The fundamental pathophysiology of CRI diverges from ARS (19). While ARS is driven by immediate, massive cellular ablation following high-dose-rate insult, CRI is the consequence of protracted exposure to LDR ionizing radiation (20). The biological outcome is governed not only by the total absorbed dose but also by DR (21). Under LDR conditions, the continuous induction of DNA damage engages adaptive response mechanisms, however, the persistent stress eventually overwhelms cellular fidelity, leading to the accumulation of sublethal damage rather than immediate cell death (22).

Populations at risk extend beyond traditional nuclear industry workers to include interventional radiologists, astronauts exposed to galactic cosmic rays and residents in high background radiation areas (23). Unlike the rapid onset of acute symptoms, CRI is characterized by a latent window, in which sub-clinical metabolic reprogramming and oxidative fluctuations occur prior to symptomatic manifestation (24). This progression complicates early risk stratification, rendering traditional biodosimetry markers ineffective and posing a challenge for precision radiation protection (Table I) (25).

CRI is a systemic disease, propagated by complex inter-organ signaling rather than isolated tissue injury. The hematopoietic system serves as the most sensitive biological indicator and primary responsive target of LDR toxicity (26). Beyond direct DNA damage to hematopoietic stem cells (HSCs), chronic exposure remodels the bone marrow microenvironment (27). Radiation-induced stromal senescence and adipogenesis (fatty marrow) create a hostile niche that fails to support HSC self-renewal, leading to HSC exhaustion, persistent cytopenia and a pre-leukemic state of clonal instability (28). The immune system undergoes a paradoxical shift termed inflammaging. While radiation depletes naïve lymphocyte pools (immunosenescence), damage-associated molecular patterns released from stressed cells trigger the cGAS/STING pathway, driving a sustained release of pro-inflammatory cytokines (IL-6, TNF-α) (29). This chronic low-grade inflammation not only impairs pathogen clearance but also creates a tumor-permissive microenvironment (30). The central nervous system (CNS), previously considered radioresistant, exhibits vulnerability through neuroinflammation and the inhibition of hippocampal neurogenesis, manifesting as cognitive rigidity (diminished mental flexibility and impaired adaptation to changing environmental demands) and autonomic dysregulation (31). Furthermore, reproductive toxicity involves germline DNA fragmentation and endocrine axis disruption, creating a cycle of systemic physiological decline amplified by oxidative stress (32).

The molecular hallmark of CRI is genomic instability, a state where the genome acquires an increased tendency to alteration. Unlike the transient reactive oxygen species (ROS) burst in acute exposure, CRI establishes a cycle of chronic oxidative stress driven by mitochondrial dysfunction (33). Leakage of electrons from damaged electron transport chains generates persistent ROS, causing continuous oxidative base damage-specifically the formation of 8-oxo-7,8-dihydro-2'-deoxyguanosine (34). Damage is not confined to directly irradiated cells (35). Through the radiation-induced bystander effect, irradiated cells transmit distress signals (via exosomes and gap junctions) to non-irradiated neighbors, propagating genomic instability (36). Studies highlight the role of epigenetic drift (aberrant DNA methylation and histone modifications) in locking cells into a dysfunctional state (37,38). These heritable epigenetic marks explain why radiation effects persist after exposure ceases, potentially affecting unexposed offspring via transgenerational inheritance (Fig. 2) (38).

A key long-term consequence of CRI is the elevation of stochastic carcinogenic risk (39). According to the LNT model, chronic accumulation of DNA mis-repairs increases the probability of malignant transformation (40).

Under continuous immune pressure and oxidative stress, somatic cells with advantageous mutations (in TP53 or DNA methyltransferase 3A) may undergo clonal expansion, a phenomenon known as clonal hematopoiesis of indeterminate potential (41). This clonal evolution serves as a precursor to radiation-induced leukemias and solid tumors (42).

Compared with spontaneous cancer, radiation-associated malignancies exhibit distinct mutational signatures and extended latency periods (43). The spectrum includes leukemia, thyroid papillary carcinoma and lung fibrosis-associated cancers (44). These mechanistic insights underscore the need for refined radiobiological models that recapitulate these stochastic events to guide international occupational safety standards (Fig. 3; Table II).

External beam irradiation is the cornerstone of radiobiological modeling, categorized by the extent of anatomical coverage. Whole-body irradiation (WBI) is the standard for simulating systemic catastrophe, such as large-scale nuclear accidents or prolonged occupational exposure in unshielded environments (49). By delivering a homogeneous dose across all physiological compartments, WBI effectively reproduces the syndromic nature of RI, capturing the interplay between bone marrow failure (hematopoietic syndrome) and systemic immune collapse (50). A challenge in WBI is achieving dosimetric homogeneity (51). Variations in animal geometry can lead to hot spots (overdose) or cold spots (underdose), potentially confounding survival data (52). Furthermore, WBI typically induces lethal acute GI toxicity before chronic fibrotic phenotypes manifest (53). To address this, fractionated low-dose regimens are employed to allow animals to survive the acute phase and develop late-onset pathology (54). To mimic scenarios such as localized radiotherapy toxicity or partial shielding, partial body irradiation selectively targets specific organs (thorax, brain) while sparing the bone marrow (55). Research has moved beyond lead shielding to use small animal radiation research platforms (56). These systems employ micro-CT guidance to deliver conformal radiation beams with sub-millimeter precision (57). This allows the study of organ-specific radiosensitivity (radiation-induced lung fibrosis) without the confounding variable of systemic hematopoietic collapse (Fig. 4) (58).

Emerging evidence suggests CRI is not solely the result of direct energy deposition but is also mediated by NTEs (59). Models of indirect irradiation focus on how radiation signals are propagated through the microenvironment (60-63). Bystander and abscopal models investigate how irradiated cells communicate with non-irradiated neighbors via gap junctions, exosomes and soluble factors (ROS, cytokines) (61). By irradiating a specific volume (such as the liver) and assessing damage in distant organs (such as the lung), researchers can elucidate the systemic propagation of inflammatory signals (62). In ecological toxicology contexts, indirect exposure also refers to the interaction with irradiated media (water or soil) (63). These models are essential for distinguishing between direct radiotoxicity and the secondary toxicity of radiation-induced chemical species (radiolysis products) (64).

While external beams simulate prompt exposure, internal contamination is the most biologically faithful model for occupational hazards (such as uranium mining) or post-accident fallout (65). The pathology is dictated by the chemical nature of the radionuclide (66). For example, osteotropic isotopes (⁹⁰Sr, ²²⁶Ra) accumulate in the bone matrix, causing chronic marrow suppression and osteosarcoma, whereas radioiodine (¹³¹I) targets the thyroid (67). These models require precise calculation of body burden (the total amount of a radionuclide present in the organism at a given time) and effective half-life (68). Internal models are key for studying high-LET radiation (α-particles from plutonium or radon) (69). Unlike low-LET γ-rays, α-particles cause dense ionization tracks and complex DNA double-strand breaks that are refractory to repair (70). This biological insult highlights the limitation of using external X-rays to mimic internal α-emitter contamination (71).

To bridge the gap between laboratory conditions and real-world catastrophes, advanced models integrate multiple stressors (72,73). Real-world exposure typically co-occurs with trauma or burns (73-75). Specialized facilities-such as gamma gardens (open-field irradiation facilities equipped with a central radioactive source) or prolonged housing near radioactive sources (⁶⁰Co or ¹³⁷Cs)-allow continuous exposure over months (0.05-0.50 Gy/week) (75). These models are suited to test the dose-rate effect (the phenomenon where the biological effectiveness of a radiation dose decreases as the rate of delivery is reduced, primarily due to ongoing cellular repair), providing key data for setting occupational safety thresholds and validating environmental risk models (Table III) (76).

Animal models of CRI are defined by a distinct phenotypic landscape that diverges from acute syndromes (77). These models capture the progressive, multidimensional failure of organ systems driven by the continuous accumulation of sublethal damage and the propagation of SASP (77,78).

The hematopoietic system serves as the primary detector of radiation toxicity (79). Under LDR exposure, the pathology shifts from acute ablation to hematopoietic exhaustion (80). Continuous irradiation imposes replicative stress on HSCs, forcing quiescent cells into the cell cycle to maintain homeostasis (81). This chronic cycling leads to telomere erosion and premature senescence, evidenced by the downregulation of stemness markers (c-Kit, CD34) and lineage skewing toward myelopoiesis (82).

Damage extends to the bone marrow microenvironment (83). Radiation damages mesenchymal stromal cells and sinusoidal endothelium, altering the cytokine milieu (decreased CXCL12, increased TGF-β) (84). This hostile niche fails to support HSC retention and promotes adipogenic differentiation, perpetuating a cycle of pancytopenia and immune incompetence akin to human myelodysplastic syndrome (85).

Chronic radiation accelerates immunosenescence, characterized by thymic involution and a functional paralysis of the adaptive immune repertoire (86). Peripheral T cell pools become oligoclonal, dominated by exhausted memory phenotypes (PD-1⁺) with decreased proliferative capacity (87). Simultaneously, the innate immune system enters a state of hyper-activation known as inflammaging (88). Irradiated macrophages activate the NLRP3 inflammasome and the cGAS/STING pathway, secreting a constant stream of pro-inflammatory cytokines (IL-1β, IL-6) (89). This immune dysregulation creates a permissive microenvironment that impairs pathogen clearance and reduces immunosurveillance against neoplastic transformation (90).

The CNS exhibits a delayed response to chronic radiation (91). The primary mechanism is neuroinflammation mediated by activated microglia (92). Chronic exposure prevents the turnover of hippocampal neural progenitor cells, leading to deficits in neurogenesis essential for memory consolidation (93). Structurally, this manifests as synaptic stripping, the loss of dendritic spines in the prefrontal cortex (94).

Physiologically, these changes translate into sickness behaviors, such as lethargy, anxiety-like thigmotaxis and cognitive rigidity (95). Furthermore, radiation disrupts the hypothalamic-pituitary-adrenal axis, leading to circadian dysregulation (96). This neuro-immune-endocrine crosstalk highlights the systemic nature of CRI, where CNS injury exacerbates peripheral immune suppression (Fig. 5) (97).

The utility of any animal model is determined by the balance between its face (phenotypic similarity to humans) and construct validity (mechanistic similarity). Compared with fractionated acute doses, LDR sources more accurately replicate the continuous radiation exposure patterns associated with occupational hazards and environmental contamination (98). LDR models capture the abscopal interactions between organs (such as the kidney-heart axis), allowing for the evaluation of holistic therapeutics rather than single-organ protectants (99). The latency of CRI requires extended housing (6-24 months), increasing costs (100). The confounding factor of aging becomes critical; distinguishing radiation effects from natural geriatric decline requires robust age-matched controls (101). Rodents possess different metabolic rates and DNA repair kinetics than humans (102). For example, the lethal dose required to kill 50% of the population within 30 days) varies, complicating the direct extrapolation of dose-response associations (103).

Selecting the appropriate model involves ensuring a match between the specific biological mechanisms under investigation and the inherent physiological attributes of the chosen animal (104).

Murine models are the workhorse of mechanistic radiobiology (105). C57BL/6 mice are preferred for fibrosis and inflammation studies (Th1-dominant), while BALB/c mice are used for solid tumor induction (Th2-dominant) (106). While useful for molecular genetics, their small size makes localized organ dosimetry challenging (107). Rats offer a larger physiological volume, facilitating surgical interventions and serial blood sampling without inducing hypovolemic stress (108). They are superior for cardiovascular and renal radiation toxicity models due to hemodynamics closer to those of humans (109). Large animal models (canine/porcine/non-human primates) are the gold standard for preclinical validation. Porcine skin and gastrointestinal tracts are anatomically nearly identical to humans, making them ideal for cutaneous and visceral RI models (110). Non-human primates are essential for pivotal studies assessing complex neurobehavioral outcomes and sophisticated immune responses due to high genetic homology (Fig. 6; Table IV) (111).

Genetic background dictates the trajectory of chronic injury (112). While inbred strains (C57BL/6) offer reproducibility, they fail to capture human heterogeneity. Strains such as ataxia telangiectasia mutated-deficient or p53-heterozygous mice accelerate the onset of stochastic effects such as carcinogenesis, compressing decades of human latency into months (113). One approach involves using diversity outbred populations to map the genetic architecture of individual radiosensitivity, mimicking the diverse human response more effectively than traditional, genetically homogeneous inbred strains (114). Humanized mouse models, such as the NSG-SGM3 strain [NOD.Cg-Prkdc^scid Il2^rgtm1Wjl Tg(Prfg-IL3,CSF2,KITLG)1Eav/MloySzJ, which expresses human cytokines to support myeloid engraftment], reconstituted with human HSCs allow the study of human immune responses to chronic radiation in vivo, bridging the translational gap (61,89,115).

Precision in CRI modeling requires rigorous dosimetry (116). The definition of chronic varies; thus, reporting the precise dose rate (Gy/h) is as critical as the total accumulated dose (117). Isodose cages ensure that animals receive uniform exposure regardless of movement (118). Furthermore, adherence to ARRIVE guidelines ensures husbandry factors (microbiome, circadian light cycles) are controlled to isolate the radiation variable (119).

Physiological metrics serve as the frontline indicators of systemic stress and sickness behavior, typically preceding clinically detectable organ failure (123). Unlike the rapid weight loss seen in acute syndrome, chronic radiation induces a wasting phenotype akin to cancer cachexia (124). Longitudinal monitoring typically reveals a blunted growth trajectory or gradual BMI decline, driven by hypothalamic inflammation and sustained catabolic signaling (125). Clinical frailty index, an aggregate score of deficits (alopecia, gait anomalies, grip strength), is a superior metric for quantifying radiation-induced accelerated aging compared with body weight alone (126). Changes in feeding behavior reflect the disruption of the gut-brain axis (127). Radiation-induced mucositis and hypothalamic dysregulation manifest as anorexia and polydipsia (128). A sustained 20-30% decline in caloric intake suggests a transition from transient stress to chronic metabolic dysregulation (129). Quantitative behavioral assays are key for validating CNS injury (130). Automated tracking using open field test and Morris water maze detects decreased locomotor velocity (lethargy), anxiety-like thigmotaxis and cognitive deficits in spatial memory (131). These behavioral phenotypes are associated with hippocampal neuroinflammation and serve as non-invasive biomarkers for successful model induction (132).

Peripheral blood analysis provides a dynamic liquid biopsy of the bone marrow microenvironment (133). In CRI models, the validation endpoint is lineage skewing (a persistent imbalance in the production of different hematopoietic cell types that reflects underlying marrow dysfunction and stem cell injury. While acute exposure causes rapid pancytopenia, chronic exposure typically leads to a myeloid-biased output (increased granulocyte/lymphocyte ratio) at the expense of lymphopoiesis, a hallmark of hematopoietic aging (134). Progressive, refractory normocytic anemia (hemoglobin <100 g/l) and elevated red cell distribution width serve as diagnostic criteria for cumulative marrow failure (135). Flow cytometric profiling must assess functional status (136). Validated models demonstrate an inverted CD4/CD8 ratio and the expansion of regulatory T cells (137,138). Upregulation of exhaustion markers on T cells-specifically programmed cell death protein 1 (PD-1) and T-cell immunoglobulin and mucin domain-containing protein 3, along with decreased natural killer cell cytotoxicity, provide definitive immunophenotypic evidence of chronic radiation stress (138).

Histopathology remains the gold standard for defining irreversible tissue remodeling (139). However, modern validation requires specific staining and digital quantification beyond standard hematoxylin and eosin staining (140). A pathognomonic feature of CRI is the replacement of hematopoietic cellularity with adipose tissue (141). Validated models demonstrate a quantifiable shift in the marrow adipocyte-to-hematopoietic cell ratio (142-144). A hallmark of late-stage CRI is the excessive deposition of extracellular matrix (143). Validation should utilize Masson's trichrome or Sirius Red staining to visualize collagen (144). Immunohistochemistry for α-smooth muscle actin (identifying myofibroblasts) is key for distinguishing active fibrogenesis from static scarring (145). To enhance reproducibility, studies employ AI-driven image analysis algorithms to score fibrosis area fraction and cellularity, eliminating inter-observer variability inherent in manual scoring (145,146).

Phenotypic observation must be corroborated by molecular mechanisms (147). The persistence of DNA double-strand breaks is the molecular footprint of radiation (148). Immunofluorescence quantification of phosphorylated histone H2AX and p53-binding protein 1 (53BP1) foci-two established biomarkers of DNA double-strand breaks-in circulating lymphocytes serves as a biodosimeter (149). Persistent upregulation of repair genes, such as ATM and X-ray repair cross-complementing protein 1, indicates sustained genomic stress and failed repair kinetics (150). Chronic injury is maintained by the SASP (151). Validated models must demonstrate a specific cytokine signature in plasma: Elevated IL-6, IL-1β and MMPs, coupled with reduced IGF-1 (152). This biochemical profile links molecular senescence to macroscopic fibrosis (153). Emerging validation strategies use cell-free DNA and radiation-responsive miRs (miR-150, miR-34a) as minimally invasive markers of tissue damage, offering a translational bridge to human clinical diagnostics (153,154).

A model is considered validated only when it satisfies this multidimensional matrix: Exhibiting the phenotypic frailty, hematological skewing, histological fibrosis and the molecular SASP signature characteristic of human CRI pathology (Fig. 7; Table V) (155).

Animal models of chronic radiation injury allow the mapping of complex signal transduction networks that drive chronic pathology (158,159). Chronic LDR exposure triggers a senescence-like arrest distinct from acute apoptosis (159). Models have been pivotal in identifying the p53-p21 axis as the gatekeeper of this arrest and the subsequent activation of NF-κB, which drives the SASP (160,161). This demonstrates how irradiated cells survive but remain metabolically active, propagating inflammation to bystander tissue (161). epigenetic scars (persistent, long-term alterations in the epigenome (such as stable changes in DNA methylation patterns) that continue to drive abnormal gene expression long after the initial radiation exposure has ceased (162,163). High-throughput bisulfite sequencing in chronically exposed rodents has identified persistent hypermethylation of tumor suppressor promoters (p16^INK4a) and global hypomethylation of retrotransposons (163). Integrative multi-omics (transcriptomics + metabolomics) has mapped the dysregulation of the tricarboxylic acid cycle and fatty acid oxidation, providing a systems-level view of radiation-induced metabolic rewiring (164).

The primary industrial application of animal models of chronic radiation injury is the rigorous screening of MCMs, distinguishing between prophylactic agents (pre-exposure) and mitigators (post-exposure) (165). In the screening phase, models quantify the dose reduction factor of candidate agents (166). Effective agents must demonstrate significant preservation of the HSC pool and a reduction in marrow adiposity (167). Research has moved beyond free radical scavengers to targeted therapy (168). Models are validating senolytics (navitoclax, quercetin) that selectively eliminate senescent cells to rejuvenate tissue (168,169). Furthermore, pharmacodynamic studies optimize dosing to mitigate DEARE, ensuring that agents do not interfere with DNA repair in a manner that promotes secondary carcinogenesis (Table VII) (57,170).

Beyond pharmaceuticals, models are key for testing holistic strategies suitable for long-term occupational exposure (171). Comparative studies demonstrate that dietary polyphenols and probiotics remodel the gut microbiome, which is often dysbiotic after radiation (171,172). This gut-bone marrow axis modulation enhances Nrf2-ARE signaling, the master regulator of antioxidant defense, thereby decreasing systemic oxidative stress (malondialdehyde levels) (173). Controlled exercise studies reveal that mechanical loading upregulates AMPK signaling, promoting mitochondrial biogenesis (174,175). This counteracts radiation-induced sarcopenia and fatigue, offering a non-pharmacological strategy for astronauts or nuclear workers (175).

Animal models provide the empirical data necessary to construct AOPs, which link a molecular initiating event (such as DNA oxidation) to an adverse outcome (such as fibrosis). By exposing large cohorts to low-dose rates, researchers test the validity of the LNT model vs. hormesis (adaptive response) (176,177). This data is critical for setting occupational limits (such as for interventional cardiologists) (178). Real-world risk involves combined stressors (179). Advanced models simulate the exposome by co-exposing animals to radiation in the presence of heavy metals (such as uranium mining) or microgravity (spaceflight) (178,179). These studies reveal synergistic toxicities that single-stressor models miss, refining ecological risk assessments (179,180).

The future of CRI animal models lies in precision radioprotection (181). By using diversity outbred mice that reflect human genetic heterogeneity, researchers can identify genetic polymorphisms (in ATM or TGF-β) that confer individual radiosensitivity (182). Furthermore, the integration of spatial transcriptomics allows the mapping of radiation injury at single-cell resolution in situ, identifying rare, radio-resistant SC subpopulations that drive tissue regeneration or fibrosis (183,184). These refined models may facilitate the development of personalized radiation health passports, guiding individual risk management based on genetic and metabolic profiles (183-185).

Future mechanistic inquiry will transcend bulk tissue analysis to identify the spatiotemporal dynamics of injury (189). By integrating single-cell RNA sequencing with spatial transcriptomics, researchers may move beyond merely identifying which cells are damaged to understanding where they are located relative to the niche (190). This will map the neighborhood effects (how a senescent stromal cell spatially corrupts adjacent stem cells via SASP) (191). Advanced genetic fate-mapping will unveil the clonal trajectories of surviving cells, distinguishing between regenerative and pre-leukemic clones (192). This high-resolution mapping may identify specific molecular switch points (such as epigenetic loci) that determine whether a tissue undergoes repair or irreversible fibrosis (193).

The therapeutic landscape is shifting from symptom management to disease modification (194). A notable frontier is the targeting of senescent cells (195). Future models will rigorously validate senolytics (drugs that induce apoptosis in senescent cells) and senomorphics (agents that suppress SASP without killing the cell) (196). The goal is to restore the irradiated microenvironment to a pre-exposure state (197). While SC therapy holds promise, the future may lie in extracellular vesicles (EVs) and exosomes (198). Derived from mesenchymal SCs, these cell-free cargoes carry regenerative miRs and proteins with lower immunogenicity and tumorigenic risk than live cells (199). Models optimize engineered EVs loaded with specific radioprotective cargos (mitochondria or mRNA) for targeted delivery (199,200). Clustered regularly interspaced short palindromic repeats-associated protein 9 and base editing in vivo will evolve toward correcting radiation-induced mutations in somatic tissue or epigenetically silencing pro-fibrotic genes (TGF-β) via modifiable promotors (201).

AI may transform experimental design and risk assessment (202). Machine learning algorithms may replace manual observation, detecting subtle neurobehavioral deficits (gait asymmetry, micro-tremors) invisible to the human eye, increasing the sensitivity of chronic injury models (203). By integrating large omics datasets from animal studies, researchers aim to create digital twins of radiation injury (204). These in silico models can simulate decades of chronic exposure in seconds, predicting long-term outcomes and screening millions of drug candidates virtually before an animal is treated (205). This may accelerate MCM discovery.

Hybrid systems may overcome the interspecies gap (206). The use of immunodeficient mice reconstituted with human hematopoietic and immune systems (MISTRG mice) may become standard for assessing human-specific immune responses to chronic radiation (207). Integration of MPSs (organs-on-chips) with animal validation may enable transition from high-throughput in vitro mechanistic screening to definitive in vivo validation (208). MPS screen for human-specific toxicity mechanisms, while the animal model confirms systemic safety (209). This aligns with the replacement, reduction and refinement principle while maximizing translational relevance for diverse populations, from astronauts to nuclear workers (210).

While the present review provides a methodological framework for CRI modeling, limitations must be acknowledged. First, due to the logistical, ethical and financial constraints of conducting life-span studies in large animal models (such as non-human primates), the majority of the mechanistic data, particularly regarding SASP and inflammaging, is derived from murine models. Extrapolating these rodent-derived mechanisms to humans requires extreme caution, given the species-specific differences in telomere dynamics, metabolic rate and immune surveillance. Second, experimental models traditionally isolate radiation as a single stressor to maintain controlled variables. However, real-world human exposure (spaceflight or nuclear accidents) typically involves complex exposome interactions, such as concomitant exposure to heavy metals, psychological stress or microgravity, which may alter the CRI trajectory. Finally, spatial multi-omics and AI-driven digital twins are currently in their early developmental or conceptual phases and require empirical validation before widespread translational application.

In conclusion, CR animal models are transitioning from descriptive pathology to predictive precision medicine (211). This may facilitate development of robust, personalized defenses against adverse health effects of ionizing radiation (212).