White Matter is the Predilection Site of Late-Delayed Radiation-Induced Brain Injury in Non-Human Primates

Abstract

Fractionated whole-brain irradiation for the treatment of intracranial neoplasia causes progressive neurodegeneration and neuroinflammation. The long-term consequences of single-fraction high-dose irradiation to the brain are unknown. To assess the late effects of brain irradiation we compared transcriptomic gene expression profiles from non-human primates (NHP; rhesus macaques Macaca mulatta) receiving single-fraction total-body irradiation (TBI; n = 5, 6.75–8.05 Gy, 6–9 years prior to necropsy) to those receiving fractionated whole-brain irradiation (fWBI; n = 5, 40 Gy, 8 × 5 Gy fractions; 12 months prior to necropsy) and control comparators (n = 5). Gene expression profiles from the dorsolateral prefrontal cortex (DLPFC), hippocampus (HC) and deep white matter (WM; centrum semiovale) were compared. Stratified analyses by treatment and region revealed that radiation-induced transcriptomic alterations were most prominent in animals receiving fWBI, and primarily affected white matter in both TBI and fWBI groups. Unsupervised canonical and ontologic analysis revealed that TBI or fWBI animals demonstrated shared patterns of injury, including white matter neuroinflammation, increased expression of complement factors and T-cell activation. Both irradiated groups also showed evidence of impaired glutamatergic neurotransmission and signal transduction within white matter, but not within the dorsolateral prefrontal cortex or hippocampus. Signaling pathways and structural elements involved in extracellular matrix (ECM) deposition and remodeling were noted within the white matter of animals receiving fWBI, but not of those receiving TBI. These findings indicate that those animals receiving TBI are susceptible to neurological injury similar to that observed after fWBI, and these changes persist for years postirradiation. Transcriptomic profiling reaffirmed that macrophage/microglial-mediated neuroinflammation is present in radiation-induced brain injury (RIBI), and our data provide novel evidence that the complement system may contribute to the pathogenesis of RIBI. Finally, these data challenge the assumption that the hippocampus is the predilection site of injury in RIBI, and indicate that impaired glutamatergic neurotransmission may occur in white matter injury.

INTRODUCTION

There have been 57 major nuclear accidents since the Chernobyl disaster (1) despite safety regulations to limit radiation exposure. Occupational, medical and malicious exposures also pose a significant threat to public health and safety. Those who survive the acute radiation syndrome are at risk of developing late-delayed sequelae, which may not appear until several years postirradiation, including pulmonary and cardiac fibrosis, cataracts, muscle wasting and cancer (2–6). The brain is considered relatively radioresistant (ED₅₀ 12–14 Gy) (7); however, recently published studies in non-human primate (NHP) long-term radiation survivors indicate that cognitive impairment may be present at lower doses (8). A clear understanding of the molecular mechanisms of radiation injury is necessary to mitigate the health risks associated with radiation exposure; however, little is known regarding the long-term consequences of single-fraction, high-dose radiation exposure to the brain.

In patients receiving fractionated whole-brain irradiation (fWBI) for the treatment of intracranial neoplasia, progressive, neuroinflammatory and neurodegenerative brain disease occurs. Characterized by multifocal cerebrovascular injury and white matter necrosis (9–24), late-delayed radiation-induced brain injury (RIBI) results in cognitive dysfunction and memory impairment, which negatively affects patient quality of life (25–27). A subset of patients will develop profound dementia, incontinence and ataxia (28).

Studies in which the long-term consequences of single-fraction, high-dose irradiation in humans are investigated are confounded by variations in radiation source, dose rate, shielding and the concurrent effects of thermal and mechanical injury (5, 29). While rodent models have been used to investigate the late effects of radiation exposure, the radiation response of rodents differs from that of humans. Dissimilarities in cerebrovascular structure and white matter investiture (e.g., rodents have a lower white:gray matter ratio) (30) may limit the translatability of findings. Rodents do not develop white matter necrosis as a component of RIBI after fWBI (31–34), thus lacking a key histologic feature of the disease. In contrast to rodents, non-human primates possess greater genomic homology to humans; and better recapitulate the histopathologic manifestations of late-delayed RIBI after fWBI (9, 11, 12, 35).

We hypothesized that non-human primates receiving single-fraction, high-dose total-body irradiation (TBI) would develop similar patterns of brain injury as seen in fWBI, but with lower severity and longer latency. Herein, we investigate patterns of change in the cerebral transcriptome from animals receiving single-fraction, high-dose TBI, and compare them to animals with histologically confirmed radiation-induced brain injury, as well as thorax-only irradiated control comparators. In our previously published targeted gene expression analysis (35), we found macrophage/microglial-mediated neuroinflammation, extracellular matrix (ECM) deposition, evidence of hypoxia, impaired glutamatergic neurotransmission within white matter and vascular maturation. We anticipated that RNA sequencing would reveal similar patterns of injury, and a broader view of contributing mechanisms.

MATERIALS AND METHODS

Animals

Three groups of young adult, post-pubertal, male rhesus macaques (n = 5 per group) were compared (Table 1). Five animals (aged 10–13 years; 6.9–19.7 kg) received high-dose TBI (median: 7.2 Gy; range: 6.75–8.05 Gy) a median of 7.8 years prior to necropsy (6.2–8.8 years). Three to four months postirradiation, these animals were enrolled in a cohort of long-term irradiation survivors. Monitoring included daily observations and annual brain magnetic resonance imaging (MRI).

TABLE 1.

Subject Demographics

Group	Age (years)	Weight (kg)	Dose to brain (Gy)	Survival interval (years)
Control	5.5 ± 0.7	6.66 ± 1.24	0.1	0.7
TBI	11.5 ± 1	11.84 ± 5.47	7.19 ± 0.53	7.8 ± 1.1
fWBI	9.7 ± 2.2	8.6 ± 0.80	40	1.2 ± 0.1

Note. Mean ± standard deviation, (△) = animal with type 2 diabetes.

Whole-brain irradiated animals (n = 5; aged 7–13 years; 8.3–9.5 kg) received 40 Gy fWBI, 12–15 months prior to necropsy as part of a separate experiment including cognitive assessments. Additional experimental procedures are described in detail Hanbury et al. (12) and Robbins et al. (36).

Control comparators (n = 5; aged 5–7 years; 5.5–8.3 kg) received 10 Gy thoracic-only irradiation as part of an unrelated experiment (37).

All animals were clinically observed daily by trained veterinary personnel. All procedures were performed in accordance with the Guide for Care and Use of Laboratory Animals, and approved by the Wake Forest University School of Medicine Institutional Animal Care and Use Committee (IACUC). The Wake Forest School of Medicine is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) and adheres to all state and federal animal welfare laws.

Diet

Animals receiving TBI and control comparators were fed a typical American primate diet [(TAD) diet no. 5L0P, LabDiett®; Land O’Lakes Inc. St. Louis, MO], which is designed to approximate the macronutrients of a Western dietary profile. Animals T2 and T5 were maintained on TAD for 3 and 2 years respectively, and then transitioned back to commercially available monkey chow (diet no. 5038, LabDiet) for the management of type 2 diabetes mellitus (hemoglobin A1C > 6.5%). Additional details are provided elsewhere (4). Animals receiving fWBI were fed diet no. 5038.

Irradiation

Irradiation procedures for control comparators and animals receiving fWBI are described elsewhere (35), and further details about fWBI are provided by Hanbury et al. (12) and Robbins et al. (36). The irradiation procedure for animals receiving TBI is described by Yu et al. (38).

Briefly, animals receiving fWBI were given a total of 40 Gy mid-plane (8 fractions × 5 Gy/fraction, 2 fractions/week × 4 weeks) of 6- MV X rays at a nominal dose rate of 4.9–5.4 Gy/min from a clinical linear accelerator (LINAC) at the WFUSM. The biologically effective dose (BED) of this regimen is 106.7 Gy, and thus comparable to a brain tumor treatment of 30 fractions of 2 Gy in 6 weeks (BED: 100.2 Gy) (39). The opposed lateral fields were designed to match human whole-brain irradiation fields, enclosing the cranial contents (nominal field size of 11 × 7 cm², with the field edge tangential to the base of the skull). The central axis of each lateral beam was placed at the respective outer canthus, and the eyes and olfactory region were shielded from radiation by cylindrical eye blocks. Non-human primates receiving fWBI were sedated with ketamine HCl (15 mg/kg body weight, intramuscularly) and maintained on isoflurane gas (3% induction, 1.5% maintenance) in 100% oxygen during irradiation.

Total-body irradiated animals received 6.75–8.05 Gy (median: 7.2 Gy; range: 6.75–8.05 Gy) administered via a 6-MV LINAC at a nominal dose rate of 0.8 ± 0.025 Gy/min. Fifty percent of the dose was delivered from the anterior-posterior direction, and the remaining dose was delivered from the posterior-anterior direction (38).

Control comparators received 10 Gy thoracic irradiation to the anterior-posterior thoracic mid-plane. X rays (6 MV) were delivered with parallel opposed anterior-posterior fields from a clinical linear accelerator at a nominal dose rate of 4 Gy/min. The field of irradiation included the heart, mediastinum and lung fields extending up to 4 cm caudal to the xyphoid. The maximum dose to the brain is estimated to be 0.1 Gy.

Tissue Collection and Histopathology

Tissue collection, processing, archival and histopathology procedures were the same for all subjects. At the time of necropsy, the animals were humanely euthanized in accordance with the American Veterinary Medical Association’s Guidelines on Euthanasia (40) by deep anesthesia with pentobarbital, followed by exsanguination and perfusion of the vascular system with 2 liters of cold normal saline. The brain was removed intact and sectioned coronally in 4-mm intervals using a stainless-steel brain matrix with cutting guides. Once removed from the matrix, all slices were photographed. Alternating sections were either immediately frozen on dry ice or immersed in 4% cold paraformaldehyde for 24 h. Fixed tissues were embedded in paraffin and sectioned coronally at 4 µm, then stained with hematoxylin and eosin. All animals were assessed by a board-certified veterinary pathologist (JMC) and received a full histopathologic assessment in accordance with the Society for Toxicologic Pathology’s Recommended Practices for Sampling and Processing the Nervous System (41), with the addition of a section of prefrontal cortex containing Brodmann area 46. Lesions were scored as described elsewhere (12, 35): absent (0), minimal (1 = inflammatory or vascular changes without disruption of the neuropil or clear neuronal loss); mild (2 = focal vascular injury and inflammation with loss of neuropil or neurons and microglial activation); moderate (3 = extensive or multifocal vascular injury, hemorrhage, disruption of the neuropil, neuronal loss and microglial activation); or severe (4 = extensive or multifocal vascular injury, hemorrhage, disruption of the neuropil, neuronal loss and microglial activation, with additionally extensive zones of necrosis within neural tissue). The presence of vascular lesions in each region was recorded as follows: a: endothelial hypertrophy; b: perivascular ECM deposition, c: perivascular edema, d: disorganized vascular morphology. The assessor was blinded to the irradiation status of the animals.

RNA Extraction Methods

Total mRNA was extracted from 100 mg of brain tissue from three brain regions [dorsolateral prefrontal cortex (DLPFC; Brodmann area 46), hippocampus (HC) and deep white matter (centrum semiovale)] for all animals. Tissue was placed into a 1.4-mm ceramic bead tube with 1 ml QIAzolt® lysis reagent (QIAGENt®, Valencia, CA), and homogenized using a Bead Ruptor 24 (Omni International, Kennesaw, GA). The tissue sample tube was processed on the Bead Ruptor for 1 cycle at a speed of 4.7 m/s for 20 s, and repeated up to three times until the sample was completely homogenized. Aliquots of homogenized lysates equivalent to 40 mg tissue were extracted for total RNA using the RNeasy Microarray Tissue Mini kit (QIAGEN). Extracted RNA was DNase-treated and purified using the RNA Clean and Concentrator-5 kit (Zymo Research Corp., Irvine, CA), then assessed for RNA quality using an Agilent 2100 Bioanalyzer and the RNA 6000 Nano Kit (Agilent Technologies Inc., Santa Clara, CA).

cDNA Library Preparation and Sequencing Methods

Total RNA was used to prepare cDNA libraries using the Illuminat® TruSeq Stranded Total RNA with Ribo-Zero Gold Preparation kit (San Diego, CA) and the SciClone NGS Work Station (PerkinElmert® Inc., Waltham, MA). RIN values for the RNA samples ranged from 7.1 to 9.3. Briefly, 750 ng of total RNA was rRNA depleted followed by enzymatic fragmentation, reverse-transcription and double-stranded cDNA purification using AMPure XP magnetic beads (Beckman Coulter® Inc., Fullerton, CA). The cDNA was end repaired, 30 adenylated, with Illumina sequencing adaptors ligated onto the fragment ends, and the stranded libraries were pre-amplified with PCR. The library size distribution was validated and quality inspected on a Bioanalyzer DNA 1000 chip (Agilent Technologies). The quantity of each cDNA library was measured using the Qubit® 3.0 (Thermo Fisher Scientific™ Inc., Rockford, IL). Three library pools were formed, each containing 15 libraries. Library pools were sequenced to a target read depth of 28M reads per library using single-end 76 cycle sequencing with the High Output 75-cycle kit on the Illumina NextSeq® 500.

Data Analysis

Raw read quality was assessed using FASTQC analysis (Babraham Bioinformatics, Cambridge, UK). Uniquely mapped reads ranged from 21M–36M reads per sample. Reads with >Q20 quality score were aligned to the Ensembl Macaca mulatta genome build Mmul_1 using the STAR aligner (42) and gene counts were determined using featureCounts software (43). Differentially expressed genes (DEGs) were identified by DESeq2 (44). Significant DEGs were conservatively defined as log₂ fold change ratios ≥ ±1 and P < 0.05 after adjustment for false discovery (Benjamini-Hochberg). Gene expression analysis data are deposited in the Gene Expression Omnibus database (accession no. GSE120901). Gene identities were preliminarily mapped in Ingenuity® Pathway Analysis (IPA®) (45) using Ensembl identification numbers for the rhesus macaque and assigned the corresponding HGNC identifier. As the rhesus gene identifiers were transcribed to human identifiers for analysis, the MHC class II molecules referred to herein as HLA-transcripts reference the MAMU sequence. Genes that did not map in IPA were identified by secondary screening in PANTHER (46, 47) and UniProt (48), then backmapped to the encoding gene by the corresponding HGNC identifier. Ensembl IDs that did not correspond to an identified gene were recorded (Supplementary Table S1; http://dx.doi.org/10.1667/RR15263.1.S1) and excluded from pathway analysis.

Enriched biological pathways and signaling networks were identified in IPA via unsupervised analysis of significant DEGs. Genes that contributed to the top 10 canonical pathways (Supplementary Table S2; http://dx.doi.org/10.1667/RR15263.1.S1)) were evaluated for contributing molecular functions in Gene Ontology using the GO Enrichment Analysis tool powered by PANTHER (46, 47) with Fisher’s exact test with false discovery rate (FDR) correction for multiple comparisons. As there were no enriched canonical pathways detected within the dorsolateral prefrontal cortex or hippocampus of TBI animals, these regions were omitted from ontologic analysis. Biological process and molecular function analyses were hierarchically clustered and the most specific subclass is reported. Hierarchies were sorted by fold-enrichment value, and the 10 greatest fold-enriched hierarchies are reported. FDR-corrected P values ≤ 0.05 were considered significant.

RESULTS

Demographics

Control comparators were younger than groups receiving fWBI (mean difference: 4.18 years; P ≤ 0.002) and TBI (mean difference: 6.06 years; P ≤ 0.0001) (Fig. 1A). The ages of animals receiving fWBI and TBI were comparable (P = 0.18). There was no difference in animal weight between groups (Fig. 1B); the TBI animals weighing >15 kg were those with type 2 diabetes mellitus (Fig. 1B, open triangles). Differences in radiation dose (Fig. 1C) and survival interval between groups (Fig. 1D) were due to differences in experimental design.

FIG. 1.

Subject demographics. Mean ± standard deviation, (△) = Animal with type 2 diabetes

Clinical Findings

Causes of death and comorbid conditions are summarized in Table 2. Neurologically, all animals (n = 15) were normal. Four out of five (4/5) animals receiving TBI developed neoplasia during the study period. Three out of four (3/4) tumors were of neuroendocrine origin, and the remaining neoplasm was a hemangiosarcoma (n = 1); there was no evidence of intracranial metastasis for any animal. Pulmonary bullous emphysema, with bulla rupture and resultant pneumothorax necessitated the euthanasia of two out of five (2/5) animal receiving TBI. Two out of five (2/5) animals receiving TBI developed type 2 diabetes mellitus four years postirradiation.

TABLE 2.

Subject Causes of Death and Comorbid Illness

Group	Animal ID	Cause of death/comorbid illness
Control	C1	Experimental euthanasia
	C2	Experimental euthanasia
	C3	Experimental euthanasia
	C4	Experimental euthanasia
	C5	Experimental euthanasia
fWBI	F1	Experimental euthanasia
	F2	Experimental euthanasia
	F3	Experimental euthanasia
	F4	Experimental euthanasia
	F5	Experimental euthanasia
TBI	T1	Clinical euthanasia
		Pulmonary bullous emphysema; with rupture and pneumothorax
		Neoplasia, hemangiosarcoma, pulmonary
	T2	Clinical euthanasia
		Fatal fasting syndrome of obese macaques
		Type 2 diabetes mellitus
	T3	Clinical euthanasia
		Pulmonary bullous emphysema; with rupture and pneumothorax
		Neoplasia, seminal vesicle, poorly-differentiated neuroendocrine tumor
	T4	Clinical euthanasia
		Neoplasia, renal tubular carcinoma with polycythemia
		Neoplasia, subcutis, chondrolipoma
		Neoplasia, subcutis, poorly-differentiated neuroendocrine tumor
	T5	Clinical euthanasia
		Type 2 diabetes mellitus
		Bacterial pneumonia
		Chronic renal disease Hypertension
		Hypertension
		Neoplasia, neuroendocrine tumor, heart base

Since control comparators had received high-dose thoracic irradiation as part of a previous study, clinical data regarding respiratory and cardiac function were reviewed prior to selection as control comparators. Respiratory rate, heart rate and pulse oximetry were within normal limits until euthanasia.

Histopathology

All animals receiving fWBI (5/5) developed multifocal white matter necrosis and cerebrovascular injury (Table 3).

TABLE 3.

Histopathologic Grading of RIBI in Animals Receiving fWBI

	F1	F2	F3	F4	F5
White matter necrosis
Forebrain: white matter	+	++++	++	+++	+++
Forebrain: cortical gray matter	−	+++	+	+	++
Prefrontal cortex	−	−	−	+	++
Basal ganglia/striatum	−	+++	−	−	−
Hippocampus	−	−	−	−	−
Thalamus	−	++	−	++	++
Midbrain	−	++	−	+	+++
Cerebellum	−	+	−	++	++
Brainstem	−	++	−	+	−
Spinal cord	−	++	na	+	−
Vascular
Forebrain: white matter	a, c	a, b, d	−	a, b, d	b, d
Forebrain: cortical gray matter	−	a, d	d	a, d	b
Prefrontal cortex	−	−	−	−	−
Basal ganglia/striatum	−	−	a	−	−
Hippocampus	c	−	−	−	−
Thalamus	−	−	−	−	−
Midbrain	−	a, d	−	−	b
Cerebellum	−	−	−	−	b
Brainstem	−	a, d	−	−	−
Spinal cord	−	−	−	−	−

Notes.

^a= Endothelial hypertrophy.

^b= Perivascular extracellular matrix deposition.

^c= Perivascular edema.

^d= Disorganized vascular morphology.

There were no significant lesions in control comparators. There was no evidence of white matter necrosis in animals receiving TBI. In animal no. T002, perivascular spaces surrounding penetrating arterioles in the rostral frontal cortex contained multifocal aggregates of hemosiderin. A few arterioles within the caudal caudate nucleus were surrounded by homogeneous eosinophilic ECM. In animal no. T004 histopathologic evaluation revealed a focal vascular malformation consistent with a cavernous hemangioma within the subcortical white matter of the right inferior occipital lobe.

Transcriptomic Profiling

In all brain regions examined, the number of DEGs was greatest in animals that received fWBI (Fig. 2). White matter contained the greatest number of DEGs within treatment groups, followed by the dorsolateral prefrontal cortex and then the hippocampus. A total of 133 DEGs were expressed within the white matter of both fWBI and TBI groups. Two DEGs (MEIS3, SCEL) were expressed within the dorsolateral prefrontal cortex of both fWBI and TBI groups (Fig. 3).

FIG. 2.

Venn diagrams of regional differences in differential gene expression by region, across irradiation groups. Both fWBI and TBI groups are normalized to the expression values of the control group. The number of genes differentially regulated in all regions in animals receiving fWBI, was greater than comparable regions in animals receiving TBI. White matter contained the highest number of DEGs within both irradiation groups.

FIG. 3.

MEIS3 and SCEL are differentially expressed within the dorsolateral prefrontal cortex of both fWBI and TBI animals. MEIS3 was downregulated within the dorsolateral prefrontal cortex (log₂ fold change: –1.02) and white matter (log₂ fold change: –1.67) of animals receiving fWBI, and the dorsolateral prefrontal cortex (log₂ fold change: –1.12) of animals receiving TBI. SCEL was upregulated within the dorsolateral prefrontal cortex in animals receiving fWBI (log₂ fold change: 1.01) and animals receiving TBI (log₂ fold change: 1.01). Bars denote mean log₂ fold change and error bars are set to log₂ fold standard error. Dashed line indicates log₂ fold change of ±1.0. DLPFC = dorsolateral prefrontal cortex; HC = hippocampus; WM = temporal white matter. *FDR adjusted P value ≤ 0.05

Canonical Pathway Analysis

The top 10 enriched canonical pathways are reported in order of decreasing statistical significance, with all pathways reaching at least P ≤ 0.05 (Fig. 4). Shared canonical pathways were detected between regions and treatments (Fig. 5A and B). Since overall patterns suggested shared processes involving inflammation, ECM remodeling and neurotransmission/signal transduction, for further analysis the contributing genes were grouped by broader function (Table 4).

FIG. 4.

Top 10 enriched canonical pathways in ingenuity pathway analysis. All comparisons are normalized to the expression values of the matching region from the control group. Pathways are listed in descending order of decreasing statistical significance, all FDR adjusted P values ≤ 0.05. The stacked bars indicate the percentage of up- or downregulated DEGs out of a reference gene set curated by IPA. Data are not shown for dorsolateral prefrontal cortex or hippocampus comparisons, as zero (0) enriched canonical pathways were detected.

FIG. 5.

Overlapping canonical pathways across brain regions and irradiation groups. Patterns in canonical pathway expression were shared across regions, and across treatments. Panel A: Shared and exclusive pathways are quantitated, and displayed by region and treatment. Panel B: Boxes are color coded to the corresponding Venn-diagram sector. White indicates that the following canonical pathway was not within the top 10 enriched pathways for that region and treatment.

TABLE 4.

Functional Grouping of Canonical Pathways

Inflammation	ECM	Neurotransmission and signal transduction
Neuroinflammation	CDC42 signaling	Calcium signaling
Granulocyte adhesion and diapedesis	Atherosclerosis signaling	ALS signaling
Hepatic fibrosis/hepatic stellate cell activation	Hepatic fibrosis/hepatic stellate cell activation	Glutamate receptor signaling
Complement system
TREM1 signaling
Antigen presentation pathway
Th1 and Th2 activation pathway
Type 1 diabetes mellitus signaling
Allograft rejection signaling
OX40 signaling
Th1 pathway
Th2 pathway
B-cell development
Dendritic cell maturation
Autoimmune thyroid disease signaling
T-helper cell differentiation
Role of hypercytokinemia/hyperchemokinemia in the pathogenesis of influenza

Canonical pathway analysis indicated enrichment of inflammation and immunologic signaling patterns within all regions in animals receiving fWBI and within the white matter of animals receiving TBI. ECM-associated pathways were enriched within fWBI animals only. Neurotransmission and signal transduction-associated pathways were enriched within the white matter of fWBI and TBI animals. Comparisons with the greatest numbers of shared pathways were between the dorsolateral prefrontal cortex and hippocampus of fWBI animals (significantly enriched pathway n = 6), and the white matter of fWBI and TBI animals (significantly enriched pathways n = 5), suggesting that the radiation response in RIBI is differentially regulated by region.

Nineteen genes were implicated in the top 10 canonical pathways within fWBI dorsolateral prefrontal cortex, 10 in fWBI hippocampus, and 192 in fWBI white matter. Thirty-one contributing genes were identified in the TBI white matter, and there were no enriched canonical pathways detected within the TBI dorsolateral prefrontal cortex or hippocampus.

Gene Ontology

Inflammation-associated canonical pathways were identified in all brain regions from animals receiving fWBI and the white matter of animals receiving TBI. The contributing DEGs (fWBI DLPFC: 15 DEGs; fWBI HC: 6 DEGs; fWBI WM: 172 DEGs; and TBI WM: 22 DEGs) were evaluated for patterns in molecular function (Table 5). Complement system transcripts were increased in the white matter of both irradiated groups. Transcripts common to major histocompatibility complex (MHC) class II-mediated peptide antigen presentation were noted in the fWBI dorsolateral prefrontal cortex, fWBI hippocampus and TBI white matter. Ontologic patterns suggesting pro-inflammatory chemokine signaling were present in the white matter of both irradiated groups. Patterns consistent with modulation of neurotransmission noted in the white matter of animals receiving fWBI (glutamate-gated calcium ion channel activity, NMDA glutamate receptor activity and calcium-dependent protein kinase C activity) suggest cross-talk between neuroinflammatory signaling pathways and neurotransmission.

TABLE 5.

Molecular Function Analysis of Neuroinflammatory Genes in Irradiated Brain

	Fold enrichment	P value
fWBI, dorsolateral prefrontal cortex
aMHC class II receptor activity	>100.00	1.50E–04
Platelet-derived growth factor binding	>100.00	1.50E–04
aMHC class II protein complex binding	>100.00	2.66E–04
aPeptide antigen binding	>100.00	5.29E–05
Integrin binding	34.91	3.63E–02
fWBI, hippocampus
aMHC class II receptor activity	>100.00	1.26E–05
aPeptide antigen binding	>100.00	1.03E–06
aMHC class II protein complex binding	>100.00	1.11E–02
fWBI, white matter
bOpsonin receptor activity	89.16	1.20E–03
Tumor necrosis factor binding	79.25	2.78E–02
ICAM-3 receptor activity	79.25	2.75E–02
Glutamate-gated calcium ion channel activity	71.33	1.76E–03
NMDA glutamate receptor activity	67.93	1.42E–04
Calcium-dependent protein kinase C activity	59.44	3.81E–02
bComplement component C3b binding	59.44	3.78E–02
Benzodiazepine receptor activity	59.44	2.39E–03
Lipoteichoic acid binding	59.44	3.75E–02
CCR2 chemokine receptor binding	59.44	3.72E–02
TBI, white matter
aMHC class II receptor activity	>100.00	2.78E–02
Platelet-derived growth factor receptor binding	>100.00	4.06E–02
CXCR chemokine receptor binding	>100.00	4.13E–02
Chemokine activity	53.68	1.06E–02
Cytokine binding	34.38	5.49E–03
Cytokine receptor activity	28.28	3.99E–02
bSerine-type endopeptidase activity	16.99	7.86E–03
aPeptide binding	13.13	4.87E–02

Notes. Analyses completed in Gene Ontology enrichment analysis tool. Maximum fold-enrichment values are truncated at 100. The top 10 enriched ontologies are reported in order of decreasing fold-enrichment and statistical significance, multiplicity adjusted P values ≤ 0.05 were considered significant. Overarching molecular function:

^aMHC class II peptide antigen presentation

^bcomplement activation.

ECM-associated canonical pathways were detected in all regions of animals receiving fWBI and none from the animals receiving TBI. The contributing genes (fWBI DLPFC: 12 DEGs; fWBI HC: 4 DEGs; fWBI WM: 43) were evaluated for patterns in molecular function (Table 6). Enriched ECM-pertinent molecular functions were not detected in the hippocampus of animals receiving fWBI. Transcripts related to the platelet derived growth factor (PDGF) signaling pathway were increased in the dorsolateral prefrontal cortex and white matter. Transcripts corresponding to ECM deposition and remodeling were detected in the fWBI white matter. Enrichment of the PDGF signaling pathway is partially due to decreased PDGF receptor alpha (PDGFRA) mRNA expression (Fig. 8), presumed reflective of oligodendrocyte progenitor depletion. However, the differential transcript expression of collagen subunits in the dorsolateral prefrontal cortex and white matter, and epidermal growth factor (EGF), epidermal growth factor receptor (EGFR), fibroblast growth factor 1 (FGF1), platelet derived growth factor beta (PDGFB) and hepatocyte growth factor (HGF) in white matter suggest that PDGF signaling may also mediate ECM deposition in RIBI.

TABLE 6.

Molecular Function Analysis of Extracellular Matrix and Remodeling-Associated Genes in fWBI Brain

	Fold enrichment	P value
fWBI, dorsolateral prefrontal cortex
aPlatelet-derived growth factor binding	>100.00	2.96E–04
fWBI, white matter
Tumor necrosis factor binding	>100.00	5.93E–03
aPlatelet-derived growth factor binding	>100.00	6.18E–06
Interleukin-1 receptor activity	>100.00	1.80E–02
Tumor necrosis factor-activated receptor activity	99.63	1.50E–06
aPlatelet-derived growth factor receptor binding	95.65	1.16E–03
Phosphatidylinositol-4,5-bisphosphate 3-kinase activity	43.48	2.51E–06
bExtracellular matrix structural constituent	42.92	5.27E–07
bIntegrin binding	29.62	1.85E–06
bProtease binding	24.98	3.89E–06
bCollagen binding	22.77	3.80E–02

Notes. Analyses completed in Gene Ontology enrichment analysis tool. Maximum fold-enrichment values are truncated at 100. The top 10 enriched ontologies are reported in order of decreasing fold-enrichment and statistical significance, multiplicity adjusted P values ≤ 0.05 were considered significant. Overarching molecular function:

^aPlatelet derived growth factor (PDGF) signaling

^bECM remodeling.

FIG. 8.

Decreases in glutamatergic neurotransmitter receptor subunits in RIBI occur independently of PDGFRA expression. Supervised analysis of oligodendrocyte precursor marker, platelet derived growth factor alpha (PDGFRA) and glutamatergic neurotransmitter receptor subunits GRIN2B and GRIA3. PDGFRA was decreased in dorsolateral prefrontal cortex, hippocampus and white matter 12 months after fWBI but not 6–9 years after TBI. Glutamatergic NMDA receptor subunit GRIN2B and AMPA receptor subunit GRIA3 expression were decreased in white matter in both fWBI and TBI groups. Bars denote mean log₂ fold change and error bars are set to log fold change standard error. *Log₂ fold changes < ±1 and multiplicity-adjusted P values (Benjamini-Hochberg) < 0.05.

Alteration of canonical pathways involved in neurotransmission and signal transduction were detected within the white matter of both irradiated groups. Analysis of the contributing genes to neurotransmission and signal transduction-associated canonical pathways (32 fWBI DEGs, 10 TBI DEGs) demonstrated alteration in mechanisms modulating glutamatergic neurotransmission and calcium channel activity (Table 7).

TABLE 7.

Molecular Function Analysis of Neurotransmission and Signal Transduction Associated Genes in White Matter

	Fold enrichment	P value
fWBI, white matter
aNeurotransmitter receptor activity involved in regulation of postsynaptic cytosolic calcium ion concentration	>100.00	2.64E–03
bGlutamate-gated calcium ion channel activity	>100.00	4.11E–05
bNMDA glutamate receptor activity	>100.00	2.13E–06
bAdenylate cyclase inhibiting G-protein coupled glutamate receptor activity	>100.00	4.97E–03
bGlutamate binding	>100.00	2.05E–04
bExtracellularly glutamate-gated ion channel activity	>100.00	7.71E–07
bHigh voltage-gated calcium channel activity	>100.00	2.34E–04
Glycine binding	>100.00	3.94E–04
bL-glutamate transmembrane transporter activity	88.41	2.30E–02
aCalcium channel regulator activity	53.05	3.19E–03
TBI, white matter
bIonotropic glutamate receptor activity	>100.00	6.99E–03
bExtracellularly glutamate-gated ion channel activity	>100.00	6.75E–03
aLigand-gated calcium channel activity	>100.00	1.01E–02
Actinin binding	97.87	2.87E–02
Neurotransmitter binding	80.93	4.01E–02

Notes. Analyses completed in Gene Ontology enrichment analysis tool. Maximum fold-enrichment values are truncated at 100. The top 10 enriched ontologies are reported in order of decreasing fold-enrichment and statistical significance, multiplicity adjusted P values ≤ 0.05 were considered significant. Overarching molecular function:

^aCalcium ion regulation

^bglutamatergic neurotransmission.

Supervised Analyses

Neuroinflammatory pathway analyses indicated that MHC class II peptide antigen presentation and the complement system play a role in RIBI. As antigen presentation may lead to T-cell activation, and the adaptive immune system may regulate complement activation, transcripts indicating T-cell activation and differentiation were also assessed.

Gene expression associated with MHC class II antigen presenting molecules (HLA-DPB1, HLA-DQA1, HLA-DMA) was increased within all brain regions in fWBI and white matter in TBI (Fig. 6A). HLA-DPA1 and HLA-DRB5 were also expressed within all brain regions in fWBI animals, and HLA-DMB within the dorsolateral prefrontal cortex and white matter of fWBI animals.

FIG. 6.

Differentially regulated gene contributors to canonically-enriched pathways of neuroinflammation are shown. Panel A: Antigen presentation. Panel B: The complement system. Panel C: T-cell activation. Gene lists were curated by Ingenuity Pathway Analysis. Gene IDs are HGCN identifiers. Red indicates upregulation and green indicates downregulation. White boxes containing ‘‘x’’ indicate the gene was not differentially regulated within that region. All log₂ fold changes < ±1 and multiplicity-adjusted P values (Benjamini-Hochberg) < 0.05.

Complement factor associated gene expression was increased within the white matter of fWBI and TBI animals (Fig. 6B). CFB and C7 transcripts were also expressed within the fWBI dorsolateral prefrontal cortex, and C3AR within the fWBI hippocampus. Complement regulatory transcripts CD59, C4BP, SERPING1, and CR1 were differentially regulated in fWBI, but not in TBI.

Gene expression related to T-cell activation and differentiation was most prominent within the white matter of fWBI animals. CD3D and CD3E gene expression was increased within the white matter of fWBI animals (Fig. 6C). RAR-related orphan gamma receptor (RORyt), a molecule involved in chronic inflammatory signaling through Th17 cells, was transcriptionally upregulated within the white matter of fWBI and TBI animals. STAT6 mRNA, involved in Th2 differentiation, was upregulated in the white matter of animals receiving fWBI. Co-stimulatory molecules essential for the activation of T cells (CD2, CD86) and B cells (CD40) were transcriptionally upregulated within fWBI white matter. These data provide preliminary evidence that MHC class II antigen presentation may induce T-cell activation in RIBI.

Ontologic evaluation suggests that ECM deposition and remodeling in the white matter may contribute to RIBI in fWBI animals. A comprehensive list of ECM contributors was compiled by literature review (48–57). Of several classes of ECM molecules, differential gene regulation was noted in basement membrane contributors, proteoglycans, fibrous proteins, glycoproteins and ECM molecules generally enriched in CNS, bone, cartilage and teeth (Fig. 7). No changes were noted in any of the keratan sulfate proteoglycans, vascular-associated fibrinogen chains, vitronectin or von Willebrand factor, small interstitial proteoglycans (decorin, biglycan, asporin), laminins, agrin, aggrecan or nidogen.

FIG. 7.

Differentially regulated ECM transcripts in white matter after fWBI. Supervised analysis of differentially expressed extracellular matrix genes in white matter in RIBI after fWBI. Gene IDs are HGCN identifiers. Red indicates upregulation and green indicates downregulation. All log₂ fold changes < ±1 and multiplicity adjusted P values (Benjamini-Hochberg) < 0.05.

Since previous analyses suggested that radiation exposure alters neurotransmission and signal transduction pathways in white matter, we completed a supervised review of DEGs in the white matter animals receiving TBI or fWBI for contributing molecular functions. The oligodendrocyte precursor marker platelet derived growth factor alpha (PDGFRA) was decreased in all regions after fWBI but not TBI (Fig. 8). Glutamatergic NMDA receptor subunit GRIN2B and AMPA receptor subunit GRIA3 were decreased in white matter but not dorsolateral prefrontal cortex or hippocampus in both irradiated groups (Fig. 8).

Additional downregulated neurotransmission-associated genes detected in white matter of both irradiated groups were bassoon presynaptic cytomatrix protein (BSN), calcium voltage-gated channel subunit alpha 1 G (CACNA1G), CACNA1I, calcium voltage-gated channel auxillary subunit beta 1 (CACNB1), calcium/calmodulin-dependent protein kinase 2 beta (CAMK2B), potassium voltage-gated channel subfamily B member 1 (KCNB1), potassium calcium-activated channel subfamily M alpha 1 (KCNMA1), potassium sodium-activated channel subfamily T member 1 (KCNT1), synaptotagamin 2 (SYT2), SYT5 and SYT6 (Supplementary Fig. S1; http://dx.doi.org/10.1667/RR15263.1.S1). CACNA1D, KCNK12, synaptic Ras-GTPase activating protein 1 (SYNGAP1), sodium voltage-gated channel beta subunit 2 (SCN2B) were differentially regulated in the white matter of animals receiving TBI. The gene encoding glutamate transporter excitatory amino acid transporter 2 (EAAT2; gene: SLC1A2) was downregulated in the white matter of animals receiving fWBI. Potassium two pore domain channel subfamily K member (KCNK5) was upregulated within the white matter of both irradiated groups.

DISCUSSION

Herein we have characterized the transcriptomic signature of radiation-induced brain injury in the rhesus macaque and demonstrate that fWBI and TBI manifest shared patterns of injury months and years postirradiation, respectively. Our analyses are in agreement with previously published studies indicating involvement of neuroinflammatory processes in RIBI (58–64) and we present novel evidence indicating that the complement system may contribute to injury. Our findings also suggest that ECM deposition and remodeling are affected after fWBI but not TBI. Lastly, regional stratification indicates that radiation-induced impairment of neurotransmission and signal-transduction transcripts is most prominent in white matter. Coupled with the greater proportion of DEGs within white matter compared to hippocampus, these data challenge the hypothesis that radiation-induced brain injury is primarily hippocampal.

Our data reaffirm that neuroinflammatory processes play a role in RIBI and agree with previously published studies in rodents, which demonstrate microglial activation present 4 h to six months after single-fraction dose (58, 59, 63, 65, 66 and fractionated doses (67) brain irradiation. Such activation includes MHC class II antigen presentation (68–72). In the absence of exogenous antigens, we presume that MHC class II molecules must present endogenously derived antigens, e.g., exposed intracellular antigens or proteins altered by radiation (73). We hypothesize that sustained microglial activation after irradiation may perpetuate RIBI via antigen presentation and direction of the neuroimmune response against endogenous targets.

RIBI-induced ECM deposition and remodeling was noted in the white matter of animals receiving fWBI but not in the white matter of those receiving TBI. It is unclear as to whether this is an effect of dose or fractionation. In an effort to understand radiation-induced alterations in the matrix-ome, we have provided an assessment of all differentially regulated ECM contributors to fWBI white matter (Fig. 7). Notably, fWBI was associated with cognitive and neurological impairment while animals receiving TBI were neurologically normal.

Regional stratification of DEGs reveals fewer gene expression changes within the hippocampus compared to white matter within both treatment groups, consistent with our previously reported gene expression analyses (35). Numerous published studies in cell culture and rodent models have demonstrated deleterious, radiation-induced effects on hippocampal structure and function, including: reduced neuronal stem cell number, survival and maturation (58, 65, 74), changes in dendritic spine morphology and density (75–78), alterations in glutamatergic ionotropic receptor subunits (32, 79–81) and neuronal cytoskeletal alterations (78, 82). Thus, hippocampal gene expression data were carefully reviewed for the involvement of neurotransmission, signal transduction and stem cell signaling pathways. Differentially regulated signaling pathways within the hippocampus were consistent with inflammatory processes, and we did not detect alteration in neurogenic or synapse markers. These data suggest that hippocampal-mediated dysfunction may contribute less to RIBI in primates than in rodents, or that hippocampal injury may be dependent on factors not examined here (e.g., neurogenesis in juvenile subjects).

An unexpected white matter-specific reduction in synaptic neurotransmission and signal transduction-associated mRNAs was observed in both groups of irradiated animals (fWBI and TBI). As mentioned previously, studies in rodent models of RIBI have also demonstrated radiation-induced synaptic pathology, including changes in: expression levels of synaptic proteins, neurotransmitter receptors and structure (32, 77, 78, 80, 81, 83–85); however, these analyses were restricted to examination of the hippocampus and/or cortex.

Although synaptic neurotransmission is generally considered restricted to gray matter, myelination may be regulated by action-potential mediated vesicular release of glutamate between unmyelinated axons (presynaptic initiator) and adjacent NG2⁺ oligodendroglial progenitors (post-synaptic recipient) (86–89). We hypothesize that the reduction in neurotransmission-associated transcripts within white matter are related to radiation-induced loss or dysfunction of these deep white matter synapses. Although radiation (90–93) and ischemia-induced (86–89) PDGFRA⁺ oligodendrocyte progenitor loss may correspond to reduced numbers of deep white matter synapses, the reduction in glutamatergic neurotransmitter receptors (GRIN2B and GRIA3) in the face of normal levels of PDGFRA in animals receiving TBI suggests that deep white matter synaptic impairment may occur independently of oligodendrocyte progenitor cell loss. Reduction in glutamatergic neurotransmitter receptors and failure of scavenging mechanisms [decreased expression of the encoding gene for glutamate scavenger EAAT2 (SLC1A2) in animals receiving fWBI; Supplementary Fig. S1; http://dx.doi.org/10.1667/RR15263.1.S1] may also contribute to RIBI via excitotoxic injury due to extracellular glutamate accumulation. A similar mechanism has been shown to result in oligodendroglial loss, demyelination and severe axonal damage in optic nerves of rats (94).

Synaptic loss and dysfunction may also explain the downregulation of ion channels in the current study (Supplementary Fig. S1; http://dx.doi.org/10.1667/RR15263.1.S1). Although KCNK channels (KCNK3 and KCNK2) are neuroprotective against excitotoxic injury (95–97), upregulation of KCNK5 exacerbates ischemic brain damage by inducing apoptosis in a rodent model of transient middle cerebral artery occlusion (98). Upregulation of KCNK5 in the white matter of animals receiving fWBI and TBI (Supplementary Fig. S1) may reflect similar physiologic processes.

Complement is a component of the innate immune system responsible for the destruction and clearance of pathogens and damaged cells from the body, and complement-mediated synaptic destruction has been implicated as a mechanism of neurodegeneration and cognitive dysfunction in aging (99) and Alzheimer’s disease (100–102). While complement is known to mediate radiation-induced cell killing (103–105), to our knowledge we are the first to demonstrate that complement activation may play a role in the pathogenesis of RIBI. Additional studies are needed to verify proteolytic activation of the complement system and to identify the cellular and molecular targets of complement fixation in RIBI.

The significance of differential regulation of MEIS3 and SCEL in the dorsolateral prefrontal cortex of animals receiving fWBI or TBI is unclear. MEIS3 encodes a homeobox protein presumed to play a role in transcriptional regulation (106, 107). SCEL is a component of the cornified envelope of keratinocytes (108–112).

The current study indicates that the brain is vulnerable to injury from single-fraction TBI (6.05–8.5 Gy), much lower than the previously estimated threshold for CNS injury (7). This suggests that even patients that do not develop the acute radiation CNS syndrome (113) after exposure may be at risk of developing more subtle long-term neurological injury. Cognitive testing in our laboratory indicates animals receiving similar TBI doses may demonstrate less cognitive flexibility (8). Thus, ARS survivors may benefit from more aggressive surveillance and monitoring of cognitive function.

We acknowledge that differing diet, age and comorbidities between groups may have affected cerebral gene expression. Nevertheless, to our knowledge, this is the first published work reporting on cerebral transcriptomic data from NHP years after single-fraction high-dose TBI. In the event of a large-scale nuclear accident or malicious exposure, anatomic sites of radiation exposure and extent of shielding within the exposed population will be heterogeneous, often with multiple organ systems affected, as has been demonstrated by published studies of the Hiroshima and Chernobyl survivors, and the NHP RSC (4, 114). Therefore, radiation-associated comorbidities are anticipated in survivor populations, and related alterations in cerebral gene expression are relevant to considerations for long-term survivors of radiation exposure.

Despite statistical differences in age, all groups of animals were within similar life stages (young adult, post-pubertal) and thus biologically comparable. Since previously published studies have demonstrated age and development-associated changes in hippocampal gene expression in rodents (115–117) and rhesus macaques (118), the agreement in hippocampal gene expression between animals receiving TBI and control animals (0 DEGs, mean age difference 6.01 years) supports that these animals were in similar life stages.

In summary, to our knowledge, we are the first to report whole transcriptomic profiling from a NHP model of RIBI after fWBI and to assess transcriptomic alterations in NHP survivors of single-fraction high-dose TBI (6.75–8.05 Gy). We present novel evidence of CNS injury years after TBI that shares gene expression patterns with fWBI. These analyses reaffirm the involvement of macrophage/microglial-mediated neuroinflammation in RIBI and indicate novel evidence that complement system activation and impaired synaptic neurotransmission within white matter may contribute to RIBI after fWBI and TBI. Our regional analyses challenge the belief that alterations in hippocampal structure and function are solely responsible for cognitive impairment in RIBI, and provide evidence that alterations in neurotransmission may be most prominent within white matter.

Supplementary Material

Supplementary file

Table S1. Unmappable Ensembl ID.

Table S2. Contributory genes to the top 10 IPA pathways, by region.

Fig. S1. Additional differentially regulated neurotransmission and signal transduction-associated transcripts in white matter after brain irradiation. Several neurotransmission and signal transduction transcripts were downregulated in white matter after brain irradiation. Potassium two pore domain channel subfamily K member (KCNK5) was upregulated within both irradiated groups. Differential expressed genes are listed in order of direction and then alphabetically within treatment groups. Bar: Mean log₂ fold change. Error bars: log fold change standard error. *Log₂ fold changes < ± 1 and multiplicity adjusted P values (Benjamini-Hochberg) < 0.05.