INTRODUCTION

Human exposure to neutron radiation is increasing with the development of new technologies. For example, astronauts are exposed to neutrons produced from protons impinging on spacecraft, cancer patients are exposed to secondary neutrons generated in therapeutic devices related to proton, heavy-particle and intensity modulated radiation therapies, and people have been exposed to neutrons as a result of severe nuclear accidents (1–5). Neutron radiation is a high-linear energy transfer (LET) values compared to gamma (γ) rays, its relative biological effectiveness (RBE) is high for biological end points including those assessed in vitro and in vivo, such as cell killing, life shortening or tumor development (6–9). The age at the time of exposure is an important modifier of various radiation effects including tumor induction (10). The recent development of proton and carbon-ion radiotherapies as treatments for pediatric cancer (11, 12) has raised substantial concerns regarding neutron-induced second-cancer risk in survivors who have received these therapies (13–16). Epidemiology on the carcinogenic effect of neutrons has been very limited with few exceptions (17). Regarding the effect of radiation exposure during infancy, the result of the atomic bomb survivor cohort with in utero/childhood exposure has been among the most informative (18), but neutrons were only a small component of the overall exposure. On the other hand, rodent experiments exploring the long-term consequences of neutron exposure during infancy have indicated its significant effects on life span and tumor induction (19–21). Still, extremely limited data have been available on the carcinogenic effect of neutron radiation during infancy.

Cancer develops as a result of various environmental and genetic factors including radiation exposure (22). Cancer-prone subpopulations include individuals who have genetic defects in DNA repair machinery or in signaling pathways related to organ development, examples of the latter being the Hedgehog and Wnt cascades (23). Among the mouse models based on deficiencies in these genes is the Ptch1^+/– mouse, a model of Gorlin syndrome, which carries a defective copy of a tumor suppressor gene involved in negative regulation of the sonic hedgehog (Shh) signaling pathway (24). Ptch1^+/– mice often develop medulloblastomas (MBs) as a result of deregulation of the Shh signal in the granule cells of the external granular layer of the cerebellum (25), and the frequency increases after mice are exposed to radiation during a narrow window of time around birth (26). Loss of the wild-type Ptch1 allele on chromosome 13 is evident in almost all spontaneously developed and radiation-induced MBs (27). We previously confirmed that the MBs which developed after radiation exposure exhibited an interstitial deletion resulting in loss of the wild-type Ptch1 allele (designated as “radiation-induced” or R-type MB), and that almost all spontaneously developed MBs exhibited loss of all chromosomal regions distal to the wild-type Ptch1 (designated as “spontaneous” or S-type MB) (28, 29). As such, MBs of Ptch1^+/– mice provide a useful experimental system for directly assessing the cancer-inducing effect of radiation by focusing on these R-type tumors.

The goal of the current study was to clarify the age dependence of the effect of fast neutrons and their RBE regarding the induction of MBs in Ptch1^+/– mice. For this purpose, we took advantage of the above-mentioned characteristics of genetic aberration specific to radiation-induced MBs. Furthermore, molecular analysis of the neutron-induced tumors revealed a complex biological mechanism distinct from that induced by γ rays and X rays.

MATERIALS AND METHODS

RESULTS

Induction of MBs in Ptch1^+/– Mice after Gamma-Ray and Neutron Irradiation

To clarify the biological effect of neutrons on the incidence of MB and its age dependence, compared to that of γ rays, we monitored the MB development up to P500 in Ptch1^+/– mice that received neutron or γ-ray irradiation at 0.025–0.5 Gy. Moribundity was reached in 860 out of 1,291 mice (67%) by P500, and all MBs therein were pathologically diagnosed ( Supplementary Table S1 (rare-196-02-09_s01.pdf);  https://doi.org/10.1667/RADE-21-00025.1.S1). Kaplan-Meier plots of MB-free survival were constructed using data before P500 ( Supplementary Fig. S2 (rare-196-02-09_s01.pdf)), and a hazard ratio was calculated for each dose using the Cox proportional hazard model. Hazard ratios increased after neutron irradiation in a dose-dependent manner, where a significant increase was noted at lower doses of 0.1 Gy (E14 and E17, both P < 0.01) and 0.05 Gy (P1, P < 0.01); the susceptibility decreased slightly at P4 (with a significant increase only at 0.5 Gy, P < 0.05), whereas a significant reduction was observed at 0.5 Gy at P10 (P < 0.01) (Fig. 1). In contrast, the γ-ray irradiated groups showed a significant increase in the hazard ratio only for 0.5 Gy at P1 (P < 0.05), and there was only a non-significant increase for 0.5 Gy at E14 and E17, with a tendency toward a decrease at P4 and P10. Because of the small increase in the γ-ray irradiated groups, it was considered unrealistic to calculate a statistically significant RBE of neutrons as relative to that of γ rays based on these hazard ratio data.

FIG. 1

Dose response of the hazard ratio of the incidence of medulloblastoma in Ptch1^+/– mice. Panels A–J: Ptch1^+/– mice received γ-ray (panels A–E) or fast neutron (panels F–J) irradiation on embryonic day (E)14 or 17 or postnatal day (P)1, 4 or 10 and were observed until P500. Data for groups γ-ray irradiated on E14 (0.1 and 0.5 Gy), E17 (0.1 and 0.5 Gy), P1 (0.1 and 0.5 Gy), and P4 (0.1 and 0.5 Gy) are from previously published work (29). Hazard ratios (dots) are superimposed with linear dose response models (solid line). Error bars show 95% confidence intervals; in panel H, the arrow with a number in parentheses indicates the upper limit. *P < 0.05 and **P < 0.01 by Cox regression.

LOH and DNA Copy Number Analyses of MB

As noted in the Introduction, MBs of Ptch1^+/– mice are classified into two molecular subtypes, which R- and S-type, based on the pattern of LOH on chromosome 13 (27, 28). Therefore, we theorized that this technique would provide a more sensitive approach to obtaining the RBE of neutrons in the current study and identified R-type MBs to quantitatively compare the effects of neutrons and γ rays. We examined the LOH patterns of chromosome 13, which harbors the Ptch1 locus, in all tumors from the experimental groups (nonirradiated, 31 tumors; neutron irradiated, 450 tumors; γ-ray irradiated, 292 tumors) except for one tumor and five tumors in the γ-ray and neutron irradiated groups, respectively, which were unavailable for molecular analysis because of their autolysis. Almost all neutron- and γ-ray-induced MBs showed loss of the wild-type Ptch1 allele on the C3H-derived copy of chromosome 13 and were classified as S- or R-type (Fig. 2A and  Supplementary Fig. S3 (rare-196-02-09_s01.pdf);  https://doi.org/10.1667/RADE-21-00025.1.S1). Furthermore, the incidence of R-type MBs increased, whereas that of S-type MBs tended to decrease, in a dose-dependent manner in all groups except P10 (Fig. 3), as observed in a previously published study on X rays (28).

FIG. 2

Chromosome 13 loss of heterozygosity (LOH) and copy number aberrations in spontaneous and radiation-induced medulloblastomas from Ptch1^+/– mice. Chromosomal regions and positions of polymorphic markers are shown on the left and right side of the schematic chromosome, respectively. Open circles indicate loss of a C3H allele, and filled circles indicate retention of both B6 and C3H alleles. Note that the wild-type Ptch1 allele is located on the C3H chromosome in these mice. Large circles indicate the results of LOH obtained from polymorphic markers, and small circles indicate the results of LOH obtained from a single-nucleotide polymorphism (SNP) in exon 23 of Ptch1. Panel A: LOH patterns of S-type and R-type tumors. S-type tumors are marked by allelic losses at all consecutive markers distal to the Ptch1 locus, indicated with gray shading, whereas R-type tumors show allelic loss(es) only at markers in the interstitial region around or within the Ptch1 locus; N-type tumors were without LOH. Panel B, left side: LOH patterns of tumors with multiple deletions having a hemi-allelic loss at interstitial sites involving the Ptch1 locus and another allelic loss at distal sites. LOH of these types existed only in neutron-induced MBs [0.5 Gy at P4 (sample no. 1), 0.1 Gy at P10 (no. 2 and 4), and 0.5 Gy at P10 (no. 3)]. Panel B, right side: Results of the DNA copy number analysis, indicating the consistency of these LOH patterns.

FIG. 3

Dose response of the incidence of S-, R- and N-type medulloblastomas in Ptch1^+/– mice. *P < 0.05, **P < 0.01 vs. nonirradiated control by Fisher's exact probability test.

Interestingly, although rare, 5 of 305 (1.6%) R-type tumors in the neutron irradiated groups (one each from the group of 0.5 Gy at E14, P4 and P10 and two from 0.1 Gy at P10) exhibited two separate LOH regions of the C3H allele on chromosome 13, one encompassing the Ptch1 locus and the other close to the telomere (Fig. 2B and  Supplementary Fig. S3A (rare-196-02-09_s01.pdf);  https://doi.org/10.1667/RADE-21-00025.1.S1). An array-CGH analysis of four of the above five tumors showed reduction in the DNA copy number in those LOH regions (Fig. 2B), indicating interstitial and terminal deletions of the C3H allele. Such patterns of LOH have not been observed in results from γ- and X-ray-induced tumors (28, 29).

Mutation of Ptch1 in MBs without LOH

In the above analysis, LOH encompassing the Ptch1 locus was not detected in 53 of the 773 tumors (6.9%) [3 of 31 (9.7%) in the nonirradiated, 18 of 450 (4.0%) in the neutron irradiated, and 32 of 292 (10.9%) in the γ-ray irradiated groups], and we tentatively referred to these tumors as “N-type” (no evidence of LOH). Most MBs in Ptch1^+/– mice show inactivation of the wild-type Ptch1 allele, with only a minority retaining it intact (34, 35). We therefore analyzed the cDNA and genomic DNA of 37 N-type MBs (3 from nonirradiated, 12 from neutron irradiated, and 22 from γ-ray irradiated mice) and found mutations in Ptch1 in 26 tumors (70%). MBs without any mutations in Ptch1 were found only among the neutron- and γ-ray irradiated tumors and occurred at frequencies [6 of 12 (50%) and 5 of 22 (23%) N-type tumors, respectively] (Table 1 and  Supplementary Fig. S4 (rare-196-02-09_s01.pdf);  https://doi.org/10.1667/RADE-21-00025.1.S1). Most of the mutations were deletions of 1–1,908 bp and resulted in de novo termination codons and amino acid changes affecting important functional sites including the Hedgehog-binding and sterol-sensing domains [3 of 3 (100%) N-type tumors in nonirradiated, 4 of 6 (67%) N-type tumors in neutron irradiated, and 9 of 17 (53%) N-type tumors in γ-ray irradiated mice]. Among them, the junctional sequences in five tumors (three in nonirradiated, one in neutron irradiated, and one in γ-ray irradiated mice) indicated that the deletions resulted from a DNA repair mechanism involving the microhomology-mediated end joining (Table 1 and  Supplementary Fig. S4 (rare-196-02-09_s01.pdf);  https://doi.org/10.1667/RADE-21-00025.1.S1), a mechanism of mutagenesis not necessarily associated with radiation. The rest of the mutations were duplications, insertions, and base substitutions at splice acceptors, which also resulted in de novo termination codons and amino acid changes.

TABLE 1

Mutations Identified in “N-type” Medulloblastomas

We further investigated whether the B6 or C3H alleles had the mutation for each tumor, taking advantage of SNPs and the neo cassette inserted in place of exons 6 and 7 in the B6 allele ( Supplementary Fig. S1 (rare-196-02-09_s01.pdf);  https://doi.org/10.1667/RADE-21-00025.1.S1); these approaches enabled identification of the allele (i.e., B6 or C3H) harboring mutations in exons 3, 4, 9, 10, 21 and 22. The result showed that mutations were on the C3H allele, indicating that these mutations affected the function of the wild-type Ptch1 allele, except for one case in the nonirradiated group that had a mutation in the B6 allele. Based on these results, the observed mutations in Ptch1 in the “N-type” tumors were unlikely to be radiation specific. Furthermore, these N-type tumors did not increase in a dose-dependent manner, unlike the R-type tumors (Fig. 3). We therefore conclude that the N-type MBs should be classified as S-type, instead of R-type, tumors.

Assessment of the Effect of Gamma Rays and Neutrons on Induction of R-type MBs

Based on the above classification of MBs into S- and R-types, we analyzed the MB development after neutron and γ-ray irradiations at various ages relative to that of the nonirradiated group (Fig. 4A and  Supplementary Fig. S5 (rare-196-02-09_s01.pdf);  https://doi.org/10.1667/RADE-21-00025.1.S1). The survival curves of mice with S-type MBs did not show significant differences between the nonirradiated and irradiated groups, except for a neutron irradiated group (0.5 Gy on P10) in which tumor development was decreased. As for the R-type MBs, neutrons significantly increased the tumor development at doses above 0.1 Gy in the E14 and E17 groups (P < 0.05), at or above 0.025 Gy in the P1 group (P < 0.05), and at 0.5 Gy in the P4 group (P < 0.05). Gamma rays also significantly increased tumor development at 0.5 Gy in the E14 and E17 groups (both P < 0.05) and at or above 0.2 Gy in the P1 group (P < 0.05). Development of R-type MBs was strikingly suppressed in neutron- and γ-ray irradiated P10 groups. These results indicate that the age of susceptibility to radiation-induced MBs is confined to a period immediately after birth and is independent of the radiation quality used in these experiments. More importantly, the use of molecular subtypes enables distinction of radiation-induced tumors from spontaneous ones and thus allows quantification of the small effects of γ rays.

FIG. 4

Incidence of R-type medulloblastomas and the relative biological effectiveness (RBE) of neutron calculated thereby. Panel A: Kaplan-Meier plots for the incidence of S- and R-type MBs in mice irradiated at P1. The data for 0.1 and 0.5 Gy of γ rays and for nonirradiated mice are reproduced from our previously published study (29). *P < 0.01 vs. nonirradiated group by log-rank test. Panel B. Dose response of the hazard ratio for R-type MBs in neutron- and γ-ray irradiated groups (black and gray, respectively). Hazard ratios for 0.1 and 0.5 Gy of γ rays were calculated using previously published data (29). The dose-response curves are fitted by linear equations. *P < 0.05, **P < 0.01 and ***P < 0.001 by Cox regression. Error bars indicate 95% confidence intervals; vertical arrows with a number in parentheses indicate the upper limit. Panel C: Slope of the linear equations in panel B and the RBE of neutrons. CI = 95% confidence interval.

Finally, we analyzed the hazard ratio of R-type MBs for all experimental groups. The hazard ratio of all irradiated groups sharply increased in a dose-dependent manner except for the P10 groups (Fig. 4B), indicating that the R-type tumor is a useful indicator of radiation carcinogenesis. We therefore calculated the RBE of neutrons, relative to γ rays, based on the slopes of the dose response of R-type tumors (Fig. 4C); the P10 groups were not included, as R-type MBs did not increase in these groups. The RBE was calculated to be 12.9 (8.6, 17.2), 9.6 (6.9, 12.3), 21.5 (17.2, 25.8) and 7.1 (4.7, 9.5) (mean and 95% confidence interval) in the E14, E17, P1 and P4 groups, respectively (Fig. 4C). The confidence interval at P1 did not overlap those at other ages. Thus, the RBE of neutrons at P1 was highest and related to the susceptibility to radiation-induced tumorigenesis of the tissue.

DISCUSSION

The age at exposure is an important modifier of various radiation effects including carcinogenesis (10). Radiation exposure is more harmful to and more carcinogenic in infants than adults. The data presented herein show an age dependence for MB induction by fast neutrons in Ptch1^+/–mice from E14 to P10, providing evidence consistent with two previously published studies reporting variation in the RBE of fast neutrons based on the age at exposure (19, 20). An advantage of the current study over previous ones is its use of radiation-specific genetic aberrations in the tumor, i.e., interstitial deletions affecting the wild-type Ptch1 allele as revealed by LOH and genomic copy number analyses. We found that 6.9% of MBs did not have LOH of chromosome 13 (designated as N-type) across the experimental groups, and most of these tumors had frameshift mutations in the functional domains of the wild-type Ptch1 allele, resulting in de novo termination codons and amino acid changes. These mutations are similar to those observed by Frappart et al. (36) in mice with neural-tissue-specific inactivation of DNA double-strand break repair proteins combined with Trp53 deficiency. MBs in those mice had lost one copy of Ptch1 through chromosomal loss or translocation, and the remaining Ptch1 allele contained mutations resembling those in the N-type tumors herein. We have also reported that Tsc2^Eker/+ rats, which have a mutation in a renal tumor suppressor gene, develop radiation-induced and spontaneous renal tumors exhibiting de novo termination codons in the wild-type Tsc2 allele due to small deletions (37). Thus, these types of mutations, including those observed herein in the N-type tumors of Ptch1^+/–mice, are not necessarily related to radiation exposure.

The susceptibility to radiation-induced carcinogenesis in Ptch1^+/– mice was highest during the perinatal period (E14 to P1) regardless of the radiation type (i.e., both γ rays and neutrons) (Fig. 3 and 4,  Supplementary Figs. S5 (rare-196-02-09_s01.pdf);  https://doi.org/10.1667/RADE-21-00025.1.S1), consistent with previously published studies in which the narrow age range of susceptibility was not affected by differences in radiation quality (26, 28). To evaluate the influence of age at exposure on carcinogenesis induced by radiation of different quality, it is crucial to extend the research to different animal models in which the age has a critical impact on carcinogenesis of organs relevant to radiological protection (e.g., bone marrow, liver, digestive tract and mammary gland). In particular, it would be ideal to use animal models that can distinguish radiation-induced tumors from spontaneous ones like those in Ptch1^+/– mice.

The RBE of neutrons herein ranged from approximately 7 to 21 depending on the age at exposure, which overlaps the previously reported ranges of 5–8 for all malignant tumors and hepatocellular carcinoma in B6 mice (19) and 7–26 for mammary carcinoma in Sprague-Dawley rats (20); high-RBE values have generally been noted to be consistent with the age of high susceptibility to radiation carcinogenesis. A limitation of the current study is the relatively small number of dose points. The dose response of radiation induction of tumor is either non-linear (e.g., downward concave or convex) or linear in previously published studies (19, 20, 38), and RBE is greatly influenced by the shape of dose response. The current study is designed to have only two to five dose points, precluding the assessment of the shape of dose response. Nevertheless, the uncertainty resulting from this issue is evaluated as the confidence interval of the RBE value in the current study. An implication of the study is that the RBE value is age dependent and within the range of those reported previously. The current methodology is considered to offer more accuracy than previous studies in that it counts tumors most likely caused by radiation; thus, the result of the current study supports the conservativeness of the radiation weighting factor of neutrons, which is a function of energy having a maximum of 20, as recommended by the International Commission on Radiological Protection (39).

Although rare, we observed some (1.6%) neutron-induced MBs that had multiple deletions on chromosome 13, in which telomeric regions were affected by interstitial or terminal deletions in addition to the interstitial regions harboring the Ptch1 gene (Fig. 2B). Cases with such coexisting deletions have not been reported in X- or γ-ray-induced tumors of Ptch1^+/– mice. Interestingly, Brown et al. have also identified such types of complex deletions on chromosome 2 in a very small number of fast neutron-induced cases of acute myeloid leukemia in mice (40). These deletions are most likely a consequence of the biophysical nature of the interaction of neutrons with DNA. Fast neutrons produce high-LET recoil protons and α particles in the absorbing tissue, contributing to dense ionization and excitation of DNA molecules, thereby generating clusters of DNA damage, which are difficult to repair (41–43). These clusters of damage are likely to lead to a spectrum of mutations such as large deletions, translocations and other chromosomal aberrations (43), suggesting that the multiple deletions detected in neutron-induced MBs herein reflect the biophysical characteristics of fast neutrons. Three of five tumors with such multiple deletions developed postirradiation at P10 ( Supplementary Fig. S3 (rare-196-02-09_s01.pdf);  https://doi.org/10.1667/RADE-21-00025.1.S1), a period of very low susceptibility as cerebellar granule cells have undergone substantial differentiation by this time (26). The mechanism of development of MBs via multiple deletions during this period of insusceptibility remains an issue for future work.

In conclusion, the age dependence of radiation induction of MBs in Ptch1^+/– mice was not affected by radiation quality. The RBE of neutrons was evaluated based on the incidence of tumors harboring genetic aberrations that are most probably the direct consequence of radiation. Multiple deletions observed in some neutron-induced tumors are most likely related to a biophysical characteristic inherent to neutron radiation.

SUPPLEMENTARY INFORMATION

 Table S1 (rare-196-02-09_s01.pdf). Experimental groups and results of MBs development.

 Table S2 (rare-196-02-09_s01.pdf). Ptch1 primers and PCR conditions.

 Table S3 (rare-196-02-09_s01.pdf). Primers for sequencing.

 Fig. S1 (rare-196-02-09_s01.pdf). Structure of mouse Ptch1 and location of primers used for mutation analyses. Panel A: Genomic DNA (upper) and cDNA (lower) of wild-type Ptch1. Topological domains (thick lines), which include glycosylation sites (triangles) and the sterol-sensing domain (dashed double-headed arrows), are indicated above the cDNA. The cDNA was divided into four segments (dashed lines with numbers) in the sequence analysis. Arrows flanking each segment indicate the primers listed in  Supplementary Table S2 (rare-196-02-09_s01.pdf). Panel B: Relevant exons of knockout (B6) and wild-type (C3H) alleles of Ptch1. Exons 6 and 7 in the genome are replaced by a neomycin resistance cassette. Dashed line reproduces segment 1 in panel A. Primers “B6_F” and “B6_R” were designed to sequence the B6 allele, whereas primers “C3H_F” and “C3H_R” are for the C3H allele. Primers “B6_R” and “C3H_R” were designed to analyze exons 3 and 4 of cases no. 4 and 12 (see Table 1), and primers “B6_F” and “C3H_F” were designed to analyze exons 9 and 10 of cases no. 8, 13 and 25 (see Table 1). Panel C: Primers for genomic sequencing to identify the cause of exon 17 deletion in cDNA from case no. 1. A 4,163-bp region between exons 16 and 18 (dashed line) was first amplified and then sequenced using primers referred to by the bracketed numbers. Alleles were identified using the two single-nucleotide polymorphisms (SNPs, triangles). Panel D: Primers for genomic sequencing to identify the cause of deletion of exons 21 and 22 in cDNA from case no. 21. A 6,167-bp region from exons 20 to exon 23 (dashed line) was first amplified and then sequenced using primers referred to by the bracketed numbers. Alleles were identified using the three SNPs (triangles).

 Fig. S2 (rare-196-02-09_s01.pdf). MB-free Kaplan-Meier plots after γ rays and neutron irradiation. *P < 0.01, **P < 0.05 vs. nonirradiated group by log-rank test.

 Fig. S3 (rare-196-02-09_s01.pdf). Loss of heterozygosity (LOH) of chromosome 13 in medulloblastomas from nonirradiated and neutron- and γ-ray irradiated groups. Panel A: Nonirradiated and neutron irradiated groups. Panel B: γ-ray irradiated group. Chromosomal regions and positions of polymorphic markers are shown on the upper left side in the schematic in panel A. Open circles indicate loss of the C3H allele, and filled circles indicate retention of both the B6 and C3H alleles. Results obtained for consecutive markers for individual MBs are aligned vertically; “×” indicates unavailable results. S-type or S = LOH in all consecutive markers distal to Ptch1; R-type or R = LOH confined to interstitial markers (big circles) or a polymorphism at position 4,016 in exon 23 of Ptch1 (small circles); N-type or N = no evidence of LOH; “No” indicates tumors that were unavailable for LOH analysis. The numbers at the top of each column indicate the number of MBs with each LOH pattern. *These MBs showed multiple stretches of LOH.

 Fig. S4 (rare-196-02-09_s01.pdf). Mutations detected in Ptch1 of medulloblastomas. Panels A–C: Characterization of mutations in MBs from nonirradiated (panel A), neutron irradiated (panel B), and γ-ray irradiated (panel C) mice. Sequences of the wild-type allele (upper) and identified mutations (lower) are indicated along with schematics for the relevant region of the cDNA (right-side bottom), in which solid boxes indicate exons, dashed boxes indicate deleted exons, and arrowheads indicate mutation sites. Numbers in the upper left side of each panel indicate the case number (see Table 1). Upper- and lowercase letters represent exonic and intronic sequences, respectively. Letters in blue = insertions; in green = duplications; in orange = base substitutions; in red = microhomology presumably used in end joining, which generated the deletion.

 Fig. S5 (rare-196-02-09_s01.pdf). Kaplan-Meier plots for the incidence of S- and R-type medulloblastomas. Separate graphs are shown for γ-ray irradiated and neutron irradiated groups. Plotted are the ages at death or moribundity of mice with MBs that were retrospectively determined to be S- or R-type; N-type tumors were treated as S-type. Data for mice irradiated with γ rays at E14 (0.1 and 0.5 Gy) and E17 (0.1 and 0.5 Gy) and P1 (0.1 and 0.5 Gy) and P4 (0.1 and 0.5 Gy) and for the nonirradiated group were reproduced from a previously reported study (29). Plots for the P1 group are the same as in Fig. 3. *P < 0.01, **P < 0.05 vs. nonirradiated group by log-rank test.