Serum microRNAs are early indicators of survival after radiation-induced hematopoietic injury

See allHide authors and affiliations

Science Translational Medicine  13 May 2015:
Vol. 7, Issue 287, pp. 287ra69
DOI: 10.1126/scitranslmed.aaa6593


Accidental radiation exposure is a threat to human health that necessitates effective clinical planning and diagnosis. Minimally invasive biomarkers that can predict long-term radiation injury are urgently needed for optimal management after a radiation accident. We have identified serum microRNA (miRNA) signatures that indicate long-term impact of total body irradiation (TBI) in mice when measured within 24 hours of exposure. Impact of TBI on the hematopoietic system was systematically assessed to determine a correlation of residual hematopoietic stem cells (HSCs) with increasing doses of radiation. Serum miRNA signatures distinguished untreated mice from animals exposed to radiation and correlated with the impact of radiation on HSCs. Mice exposed to sublethal (6.5 Gy) and lethal (8 Gy) doses of radiation were indistinguishable for 3 to 4 weeks after exposure. A serum miRNA signature detectable 24 hours after radiation exposure consistently segregated these two cohorts. Furthermore, using either a radioprotective agent before, or radiation mitigation after, lethal radiation, we determined that the serum miRNA signature correlated with the impact of radiation on animal health rather than the radiation dose. Last, using humanized mice that had been engrafted with human CD34+ HSCs, we determined that the serum miRNA signature indicated radiation-induced injury to the human bone marrow cells. Our data suggest that serum miRNAs can serve as functional dosimeters of radiation, representing a potential breakthrough in early assessment of radiation-induced hematopoietic damage and timely use of medical countermeasures to mitigate the long-term impact of radiation.


Exposure to high doses of radiation in the event of industrial accidents, terrorist attacks, or use of nuclear weapons in military settings poses a significant threat to human life (13). Although substantial advancements in characterizing the effects of radiation on different organs and systems have been made, treatment options to exposed individuals are still dependent on a slow manifestation of symptoms (2). For example, delayed damage to the hematopoietic system at moderately high doses of radiation can take several weeks or months to appear, and existing biodosimetry techniques do not effectively predict the severity of injury sustained. In such situations, medical intervention is complicated by the difficulty in triaging individuals exposed to low, moderate, or high doses of radiation.

The radiation median lethal dose (LD50) for untreated humans is about 4 Gy (4). Low to moderate radiation exposures in humans lead to progressive development of acute radiation syndrome (ARS) consisting of dose-dependent hematopoietic, gastrointestinal, and cerebrovascular malignancies (2). The hematopoietic system is the most vulnerable tissue to the damaging effects of radiation (5). Exposure to low or moderate doses of radiation leads to a rapid decrease in blood cell counts, including loss of lymphocytes, neutrophils, and thrombocytes, and a severe decrease in hematopoietic progenitors. Radiation injury is also linked to increased risk of cancer and infection. Exposure to high doses of radiation causes severe, nonrecoverable bone marrow damage, resulting in pancytopenia owing to complete loss of hematopoietic stem cell (HSC) populations eventually leading to death. At 2- to 6-Gy exposure in humans, the hematopoietic component of ARS appears in a few weeks to 2 months. At higher doses of 8 to 12 Gy, lethal gastrointestinal as well as bone marrow toxicity are observed, and death is probable in 1 to 3 weeks (2, 3).

Existing technologies used to assess the extent of radiation have notable limitations. For example, one of the hallmarks of the lymphocyte depletion kinetics assay is the fact that it can be performed outside the laboratory, but a drawback is that several measurements are needed to get a dose estimate (6, 7). The ideal time frame for DNA damage assays using γ-H2AX is 0.5 to 2 hours after exposure (810), which may not be enough time for individuals to report to a medical countermeasures facility. The dicentrics chromosome assay is very specific to radiation and is considered the “gold standard” for determining the dose, but it has a long processing time, tedious scoring methods, and relatively narrow range for dose determination (9, 11). Thus, there is a critical need to develop radiation-specific indicators that are capable of predicting latent damage to various organs and systems immediately after radiation exposure.

MicroRNAs (miRNAs) have recently emerged as promising biomarkers for different pathological conditions. Deregulation of miRNAs has been implicated in the pathogenesis of various conditions ranging from cancer to autoimmune and cardiovascular disease (12). miRNAs are present in body fluids, such as serum and plasma (12, 13), and several studies have correlated levels of specific serum/plasma miRNA with various pathological conditions, including post-exposure to ionizing radiation (14, 15). The inherent stability of serum miRNAs under harsh conditions and reproducible levels in individuals of the same species make miRNAs attractive candidates for use as noninvasive biomarkers (16).

This study was designed to assess whether changes in serum miRNAs immediately after exposure may accurately predict the long-term impact of radiation-induced hematopoietic injury in murine models. We discovered unique miRNA signatures that effectively distinguished between control, sublethal (low dose), sublethal (high dose), and lethal total body irradiation (TBI) cohorts within 24 hours after radiation and correlated these signatures with the radiation-induced loss of HSCs. Twenty-four hours after exposure, this serum miRNA signature distinguished fatal versus nonfatal hematopoietic injury. Treatment with a radioprotective agent before lethal TBI or radiation mitigation by bone marrow transplant after lethal radiation exposure produced serum miRNA signatures that correlated with the functional impact of radiation. The human relevance of the serum miRNA signature was observed using a humanized mouse model. The serum miRNA signature indicated radiation-induced injury to human bone marrow cells in this model system, suggesting that miRNAs can be used to detect radiation-induced hematopoietic injury in affected human populations within hours of radiation exposure.


Dose-dependent hematopoietic injury occurs in mice after exposure to different doses of radiation

To test our hypothesis that serum miRNAs may be used to predict long-term impact of radiation-induced damage to the hematopoietic system, we systematically assessed radiation-induced hematopoietic injury by exposing C57BL/6J mice to different doses of radiation. Consistent with previous reports (14, 17), at 2- and 6.5-Gy TBI doses, all animals survived, and the 8-Gy dose was lethal for most (65%) of the mice (Fig. 1A). Therefore, 2 and 6.5 Gy were chosen as the sublethal low and sublethal high doses, respectively, and 8 Gy was considered the lethal dose for subsequent experiments.

Fig. 1. Characterization of hematopoietic injury in C57BL/6J mice after exposure to TBI.

(A) Establishment of sublethal and lethal dose in C57BL/6J mice. Kaplan-Meier survival curves of C57BL/6J male mice exposed to 0 (control), 2, 6.5, or 8 Gy of TBI (n = 20 per group). P value determined by log-rank (Mantel-Cox) test. (B to E) C57BL/6J mice were exposed to TBI at the indicated doses and allowed to recover for up to 3 months (B). At each time point, animals were sacrificed and bone marrow was analyzed for number of BM-MNCs (C), CFU-C content (D), and frequency of LKS HPCs (E) per hind limb. Data in (C) and (D) are means ± SEM (n = 5 per group; two independent experiments). Data in (E) are individual animals, and horizontal bars are means. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; n.s., not significant; one-way analysis of variance (ANOVA) followed by Tukey’s test.

Complete blood count (CBC) of irradiated animals displayed a reduction in white blood cells (WBCs) when compared with unirradiated controls at days 1, 7, and 15 after TBI. By day 30, the 2- and 6.5-Gy cohorts showed a near complete recovery in their WBC levels, whereas the 8-Gy animals did not survive (fig. S1). Red blood cell (RBC) and hemoglobin levels plummeted in the 6.5- and 8-Gy animals until day 15; however, by day 30, these levels returned to normal in the 6.5-Gy cohort (table S1). Platelet levels followed a similar trend, reaching a nadir at day 7 for 6.5-Gy and at day 15 for 8-Gy animals. By day 30, platelet levels in the 6.5-Gy cohort improved and were comparable to the 2-Gy cohort (fig. S1).

Both the 6.5- and 8-Gy irradiated animals displayed severe lymphopenia and anemia at day 7. Although there was no significant difference in WBC counts at day 15 (P = 0.65, Tukey’s test) between these two cohorts of animals, RBCs and platelets were significantly higher in the 6.5- than in the 8-Gy group (P = 0.002 for both RBC and platelets at day 15,Tukey’s test) (fig. S1 and table S1). Animals exposed to 8 Gy succumbed to bone marrow failure by day 30, but there was complete recovery in CBC levels in animals exposed to lower doses of irradiation (fig. S1).

Decrease in bone marrow cellularity is an important measure of hematopoietic injury. Therefore, we examined kinetics of bone marrow mononuclear cell (BM-MNC) depletion after exposure to different doses of TBI (2 to 8 Gy). We observed a dose-dependent reduction in the cellularity at 24 hours after radiation (Fig. 1C). By days 7 and 15, we observed complete recovery of BM-MNCs in the hind limbs of mice exposed to 2 Gy, whereas for mice exposed to 6.5 and 8 Gy, the BM-MNC count remained very low. Mice exposed to 6.5 Gy showed significant recovery of BM-MNCs by 1 month, and complete recovery by 3 months compared to the 15-day time point (P < 0.01, paired t test) (Fig. 1C). After a similar pattern of injury and recovery, by day 15, hematopoietic progenitor function after TBI had significantly decreased for all doses, and the 6.5- and 8-Gy cohorts were indistinguishable from each other, with very low colony-forming unit in culture (CFU-C) counts (Fig. 1D). At subsequent time points, bone marrow from both the 2- and 6.5-Gy groups displayed improved progenitor function, but mice in the 8-Gy group failed to recover.

To further evaluate the bone marrow hematopoietic progenitor population in control and irradiated animals, we quantified the LKS (lineage/c-kit+/Sca-1) population, which is enriched in hematopoietic progenitor cells (HPCs), and the LKS+ (lineage/c-kit+/Sca-1+) population, which is enriched in HSCs. Severe reduction in the HPC content was observed at 24 hours after TBI in all irradiated animals (Fig.1E). The kinetics of recovery for the HPC population (LKS cells) in the weeks and months after TBI were consistent with hematopoietic function (Fig. 1, D and E). The numbers of HPCs in the 6.5- and 8-Gy cohorts remained comparably low and indistinguishable from each other at day 15 after TBI.

Collectively, our data reveal that dose-dependent hematopoietic injury occurs after TBI, but animals exposed to sublethal high (6.5 Gy) and lethal (8 Gy) TBI doses remain largely indistinguishable up to 15 days after TBI. Animals exposed to sublethal high doses do show substantial recovery, unlike their 8-Gy counterparts.

Residual HSCs in sublethally irradiated mice retain the capacity to repopulate the bone marrow

Sublethal doses of TBI can cause permanent damage to the stem cell compartment, leading to stem cell senescence and decrease in the engraftment potential of HSCs (17, 18). Our hematopoietic analysis in Fig. 1 suggested that TBI at sublethal doses caused a severe reduction but not complete depletion of HPCs in the 2- and 6.5-Gy irradiated animals. We observed a similar trend in the HSC population (LKS+ cells) with striking ablation until 7 days in all TBI cohorts, and detectable recovery at the 15-day time point in the 2-Gy irradiated animals (Fig. 2A). By comparison, HSC content in 6.5- and 8-Gy irradiated animals 15 days after TBI was significantly lower.

Fig. 2. Stem cell transplantation from irradiated mice reveals a defect in short- and long-term repopulating ability.

(A) Total number of HSCs (LKS+ cells) per hind limb measured by FACS. Data are individual animals, and horizontal bars are means. (B) Schematic to describe transplantation (Tx) of HSCs or unfractionated whole bone marrow (WBM) into lethally irradiated recipients. BM, bone marrow. (C) Representative FACS profiles of stained bone marrow from control and irradiated donor mice used to sort HSCs for transplant. Individual profiles show total scatter, lineage (lin), and LKS+ gates. FSC, forward-scattered light; SSC, side-scattered light. (D) Donor cell engraftment in peripheral blood of recipients transplanted with either HSCs or WBM. Total leukocyte engraftment at 1 and 4 months posttransplant is shown. Data are means ± SEM (n = 5 per group for HSC, n = 4 per group for WBM). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; n.s., not significant; compared to control (0 Gy) mice; one-way ANOVA followed by Tukey’s test.

To confirm that the recovered residual HSCs from 2- and 6.5-Gy animals were indeed functional in their ability to repopulate the hematopoietic system, we performed HSC transplantation experiments and measured engraftment by peripheral blood chimerism at 1 and 4 months posttransplant (Fig. 2B). Bone marrow HSCs from TBI-treated mice were taken at 3 months after irradiation and transplanted in unirradiated animals. Representative fluorescence-activated cell sorting (FACS) profiles of control, 2-Gy, and 6.5-Gy donor bone marrow at 3 months after irradiation are presented in Fig. 2C. Sorted cells were gated on the LKS+ population. Donor cell engraftment (total leukocytes) at 1 and 4 months posttransplant showed an about 4- and 15-fold decrease in irradiated recipients transplanted with sorted HSCs from the 2- and 6.5-Gy irradiated animals, respectively, compared to control (Fig. 2D). When multilineage reconstitution of T cells, B cells, and myeloid cells was investigated, a similar defect in peripheral blood chimerism was observed (figs. S2 and S3). Competitive repopulation assays performed with unfractionated whole bone marrow showed similar defects in the chimerism of total leukocytes (Fig. 2D) and lineage-restricted cells in peripheral blood (fig. S4). These data suggest that, although most of the HSCs in the sublethally irradiated animals are severely impaired in their repopulating potential, rare functional HSCs do exist and maintain the hematopoietic system.

Serum miRNA profiling identifies dose-specific miRNA signatures

To test whether changes in serum miRNA levels after TBI correlate with hematopoietic injury and predict long-term damage, we profiled serum miRNAs in mice exposed to 0 (control), 2, 6.5, or 8 Gy of TBI 24 hours after radiation (n = 10 animals per group). The number of miRNAs detected per sample (count) and the average amplification threshold (Cp) value for miRNAs were evaluated, and reproducibility was confirmed (fig. S5, A and B). RBC contamination was ruled out by calculating ΔCp of miR-451 (expressed in RBCs) and miR-23a-3p (relatively stable in serum) (19) (fig. S5C).

Of 170 miRNAs detected in each sample, 68 were found to be significantly altered by radiation (table S2) and were considered dose-responsive (fig. S7). A heat map representing the top 8 of the 68 miRNAs that effectively separated irradiated from nonirradiated samples is shown in fig. S6. These miRNAs—namely, 150-5p, 142-5p, 142-3p, 136-5p, 33-5p, 320-3p, 30c-5p, and 126-3p—were the top significant hits in ANOVA analysis consisting of all irradiated samples (table S2).

We investigated whether serum miRNA expression can distinguish the 0-Gy versus 2-Gy cohorts 24 hours after exposure, and whether these miRNAs correlated with BM-MNC or HPC/HSC counts at the 7-day time point. On the basis of our profiling, five serum miRNAs were effective in segregating the two groups 24 hours after radiation exposure (Fig. 3A). miR-130a-3p showed an increase, whereas miR-150-5p, miR-142-5p, miR-706, and miR-342-3p showed significant decreases. We validated this signature using an independent set of animals that were left untreated or exposed to a TBI dose of 2 Gy and observed that all but one miRNA (miR-706) distinguished control and 2-Gy irradiated samples after 24 hours (Fig. 3B). The serum miRNA pattern continued to distinguish the 0-Gy versus 2-Gy cohorts at the 7-day postradiation time point (Fig. 3C), and was consistent with the diminished numbers of HSCs and HPCs in the 2-Gy cohort compared with the unirradiated controls (Figs. 1E and 2A). It is noteworthy that, unlike the serum miRNA signature, BM-MNC counts 1 week after radiation exposure did not differentiate the unirradiated versus 2-Gy cohorts (Fig. 1C).

Fig. 3. Serum miRNA profiling and identification of radiation dose–specific miRNA signatures.

C57BL/6J mice were exposed to 0, 2, 6.5, or 8 Gy of TBI (n = 10 per group). Serum collected from these animals 24 hours after TBI was subjected to miRNA profiling. Signatures consisting of the most highly altered five miRNAs were generated. (A) Control (0 Gy) versus 2-Gy miRNA signature with hierarchical clustering depicting relationship between individual samples. (B) Validation of the 0-Gy versus 2-Gy signature in (A) with an independent set of animals. (C) miRNA fold changes in 2-Gy irradiated animals compared to 0-Gy controls at 24 hours and 7 days after TBI. (D) Signature for 2 Gy versus 6.5 Gy with hierarchical clustering depicting relationship between individual samples. (E) Validation of the 2-Gy versus 6.5-Gy signature in (D) with an independent set of animals. (F) miRNA fold changes in 6.5-Gy irradiated animals compared to 2 Gy at 24 hours and 7 days after TBI. Data in (B), (C), (E), and (F) are normalized to miR-101a or miR-19b. Data in (B) and (E) are individual animals with means ± SEM (n = 6 to 10 per group, two independent experiments). Data in (C) and (F) are means ± SEM (n = 4 to 5 per group). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; n.s., not significant; paired t test.

Next, we asked whether serum miRNAs can be used to distinguish between low and high sublethal doses of radiation. Such an miRNA signature may be useful in distinguishing individuals that incurred mild radiation-related toxicity at low sublethal doses and more severe (often nonrecoverable) bone marrow damage at high sublethal doses (20, 21). To address this question, we revisited our profiling data to identify miRNAs differentially expressed in the sera of mice exposed to 2 and 6.5 Gy of TBI. Five miRNAs—miR-136-5p, miR-17-3p, miR-126-3p, miR-322-3p, and miR-34b-3p—showed the highest differential expression (Fig. 3D). We independently confirmed that miR-34b-3p, miR-126-3p, and miR-17-3p effectively distinguished between low and high sublethal TBI doses at 24 hours (Fig. 3E). These three miRNAs continued to separate the two sublethal groups at day 7 (Fig. 3F).

Serum miRNA signature distinguishes animals exposed to sublethal and lethal doses

Similar to the 0-Gy versus 2-Gy scenarios, analysis of hematopoietic damage alone was unable to differentiate between animals exposed to high sublethal (6.5 Gy) and lethal (8 Gy) doses of TBI until 15 days after radiation (Figs. 1 and 2). Although a relatively small difference (1.5 Gy), this increase in radiation dose is lethal. We therefore mined our profiling data in search of a set of serum miRNAs that were significantly different between 6.5- and 8-Gy treatment groups at 24 hours after radiation. The five differentially expressed miRNAs are in Fig. 4A; two were higher at 8 Gy (miR-30a-3p and miR-30c-5p) and three were lower at 8 Gy (miR-187-3p, miR-194-5p, and miR-27a-3p). In a separate cohort of animals, four of the five miRNAs remained significant (Fig. 4B). None of the miRNAs that distinguished 0 from 2 Gy, 2 from 6.5 Gy, or 6.5 from 8 Gy overlapped. The focus of our study was to identify distinct sets of miRNAs with the highest differences between certain TBI comparisons, so our signatures represent the most highly altered miRNAs. However, there may be other miRNAs that continue to change with increasing doses (table S2).

Fig. 4. An miRNA signature can differentiate between sublethal and lethal radiation exposure.

(A) The 6.5-Gy versus 8-Gy miRNA signature presented with hierarchical clustering showing relationship between individual samples. (B) Validation of the 6.5-Gy versus 8-Gy signature in (A) with an independent set of animals. Data are individual animals with means normalized to miR-101a ± SEM (n = 7 to 9 per group, two independent experiments). (C) miRNA fold changes in 8-Gy irradiated animals compared to 6.5 Gy at 24 hours, 3 days, and 7 days after TBI. Data in (C) are means normalized to miR-101a ± SEM (n = 4 to 5 per group). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; n.s., not significant; paired t test.

We further characterized the sublethal versus lethal serum miRNA signature to ask whether the miRNAs differentially expressed between the two groups continued to show differences at later time points. Sera were collected from mice at 24 hours, 3 days, and 7 after TBI doses of 6.5 and 8 Gy. Only miR-30a-3p and miR-30c-5p continued to differentiate 6.5 Gy versus 8 Gy at days 3 and 7 after TBI (Fig. 4C), suggesting that these miRNAs may allow differentiation of sublethal versus lethally irradiated samples up to 7 days after initial exposure.

Serum miRNAs distinguishing impact of sublethal and lethal radiation, not only radiation dose, correlate with survival and radioprotection

Next, we determined whether the sublethal versus lethal serum miRNA signature correlated only with the radiation dose, or if it was also linked to overall survival. To address this question, we used a radioprotective agent amifostine, which is known to extend survival in mice and in humans by decreasing radiation-related cytotoxicity (22, 23). We hypothesized that if serum miRNAs are indeed linked to the impact of radiation on health, then serum miRNA levels in animals treated with amifostine before lethal TBI should correlate with the protective effect of this drug.

Cohorts of mice were treated with saline or amifostine before 8.5 Gy of TBI, and sera were collected 24 hours after exposure (Fig. 5A). All mice exposed to lethal radiation and injected with saline died around day 16, whereas 100% of animals treated with amifostine survived (Fig. 5B). All miRNAs in the sublethal versus lethal signature from Fig. 4A were significantly altered when comparing differences between groups (table S3) and when comparing saline-treated versus amifostine-treated irradiated animals (table S4). The serum miRNA signature only changed in response to amifostine treatment after lethal radiation (not after control 0 Gy) (Fig. 5C). The serum miRNA levels correlated with the radioprotective function of amifostine; that is, the amifostine-treated lethal radiation cohort resembled the control sublethal radiation cohorts (Fig. 5C). This result strongly suggests that a serum miRNA signature distinguishing sublethal versus lethal radiation correlates with impact of radiation rather than dose.

Fig. 5. The sublethal versus lethal miRNA signature predicts impact of the radioprotective agents.

(A) Schematic of experiment with C57BL/6J mice given amifostine (ami) or saline (sal) 1 hour before 0 or 8.5 Gy TBI (n = 10 per group). Serum was isolated for miRNA profiling 24 hours later. (B) Kaplan-Meier survival curves of mice. P value determined by log-rank (Mantel-Cox) test. (C) Relative levels of indicated miRNAs in the sera of mice. Data are individual animals with means normalized to miR-101a ± SEM (n = 5 per group; two independent experiments). (D) Correlation of relative expression ratios of miRNAs in the 6.5-Gy versus 8-Gy signature from two separate experiments: those described in (A) and in Fig. 4A (r = 0.97; P = 0.0067, Pearson’s correlation). **P < 0.01; ***P < 0.001; ****P < 0.0001; n.s., not significant; one-way ANOVA followed by Dunnett’s test.

To further substantiate this point, we correlated relative miRNA expression for animals irradiated with 8.5 Gy in the radioprotection study (amifostine) with relative expression ratios of 8- and 6.5-Gy TBI cohorts in the original profiling study (Figs. 4A and 5A). We hypothesized that if the serum miRNA signature distinguishing sublethal versus lethal radiation correlates with viability rather than radiation dose, the changes in these miRNAs in the radioprotection experiment will statistically correlate with the changes in the 8- and 6.5-Gy TBI cohorts. That is, the amifostine-treated lethal radiation cohort will truly represent the sublethally radiated 6.5-Gy cohort. Consistent with our hypothesis, we uncovered a significant correlation (r = 0.97; P = 0.0067, Pearson’s correlation) (Fig. 5D), suggesting that the five miRNAs (187-3p, 194-5p, 27a-3p, 30a-3p, and 30c-5p) may serve as markers of radiation-induced loss of viability in mice.

We next asked whether this miRNA signature also corresponds to the protective effects of radiomitigating agents introduced after radiation. Transplantation of bone marrow stromal cells (BMSCs) after lethal radiation exposure promotes recovery of bone marrow cellularity, HPCs, and HSCs in mice (24). We therefore transplanted lethally irradiated (10.4 Gy) C57Bl/6J mice with two doses of BMSCs after radiation and monitored survival for 30 days (Fig. 6A). All animals in the TBI group succumbed to lethal radiation by day 10, but 100% of animals transplanted with BMSCs survived, exhibiting the protective effect of BMSC transplantation.

Fig. 6. The sublethal versus lethal miRNA signature predicts impact of radiomitigating agents.

(A) Survival curve of C57BL/6J mice exposed to 10.4 Gy of TBI followed by transplantation of BMSCs (n = 5 per group). P value determined by log-rank (Mantel-Cox) test. (B) Fold change of indicated miRNAs from sera of animals in (A) at day 5 after TBI. Data are means normalized to miR-101a ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001; n.s., not significant; one-way ANOVA followed by Dunnett’s test.

Furthermore, four of five miRNAs in the sublethal versus lethal signature (not miR-194-5p) indicated radiomitigation by BMSCs at day 5 after exposure (Fig. 6B). Decrease in miR-150-5p has emerged as a consistent marker of radiation exposure in mice (14, 25), including our 0-Gy versus 2-Gy miRNA signature (Fig. 3A). However, as anticipated from our data, miR-150-5p did not differ between TBI and TBI with BMSC treatment.

Serum miRNA signature is partially conserved in humanized mice

To test the relevance and potential applicability in humans, we validated miRNA levels in a second, “humanized” model system. We used NOD (nonobese diabetic) scid gamma (NSG) mice engrafted with human CD34+ (huCD34+) HSCs. NSG mice support robust long-term engraftment of human HSCs and their multilineage differentiation (26). Initial engraftment percentages of human CD45+ cells in peripheral blood of mice are in table S5. CD45 staining of bone marrow and peripheral blood in animals about 12 weeks after initial assessment of engraftment showed similar or higher percentages of human cells (Fig. 7A).

Fig. 7. Sublethal versus lethal miRNA signature correlates with the protective effect of amifostine in NSG mice engrafted with human CD34+ HSCs.

(A) Representative (n = 5) FACS plots showing percent engraftment of human CD45+ cells in the bone marrow and peripheral blood of NSG mice transplanted with human CD34+ HSCs. (B to E) Humanized NSG mice were pretreated with amifostine or saline and exposed to 4.5 Gy of TBI. Control animals received no pretreatment and 0 Gy of irradiation. Moribund animals were analyzed for total bone marrow cellularity (B), number of human CD45+ cells per hind limb (C), and CFU-Cs measured at 7 days after plating in human methylcellulose media (D). Data are individual animals with means ± SEM (n = 5 to 8 per group). Relative miRNA levels in the sera of humanized mice at 24 hours after exposure to TBI (E). Data are individual animals with means normalized to miR-101a ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001; n.s., not significant, by paired t test (B to D) or one-way ANOVA followed by Dunnett’s test (E).

Four to 4.5 Gy of TBI causes 100% mortality in NSG mice (27). Therefore, we exposed huCD34+ NSG mice to a lethal dose of 4.5 Gy of TBI, and considering the reconstitution of the hematopoietic compartment with human cells, we focused on bone marrow injury, peripheral blood counts, and the potential for marrow recovery by amifostine. Exposure to lethal radiation showed near complete depletion of BM-MNCs in the saline-treated TBI cohort, whereas previous treatment with amifostine showed rescue of bone marrow cellularity, and moderate recovery of human CD45+ cells and bone marrow CFU-Cs (Fig. 7, B to D). Peripheral blood counts of moribund animals indicated a significant decrease in all blood cell parameters except WBCs, and a substantial improvement in hemoglobin, RBC levels, and hematocrit levels which was observed in the amifostine-treated group (table S6). These data suggest that huCD34+ NSG mice are sensitive to the effects of radiation and that amifostine treatment led to protection of the human cells.

The sequences of the miRNAs in the sublethal versus lethal signature in mice were identical to those of humans (fig. S8) (28). To assess whether the effect of lethal radiation with amifostine treatment is reflected in the levels of miRNAs in the sublethal versus lethal signature, we isolated serum from these humanized animals 24 hours after radiation and measured miRNA levels (Fig. 7E). Consistent with previous results in C57Bl/6J mice, miR-150-5p and four of five miRNAs in the 6.5-Gy versus 8-Gy signature (miR-27a-3p, miR-187-3p, miR-30a-3p, and miR-30c-5p) were altered in response to radiation in huCD34+ NSG mice. Furthermore, amifostine pretreatment rescued levels of three of the miRNAs (miR-187-3p, miR-27a-3p, and miR-30a-3p). Together, these results suggest that serum miRNA signatures may be conserved between mice and humans, and thus have the potential to serve as indicators of radiation injury in humans.


Current use of diagnostic screening to estimate dose of accidental radiation exposure is mainly based on three factors: time to onset of radiation sickness, kinetics of lymphocyte depletion, and analysis of chromosomal abnormalities (2, 29). These techniques are time-consuming and are often not quantitative enough to draw definite conclusions. Serum miRNAs fall under the emerging “omic” biodosimetry assays and represent a simple technology that may effectively determine whether an individual was exposed to radiation (if so, whether the dose was sublethal or lethal), and predict long-term survival of exposed individuals. With current progress in miniaturization of quantitative polymerase chain reaction (qPCR), an miRNA-based assay has the potential to become a point-of-care technology. Compared to lymphocyte depletion kinetics and DNA damage assays using γ-H2AX, miRNA-based assays could allow longer lag time after exposure before the first sample is taken and provide results within 12 to 24 hours after exposure.

Here, we successfully profiled and identified serum miRNAs that were differentially expressed in response to TBI and correlated well with injury at sublethal and lethal doses. Recent studies (14, 15, 25) have identified circulating (serum/plasma) miRNAs that are altered in response to TBI, and there is partial overlap with some of the candidate miRNAs (miR-126-3p, miR-150, miR-342-3p, miR-151-3p, miR-139-3p, and miR-142) that emerged from our analysis. A possible cause for the differences is the miRNA expression profiling platform. Recently, a study (30) systematically compared 12 different miRNA expression platforms. Specifically for serum miRNAs, there was a 12-fold difference between the highest and lowest numbers of detected miRNA when identical samples were profiled by different platforms. According to this report, the LNA (locked nucleic acid)–based platform from Exiqon—which we used in our study—had the highest specificity.

A key issue we addressed here is the correlation of serum miRNAs with the impact of radiation, specifically with hematopoietic injury and animal viability. Working with a narrow dose range, we identified serum miRNAs that distinguish between sublethal (6.5 Gy) and lethal (8 Gy) exposure. This is of paramount importance because during a radiologic emergency, doses sustained will almost never be in specific increments and distinction between lethal and sublethal doses is the key challenge. Moreover, this miRNA signature also predicted the impact of radiation on animal survival after pretreatment with the radioprotective agent amifostine or mitigation using BMSC. Most of the serum miRNAs (miR-187-3p, miR-27a-3p, miR-30a-3p, and miR-30c-5p) that correlated with amifostine radioprotection also predicted survival of animals transplanted with BMSCs. Finally, our experiments with huCD34+ NSG humanized mice suggest that the miRNAs identified in the C57BL/6J mouse model may also be relevant in humans. Together, these findings highlight the use of miRNAs in predicting the functional impact of both radioprotective and radiomitigating agents and broadly suggest a potential application of serum miRNAs in prognosticating the impact of radiation. Future studies, however, will be necessary with human samples to validate these miRNAs or discover new signatures that differentiate lethal from sublethal radiation effects.

Although the miRNAs identified here can forecast the extent of hematopoietic injury and predict for survival immediately after exposure, it is possible that radiation victims may not report to a medical countermeasures facility in the first 24 hours. In such a scenario, biomarkers that persist longer will need to be identified. Profiling of serum miRNAs at later time points, such as 5 or 7 days, will allow detection of more persistent miRNAs that continue to follow a specific trend after radiation. As a proof of principle, miR-30a-3p and miR-30c-5p in our sublethal versus lethal signature continued to display an increase until day 7. Future studies should focus on defining the cell of origin of specific serum miRNAs and investigate their physiological relevance. This will help determine whether serum miRNAs are passive bystanders secreted in the wake of radiation injury or act as potential alarm signals communicating a state of distress to different parts of the body (12, 13). Indeed, several groups have reported identification of circulating miRNAs found in exosomes (31), apoptotic bodies (32), high-density lipoprotein (33), and RNA binding proteins (34) as a form of cell-to-cell communication (14, 35), providing a rationale for future experimentation. Validation of serum miRNAs identified here in nonhuman primate and patient samples will allow the application of these signatures in humans.

Broadly, our results provide the first evidence that serum miRNAs may effectively predict the impact of radiation on long-term viability of animals. Our work represents an advance in early assessment of radiation damage, which can help alleviate hematopoietic symptoms, facilitate timely intervention after exposure, and improve overall survival of exposed individuals.


Study design

This study was designed to investigate whether serum miRNAs can predict radiation-induced hematopoieitic damage early after radiation exposure. Differential expression of miRNAs in response to radiation was investigated in two model systems, strongly suggesting the applicability of miRNAs as radiation-specific markers of latent hematopoietic injury. Animals were also treated with radioprotective and radiomitigating agents to correlate levels of specific miRNAs with animal survival at 30 days after irradiation. All animal procedures performed were approved by the Institutional Animal Care and Use Committee (IACUC) at Dana-Farber Cancer Institute (DFCI). Animals were maintained in the animal facility and given ad libitum access to food and water. Body condition score (BCS) as described in (36) was used to standardize endpoints. A BCS of 3 was regarded as the endpoint for all irradiated animals irrespective of treatment group, at which point animals were considered moribund and euthanized. Mouse serum samples used for miRNA profiling were randomized before analysis, and experiments were generally repeated three times.

Power analysis was performed using the Hierarchical Clustering Explorer 3.5 tool (37). The number of samples was estimated to be sufficient to provide statistical power of at least 80% needed to obtain a P value of less than 0.01 for differentially expressed miRNAs with a fold change 0 > 1.5 or < 0.67 in between group comparisons. The P value threshold was lowered from 0.05 to account for multigroup post hoc testing. A sample size of 10 per group was thus calculated to allow us to confirm statistically significant differences for the top 95 differentially expressed miRNAs with the predetermined effect sizes.

Mice and TBI

C57BL/6J male mice (10 weeks old) were obtained from Jackson Labs and acclimated in the Animal Research Facility at DFCI before irradiation at the age of 12 to 13 weeks. All procedures performed were approved by IACUC at DFCI. Animals were exposed to TBI in an irradiation pie cage (Braintree Scientific). Irradiation was performed using a 137Cs source at a dose rate of 110 cGy/min using a 137Cs source (Gammacell 40 Exactor, Best Theratronics). Instrument calibration was performed according to vendor instructions in accordance with the DFCI Office of Radiation Safety. After irradiation, bone marrow was harvested and cells were counted by flow cytometry, as described in the Supplementary Materials and Methods.

HuCD34+ NSG mice

HuCD34+ humanized NSG mice were obtained from Jackson Labs and housed in a BL2/N facility at DFCI. All procedures were approved by IACUC at DFCI. During the generation of these mice at Jackson Labs, 3-week-old female NSG mice were irradiated at 1.4 Gy to deplete their bone marrow and injected with CD34+ human HSCs. At 12 weeks after transplant, each mouse was tested by FACS for engraftment of human CD45+ and murine CD45+ cells. Animals were obtained at DFCI about 10 weeks after engraftment confirmation at Jackson Labs. Peripheral blood and bone marrow from animals in the untreated control arm were reconfirmed for the presence of human CD45+ cells at DFCI with antihuman CD45 FITC (fluorescein isothiocyanate) (clone 2D1 from BD Biosciences).

Murine miRNA profiling

miRCURY LNA Universal RT miRNA PCR Rodent Panel I&II containing 742 assays was used to profile miRNAs differentially expressed in mouse serum from animals exposed to different doses of radiation (Exiqon). Normalization of data was performed using the global mean of 170 miRNAs most commonly expressed in all samples. RNA spike-in control (UniSp6) was used to test the efficiency of the complementary DNA synthesis reaction, whereas DNA spike-in control (UniSp3) tested the efficiency of qPCR amplification. Spike-in controls were used throughout the study for profiling and validation. To negate the possibility of hemolysis, ΔCp for miR-451 and miR-23a-3p was computed for each sample as previously reported (19).

Statistical analysis

miRNA profiling. Normalization of miRNA serum levels was performed using 170 commonly expressed miRNA. ANOVA was used to determine which miRNAs differed significantly between groups. To adjust for multiple comparisons testing, the Benjamini-Hochberg correction was applied. A threshold of P < 0.05 in ANOVA was selected as the level of statistical significance. Tukey’s test was used to determine between-group significance in post hoc comparisons. miRNAs with P values < 0.05 in ANOVA were used in hierarchical clustering analysis to visualize expression patterns. Differentially expressed miRNAs were tested in pairwise comparisons with a Benjamini-Hochberg adjusted Student’s t test to determine between-group differences.

Validation with real-time qPCR. One-way ANOVA was used to confirm global significance. Dunnett’s post hoc testing procedure was used to compare miRNA levels in the 8.5 Gy + saline group against the other three experimental groups. Univariate comparisons were performed using the Student’s t test or the Student’s t test for paired samples. Pearson’s correlation coefficient was used for correlation testing. Survival analysis was performed using the log-rank (Mantel-Cox) test.


Fig. S1. Analysis of peripheral blood CBC parameters after TBI-induced hematopoietic damage.

Fig. S2. Repopulation analysis at 1 month after HSC transplant.

Fig. S3. Repopulation analysis at 4 months after HSC transplant.

Fig. S4. Repopulation analysis after unfractionated whole bone marrow transplant.

Fig. S5. Serum miRNA profiling.

Fig. S6. miRNAs significantly altered in all irradiated samples.

Fig. S7. Radiation dose dependence of select miRNAs.

Fig. S8. Identical mature miRNA sequences in human and mouse.

Table S1. Peripheral blood CBC levels in individual mice.

Table S2. Fold changes of 68 statistically significant miRNAs at different doses of TBI with respect to controls.

Table S3. Significantly altered miRNAs between 6.5 and 8 Gy of irradiation and pretreatment with amifostine.

Table S4. miRNAs in the 6.5-Gy versus 8-Gy signature as indicators of survival after TBI.

Table S5. Percent human CD45+ cell engraftment in individual huCD34+ NSG (humanized) mice.

Table S6. Peripheral blood CBC in individual huCD34+ NSG (humanized) mice.

Table S7. Target sequences of individual miRNAs.

Reference (38)


  1. Acknowledgments: We thank A. Andrea for reading the manuscript. Funding: D.C. is supported by R01 AI101897-01 (National Institute of Allergy and Infectious Diseases), Basic Scholar Grant (American Cancer Society), Leukemia & Lymphoma Society Scholar Grant, Ann-Fuller Foundation, and Mary Kay Foundation. C.G. is supported by 1U19AI091175-05. W.F. is funded by the National Science Center of Poland (grant number 2012/05/E/NZ5/02130) and the INTER program of the Foundation for Polish Science. Author contributions: S.S.A. conducted most of the experiments with help from A.H., J.W., Y.P., P.M., and E.G. W.F. did all the statistical analysis. miRNA analysis was performed by S.S.A with help from P.M. and Y.P. P.B. and S.S. conducted the radiation mitigation experiment using BMSCs. C.G., K.P., and D.C. designed the experiments. The manuscript was written by S.S.A. and D.C. with input from K.P. and C.G. Competing interests: The authors declare that they have no competing interests. Data and materials availability: No material transfer agreement required for materials used in this study.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article