Total-Body Irradiation Alters White Matter Volume and Microstructural Integrity in Rhesus Macaques

Abstract

Purpose:

Long-term survivors of brain irradiation can experience irreversible injury and cognitive impairment. T1-weighted and diffusion tensor magnetic resonance imaging (MRI) are used to evaluate brain volume and white matter (WM) microstructure in neurodevelopmental and neurodegenerative conditions. The goal of this study was to evaluate the long-term effects of single-dose total-body irradiation (TBI) or TBI with 5% partial-body sparing on brain volumetrics and WM integrity in macaques.

Methods and Materials:

We used MRI scans from a cohort of male rhesus macaques (age range, 3.6–22.8 years) to compare global and regional brain volumes and WM diffusion in survivors of TBI (T1-weighted, n = 137; diffusion tensor imaging, n = 121; dose range, 3.5–10 Gy) with unirradiated controls (T1-weighted, n = 48; diffusion tensor imaging, n = 38).

Results:

In all regions of interest, radiation affected age-related changes in fractional anisotropy, which tended to increase across age in both groups but to a lesser extent in the irradiated group (interaction P < .01). Depending on the region of interest, mean diffusivity decreased or remained the same across age in unirradiated animals, whereas it increased or did not change in irradiated animals. The increases in mean diffusivity were driven by changes in radial diffusivity, which followed similar trends across age. Axial diffusivity did not differ by irradiation status. Age-related changes in relative volumes in controls reflected normal trends in humans, with increasing WM and decreasing gray matter until middle age. Cerebrospinal fluid (CSF) volume did not differ across age in controls. WM volume was lower and CSF volume was higher in young irradiated macaques. WM volume was similar between groups, and CSF volume lower in older irradiated macaques. Gray matter volume was unaffected by radiation.

Conclusions:

TBI results in delayed WM expansion and long-term disruption of WM integrity. Diffusion changes suggest that myelin injury in WM is a hallmark of late-delayed radiation-induced brain injury.

Introduction

Historically, the brain has been considered a radioresistant organ, requiring tens of gray (Gy) to induce clinical abnormalities.¹ Radiation-induced brain injury (RIBI) is associated with high cumulative doses as seen with radiation therapy and high single-dose irradiation (>15 Gy) that results in brain lesions² and cognitive impairment in humans, rodents, and nonhuman primates (NHPs).^3–5 However, there is increasing evidence that exposures of less than 10 Gy also result in cognitive deficits.^6,7 In the event of a nuclear accident or attack, many people would be exposed to radiation at survivable doses (<2 Gy), less than those traditionally believed to result in neurologic impairment,⁸ highlighting the importance of long-term neurologic evaluation after low dose exposures.

Neurologic deficits after radiation exposure can occur in the absence of major anatomic abnormalities.⁹ Liquidators of the Chernobyl disaster developed electroencephalographic changes,¹⁰ cerebrovascular disease,¹¹ and cognitive dysfunction¹² years after the cleanup, during which they were exposed to estimated doses of 0.1 to 5.9 Gy. Previous work involving NHPs has shown cognitive impairment in NHPs exposed to total-body doses between 4 and 8 Gy.¹³ Additionally, NHPs exposed to 4 to 8.5 Gy develop a spectrum of microvascular lesions months to years after total-body irradiation (TBI), similar to human radiation therapy patients, which are detectable with susceptibility-weighted magnetic resonance imaging (MRI).¹⁴ These animals also develop transcriptomic signatures similar to but less severe than those exposed to high doses (40 Gy fractionated whole brain irradiation) but show no MRI-detectable macroscopic changes.¹⁵ This demonstrates that the brain is susceptible to injury at doses that do not result in overt radiographic lesions and that quantitatively evaluating the brain using advanced imaging techniques may have utility in detecting RIBI.

Structural MRI uses noninvasive techniques to evaluate the anatomy and microstructural integrity of the central nervous system. Modalities including T1-weighted and diffusion tensor imaging (DTI) have allowed researchers and clinicians to better understand the effects of high-dose brain irradiation (>30 Gy) on peritumoral brain tissue.^16–18 However, studies of human patients are confounded by the effects of chemotherapy and primary or metastatic cancers^16,18 Use of NHPs allows evaluation of RIBI in a model anatomically and physiologically similar to humans without the confounding effects of chemotherapy and tumor progression.

Volumetric analysis leveraging T1-weighted MRI can unveil shifts in gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) proportions, providing assessments of cerebral macrostructure. These compartments can also be divided into subregions to provide data on functional areas of interest (eg, hippocampus and prefrontal cortex). GM and WM volume reduction is associated with impaired cognitive function^19,20 and has been demonstrated to occur after radiation therapy.^21,22 No significant reduction in hippocampal volume has been observed at estimated doses of <10 Gy,^17,23 but the late effects of TBI on global and regional volumetric changes have not been investigated in juvenile and adult macaques.

Changes in WM integrity are commonly evaluated using DTI, an MRI modality used to map the 3-dimensional diffusion of water through a tissue, providing a measure of the orientation of diffusion and the degree of anisotropy. Anisotropy is the directional restriction of diffusion, whereas isotropy is unrestricted diffusion in all directions. Radiation-induced WM microstructural change, including reduced fractional anisotropy, increased mean diffusivity and increased radial diffusivity, occurs in humans^24–26 and in NHPs at doses greater than 12.5 Gy.^2,27,28 Measurements of diffusion provide information regarding WM organization, integrity, and connectivity to predict cognitive deficits in patients undergoing radiation therapy with normal-appearing WM.²⁴

In this cross-sectional study, we analyzed MRI scans (T1-weighted and DTI) from 185 male rhesus macaques in the Wake Forest Radiation Late Effects Cohort (RLEC) for volumetric and diffusion changes after single-dose TBI. As 82% of the animals that were irradiated were juveniles, their results provide valuable insight into radiation effects on pediatric patients. We hypothesized that we would observe lower WM and GM volume and higher CSF volume across age in irradiated animals compared with unirradiated controls. Our prior work suggested that WM regions would be particularly sensitive to radiation, and we anticipated we would find changes similar to those in human patients undergoing radiation therapy (decreased fractional anisotropy, increased mean diffusivity, and increased radial diffusivity).²⁵ Our study is the first to evaluate RIBI in rhesus macaques using DTI. Future work will incorporate longitudinal imaging data to evaluate intraindividual radiation effects.

Methods and Materials

Subjects

MRI scans of 185 male rhesus macaques (Macaca mulatta) aged 3 to 22 years acquired between 2013 and 2021 were evaluated. Demographic information is provided in Table 1. As part of their long-term follow-up, animals within the RLEC undergo MRI on a triennial basis and their images are stored in a repository, thus reducing the number of animals needed to study radiation late effects. One MRI data set from each animal in the RLEC was included. One-hundred thirty-seven animals received single-dose, total-body gamma irradiation (TBI, 3.5–8.5 Gy) and 4 animals received single-dose, partial-body irradiation (PBI) of 10 Gy (5% sparing of lower limbs) as part of separate studies conducted at various institutions (Table E1). All studies conducted at previous institutions were approved by an institutional animal care and use committee (IACUC). Forty-four percent of irradiated animals (46% of animals with DTI data) were given mitigators during the acute hematopoietic phase of radiation injury consisting primarily of hematopoietic growth factors. Additional demographic data, including radiation source and dose rate, are listed in Table E1. These animals were brought into the RLEC for long-term monitoring of the effects of radiation. Forty-eight animals with no previous exposure to radiation were adopted as control comparators in the RLEC. All 185 animals included had volumetric analyses performed on structural (T1-weighted) scans; DTI data were available for 159.

Table 1.

Demographic data of animals undergoing T1-weighted and diffusion magnetic resonance imaging

Imaging modality	Group	Number	Age at MRI, median (range), y	Age at irradiation, median (range), y	Time since irradiation, median (range), y	Radiation dose, median (range), Gy
T1-weighted	Control	48	9.1 (3.6–22.8)	–	–	–
	Irradiated	137	7.7 (3.6–20.2)	4.0 (2.3–15.4)	3.6 (0.5–14.3)	6.6 (3.5–10)
DTI	Control	38	8.5 (3.6–22.8)	–	–	–
	Irradiated	121	7.1 (3.6–18.8)	4.1 (2.3–8.2)	2.9 (0.5–14.3)	6.6 (3.5–10)

Abbreviations: DTI = diffusion tensor imaging; MRI = magnetic resonance imaging.

Animals were housed in indoor-outdoor pens or stainless steel cages in temperature and humidity-controlled rooms with a 12:12 hour light-dark cycle. Ninety-four percent of animals were socially housed at the time of MRI. Animals were fed a typical American primate diet with commercially available food (LabDiet 5L0P; Land O’Lakes Inc) designed to resemble a Western human diet. All animals received daily cage-side clinical assessments by trained veterinary personnel that included cage-side neurologic assessments (evaluation for signs of weakness, asymmetry, or ataxia); all animals were neurologically normal during the study period. Additionally, all animals underwent an extensive panel of clinical assessments (physical examination, complete blood count, blood chemistry, and ultrasound examination) and a complete necropsy at the end of life.

Animals in the RLEC typically present with multiple comorbidities as they age. A preliminary report of the prevalence of disease in this cohort has been published.²⁹ All procedures were approved by the Wake Forest University School of Medicine (WFUSM) IACUC and performed in accordance with all state and federal animal welfare laws. WFUSM is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care.

Magnetic resonance imaging

Animals were sedated with ketamine HCl (15 mg/kg body weight, IM), intubated, and maintained on inhaled isoflurane (5% induction, 1.5% maintenance) in 100% oxygen anesthesia for the duration of the MRI procedure. Images were acquired on a 3.0-Tesla Siemens Skyra Clinical Magnetic Resonance Scanner (Siemens AG) with a maximum gradient field strength of 45 mT/m. Sequences were optimized for NHPs and acquired using 1 of 4 dedicated NHP radiofrequency coils (two 8-channel and two 32-channel coils). T1-weighted anatomic images were acquired using a 3-dimensional, T1-weighted, gradient echo (magnetization-prepared rapid gradient echo) sequence with the following parameters: repetition time = 2700 ms, echo time = 3.39 ms, inversion time = 880 ms, field of view = 128 mm, matrix = 256 × 256, slice thickness = 0.5 mm, and number of excitations = 1. A single-shot 2-dimensonal spin echo echo-planar imaging pulse sequence was used to acquire DTI data with full brain coverage. The echo-planar imaging DTI parameters were repetition time = 5800 ms, echo time = 94 ms, field of view = 204 × 164 mm, matrix = 108 × 82, slice thickness = 2.0 mm, no slice gap (yielding 2 mm isotropic voxels), and b-value = 1000 s/mm² in 30 noncolinear directions with 10 b0 images. All diffusion series in this study were acquired with opposing anterior and posterior phase encoding directions, facilitating the use of FSL topup for correcting susceptibility induced field distortions.

Automated image analysis pipeline

Each preprocessed (skull-stripped, intensity-normalized, denoised) brain was rigidly aligned to the INIA19 rhesus template (https://www.nitrc.org/projects/inia19/) using the antsAffine.sh tool from the Advanced Normalization Tools (ANTs) toolkit (Penn Image Computing and Science Laboratory).³⁰ A study-specific template was created from these aligned native images using the antsMultivariateTemplate-Construction.sh tool. Template tissue segmentation priors were inverse warped to the native space and used as prior inputs into the antsAtroposN4.sh tool to perform tissue segmentation of the skull-stripped brain. Volume estimation of total intracranial volume (ICV = GM + WM + CSF), GM, WM, CSF, total brain volume (TBV = GM + WM), and 34 cortical and subcortical regions of interest was then performed for each tissue segmentation map using the fslstats tool from the FMRIB Software Library (University of Oxford) toolkit.³¹ The reference atlas for mapping was the University of North Carolina Primate Atlas (Fig. E1; https://www.nitrc.org/projects/primate_atlas).

Diffusion image processing was performed by the Wake Forest School of Medicine Radiology Informatics and Image Processing Laboratory (RIIPL) leveraging the DEMON Computing Cluster provided by the WFUSM. All diffusion images were converted from the Digital Imaging and Communications in Medicine (DICOM) image format to the NifTi format using dcm2niix.³²

FSL eddy was used to correct eddy current distortions and subject movements. The DIPY dipy_dti_fit function³³ was used to compute DTI metrics and create fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD), and axial diffusivity (AD) maps. These maps were then normalized to a DTI brain template for rhesus macaques (https://www.nitrc.org/projects/rmdtitemplate/) using the ANTS antsRegistrationSyN.sh tool for nonlinear image registration.³⁰ An FA threshold of 0.20 was applied to the normalized FA map to create a WM mask that was applied to the template regions of interest (ROIs) to exclude possible non-WM voxels. The mean value for all nonzero voxels was collected for each ROI using FSL fslstats. Diffusion metrics were provided for total hemispheric white matter and 5 functional ROIs (projection tracts, association tracts, commissural tracts, brain stem WM, and short range WM; Fig. E2) as described by Zakszewski et al.³⁴

Statistical analyses

All statistical analyses were performed using SAS software version 9.4 (SAS Institute Inc) or GraphPad Prism version 9.3.1 (GraphPad Software). Animals were stratified by irradiation status, and age differences between groups were assessed using a 2-group t test. The effects of age on brain volume in control animals were determined by fitting either a linear or a quadratic model. The effect of age on diffusion metrics in control animals was determined using quadratic regression analysis for FA and linear regression analysis for MD, AD, and RD. To determine whether radiation affected age-related trends or whether an interaction existed between age and irradiation on volume and diffusion metrics, quadratic regression analysis was performed, fitting age as either a linear or quadratic effect and using an F test to determine the best model fit (alpha = 0.05).

For linear regression analysis, we used the following equation:

$$\text{Y}=\beta 0+\beta 1^{\star }\text{age}+\beta 2^{\star }\text{irradiation}+\beta 3^{\star }{\text{age}}^{\star }\text{irradiation}+\epsilon .$$

For quadratic regression analysis, we included the terms β4*age² + β5*age²*irradiation. For large ROIs (GM, WM, CSF, ICV, TBV, and hemispheric WM), a P value <.05 was considered significant. Bonferroni correction was used for analyses of small ROIs. For volumetric assessments of cortical and subcortical ROIs (34 ROIs), an uncorrected P value <.0015 was considered significant; for the diffusion analyses evaluating 5 functional ROIs, an uncorrected P value <.01 was considered significant.

Results

Volumetric analysis

Effect of age on global and regional brain volumes

To account for normal variations in ICV within our population, volume values except TBV are reported as percentages of ICV.³⁵ Comparison between unirradiated controls and irradiated groups showed that unirradiated controls were older (difference between medians, 1.4 years; P = .0059; Fig. E3A). Examination of the effect of age on relative brain volume in controls revealed that relative GM volume decreased until middle age (~15 years) before increasing through late adulthood (R² = 0.5868, P < .0001; Fig. E4A). Relative WM volume increased until middle age before gradually declining throughout late adulthood (R² = 0.3627, P < .0001; Fig. E4B). Relative CSF volume did not change across age (R² = 0.000005, P = .9880; Fig. E4C). The changes in GM and WM resulted in the GM:WM matter ratio decreasing until middle age before increasing throughout late adulthood (R² = 0.4572, P < .0001, Fig. E4D). ICV (R² = 0.0408, P = .1686) and TBV (R² = 0.0381, P = .1840) did not vary across age. The left half of the corpus callosum demonstrated a significant increase across age (R² = 0.2650, P = .0010); a similar trend was seen in the right half, but it did not reach the cutoff for significance (P = .0019). Relative right and left prefrontal ROI volumes demonstrated a nonlinear change across age, decreasing slightly until middle age before increasing slightly in late adulthood (right, R² = 0.3992; P < .0001; left, R² = 0.4824; P < .0001). Thirty-one cortical and subcortical ROIs demonstrated no significant change across age (Table E2).

TBI differentially affects brain volume across age

Multiple linear or quadratic regressions were used to assess whether age and irradiation status were significant predictors of regional brain volumes. Total ICV and TBV significantly differed by irradiation status (P = .003 and P = .0072, respectively) but the effect differed across age (interaction: P = .0129 and P = .0277, respectively). Age had a significant effect on relative GM volume, with a consistent decrease in the control and irradiated groups (P < .0001; Fig. 1A). Relative WM volume was affected by age (P < .0001) and irradiation (P = .0391), with young irradiated animals having a lower percentage of WM than controls. In both groups, relative WM volume increased until middle age, at which time no difference by irradiation status was observed (Fig. 1B). Relative CSF volume was significantly affected by age (P = .0007) and irradiation (P = .0045), with the effect of irradiation differing across age (P = .0069). In irradiated animals, relative CSF volume was lowest around middle age, whereas CSF volume did not differ across age in control animals (Fig. 1C).

Fig. 1.

Quadratic regression analyses of relative gray matter volume (A), white matter volume (B), cerebrospinal fluid volume (C), and gray matter:white matter ratio (D) across age by irradiation status (n = 48 no irradiation, n = 137 irradiation). Relative white matter was significantly lower in irradiated juveniles and young adults compared with unirradiated controls. Relative cerebrospinal fluid volume was higher in young irradiated animals but lower than controls in middle-aged irradiated animals. P values <.05 were considered significant. Shaded areas represent 95% CIs.

Of the 34 cortical and subcortical ROIs, only the relative volume of the left occipital lobe was affected by irradiation (P = .0003). Across age, the volume increased in irradiated animals (P = .0008), whereas the volume exhibited no difference across age in controls (P = .2369; Fig. E5). Twelve ROIs were significantly affected by age (Table E3; right and left prefrontal, right and left corpus callosum, right and left temporal auditory, right and left insula, left frontal, right subcortical, right pons and medulla, and left putamen). Additional information regarding the best fit models is presented in Table E4.

Diffusion tensor imaging

Effect of age on metrics of WM integrity

The effect of age on WM diffusion parameters was evaluated in hemispheric WM and all functional ROIs in control animals (Table 2). FA increased until middle age before starting to decline in hemispheric WM and in all functional ROIs except commissural tracts. MD and RD decreased linearly with age in hemispheric WM and all other ROIs except the brain stem and commissural tracts. RD decreased significantly in short range WM. Age did not affect AD in any ROI within our cohort. Additional information regarding the models is provided in Table E5.

Table 2.

Age-related patterns in FA, MD, AD, and RD in unirradiated male rhesus macaques (n = 38) from 3.6 to 22.8 years of age

ROI	FA	MD	AD	RD
Hemispheric white matter	↑ to middle age, then ↓	↓	NS	↓
Association tracts	↑ to middle age, then ↓	↓	NS	↓
Brain stem white matter	↑ to middle age, then ↓	NS	NS	NS
Commissural tracts	NS	NS	NS	NS
Projection tracts	↑ to middle age, then ↓	↓	NS	↓
Short-range white matter	↑ to middle age, then ↓	NS	NS	↓

Arrows represent the direction of significant age-related changes in diffusion. P < .05 is considered significant for hemispheric white matter. P < .01 is considered significant for all other regions of interest.

Abbreviations: AD = axial diffusivity; FA = fractional anisotropy; MD = mean diffusivity; NS = no significant effect of age; RA = radial diffusivity; ROI = region of interest.

TBI alters age-related trends in WM diffusion

There was no significant difference in age distribution between the groups (difference between the medians, 1.4 years; P = .19; Fig. E3B). Radiation modulated the age-related differences in FA, MD, and RD (Figs. 2–4, Table E6), but the effect was regionally dependent. Hemispheric WM FA increased initially before peaking around middle age in both groups, but the increase in FA across age was less pronounced in irradiated animals (interaction P < .0001; Fig. 2A). MD and RD in hemispheric WM decreased across age in unirradiated animals (P = .0004 and P < .0001, respectively; Fig. 2B, D), whereas there was no significant change in MD or RD with age in the irradiated animals’ hemispheric WM (P = .8824 and P = .7097, respectively). These findings revealed a significant interaction between age and irradiation in their effect on MD and RD in hemispheric WM (MD, P = .0042; RD, P = .0003). Neither age (P = .8414) nor irradiation (P = .3018) significantly affected hemispheric WM AD (interaction P = .1778; Fig. 2C).

Fig. 2.

Hemispheric white matter diffusion parameters fractional anisotropy (A), mean diffusivity (B), axial diffusivity (C), and radial diffusivity (D) across age by irradiation status. Blue circles with dashed line represent unirradiated controls (n = 38) and open red triangles with solid line represent irradiated animals (n = 121). P values <.05 were considered significant. Shaded areas represent 95% CIs.

Fig. 4.

Annual differences in mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD) across age between control (blue circles, n = 38) and irradiated (red squares, n = 121) animals in hemispheric white matter and 5 functional regions of interest. Data are presented as slope +/− 95% CI. **P < .01, ***P < .001, ****P < .0001 (n = 48 no irradiation, n = 137 irradiation.

Aside from overall hemispheric WM, all functional ROIs demonstrated radiation-induced differences in FA across age. Although radiation alone was not a significant predictor of FA in functional ROIs, significant differences in FA between the groups were observed in all ROIs across age (P < .01 for the interaction in each ROI; Fig. 3). Irradiated animals demonstrated a less pronounced change in FA across age, whereas FA increased until middle age in all ROIs in the control animals. Significant differences in MD across age were found in association tracts; within this ROI, MD increased with age in irradiated animals but decreased in control animals (interaction P = .0017; Fig. 4). Differences in RD were found in association and commissural tracts; irradiated animals had no change across age in association tracts but increases in RD in commissural WM, whereas controls had decreasing RD with age in association tracts and no change in commissural WM (P < .01 for the interaction in each ROI; Fig. 4). The annual difference in AD did not differ by irradiation status in any functional ROI assessed. Additional information regarding the models is presented in Table E7.

Fig. 3.

Quadratic regression analyses evaluating the effects of age and irradiation on fractional anisotropy within functional regions of interest. Blue circles with dashed line represent unirradiated controls (n = 38) and open red triangles with solid line represent irradiated animals (n = 121). P values <.01 were considered significant. Shaded areas represent 95% CIs.

Discussion

Using T1-weighted and diffusion MRI, we demonstrated that single-dose TBI diffusely alters WM expansion and diffusion across age in rhesus macaques. One important consideration with our cohort is age at the time of irradiation; 82% of the animals were irradiated as juveniles (age 2–5 years), a critical period of neurodevelopment. Young radiation therapy patients are prone to RIBI,³⁶ so our findings of altered WM volume and diffusion support abnormal neurodevelopment postirradiation in pediatric patients.

Intracranial volume differed by irradiation status, with volume decreasing across age in irradiated animals. Given the cross-sectional nature of this study, this is likely the result of intersubject variability and supports the use of relative brain volumes. Relative WM volume was lower in young irradiated animals; however, the difference was not apparent in older animals, as WM volume appeared to normalize around middle age. Delayed maturation and repair or inflammation and tissue expansion by edema and microglia could explain the rebound in WM volume in the irradiated animals. However, examination of the T2*FLAIR images acquired for all animals revealed that only one animal demonstrated clear WM hyperintensity indicative of edema. No significant radiation-induced MRI changes in relative GM volume were observed. Given the total-body exposure of these animals, GM and WM received equivalent doses of radiation; thus, our findings suggest that WM is more radiosensitive than GM, which is consistent with previous studies.^28,37,38

One possible explanation for the differences in the response to radiation between GM and WM is the age at which the animals were exposed to radiation. Human relative GM volume increases rapidly during the first few years of life, peaks around 6 years of age (~2-year-old macaque), and then gradually declines through adolescence and adulthood.³⁹ After GM peaks, a period of WM-driven growth occurs until middle age before gradually declining through late adulthood.³⁹ The gradual age-related decline in GM volume in both groups is consistent with aging in humans of comparable ages.^39,40

As we did not include any scans before 3.5 years (~10-year-old humans), we could not examine an expected period of GM expansion between birth and 24 months.⁴¹ WM change in our cohort also resembles humans, with the largest relative WM volume observed in middle-aged animals (~15-year-old macaques).^39,40 Eighty-two percent of the animals in this cohort were irradiated between 2 and 5 years of age (~6- to 15-year-old humans) after the greatest period of GM expansion but during WM expansion. Considering the differential development patterns in GM and WM, and the ages at irradiation, the insult caused by radiation likely occurred after peak GM expansion but at a critical time during WM development.

Changes in WM diffusion are associated with radiation therapy in humans and rodents, including decreased FA and increased MD and RD occurring as early as a few weeks postirradiation.^3,24,25 Similar to measures of WM volume, measures of WM diffusion normally follow a curvilinear trend with age, with the maximum or minimum value reached at approximately 11 to 12 years of age in rhesus macaques.⁴² FA increases as WM matures and fibers become more linearly oriented throughout adulthood and decreases as diffusion becomes more isotropic. Correspondingly, MD, AD, and RD decrease during WM maturation as axonal density increases and myelination occurs.⁴⁰ These trends end around middle age before age-related degeneration causes them to slowly rise through late adulthood. We found increasing FA until middle age and decreasing MD and RD across age in controls in nearly all ROIs; no age effect was seen on AD. Our finding of increasing FA until middle age is consistent with aging trends in rhesus macaques and humans,^40,42 whereas the age-related trends in MD and RD are consistent with aging until middle age but not beyond middle age, likely due to the age distribution of our control animals at the time of MRI (84% younger than 15 years).

Radiation altered the age-related changes in FA, MD, and RD measured in hemispheric WM. FA increased in irradiated and unirradiated groups, but the change was less pronounced in irradiated animals, suggesting that WM organization occurs but is disrupted by TBI. MD and RD changed little across age in irradiated animals unlike the control group, which demonstrated significant decreases across age. Hemispheric WM AD did not differ by irradiation status in our cohort. Increased RD is associated with demyelination,⁴³ a consequence of brain irradiation.²⁸ Our data indicate that reduced myelin maturation or injury occurs globally after TBI at doses <10 Gy and persists throughout the late-delayed phase of RIBI. As mean diffusivity is derived from AD and RD, the changes observed in MD after irradiation reflect reduced myelin integrity.

In addition to global alterations in WM diffusion, we found similar radiation-induced changes in FA in all functional ROIs, MD and RD in association tracts, and RD in commissural tracts. Projection tracts, which connect cortical GM with deep GM and the brain stem and cerebellum, include structures such as the corona radiata, cerebral peduncles, and internal capsule. Short range WM corresponds to peripheral WM regions named for nearby landmarks, such as adjacent amygdala, inferior frontal gyrus, and anterior cingulum WM. Brain stem WM includes corticospinal tracts and cerebral and cerebellar peduncles, regions critical for motor function and neurotransmission to the spinal cord.

Association tracts connect cortical regions within the same hemisphere and include the superior longitudinal fasciculus, uncinate fasciculus, and cingulum bundle fibers. Impaired diffusion through association tracts impairs communication between large cortical regions essential for many functions, including motor control and cognitive function. Commissural tracts include the corpus callosum, tapetum, and anterior commissure, which connect the 2 hemispheres. Radiation-induced damage to the corpus callosum, internal capsule, frontal WM, and para-hippocampal cingulum is associated with alterations in diffusion and neuropsychological deficits in human radiation therapy patients^24,44,45; thus, our findings of altered diffusion in the projection tracts, corpus callosum, and association tracts corroborate studies in human radiation therapy patients. Moreover, age-related reduction in FA and elevation in MD are associated with impaired executive function and motor skill in rhesus monkeys.⁴⁶ Given the global and regional occurrence of these changes, associated cognitive or motor deficits are expected. These data demonstrate that TBI with doses <10 Gy result in long-term disruption of WM organization (lower FA) and diffusion due to myelin injury (higher MD and RD), in rhesus macaques.

Increases in CSF volume are associated with aging, decreased brain volume, and impaired cognitive function. Unexpectedly, unirradiated animals in our cohort did not experience the age-related increase in CSF volume reported by others,^39,47 which may have been due to the small number of animals >20 years old (~60-year-old human), the approximate age when CSF volume begins to increase expoentially.³⁹ Relative WM volume was lower in irradiated animals in the young group, supporting our finding that TBI results in WM injury, which may delay WM expansion and increase CSF volume. As WM normalized around middle age, a corresponding reduction in relative CSF volume was observed. This finding differs from studies of humans that have observed decreased brain volume and increased CSF volume in WBRT patients with brain metastases.^16,18 No study has reported findings equivalent to ours using comparable doses in normal human brain. Since CSF occupies a small percentage of ICV, the changes in GM and WM volumes are proportionally lower. Although we did not expect ICV to change with age in control animals, our finding of no age-related change in TBV was unexpected, as TBV is expected to decrease with increasing age.⁴⁷

The occipital lobe demonstrated an asymmetrical response to irradiation, with only the left side demonstrating changes associated with irradiation. In addition to the radiation effect, an interaction between age and irradiation status was seen in this region, with the volume in irradiated animals increasing across age but remaining the same in control animals. An increase in volume has not been demonstrated years after brain irradiation in humans. Andrews et al demonstrated a prevalence for radiation-induced vascular lesions in the occipital lobe of animals within their cohort at <8.5 Gy¹⁴; thus, expansion of GM or WM by edema or inflammatory cells could explain the trend seen in our cohort. Our findings suggest the occipital lobe is prone to radiation injury at doses <10 Gy, a finding that warrants further investigation in future studies of radiation-induced brain injury.

One limitation of our study was its cross-sectional nature, which prevented evaluation of intraindividual changes over time. Another limitation was the lack of imaging data from geriatric animals. As this cohort ages, longitudinal data and additional data from late life stages will be available for incorporation into future analyses, providing valuable information regarding the progression of RIBI and the differences between animals irradiated as juveniles or as adults. An additional limitation to diffusion MRI in our study was the use of different head coils. Given the long-term nature of this study (9 years) and technological advancement, diffusion imaging used 4 head coils designed for pediatric human and monkey use. While similar, these could affect WM diffusion analyses; due to the subject variation between the head coil groups, an accurate assessment of the variation in diffusion due to head coil could not be made. Imaging data from 75 females has been obtained; however, a lack of control comparators (n = 2) prevented an assessment of sex-related differences in brain radiosensitivity. Future studies will incorporate female data into the analysis.

Conclusion

T1-weighted and diffusion MRI are important and noninvasive modalities for evaluating brain anatomy and microstructure that help predict cognitive decline and other neurologic deficits. In RIBI, they provide insight into the effect of irradiation on grossly normal-appearing tissue, allowing for early intervention to ameliorate adverse outcomes. Our findings suggest that TBI induces lasting damage to the brain and provides evidence that the brain is sensitive to doses <10 Gy. Moreover, they suggest that WM is particularly sensitive to doses in this range and support the use of diffusion MRI to evaluate RIBI.

Supplementary Material

Supplementary material associated with this article can be found in the online version at doi:10.1016/j.ijrobp.2023.11.014.

Data Sharing Statement:

Research data are stored in an institutional repository and will be shared upon request to the corresponding author.