An automated platform for cytogenetic biodosimetry, the “Rapid Automated Biodosimetry Tool II (RABiT-II),” adapts the dicentric chromosome assay (DCA) for high-throughput mass-screening of the population after a large-scale radiological event. To validate this test, the U.S. Federal Drug Administration (FDA) recommends demonstrating that the high-throughput biodosimetric assay in question correctly reports the dose in an in vivo model. Here we describe the use of rhesus macaques (Macaca mulatta) to augment human studies and validate the accuracy of the high-throughput version of the DCA. To perform analysis, we developed the 17/22-mer peptide nucleic acid (PNA) probes that bind to the rhesus macaque's centromeres. To our knowledge, these are the first custom PNA probes with high specificity that can be used for chromosome analysis in M. mulatta. The accuracy of fully-automated chromosome analysis was improved by optimizing a low-temperature telomere PNA FISH staining in multiwell plates and adding the telomere detection feature to our custom chromosome detection software, FluorQuantDic V4. The dicentric frequencies estimated from in vitro irradiated rhesus macaque samples were compared to human blood samples of individuals subjected to the same ex vivo irradiation conditions. The results of the RABiT-II DCA analysis suggest that, in the lymphocyte system, the dose responses to gamma radiation in the rhesus macaques were similar to those in humans, with small but statistically significant differences between these two model systems.
The dicentric chromosome assay (DCA) is an established method of measuring radiation exposure and is considered by the International Organization for Standardization (ISO), International Atomic Energy Agency (IAEA) and U.S. Food and Drug Administration (FDA) to be the gold standard for biological biodosimetry (1–4). However, this assay is not very practical for mass screening, as it has a low throughput and is restricted to a certain time window for application (5). Recently, we achieved a high-throughput version of the DCA by miniaturization and parallelization of all procedures performed in multiwell plates by an automated robotic device, the Rapid Automated Biodosimetry Tool II (RABiT-II) (6). The next logical step, recommended by the FDA for the development of the RABiT-II DCA, is to ensure that this high-throughput assay correctly reports the dose for both in vitro and in vivo irradiated blood samples.
In vivo human studies are primarily limited to radiotherapy patients who have received fractionated doses, which often fail to cover doses of interest for dosimetry and do not necessarily mimic relevant exposure situations. Moreover, the blood samples from these patients may provide uncertainties in the dicentric rates because the lymphocytes are likely to get damaged due to the disease, or chemotherapy, in addition to the radiation treatment plans (7). When human validation studies are not possible, the FDA guidelines recommend using an appropriate animal model for assessment of the biodosimetry device performance (8). Successful extrapolation of the data from experimental animals to humans can be used to bridge the results of the assay from clinical doses to more realistic exposure scenarios (9).
Most information on aberration frequency in vitro versus in vitro irradiated blood has been obtained from animal studies (10–13). However, not all results derived using experimental animals are well applicable to humans. In some species, the dicentric yields were found to be related to the effective chromosome arm number in a way that a human male with an effective arm number of 81 has twice as many dicentrics as a mouse with the arm number 40 (14). In contrast, in the rabbit model, the dicentric yields were less than half of those in humans, even though the arm numbers were almost comparable (13). Moreover, not all mammalian species are appropriate for testing an automated version of the DCA, as their chromosomes, with different morphology from those of humans, will be outside of the detection range of the custom software. This makes certain animal studies of limited utility for further validation of the RABiT-II DCA.
Rhesus macaques (Macaca mulatta), the most widely used non-human primate (NHP) species, have been recommended as a human substitute for various research applications (15), including practical biodosimetry (16). The manual DCA works just as well on the NHP model as on humans and has already demonstrated agreement with humans in the frequency of both stable (translocations) (12) and unstable (dicentric) (13) aberrations after irradiation. Concerning chromosomal size and morphology, in both species, the majority of chromosomes are bi-armed with almost equal numbers of effective chromosomal arms (84/ 83 vs. 82/81 for female/male rhesus macaques vs. humans, respectively) (14, 17). Remarkable genetic similarity (18, 19) and similar cell cycle kinetics of cultured primate lymphocytes to that in man (13) minimizes the impact of confounding factors [e.g., chromatin state and DNA repair pathway choice (20, 21)] that may influence radiosensitivity and the yield of dicentrics, reinforcing the suitability of this model for further validation of the high-throughput DCA.
The peptide nucleic acid (PNA) fluorescent in situ hybridization technique has radically improved chromosome aberration scoring (22, 23). Based on its ability to cover the alpha satellite sequences flanking the centromeres, PNA probes are an excellent tool for the automated discrimination between normal (one centromere) and aberrant (multiple centromere) chromosomes. However, as was shown in other primate species (24), the commercially available PNA probes, suitable for detection of human centromeres, do not hybridize to the rhesus macaque's centromeres. Therefore, prior to analysis, it was necessary to generate custom centromere probes for the detection of radiation damage in the rhesus macaques. We constructed PNA probes that allowed us to perform the DCA on the rhesus macaques using the automated RABIT-II system. To improve quality of automated chromosome analysis, we optimized our non-classical (37°C) PNA FISH staining (6) using both centromere and telomere probes, and updated our custom software to score entities with excess telomeres as chromosome clusters as opposed to true multicentric chromosomes. These improvements enabled automated comparison of the radiation responses of rhesus macaque blood samples to that of human blood samples subjected to the same ex vivo irradiation conditions, to establish equivalence between two species.
MATERIALS AND METHODS
Unless otherwise noted, all reagents and plasticware were purchased from Thermo Fisher Scientific™ Inc. (Waltham, MA).
Sequence Alignments and Probe Design
The PNA probes used in this study are listed in Table 1. The rhesus macaque centromere probes were designed using a 343-bp-long fragment of M. mulatta highly repeated DNA (PubMed nucleotide database, gene bank accession X04006.1, p Rh2 27). All sequence alignments were performed using A Plasmid Editor (ApE) software. Properties of the candidate oligomers (purine nucleotide composition, binding specificity and cross dimers) were analyzed using the PNABio tool ( https://www.pnabio.com/support/PNA_Tool.htm) and Oli2go oligonucleotide design tool ( http://oli2go.ait.ac.at/). Refined oligonucleotides were synthesized by Panagene (Daejeon, South Korea). The centromere PNA probes for human chromosomes and telomere PNA probes were ordered from a licensed Panagene distributor (PNA Bio Inc., Newbury Park, CA). All probes were received in lyophilized form and dissolved in formamide to a final concentration of 10 µM (50 µg/ml). The aliquots were stored at –20°C in the dark.
The Nucleotide Sequences of the PNA Probes Used in this Study
Animal blood samples (from 5 healthy males and 5 healthy females) were purchased from AlphaGenesis®, Inc. (Yemassee, SC), shipped at 20 ± 2°C to Columbia University (New York, NY), and processed immediately on arrival (within 24 h of blood draw). Human blood samples (from 5 healthy male and 5 female donors) were collected under Columbia University Institutional Review Board Protocol no. AAAR0643. After informed consent, the blood was drawn by venipuncture into BD Vacutainer™ plastic blood collection tubes with lithium heparin anticoagulant and processed immediately.
The blood aliquots were pipetted into 2D-barcoded tubes (Matrix Storage Tubes) and transported to a Gammacell®-40 137cesium (137Cs) irradiator (Atomic Energy of Canada Ltd., Chalk River, Canada). The tubes were γ-ray irradiated (137Cs) at 0, 2, 4, 6, 8 and 10 Gy at a dose rate of 0.73 Gy/min (measured at the height of 2 mm in the chamber). Annual calibration of the Gammacell-40 is performed by the chief medical physicist at Columbia University Irving Medical Center's Department of Radiation Oncology, using thermoluminescent dosimeters (TLDs).
RABiT-II Automated Dicentric Assay
After irradiation, human blood samples were processed automatically using the RABiT-II DCA, as described elsewhere (6). Several modifications to this protocol were applied for culturing the rhesus macaque whole blood samples. For lymphocyte activation and expansion, RPMI-1640 medium (Mediatech Inc., Herndon, VA), supplemented with 20% heat-inactivated fetal bovine serum (FBS; HyClone™, Logan, UT) and 25 µg/ml of the mitogen concanavalin A (Sigma-Aldrich® LLC, St. Louis, MO) was used. After 44 h in culture, 0.1 µg/ml colcemid (Sigma-Aldrich) was dispensed into the multiwell plates. After a further 4 h incubation, cells were harvested, swollen in a hypotonic solution (0.075 M KCl), then fixed using a 3:1 methanol:acetic acid fixative. Chromosomes were released from mitotic nuclei and stained in multiwell plates using PNA FISH. Before staining, samples were blocked with 200 µl of 1% bovine serum albumin (BSA; Sigma-Aldrich) in PBS for 30 min at room temperature. Then, BSA was replaced with 200 µl of hybridization cocktail consisting of 90% formamide, 2× saline sodium citrate buffer [0.33 M sodium chloride in a 0.03 M sodium citrate buffer (pH 7.0)], 0.25 µg/ml of centromere PNA probe, and 0.05 µg/ml of telomere TelG probe. Hybridization was performed for 3 h at 37°C. After staining, 200 µl of PBS containing 1.5 µg/ml 4′,6-diamidino-2-phenylindole (DAPI) was dispensed in each well for counterstaining. After 10 min, 400 µl of staining mixture (PNAs and DAPI) was aspirated, the plates were then washed two times with 200 µl of PBS and imaged.
Image Acquisition and Analysis
Image acquisition was performed using the RABiT-II plate imager, a BioTek® Cytation 1 Cell Imaging Multi-Mode Reader (Winooski, VT), with low magnification (20× objective). From each well, a total of 450 image sets (size 380 × 280 µm per image, 1,152 × 832 pixels, 16 bit) from DAPI (blue), FITC (green), and RFP (red) channels were obtained and stored in an image gallery. Upon completion of the acquisition, the images were automatically analyzed for the presence of normal and aberrant chromosomes using our home-written software, FluorQuantDic V4 (6). Individual chromosomes were identified in the DAPI channel and only those falling within acceptable morphology were scored for bright foci in the corresponding region of the GFP (centromere) and RFP (telomere) images. Chromosomes with telomere spots more than 5 pixels from the edge of the chromosome were rejected. The software classified validated chromosomes as acentric, monocentric, dicentric, or multicentric chromosomes based on the number of detected centromere spots.
Dose-Response Curves and Statistical Analysis
Dose-response calibration curves were constructed using DoseEstimate software version 5.2 (25). The coefficients of the fitted curves were derived for collectively pooled individuals (2 replicates per animal/donor). For statistical analysis, we performed a customized fit to the dicentric fraction data using a linear-quadratic (LQ) dose dependence, with binomially-distributed errors. This was done using Maple 2018 software (MapleSoft, Waterloo, Canada) by the maximum likelihood of pooled data for each species separately. Additionally, to examine variability in radiation responses between macaques and humans, we performed mixed-effects modeling using logistic regression in the lme4 R package. Equation (1), below, was used for the dicentric probability (P = number of dicentrics/number of scored chromosomes), where b is the background parameter, α and β are standard LQ dose-response parameters and D is the dose:
The 1 – exp structure was used to prevent P from becoming >1 at high doses, thus allowing it to be treated as a probability rather than a yield. At relatively small P values (<0.12) this adjustment is not very important numerically but facilitates the analysis. Comparisons between actual and reconstructed doses were performed by solving Eq. (1) for D, using P values from the measured human data as the input. This produced two sets of reconstructions: using macaque parameters and using human parameters. A real solution for D becomes complex and cannot be calculated if P < b, so in those instances D was set to zero. A paired t test was used to compare dose reconstructions based on human dicentrics data, but using best-fit LQ model parameters for either macaques or humans.
Optimization of the RABiT-II DCA Protocol for Rhesus Macaques
Compared to human lymphocytes, the PB-MAX karyotyping media containing phytohemagglutinin (PHA) as a mitogen had minimal effectiveness to trigger the proliferation of the rhesus macaque's lymphocytes (26). To increase the yields of rhesus chromosomes scored using RABiT-II DCA, we tested several culture media and mitogen combinations (Fig. 1). The effect of mitogens was assessed by manual scoring of mitotic indexes throughout the observation of images captured on a BioTek plate imager. The highest mitotic index was registered in RPMI-1640 media containing 20% FBS supplemented with 25 to 50 µg/ml concanavalin A (Fig. 1). Thus, the rhesus experiments in this study were performed with a 25 µg/ml concanavalin A.
Design of the Centromere PNA Probe for Use in Rhesus Macaque DCA and Optimization of Centromere-Telomere PNA FISH Staining at 37°C
Preliminary staining tests revealed that the commercial human centromere PNA probe does not hybridize to rhesus macaque centromeres. To our knowledge, alternative PNA probes for staining of M. mulatta centromeres are not available and were not previously reported by other researchers. To generate centromere PNA probes for the dicentric analysis of rhesus macaque lymphocytes, we analyzed the nucleotide sequence of the 343-base-pair highly repeated fragment of the rhesus macaque DNA and selected 15 candidate regions with a length not longer than 22-bp each. Subsequently, the chemical properties of these oligonucleotides were analyzed to select a total of 3 candidate PNA probes (Fig. 2 and Table 2). These probes were ordered from a commercial distributor. The efficiency of the probe hybridization to the rhesus macaque centromeres and the stability of the duplexes was analyzed at 75°C (classical PNA FISH) and 37°C (the RABiT-II DCA) hybridization. We found that at 37°C, the pMm1 PNA probe did not produce any fluorescent signal, whereas the pMm2 and the pMm3 PNA probes have sufficiently stained the rhesus centromeres. In this work, for staining the rhesus macaque centromeres, we used the 17-mer pMm2 PNA probe (Fig. 2).
Selected Properties of the Designed Rhesus Macaque Centromere Probes
For optimization of telomere staining without heat denaturation (37°C), we tested both the C-rich and the G-rich telomere probes. We observed a strong dependence of telomere probe hybridization performance on the concentration of formamide in the hybridization mixture (Fig. 3). Specifically, the TelC probe did not hybridize at 90% formamide, which can be due to probe detachment during washing (27). In contrast, the TelG probe demonstrated excellent hybridization at 37°C in 80% to 90% formamide for both species (humans and rhesus macaques) (Fig. 4). The TelG probe was chosen for further experiments with both species.
Comparison of Rhesus Macaque and Human Responses to Gamma Radiation
After culturing and staining were optimized for both species, the number of chromosomes and yields of dicentrics in ex vivo irradiated human and animal blood samples were automatically scored using the FluorQuantDic V4 software. The raw results of scoring for macaques and humans are given in Supplementary Tables S1 and S2 ( https://doi.org/10.1667/RR15547.1.S1), respectively. The average fraction of dicentrics is shown in Table 3. To characterize statistical variation attributable to both model systems, we used LQ and logistic regressions, including an assessment of inter-donor variability in the dose-response parameters.
Dicentric Chromosome Frequencies in Macaques and Human Samples Scored by the FluorQuantDic V4
The customized LQ model (Eq. 1) fits are shown in Fig. 5. These data summarize the fraction of the dicentrics in each sample obtained from 10 animals and 10 human individuals (two technical repeats for each dose). Macaque best-fit parameters were: b= 8.71 × 10–3, α = 0 Gy–1 and β = 7.62 × 10–4 Gy–2. Human best-fit parameters were: b = 1.13 × 10–2, α = 1.46 × 10–3 Gy–1 and β = 5.36 × 10–4 Gy–2. Table 4 shows the uncertainties for each model parameter, calculated using profile likelihood. The 0 Gy data from animal 4 has an abnormally high probability of dicentrics, probably due to the low number of chromosomes (a total of 405) scored for this dose point (Supplementary Table S1; https://doi.org/10.1667/RR15547.1.S1). However, exclusion of this animal from the analysis had only small effects on model parameters (Table 4), probably because the “weight” of the outlier data at 0 Gy for animal 4 was very small due to the small number of scored chromosomes.
The Fit Parameters for the LQ Probability Model with Errors 95% Confidence Intervals (CIs) for Each Model Parameter Calculated Using Profile Likelihood
Graphs of dose reconstructions of human doses based on the human calibration curve (R2 = 0.8) and the macaque calibration curve (R2 = 0.8) are shown in Fig. 6. A paired t test, however, showed a significant difference between the dose reconstructions: those using human parameters were, on average, lower than those using macaque parameters (P value 1.85 × 10–6, 95% CI: –0.59 to –0.26 Gy). Additional information on the inter-donor variability analysis performed without pooling the data is shown in the Supplementary Text, Figs. S1 and S2, and Tables S3 and S4 ( https://doi.org/10.1667/RR15547.1.S1)
The RABiT-II is the fully-automated biodosimetry platform for high-throughput emergency screening of a large number of individuals exposed to an unknown level of ionizing radiation/s (28). Using this robotic system, we recently implemented the RABiT-II DCA (6) to score aberrant chromosomes in human blood automatically. The results of the RABiT-II DCA determine the dose absorbed by the whole body to detect significant radiation exposures during triage assessment. While no difference between in vitro and in vivo measurements of chromosomal aberrations is expected (10–13), the FDA requires this to be demonstrated for each specific assay prior to licensure. In vivo studies in humans using radiotherapy patients are a less suitable model for validation of the RABiT-II DCA since some treatment plans can elevate chromosome breakage, and the dicentric chromosomes can arise as a direct result of the previous therapies (7). Thus, the animal study using the rhesus macaque model has been performed instead.
Taking into account various factors operating during automated sample preparation, all essential experimental procedures and cytogenetic techniques that had been previously used for human samples were well applicable to the rhesus macaques. None of the technical issues, reported earlier, including difficulties with obtaining sufficient yields of mitotic cells or unsatisfactory quality of mitotic spreads, such as overlapping chromosomes in metaphase cells (26, 29), were observed in our established workflow. The quality of mitotic spreads is not an issue in the RABiT-II DCA since it extracts the chromosomes from metaphase cells and performs the analysis on isolated chromosomes; high mitotic yields were achieved in all experiments using concanavalin A. The culture conditions and fixation time of the blood cultures obtained from both species were in good agreement and were suitable for inter-species comparison of the dicentric yields.
The molecular target for binding of the commercial centromere PNA probe is the 17-mer centromere protein (CENP) B box sequence on the alpha satellite DNA. In humans, the alpha satellite DNA consists of tandem repeats of 171-bp-long monomer units (30). Our sequence alignment has revealed no identical 17-mer CENP-B box sequence on the rhesus macaque's alphoid DNA. The rhesus macaque satellite DNA was longer (343-bp long dimer), comprised of the two monomer units (172-bp and 171-bp) that are quite divergent from each other (31). Indeed, the consensus sequence of the rhesus macaque satellite is more than 98% homologous to baboons but is only partially homologous (less than 70%) to the most diverged alphoid sequence of humans (31). Therefore, to realize dicentric analysis for the rhesus macaques using PNA probes, we developed custom M. mulatta centromere probes. To our knowledge, these probes are the first reported probes that allow for performing dicentric analysis in rhesus macaques using PNA FISH staining.
The developed PNA probes enabled a fully-automated comparison of the dicentric chromosome rate of rhesus macaque to that of human blood samples that were subjected to the same irradiation conditions, mimicking acute whole-body exposure to gamma rays. It is noteworthy that dicentric yields in the lymphocytes of the rhesus macaques examined in this study were similar to that of humans (Fig. 5). These findings are in agreement with previously reported studies that used manual DCA (13) or other techniques for comparison of both species. Specifically, rhesus macaques and humans have shown similar changes in expression of the radiation response genes (32) and kinetics of H2AX phosphorylation in the sites of nascent DNA double-strand breaks (33).
More rigorous statistical analysis for the RABiT-II DCA has demonstrated that there are small differences in best-fit parameters between macaque and human dicentric yields, although inter-donor variability ranges were overlapping. Regression fits were used to reconstruct doses. Doses reconstructed using macaque parameters applied to human data were not dramatically different from those reconstructed using human parameters, although the difference between means did reach statistical significance. The dose reconstruction accuracies tended to be better at higher doses than at lower doses (Fig. 6) because of the curved shape of the dose response (Fig. 5): at lower doses, there are smaller differences with baseline values, magnifying the errors in dose reconstruction. In these circumstances, several factors determining these small statistical differences on the yields of dicentrics can be considered, but the actual cause needs to be clarified. We can assume that an automated DCA requires more samples to be analyzed to obtain more precise statistical data. Alternatively, the chromosomal radiosensitivity can be essentially the same for animal and human cells (13) but can be modified by the efficacy of detection of DNA damage at lower doses or by the DNA repair mechanisms, which can differ between these two species (14). This remains to be elucidated by an appropriate statistical study in the future.
Supplementary Text. Assessment of Differences in Radiation Dose Response Between Species and Between Individuals Within Each Species by Logistic Regression. Effects on Dose Reconstruction.
Table S1. RABiT-II DCA yields after in vitro γ-ray irradiation of rhesus macaque blood samples derived from 10 healthy animals.
Table S2. The RABiT-II DCA yields after in vitro γ-ray irradiation of human blood samples derived from 10 healthy donors.
Table S3. Dose-response parameters and intercepts across animals 1–10.
Table S4. Dose-response parameters and intercepts across donors 1–10.
Fig. S1. Logistic regression fits for (panel A) rhesus macaque and (panel B) human data. Best fits are represented by the curves; samples by the symbols.
Fig. S2. Comparisons of actual vs. reconstructed doses. Panel A: Human parameters applied to human data. Panel B: Rhesus macaque parameters applied to human data. Dashed lines show a theoretically perfect 1:1 correlation.
We thank Maria Taveras for blood collection. This work was supported by the National Institute of Allergy and Infectious Diseases, National Institutes of Health, (NIAID/NIH contract no. HHSN272201600040C).