previous studies developing the CBMN assay for the IFC, 2 mL of blood were used; however, it has been demonstrated that less volume is required to collect adequate cells for accurate dose estimation. The use of smaller volumes reduces the time needed for sample processing and the reagents required as fully automated sample processing becomes more feasible in multi-well plates [16]. This in turn reduces the storage costs and waste, increasing efficiencies for emergency preparedness. It can, however, increase the sample acquisition time on the IFC as it results in a lower density of cells per sample in order to collect sufficient cells to maintain the accuracy of the dose estimates. As samples can be run on the IFC using a 96-well plate loader, the sample collection time is of lower importance than the sample processing time. Although time of processing and analysis may not both be reduced, the shifting of hands-on time to fully automated procedures will be of great benefit during a mass casualty event.

Traditionally, the CBMN microscope method has required stimulating the lymphocytes to divide and culturing for 72 h to optimize the number of cells in cytokinesis. Decreasing the time of cell culture helps shorten the time from sample receipt to dose estimation for the individual. Other laboratories have explored shortening culture times and found that a 68 h culture time had minimal effect on the dose estimates but the culture time in this study has been further reduced to 48 h [6,10].

To address both decreasing the blood volume and reducing the culture time, three culture conditions were compared to the original benchmark IFC method of 2 mL of blood and 68 h of culturing. These included combinations of reduced culture time (48 h) and/or reduced blood volume (200 μL) conditions. For each condition, dose response models were generated up to 10 Gy and validated with blinded, irradiated samples in the same dose range. The method for estimating the dose has been described in detail previously [9] but briefly involved two steps: classifying the dose as either low or high based on proliferation-based parameters and then determining more precisely the dose based on the MN/BNC.

A limitation of the CBMN assay has always been to detect sufficient BNCs for a good estimation of dose. Radiation damage is known to slow down or even block the cell cycle such that the more highly damaged cells take longer to reach cytokinesis or are not even able to progress through mitosis, reducing the number of BNCs and reducing the MN/BNC as a function of dose. Using the benchmark IFC method (Condition 1) for comparison, decreasing the culture time to 48 h resulted in a decrease in the number of MN/BNC as expected. However, even at 48 h with 2 mL of blood (Condition 2), 100% of the samples were correctly classified as above or below the dose at which the maximum MN/BNC occurred, outperforming all other conditions. The dose predictions for this condition were almost as successful as the benchmark condition with only 1 fewer within 30% of the actual dose. Of those misidentified, all were still within 1 Gy of the true dose. The resulting lower dose response calibration model was still sufficient as blinded samples were processed under the same conditions and were affected similarly by the shortened culture time. t

When a smaller volume of blood was used with varying culture times, the more accurate dose estimates were made for the 200 μL samples with the longer culture time (68 h) as compared to the 200 μL samples at 48 h. When comparing the small volume samples to the large volume samples for each time point, the larger sample volumes outperformed the smaller volume samples. One reason for poorer results with smaller volumes was the number of BNCs collected in theses samples. Even with long acquisitions times, some of the samples had very low numbers of BNCs (some as low as 50 cells). Substantially fewer cells were analysed with the small volume samples, even for the 68 h time point. The BNC numbers also dropped with dose, exacerbating the lower accuracy of the dose estimates at the high doses.

Overall, Conditions 2 (48 h, 2 mL) and 3 (68 h, 200 μL) performed with sufficient accuracy for determining whether medical intervention is required for the exposed individuals. Only Condition 4 (48 h, 200 μL) with both the small volume and short culture time demonstrated a reduced accuracy.

These results demonstrate that reasonable dose estimates could be made under different culturing conditions. If a situation benefited from the collection of a smaller blood volume, it is recommended that the standard culture time (68 h) be used. An example would be in a situation where there were thousands of samples that would benefit from automated sample processing and analysis that could be performed using a system like the RABiT system [8]. In a situation with smaller numbers of samples, one might choose to draw larger volumes and then decrease the culture time to 48 h to enable the provision of valuable information to the medical community as quickly as possible. The results, however, clearly indicate that one should not decrease both time and culture volume. In order to take advantage of these modifications, it is important to ensure that the laboratory has created calibration models specific to the conditions. Even just reducing the volume of the blood used without shortening the culture time drastically changed the shape of the calibration model such that they cannot be used interchangeably.

5. Conclusion

The CBMN method for biodosimetry has previously been adapted to the IFC and has been demonstrated to be a high-throughput option for large scale radiation events over the dose range of 0–10 Gy. Further optimization of the assay can be accomplished through either the shortening of the culture time or with the reduction of the sample volume. Even in these sub-optimal conditions, as long as the processing methodology of the calibration model and the unknown samples are matched, adequate dose estimates can be made. This results in an flexible assay that can be modified according to the needs of the situation to improve its efficiency and through put when large numbers of samples need to be processed or there is an urgency in reporting dose estimates to the medical community.

HIGHLIGHTS.

• The cytokinesis-block micronucleus assay is a standardized assay for radiation biodosimetry

• This assay has been adapted to imaging flow cytometry

• Reducing the culture time or blood volume can improve the efficiency of the assay

• Reducing either the culture time or blood volume can maintain accurate dose estimates