Clonal hematopoiesis related to ionizing radiation in atomic bomb survivors

Clonal hematopoiesis (CH) is characterized by the clonal expansion of hematopoietic stem and progenitor cells harboring somatic genomic alterations that confer a competitive fitness advantage. The prevalence of CH increases with age, and CH harboring mutations in genes commonly recognized as drivers of myeloid neoplasms (MNs) is known to predispose individuals to the subsequent development of hematologic neoplasms [ 1 ]. Exposure to ionizing radiation is associated with an increased risk of carcinogenesis. Epidemiological studies of atomic bomb survivors (ABS) have demonstrated elevated risks of leukemia and myelodysplastic neoplasms (MDS), with the excess risk persisting for more than 50 years after exposure, particularly for acute myeloid leukemia (AML) and MDS [ 2 , 3 ]. Given the persistently elevated risk of MDS among ABS decades after exposure, we hypothesized that the characteristics of CH in otherwise healthy ABS differ from those in non-exposed individuals. To investigate this possibility, we evaluated the landscape of CH in ABS. Whole-exome sequencing (WES) was performed on peripheral blood samples obtained from 80 proximally exposed survivors (within 1 km of the hypocenter) and 20 non-exposed individuals residing in Nagasaki Prefecture (Control-1, lower recruitment than anticipated). Single-nucleotide variants (SNVs) and insertions/deletions (INDELs) with a variant allele frequency (VAF) of ≥2% were compared between the two groups. In particular, for 171 candidate driver (CD) genes implicated in MNs [ 4 , 5 , 6 ], we assessed the prevalence of mutations and the association with the clone size ( Supplementary Methods ). None of the participants in either group had a history of malignancy or hematologic disease. Among the 80 ABS, 48 (60%) had experienced acute radiation-related symptoms (ARS) following exposure, including nausea, vomiting, diarrhea, alopecia, and fever (Table 1 ). Table 1 Main features of each group. Full size table Copy-number alteration (CNA) analysis using WES data was feasible in ABS but failed to generate adequately reliable results in Control-1 despite the application of multiple algorithms, and the underlying cause could not be identified. Therefore, we established an additional comparison cohort (Control-2) by randomly selecting 2400 age- and sex-matched individuals from BioBank Japan, a large-scale Japanese biobank resource [ 7 , 8 ]. Individuals with a history of hematologic disease were excluded from this cohort. For Control-2, data were available for targeted sequencing of 22 representative driver genes (all included among the aforementioned CD genes) as well as SNP array–based CNA analysis ( Supplementary Methods ) [ 8 ]. The number of putative somatic SNVs and INDELs identified by WES was 530 in the ABS cohort and 109 in the Control-1 cohort, with DNMT3A being the most frequently mutated gene (Table 1 , Supplementary Tables 2 , 3 ). The prevalence of CH (VAF ≥ 2%) was comparable between the two cohorts (85% vs. 75%, respectively; Table 1 ). Likewise, the median number of mutations per individual did not differ significantly (4 [range, 0–32] vs. 3 [0–21], respectively; p = 0.51; Supplementary Fig. 1A–C ). Because CH with CD gene mutations is associated with an increased risk of MNs, we expected a higher prevalence of such clones in ABS. However, CD-mutated CH was detected in 22.5% of ABS and 20.0% of Control-1 individuals, with no significant difference between the cohorts (Table 1 ). The spectrum of mutated CD genes was also similar between the cohorts. Among ABS, the experience of ARS was not associated with the prevalence of CH or CD gene mutations (Supplementary Fig. 1D–F ). Although CH prevalence was similar between the groups, clone sizes were significantly larger in the ABS cohort than in the Control-1 cohort (median VAF, 4.9% [range, 2.7–31.1%] vs. 4.4% [2.7–17.1%], p = 0.033; Fig. 1A ). In the Control-1 cohort, CH harboring CD gene mutations exhibited significantly larger clone sizes than CH lacking such mutations (Fig. 1B ; p = 0.016), whereas this difference was not observed in the ABS cohort ( p = 0.2). Instead, the increase in clone size observed in the ABS cohort was largely attributable to CH lacking CD gene mutations, suggesting that mechanisms independent of CD gene mutations may promote clonal expansion. We next compared CH in ABS with that in Control-2 using a panel of 22 selected driver genes. Mutations in these genes were identified in 10% of ABS and 20% of Control-2 individuals, representing a significant difference (Table 1 ). Examination of individual genes revealed that mutations in DNMT3A and TET2 were more frequent in Control-2 than in ABS, a trend that was also observed in Control-1 (Table 1 , Supplementary Fig. 2A, B ). Consistently, the frequency of DTA mutations ( DNMT3A , TET2 , and ASXL1 ) was significantly higher in Control-2 than in ABS (Table 1 ). This pattern is reminiscent of the molecular characteristics that we previously reported in MDS among ABS [ 9 ]. Fig. 1: Characteristics of clonal hematopoiesis (CH) in atomic bomb survivors (ABS). Full size image A Distribution of clone sizes, measured by variant allele frequency (VAF), in the ABS and Control-1 cohorts, which was significantly larger in the ABS group ( p = 0.033). Variants on sex chromosomes and within copy number-altered regions were excluded to enable accurate VAF comparison. Horizontal lines indicate median VAFs. B Clone sizes according to the presence or absence of candidate driver (CD) gene mutations (mut) in the ABS and Control-1 cohorts. C Genomic landscape of CH in the ABS cohort. The heatmap shows copy number alterations and selected mutated genes. The bar plot above the heatmap indicates the number of SNVs and INDELs in each subject. Subjects who subsequently developed myeloid neoplasms and those harboring CD gene mutations are indicated. USN is shown on the x-axis. D Longitudinal changes in hemoglobin levels in individuals by the presence of CH, further stratified by CD gene mutation status. Subjects with CH exhibited a significant decreasing trend ( p = 0.045), which was more pronounced in those with CH harboring CD gene mutations ( p < 0.001), but not in those with only non-CD gene mutations ( p = 0.278). The x-axis shows time (years), with 0 representing the time of sample collection; negative and positive values indicate time before and after sample collection, respectively. E Overall survival of the ABS cohort from the time of sample collection according to CH status. VAF variant allele frequency, ABS atomic bomb survivors, CD candidate driver, mut mutation, CH clonal hematopoiesis, CNA copy number alteration, USN unique subject number, SNV single nucleotide variant, INDEL insertion/deletion, MDS myelodysplastic neoplasms, CN-LOH copy-number neutral loss of heterozygosity, MN myeloid neoplasms, ET essential thrombocythemia, Chr chromosome, Hb hemoglobin. Because of the limitations described above, CNA analyses were performed by comparing ABS with Control-2. Overall, CNAs were detected significantly more frequently in ABS than in Control-2 (18.8% vs. 1.6%, respectively; Table 1 ). Among individual lesions, del(20q) showed the largest difference, being identified in 8.8% of ABS compared with 0.2% of Control-2 individuals ( p < 0.001; Fig. 1C , Supplementary Fig. 2C ). Significant differences were also observed for del(5q) and del(11q) (Table 1 ). When CH was defined by the presence of SNVs/INDELs and/or CNAs, the overall prevalence of CH in ABS was 86.3% using WES-derived mutation data combined with CNA data. Using data for the selected 22 genes together with CNA data, CH was identified in 28.8% of ABS and 21.0% of Control-2 individuals, indicating no substantial difference in overall prevalence. However, the relative contributions of SNVs/INDELs and CNAs differed markedly between the two groups, suggesting distinct patterns of genomic alterations underlying CH in ABS and non-exposed individuals (Table 1 ). During a median follow-up period of 13 years, three ABS developed MNs: two cases of MDS (unique subject number [USN]-010 and USN-026, diagnosed 12 years and 2 years after sample collection, respectively) and one case of essential thrombocythemia (ET; USN-033, diagnosed 12 years after sample collection) (Fig. 1C ). No CD gene mutations were detected in any of these individuals at the time of CH assessment, including the JAK2 V617F mutation in USN-033, which was identified at the time of ET diagnosis. Previous studies have shown that JAK2 -mutated clones may be detectable many years before the diagnosis of ET [ 10 ]; therefore, the absence of detectable JAK2 mutation in our analysis may reflect a very low VAF at the time of sampling. In Control-2 ( n = 2110), eight hematologic malignancies were reported during a median follow-up of 5.9 years, including three cases of MDS (diagnosed 12.2, 6.8, and 3.6 years after sample collection), one AML (4.9 years), one chronic myeloid leukemia (3.2 years), and three malignant lymphomas (9.6, 5.3, and 1.6 years). For the ABS cohort, longitudinal health examination data obtained before sample collection were available for up to 27 years in many individuals. Analysis of serial hematologic parameters ( Supplementary Methods ) revealed a gradual but significant decline in hemoglobin levels over time among CH-positive individuals. This decline was most pronounced in those harboring CD gene mutations, whereas no significant change was observed in individuals with only non-CD gene mutations (Fig. 1D ). In contrast, no comparable longitudinal changes were observed in WBC or PLT (Supplementary Fig. 3 ). We further examined the association between CH and survival in ABS, demonstrating a trend toward inferior overall survival ( p = 0.055; Fig. 1E ). Among the 23 recorded deaths, causes were available for 19 individuals and were largely unrelated to hematologic malignancies; age-related events ( n = 5; bathing accidents and aspiration), cardiovascular events ( n = 5; heart failure, arrhythmia, myocardial infarction, and cerebral infarction), solid malignancies ( n = 4), infections ( n = 3), and others ( n = 2; interstitial pneumonia and acute pancreatitis). In this study, the prevalence of CH in ABS was comparable to that observed in both control cohorts, despite the long-term excess risk of MNs in ABS. Likewise, WES did not reveal an increased burden of SNVs or INDELs in ABS. These findings suggest that the persistent susceptibility to MNs in ABS cannot be explained simply by an increased frequency of conventionally defined CH, or mutations of CD genes. In contrast, CNAs were markedly enriched in ABS, particularly del(20q), highlighting structural genomic alterations as a distinctive feature of CH in this population. Given that ionizing radiation induces DNA double-strand breaks, the increased prevalence of CNAs may represent a long-lasting genomic consequence of radiation exposure and a potential mechanistic link to the elevated risk of MNs. Although del(20q) is commonly observed in age-related CH and MDS and has been associated with subsequent MNs [ 11 , 12 , 13 ], the basis for its selective enrichment in ABS remains unclear. Another notable finding was the low frequency of DTA mutations in ABS. This pattern mirrors the molecular characteristics previously observed in MDS arising among ABS [ 9 ] and suggests that radiation-associated clonal evolution may differ from the canonical age-related pathway dominated by DTA mutations. Whether this reflects distinct selective pressures on hematopoietic stem cells or differences in the spectrum of radiation-induced genomic lesions warrants further investigation. Together, these findings suggest that radiation-associated CH follows a distinct evolutionary trajectory characterized by enrichment of structural genomic alterations rather than canonical age-related driver mutations. Future studies incorporating ultra-deep sequencing and lower VAF thresholds may further clarify the unique clonal architecture and clinical consequences of CH in radiation-exposed individuals.