Research ArticleTraumatic Brain Injury

Repetitive blast exposure in mice and combat veterans causes persistent cerebellar dysfunction

See allHide authors and affiliations

Science Translational Medicine  13 Jan 2016:
Vol. 8, Issue 321, pp. 321ra6
DOI: 10.1126/scitranslmed.aaa9585

The cerebellum is vulnerable to blast injury in mice and combat veterans

Mild traumatic brain injury (TBI) is often referred to as the “signature injury” of the wars in Iraq and Afghanistan. Most of these TBIs are blast-related. Currently, there is limited understanding of how mild blast causes persistent brain injuries. There is also limited insight into how blast-induced brain injuries in animal models correspond to humans with mild TBI. Meabon et al. report that the cerebellum, a brain structure important for integrating sensory information and movement, is injured by blast exposure in mice in specific areas that correspond to abnormal brain imaging findings obtained in similar cerebellar regions in blast-exposed combat veterans.


Blast exposure can cause mild traumatic brain injury (TBI) in mice and other mammals. However, there are important gaps in our understanding of the neuropathology underlying repetitive blast exposure in animal models compared to the neuroimaging abnormalities observed in blast-exposed veterans. Moreover, how an increase in the number of blast exposures affects neuroimaging endpoints in blast-exposed humans is not well understood. We asked whether there is a dose-response relationship between the number of blast-related mild TBIs and uptake of 18F-fluorodeoxyglucose (FDG), a commonly used indicator of neuronal activity, in the brains of blast-exposed veterans with mild TBI. We found that the number of blast exposures correlated with FDG uptake in the cerebellum of veterans. In mice, blast exposure produced microlesions in the blood-brain barrier (BBB) predominantly in the ventral cerebellum. Purkinje cells associated with these BBB microlesions displayed plasma membrane disruptions and aberrant expression of phosphorylated tau protein. Purkinje cell loss was most pronounced in the ventral cerebellar lobules, suggesting that early-stage breakdown of BBB integrity may be an important factor driving long-term brain changes. Blast exposure caused reactive gliosis in mouse cerebellum, particularly in the deep cerebellar nuclei. Diffusion tensor imaging tractography of the cerebellum of blast-exposed veterans revealed that mean diffusivity correlated negatively with the number of blast-related mild TBIs. Together, these results argue that the cerebellum is vulnerable to repetitive mild TBI in both mice and humans.


There is mounting concern that blast exposure may initiate latent pathological processes leading to chronic traumatic encephalopathy or other related neurodegenerative disorders (16). Detonation of high explosives can inflict brain injury in at least three ways: primary blast effects due to blast overpressure, secondary blast effects caused by fragmented objects or shrapnel inflicting trauma when hitting the head, and tertiary effects caused by the head striking other objects (7). Even without direct blunt impacts or significant head acceleration/deceleration from secondary and tertiary impacts (8), blast overpressure is capable of injuring the brain (7, 915).

The prevalence of repetitive blast-related mild traumatic brain injuries (TBIs) among Iraq and Afghanistan veterans (16, 17) has spurred intensive efforts to understand its cognitive and pathological consequences in both humans and animal models. Studies in mice, rats, swine, and other mammals have demonstrated that the intense energy imparted by blast overpressure is sufficient to induce an array of neuropathological and behavioral disturbances (7, 914, 1821).

Neuroimaging studies of military service members and veterans have documented chronic abnormalities on 18F-fluorodeoxyglucose (FDG) positron emission tomography (PET) imaging, which assesses glucose metabolic activity in the brain (4, 22, 23). Abnormalities of brain white matter tract structural integrity and myelin density in these veteran populations have also been characterized using magnetic resonance diffusion tensor imaging (DTI) and macromolecular proton fraction mapping (5, 2426).

Multiple research groups, including ours, have shown that blast exposure in mice permits study of well-characterized mild to moderate effects of blast overpressure with attenuated secondary and tertiary injury components, thereby affording new opportunities for mechanistic pathological and genetic investigations that are difficult to achieve in any other species (7, 915). Nonetheless, it remains challenging to assess potential correlations between this animal model and neuroimaging findings in patients with blast-related mild TBIs and blast-induced neuropathology in specific anatomical brain regions. This is due in part to interspecies differences in cerebral cortical gray matter and subcortical white matter anatomy between humans (a gyrencephalic species) and rodents (a lissencephalic species).

The cerebellum, however, is gyrencephalic in both humans and rodents and has broadly similar structural organization in both species. Findings from previous neuropathological studies in rodents (7, 915, 27) and neuroimaging studies in human subjects (35, 26, 2831) suggest that the cerebellum is susceptible to blast injury. A growing body of data suggest that the cerebellum, in addition to controlling motor-related functions, subserves important cognitive and affective functions that may be specifically related to cognitive and behavioral symptoms in blast-exposed individuals (3235). Therefore, we investigated the consequences of repetitive blast exposure in the cerebellum of mice and humans.


An increase in the number of blast exposures is associated with lower cerebellar glucose metabolism in veterans with blast-related mild TBI

Several FDG-PET studies have demonstrated lower metabolic activity in a number of brain regions including the cerebellum in veterans with blast-related mild TBI compared to veterans without blast exposure (5, 22) or nonmilitary controls (4, 36). Nonetheless, the significance of graded increases in the number of blast exposures on subsequent chronic brain glucose hypometabolism has not been addressed directly. Among our study cohort of blast-exposed veterans with mild TBI (Table 1), the number of blasts experienced by each individual varied considerably, ranging from 1 to >100 reported blast events that caused acute symptoms consistent with mild TBI. This wide range in the number of blast-related mild TBIs in a well-characterized cohort (5) presented an opportunity to test the hypothesis that an increasing number of blast events would be associated with a greater degree of chronic brain glucose hypometabolism and would allow us to address the possibility that the cerebellum may be vulnerable to injury from repetitive blast exposure.

Table 1. FDG-PET (n = 33). PSQI, Pittsburgh Sleep Quality Index; NSI, Neurobehavioral Symptom Inventory.
View this table:

Here, blast-exposed veterans reported experiencing an average of 21 (median, 11; range, 1 to 102) blast-related mild TBIs during the entire period of their military service (Table 1). Among this group, 21, 52, and 27% reported 1 to 5, 6 to 20, and 21 to 102 blast-related mild TBIs, respectively. The number of mild TBIs associated with loss of consciousness from any cause was relatively low (average of 2.3 lifetime mild TBIs; Table 1). The average time from the last blast-related mild TBI to imaging with FDG-PET was 3.6 years.

Figure 1A shows a statistical parametric map of the mean voxel-wise correlations between brain metabolic activity (normalized to global activity) and log10-transformed number of blast exposures among the 33 blast-exposed veterans. The brightest colors (yellow, z score range from −4 to −5) denote brain regions with the strongest negative correlations between blast number and glucose metabolism. Statistical significance following correction for multiple comparisons was achieved in three brain areas, all of which were localized within the right inferior semilunar lobule of the cerebellum (z scores, 4.02, 4.09, and 4.37, respectively; P ≤ 0.05; see Supplementary Methods).

Fig. 1. Number of blast exposures is associated with decreased FDG uptake in the cerebellum of veterans with mild TBI.

(A) Distribution and relative magnitude of negative correlations between log10 number of blast exposures and FDG uptake by the human brain (normalized to global activity) overlaid on magnetic resonance template images for anatomical orientation. Shown are right lateral (RT LAT), left lateral (LT LAT), right medial (RT MED), left medial (LT MED), and inferior (INF) aspects of the human brain. Scale bar indicates z scores. A greater number of log10 blast exposures were associated with decreased metabolic activity in three adjacent loci within the right inferior semilunar lobule by whole-brain voxel-wise analyses. (B) Scatter plots of the mean activity in predefined cerebellar hemisphere volumes of interest (VOIs; left and right cerebellum) versus log10 number of blast exposures during military service. There was a statistically significant negative correlation between the number of blast exposures (log10) and FDG uptake in each cerebellar hemisphere VOI (Spearman r correlations, n = 33).

Using an alternative, independent analytical approach, we also correlated blast number (log10) versus mean metabolic activity in regions of interest circumscribing the entire left and right cerebellar hemispheres. Figure 1B shows that this method also revealed a statistically significant negative correlation between the number of blast-related mild TBIs and cerebellar metabolism [Spearman r = −0.433 (P ≤ 0.012) and −0.406 (P ≤ 0.019) for left and right cerebellar hemispheres, respectively].

Because posttraumatic stress disorder (PTSD), depression, and alcohol use are common in this cohort of blast-exposed veterans (4, 37) and may contribute to hypometabolism in the central nervous system (CNS), we performed multivariate regression analysis to control for potential confounding effects. We modeled FDG uptake in the left and right cerebellar hemispheres as a function of blast number (log10) and controlled for PTSD with the Clinician-Administered PTSD (CAPS) total score, Patient Health Questionnaire-9 (PHQ-9), and alcohol use with the Alcohol Use Disorders Identification Test—Consumption (AUDIT-C). The associations between blast number and decreased FDG uptake in both left and right hemispheres of the cerebellum remained statistically significant (log10 blast exposure, P ≤ 0.01, and log10 blast exposure, P < 0.02, left and right cerebellar hemispheres, respectively). Also, none of these factors was significantly correlated with metabolic activity in either cerebellar hemisphere (P > 0.05). Thus, the associations between the number of blast-related mild TBIs and reduced normalized FDG uptake in the cerebellum of veterans exposed to mild TBIs were not likely to be due to comorbid PTSD, depression, or alcohol use.

Repetitive blast exposure causes chronic neuronal loss and persistent synaptic disturbances in murine cerebellum

Our FDG-PET findings (Fig. 1) suggested the possibility that the cerebellum may be especially vulnerable to the cumulative effects of repetitive blast exposure. In addition, the correlation of the number of blast exposures and glucose hypometabolism in the cerebellum (Fig. 1A) raised the possibility of subregional vulnerability within the cerebellum to blast exposure.

To address these issues in more detail, we used a mouse model of blast-induced TBI that mimics battlefield-relevant blast overpressure in keeping with well-established approaches. We used blast overpressure parameters consistent with mild to moderate blast exposure (9, 38, 39). Briefly, 3-month-old male C57BL/6 mice were exposed to blasts using a shock tube and exposure parameters described in detail elsewhere (9). Here, mice were exposed either to one (1×) or three (3×) blast overpressure events (24-hour interblast intervals), which were highly reproducible, with a mean peak intensity of 19.2 psi ± 0.21, initial peak wave duration of 5.8 ± 0.028 ms, and a calculated impulse of 30.7 pressure per square inch per millisecond (Fig. 2A). Under these experimental conditions, the overall survival rate was 99%, with blast-exposed mice appearing comparable to sham-exposed animals by inspection 2 hours after blast exposure. Sham-exposed control mice were treated identically to animals exposed to blast overpressure events: animals were anesthetized for 3 to 5 min with isoflurane and mounted in the same shock tube restraint harness for the same amount of time. Each sham-exposed control animal was yoked to a blast-exposed animal, thereby receiving either one or three sham treatments.

Fig. 2. A single blast causes inferior cerebellar BBB dysfunction and Purkinje cell injury in a mouse model of mild TBI.

(A) Shock tube–generated highly reproducible battlefield-relevant blast overpressure. The red trace shows the time (ms) versus pressure (psi) measured 5 cm above the animal averaged among n = 32 blast overpressure waveforms that were collected over a randomly selected 2-day period (black bars indicate ±SEM). Superimposed blue Friedlander waveform illustrates the close correspondence of the shock tube–generated blast overpressure to the waveform expected from an open-field detonation of about 22.4 kg of trinitrotoluene (TNT) at a distance of 6.1m. (B) Blast-exposed mice were retro-orbitally loaded with 10-kD dextran labeled with Texas Red; they were allowed to recover for 1 or 4 hours after a single blast exposure, and brain slices were observed using confocal microscopy. Blast-specific dextran (red) labeling was most frequently observed in the cerebellum (white arrowheads). Shown is an example of a mouse brain slice (50 μm thickness) 1 hour after treatment. (C) Texas Red–labeled dextran in the blood that crossed the BBB into the cerebellar parenchyma was observed at both 1 and 4 hours after blast exposure (red highlights). Extravasated dextran was most noticeable in the cerebellar molecular layer. DAPI, 4′,6-diamidino-2-phenylindole. (D) Blast-exposed Purkinje cells that internalized dextran that had crossed the BBB were associated with morphologically distinct microglia (green) that appeared to be activated. (n = 20 blast-treated animals, 19 of 73 brain slices had dextran-labeled Purkinje cells with only one cell (a granule cell) labeled in the upper half of the cerebellum; n = 20 sham-treated animals, with 0 of 75 slices showing dextran-positive Purkinje cells. GFAP, glial fibrillary acidic protein. (E) To assess regional BBB breakdown, the linear pixel-by-pixel mean dextran fluorescence intensity within the molecular layer of the cerebellum was measured across each lobe and plotted. Results show the mean normalized with respect to values for seven sham-treated and seven blast-treated animals. Only lobule 9 showed evidence of increased dextran fluorescence (P ≤ 0.05, two-tailed t test; mean = 1.65 and 1.0 for blast- and sham-treated animals, respectively). (F) Confocal microscopy revealed elevated aberrant tau phosphorylated at Ser396 (detected with mouse monoclonal anti-tau396 antibody) in blast-treated versus sham-treated cerebellum. (G) Confocal microscopy revealed elevated aberrant tau phosphorylated at Ser396 (detected with rabbit anti-tau396 antibody) in blast-treated versus sham-treated cerebellum. (H) In mouse cerebellum, multiple pathological forms of phosphorylated tau were observed 1 to 4 hours after blast exposure. Representative Western blots illustrate blast-induced elevation of phosphorylated tau in mouse cerebellum. Each lane corresponds to one sham-treated or blast-exposed animal (1 hour: n = 4, 4 hours: n = 4; n = 3 sham-treated and n = 4 blast-exposed). Pyruvate kinase (PK) was the gel protein loading control. (I) Three days after experimental treatment, Purkinje cell injury (black arrowheads) was observed in the form of swollen cell bodies, cell membrane breakdown, and aberrant IP3R1 staining in two of three blast-exposed animals. Similar Purkinje cell dysmorphology was not observed in sham-treated animals (n = 0/3). Images are representative examples from lobule 9, taken at 20×. Slides were prepared using antibodies against IP3R1 and standard immunohistochemistry. Scale bars, 500 μm (B), 250 μm (C), and 50 μm (D, F, and G). (F) and (G) represent data from five sham-treated and five blast-exposed animals. Error bars represent SEM. *P ≤ 0.05, two-tailed t test.

First, we examined mice exposed to one blast that were administered Texas Red–conjugated 10-kD dextran immediately before treatment. A sagittal section of the brain following clearance of peripheral circulating dextran revealed domains of breakdown in the blood-brain barrier (BBB) as shown by regions of parenchymal dextran accumulation (Fig. 2B). At lower magnification, dextran accumulation was most obvious in the cerebellum, particularly in the more ventral regions (Fig. 2B). Figure 2C shows cerebellar lobule 9 of a blast-exposed mouse displaying distinct regions of BBB breakdown.

Four-channel confocal microscopy (Fig. 2D) carried out in these domains of BBB breakdown revealed that within 1 to 4 hours after a single blast, individual cerebellar Purkinje cells took up dextran (red staining) that had crossed the BBB. It has been reported previously that neuronal uptake of dextran exogenously applied directly onto the brain following TBI is associated with disturbances of the neuronal plasma membrane (40). Consistent with this, in our study, dextran uptake by Purkinje cells was evident in blast-exposed but not sham-treated animals (Fig. 2D). In addition, within the same domain in which dextran had crossed the BBB, other Purkinje cells immediately adjacent to the dextran-positive neuron were also exposed to parenchymal dextran but did not accumulate it. Although the dextran-positive Purkinje cell in Fig. 2D appeared morphologically normal following blast exposure, accumulation of extracellular parenchymal dextran indicated early-stage neuronal dysfunction. To obtain initial insight into the loss of BBB integrity in different cerebellar lobules, Fig. 2E shows the results of an analysis where a line of interest (see Supplementary Methods) was traced along the cerebellar molecular layer of sagittal brain sections from dextran-treated blast-exposed and sham-treated animals. Figure 2E shows that lobule by lobule, relative to sham-treated mice, the amount of Texas Red–conjugated 10-kD dextran that crossed the BBB into the brain parenchyma was greater in blast-exposed animals, with lobule 9 showing the greatest effect (t12 = 2.15, P ≤ 0.05, n = 7, blast-exposed and sham-exposed, respectively). Figure 2F shows representative confocal images from a sham-exposed and a blast-exposed animal, where Texas Red–conjugated dextran crossed the BBB into the cerebellar parenchyma with subsequent Purkinje cell dextran uptake 1 hour after a single blast exposure. This was accompanied by aberrant expression of phosphorylated tau in brain parenchyma. Aberrant expression of phosphorylated tau was further confirmed using a different rabbit anti-tau antibody in the same mice (Fig. 2G). In addition, Western blot analysis confirmed that a single blast induced increased expression of multiple phosphorylated tau species in the cerebellum within 1 hour (Fig. 2H). Three days after a single blast exposure, 1,4,5-trisphosphate receptor type 1 (IP3R1) immunostaining (a morphological marker of Purkinje cells) in lobule 9 (Fig. 2I) revealed dystrophic Purkinje cells (black arrowheads). Together, these data suggest that blast exposure disturbs cerebellar BBB integrity, which is associated with early accumulation of aberrant phosphorylated tau and neuronal plasma membrane dysfunction.

The FDG-PET findings in the blast-exposed veterans argue that the chronic consequences of repetitive blast exposure are demonstrable in the cerebellum (Fig. 1). The findings in blast-exposed mice indicate that even a single blast is sufficient to evoke aberrant phosphorylated tau accumulation and uptake of microvessel-derived dextran indicative of neuronal injury (Fig. 2). To address directly whether repetitive blast exposure causes permanent and specific subregional cerebellar Purkinje cell loss, we quantified “empty baskets” (a term referring to the dense axonal terminals from basket cells enveloping Purkinje cell bodies that persist long after Purkinje cells have been lost) in mice 30 days after they were exposed to either one or three blasts.

Fixed cerebellar sections (50 μm) were serially imaged by confocal microscopy (22 scans at 2.27-μm intervals in the z plane). Figure 3A shows a merged maximum field projection with orthogonal views of lobule 9 in the x-y plane with basket cell processes (immunostained green with anti-neurofilament heavy chain antibodies) enveloping the soma of Purkinje cells (immunostained red with anti-IP3R1 antibodies). The corresponding serially reconstructed z-plane image stacks along the indicated image planes (white lines) are depicted above and to the right of merged x-y images. We observed distributed, patchy microdomains of empty baskets in mice exposed to three blasts compared to sham-exposed animals 30 days after exposure.

Fig. 3. Purkinje cell loss increases with the number of blast exposures in mice.

(A) Representative confocal images of cerebellar lobule 9 in sham- and 3× blast–treated mice 30 days after treatment. Images show a merged maximum projection of 22 scans acquired at 2.27-μm intervals in the z plane immunostained for neurofilament heavy chain (NF-HC)–positive basket cells (green) and IP3R1-positive Purkinje cells (red). Blast-exposed cerebellum displayed prominent microdomains of Purkinje cell loss (empty baskets) that were also apparent when viewed along the z plane for highlighted regions (white boxes) projected above and to the right of the merged x-y images. (B) For illustrative purposes, this sagittal section denotes cerebellar lobule anatomy (lobules 1/2 to 10). (C) Histogram shows empty basket frequencies in 1× blast– and 3× blast–exposed mice normalized with respect to the number of empty baskets observed in sham-treated mice (n = 10, 7, and 12 for sham-treated, 1× blast–exposed, and 3× blast–exposed mice, respectively). In both dorsal/superior (lobules 1 to 6) and ventral/inferior (lobules 7 to 10) regions of the cerebellum, blast exposure caused an overall increase in empty baskets [P < 0.01 and P < 0.0001, analyses of variance (ANOVAs), respectively]. 3× blast exposure caused marked Purkinje cell loss throughout the cerebellum. However, the ventral/inferior lobules appeared more vulnerable to Purkinje cell loss after a single blast exposure. **P < 0.01, Helmert test. (D) Histogram shows normalized empty baskets for each lobule in the same sham-treated, 1× blast–exposed, and 3× blast–exposed mice. *P < 0.05, **P < 0.01, ANOVA. n.s., nonsignificant. Scale bars, 50 μm (A) and 500 μm (B). Data represent mean normalized to values for sham-treated mice, with error bars indicating ±SEM.

Thirty days after treatment, empty baskets were quantified in each of the eight major cerebellar lobules (Fig. 3B) in 1× blast– and 3× blast–exposed mice. Quantification of empty baskets in sham-exposed mice revealed no statistically significant differences between 1× and 3× animals in any cerebellar lobule. Therefore, these groups were combined as a single composite sham group for each lobule. Results among the 1× blast– and 3× blast–exposed mice are presented as a ratio of empty baskets in blast-exposed versus sham-exposed mice (Fig. 3, C and D). Analysis of the results shown in Fig. 3C revealed a statistically significant effect of repetitive blast exposure in both the ventral cerebellar lobules (7 to 10) and the dorsal cerebellar lobules (1 to 6) (F2,26 = 14.997, P ≤ 0.0001, and F2,26 = 6.356, P ≤ 0.01, respectively). This confirmed that repetitive blast exposure produced chronic cerebellar Purkinje cell loss. Further inspection also revealed that although 3× blast exposure caused Purkinje cell loss throughout the cerebellum, the ventral lobules appeared to be more vulnerable to Purkinje cell loss after a single blast exposure. This was confirmed by a planned Helmert comparison test of blast (1× and 3×) versus sham, which showed statistically significant differences in Purkinje cell loss in the ventral (P < 0.01) but not the dorsal cerebellum.

ANOVA tests of empty baskets in cerebellar lobules of sham-, 1× blast–, and 3× blast–exposed mice (Fig. 3D) further confirmed that repetitive blast exposure caused Purkinje cell loss throughout the cerebellum. However, the more ventral subregions of the cerebellum were more vulnerable to Purkinje cell loss: lobules 1/2, F2,22 = 2.678, n.s.; lobule 3, F2,25 = 3.617, P ≤ 0.042; lobules 4/5, F2,23 = 2.372, n.s.; lobule 6, F2,26 = 1.857, n.s.; lobule 7, F2,23 = 3.562, P ≤ 0.045; lobule 8, F2,26 = 4.369, P ≤ 0.023; lobule 9, F2,22 = 6.738, P ≤ 0.005; lobule 10, F2,24 = 7.298, P ≤ 0.003.

In keeping with the Purkinje cell loss findings, Fig. 4 shows that 30 days after blast exposure, the mice showed impaired rotarod performance in the 3× but not the 1× blast group (F1,37 = 0.423, n.s., 1× blast; F1,6 = 7.477, P ≤ 0.034, 3× blast). In addition, a two-way mixed between/within-subjects ANOVA comparing 1×, 3×, and sham controls (1× sham– and 3× sham–treated mice were not significantly different and thus were pooled) further confirmed that repetitive blast exposure decreased motor performance on the rotarod test (F2,44 = 5.603, P ≤ 0.007).

Fig. 4. Repetitive blast exposure impairs motor performance in mice.

(A) Single blast exposure (1×) did not alter rotarod performance in mice at 30 days after treatment (n = 20 sham-treated, n = 19 blast-exposed mice). (B) In contrast, mice exposed to 3× blast exposure showed significant impairment in rotarod performance compared to sham-treated mice (n = 4 sham-treated, n = 5 blast-exposed; P ≤ 0.034, ANOVA). Data represent mean, with error bars indicating ±SEM.

To address the consequences of blast exposure on synaptic proteins, we examined the expression of the glutamatergic synaptic scaffolding protein PSD-95 (postsynaptic density-95) (41) and of GABAB-R1 (γ-aminobutyric acid B receptor 1) by Western blotting (Fig. 5, A and B). One month after a single blast exposure, PSD-95 protein expression, but not GABAB-R1 expression, was significantly reduced compared to sham-treated animals (t8 = −5.988, P ≤ 0.0003; t8 = −0.003, n.s., respectively). Additional experiments in Fig. 5C showed that blast-induced loss of PSD-95 expression in the cerebellum developed slowly. There were no significant differences between sham-treated and 1× blast–exposed animals at 1 day after exposure (F1,7 = 0.262, n.s.). However, at the 30-day time point, PSD-95 expression was significantly reduced in 1× blast– and 3× blast–exposed animals compared to sham-treated controls (F2,26 = 6.392, P ≤ 0.006). Confocal microscopy carried out 120 days after 3× blast exposure also indicated that PSD-95 expression within the cerebellar molecular layer was reduced compared to 3× sham controls (Fig. 5, D and E).

Fig. 5. Blast exposure reduces PSD-95 expression in mouse cerebellum.

(A) Western blots for PSD-95 and GABAB-R1 performed at 30 days after treatment showed reduced PSD-95, but not GABAB-R1, in 1× blast-exposed mice compared to sham-treated animals. Each lane shows Triton X-100–soluble cerebellar protein lysates (20 μg), with one animal per lane using PK as a gel protein loading control. (B) Histograms show densitometric quantification of Western blots in (A) normalized to sham-treated mice, which are considered to be 1.0 (that is, 100%). Expression of PSD-95, but not GABAB-R1, was reduced by blast exposure (P ≤ 0.0003, t test; n = 5 sham-treated and n = 5 blast-exposed mice). (C) Histogram shows normalized PSD-95 expression from Western blot experiments. One day after 1× blast exposure, PSD-95 expression was not different from sham controls (ANOVA, n.s.; n = 5 sham-treated and n = 4 blast-exposed mice). However, 30 days after blast exposure, PSD-95 expression was significantly reduced in blast-exposed animals compared to sham controls [P ≤ 0.006, ANOVA; n = 14, 11, and 4, sham-treated, 1× blast–exposed, and 3× blast–exposed mice, respectively; includes data pooled from (B) and (C)]. (D) Confocal microscopy confirmed decreased PSD-95 immunostaining in cerebellum of 3× blast–exposed mice 4 months after treatment. (E) At higher magnification, different sections from (D) showed reduced PSD-95 immunopositive puncta in the cerebellar molecular layer (white arrows) of 3× blast–exposed mice 4 months after treatment compared to sham-treated mice. Results are representative of n = 5 blast-exposed and n = 5 sham-treated mice. Scale bars, 500 μm (D) and 50 μm (E).

Repetitive blast exposure causes persistent cerebellar white matter and glial neuropathology

To investigate the consequences of repetitive blast exposure on cerebellar white matter tracts in mice, we first performed Bielschowsky silver staining, which revealed aberrant axonal varicosities and thickened neuronal processes within the granule layer adjoining the deep white matter tracts and the Purkinje cell layer 24 hours after a single blast exposure (Fig. 6, A and B). To further investigate the effects of repetitive blast exposure in cerebellar white matter, we also measured amyloid precursor protein (APP) expression, which was easier to quantify than Bielschowsky staining (42, 43). At both 24 hours and 7 days after a single blast exposure, APP expression was increased compared to sham-exposed animals but then returned to sham levels by 30 days after exposure (Fig. 6, C and D). However, 30 days after a 3× blast exposure, APP expression in the cerebellum remained elevated. A two-factor ANOVA comparing treatment groups (sham-treated, 1× blast, and 3× blast) with respect to time after exposure (1, 7, and 30 days) confirmed that blast exposure significantly increased APP expression in the cerebellum (treatment group factor: F2,31 = 9.923, P ≤ 0.0005; time post-exposure factor: F2,31 = 1.310, n.s.). In addition, a one-factor ANOVA comparing the overall effects of the number of blast exposures (sham, 1× blast, and 3× blast) was also statistically significant (F2,35 = 8.048, P ≤ 0.001).

Fig. 6. Blast exposure induces persistent axonal injury in mice.

(A) Bielschowsky silver staining of cerebellar slices from single blast-exposed mice at 24 hours shows thickened axonal varicosities evident in axons emanating from the Purkinje cell layer (n = 5/6 blast-exposed and n = 1/5 sham-treated mice). (B) Enlarged insets from (A) numbered 1 to 4 (arrowheads denote axonal varicosities). (C) Representative Western blot analyses from individual blast-treated and sham-exposed mice showed that blast exposure increased APP expression in cerebellum. Each lane represents Triton X-100–soluble protein lysates (20 μg total protein per lane) for a single animal using PK or SYPRO-Ruby (SR) as a total protein loading control. (D) Histogram indicates APP normalized to values for sham-treated mice, which are considered to be 100%. A two-factor ANOVA comparing all groups (sham-treated, 1× blast–exposed, and 3× blast–exposed mice) with respect to posttreatment delay (1, 7, and 30 days) confirmed that blast exposure increased APP in the cerebellum (ANOVA: treatment group, P ≤ 0.0005). Each group n = 5, except 1× blast–exposed and 1× sham–treated, which were n = 4. Error bars indicate ±SEM. Scale bars, 50 μm (A) and 10 μm (B).

To investigate prolonged consequences of blast exposure, we next examined GFAP immunoreactivity in the mouse cerebellum by confocal microscopy 30 days after one or three blast exposures. Persistently elevated GFAP, an indication of inflammatory reactive astrocytes, was apparent throughout the cerebellar white matter tracts (Fig. 7A). Quantification of relative GFAP immunoreactivity (Fig. 7B) circumscribing both the lobular and peri-deep cerebellar nuclei white matter tracts confirmed that GFAP expression was significantly increased 30 days after blast exposure (F2,18 = 21.055, P < 0.001). In addition, at higher magnification (Fig. 7C), GFAP expression was found in white matter regions adjacent to deep cerebellar nuclei with statistically significant increases in blast-exposed mice compared to sham-exposed animals (F2,18 = 25.990, P < 0.001) (Fig. 7D).

Fig. 7. Blast exposure induces persistent reactive gliosis in mouse cerebellum.

(A) Representative confocal images (20 merged scans at 2.5 μm along the z axis) show persistently elevated GFAP expression (red) in cerebellar white matter 30 days after 1× and 3× blast–exposure. White lines indicate areas of interest (AOI) delineated by the interior border of the cerebellar granule cell layer. (B) GFAP immunoreactivity (normalized with respect to sham-treated mice) in cerebellar AOIs circumscribing overall cerebellar white matter was increased in the blast-exposed animals (P ≤ 0.001, ANOVA; n = 8, 10, and 3 sham-treated, 1× blast–exposed and 3× blast–exposed mice, respectively). (C) In these slices, GFAP immunostaining in white matter surrounding deep cerebellar nuclei/dentate nuclei (DCN/DN) was increased (see insets). (D) GFAP immunoreactivity (normalized with respect to sham-treated) in cerebellum surrounding the deep cerebellar nuclei/dentate nuclei was increased in the blast-exposed animals (P ≤ 0.001, ANOVA; n = 8, 10, and 3 sham-treated, 1× blast–exposed, and 3× blast–exposed mice, respectively). (E) Fourteen days after treatment, Iba-1 staining (green) in the brains of 3× blast–exposed mice shows clusters of Iba-1–positive cells (green). Highlighted regions (white boxes) are shown at higher magnification (far right panels). Iba-1 immunoreactivity does not colocalize with GFAP-positive astrogliosis (red). (F) The cumulative distribution of microglial/macrophage clusters was observed in the brains of individual animals (with sample size noted in panels) at 14 and 30 days after exposure. At 14 days, microglial clusters were observed with a predominant distribution in the cerebellar nuclei and white matter, with occurrences increasing near the peduncle. The number of clusters increased with the number of blast exposures and diminished 30 days after exposure. (G) The overall volume of microglial processes was quantified by calculating the convex hull volume (right panels) of the Iba-1 immuno-positive cells in 285-μm2 (x, y plane) × 50-μm (z plane) images from deep cerebellar white matter from sham-treated, 1× blast–exposed, and 3× blast–exposed animals 14 days after treatment (left panel shows representative maximum field projections). (H) Histogram shows that microglial convex hull volume was reduced in blast-exposed animals compared to sham-treated mice (P ≤ 0.04; n = 8, 6, and 12, 1× blast, 3× blast, and sham-treated mice, respectively). (I) Histogram of Western blot densitometry of CD68 expression (an indication of microglial/macrophage activation) in cerebellar protein lysates in blast-exposed mice (ANOVA: P ≤ 0.01; n = 9, 5, and 5, sham-treated, 1× blast–exposed, and 3× blast–exposed mice, respectively) 14 days after exposure. Total protein was normalized with respect to reprobing of PK immunoreactivity. (J) Double-label confocal microscopy shows Iba-1–positive microglial cells (green) coexpressing CD68 (red). Data represent mean (normalized to sham values), with error bars indicating ± SEM. Scale bars, 250 μm (A and C), 100 μm (E, left panels, and G), 50 μm (E, right panels), and 25 μm (J). WM, white matter; ML, molecular layer; GCL, granule cell layer.

To further investigate blast-induced neuroinflammation-related responses, we also carried out Iba-1 immunostaining of the mouse cerebellum. Figure 7E shows that 3× blast exposure evoked distinct patches of microglia/macrophages with a morphology indicative of activation. Figure 7F shows the cumulative anatomical distribution of these microglial patches in 1× blast, 3× blast, and sham controls at 14 and 30 days after treatment. Blast-induced microglial responses were particularly prominent in regions near the cerebellar peduncles of mice exposed to 3× blast at 14 days after exposure.

To address this in more detail, a neuropathologist blind to treatment condition evaluated 13 pairs of Iba-1–immunostained confocal images from the white matter superior to the deep cerebellar nuclei (1× blast, 3× blast, and corresponding shams, n = 5, n = 8, and n = 13, respectively). Blinded pairwise (sham versus blast) comparisons were judged on the basis of an expected overall appearance of reduced microglial process length/arborization and increased soma size in blast-exposed mice. The rater correctly assigned condition (blast versus sham) in 84.6% of comparisons (11 of 13 correct group assignments; χ2 = 4.9231, df = 1, P ≤ 0.027; with 4 of 5 and 7 of 8 correct responses for the 1× and 3× blast/sham groups, respectively). As a further confirmation, overall size of microglial processes was also examined by convex hull volume analysis that calculated the volume created by forming a convex envelope containing all of the cellular filaments associated with a given Iba-1–immunostained microglial cell. Figure 7G (right panel) shows convex hull volumes calculated using Imaris software from the representative Iba-1–immunostained confocal images in the left panel from white matter tracts superior to the deep cerebellar nuclei 2 weeks after sham, 1× blast, or 3× blast exposure. The summary histogram (Fig. 7H) indicates that convex hull volume was statistically significantly reduced in Iba-1–immunostained cells in blast-exposed animals (1× and 3×) compared to sham controls [F1,24 = 4.718, P ≤ 0.040; average number of cells examined per animal = 23 (range, 15 to 34); n = 8, n = 6, n = 12; 1× blast, 3× blast, and sham controls, respectively; with sham controls from 1× and 3× groups pooled together because they were not statistically different].

Together, these data argue that blast exposure provokes a more contracted morphology of microglia, suggesting that the microglia were activated. Consistent with this, we also observed a significant increase in CD68 expression, an indication of microglia/macrophage activation, on Western blots of cerebellum protein lysates obtained 2 weeks after blast exposure (Fig. 7I) (F2,16 = 6.359, P ≤ 0.01). Double-labeled fluorescent immunostaining also showed that Iba-1–immunopositive microglia/macrophages coexpressed CD68, further suggesting that blast exposure provoked microglial activation (Fig. 7J).

Number of blast-related mild TBIs correlates with mean diffusivity in deep cerebellar tracts of combat veterans

The findings in mice indicate that blast exposure can give rise to persistent neuropathology in cerebellar deep peri-peduncular/deep cerebellar nuclei regions. The results in Fig. 7 suggest that this area, which is the major route of fiber passage in and out of the cerebellum, is susceptible to injury caused by blast exposure. This prompted us to hypothesize that an increase in the number of blast exposures in combat veterans might correlate with structural neuroimaging measures that are sensitive to disturbances in the deep white matter tracts traversing the cerebellum.

To test this hypothesis, DTI tractography was carried out to define fiber tracts derived from a seed region located in areas surrounding the bilateral dentate nuclei. Out of the total cohort of blast-exposed veterans in this study, 19 participants had magnetic resonance imaging (MRI) scans that were suitable for tractography analysis (Table 2). Mean diffusivity, fractional anisotropy radial diffusivity, and axial diffusivity were calculated for three spherical 1-cm3 volumes of interest (VOIs) within fiber tracts arising from the seed point near the dentate nucleus (see Materials and Methods). In keeping with the findings in blast-exposed mice, VOI1 and VOI2 were located in the white matter tracts near the left and right dentate nuclei, and VOI3 was located in the middle cerebellar peduncle, a region that contains white matter tracts contiguous with those crossing through VOI1 and VOI2 (Fig. 8A) (44). The three VOIs were averaged together for each subject. Figure 8B shows that there was a statistically significant negative correlation between average mean diffusivity–VOI(1–3) and log10 number of blast exposures (rank-order Spearman r = −0.487, P ≤ 0.034). Fractional anisotropy, radial diffusivity, and axial diffusivity, as well as average track density were not significantly correlated with the number of blast exposures (Spearman r = −0.201, −0.340, −0.278, and −0.106, respectively; all n.s. P > 0.05).

Table 2. DTI tractography (n = 19).
View this table:
Fig. 8. Decreased mean diffusivity in human cerebellum correlates with an increase in the number of blast-related TBIs.

(A) Approximate spatial locations of three VOIs (1 to 3) used for quantification of DTI parameters are shown on a reference T1-weighted MRI scan (Montreal Neurological Institute −32 mm in z axis). (B) Scatter plot reveals a statistically significant negative correlation between log10 number of blast-related mild TBIs and mean diffusivity calculated as the average of VOI 1 to 3 (Spearman r, P ≤ 0.034, n = 19). (C) Examples of DTI fiber tracts arising from the seed point defined near the dentate nucleus for one VOI in a subject with 5 reported blast exposures (upper panel) and a subject with >100 reported blast exposures (lower panel). Color scale represents the maximum eigenvalue of the tensor that is related to mean diffusivity (MD). Compared to the subject with five blasts, the greater amount of green color in the highlighted regions (white arrows) in the >100 blast-exposed subject indicates a lower mean diffusivity. (D) Mean diffusivity axial images from the same two subjects shown in (C) depicted as a mean diffusivity map. The ovals highlight the same regions shown in (C), demonstrating lower mean diffusivity (more green) in the subject with >100 blast-related mild TBIs compared to the subject with 5 blast exposures. Color bars represent mean diffusivity values (mm2/s).

Whereas increased mean diffusivity is most often associated with demyelination and neuronal loss, recent reports suggest that presymptomatic inflammation-related processes, including gliosis occurring in advance of overt neurodegeneration, can restrict water diffusivity, thus resulting in reduced mean diffusivity (45) (see Discussion). Figure 8 (B and C) shows representative DTI tractography images from two blast-exposed individuals, one with 5 and the other with >100 blast exposures. Tract colors correspond to mean diffusivity values, with green indicating a lower mean diffusivity. In this example, the white matter tracts depicted in the subject exposed to more blasts appear to be more green in the highlighted region (arrow), indicative of lower mean diffusivity. As noted above, overall average track density for VOI(1–3) was not significantly correlated with the number of blast exposures. This suggests that apparent tract density differences in these two examples in this angle of view may not be representative of tract density in the group as a whole. An axial presentation of mean diffusivity with the tractography removed in the same two subjects (Fig. 8D) further corroborates that the tractography mean diffusivity observations (Fig. 8C) in the subject with >100 blast exposures were lower compared to those in the subject reporting only 5 blast exposures. The number of blast exposures (log10) did not correlate significantly with scores measuring PTSD (CAPS), depression (PHQ-9), and alcohol use (AUDIT-C) (Spearman correlation coefficients = −0.012, −0.027, and −0.199, respectively; all P > 0.05; Supplementary Materials). Multivariate regression analysis modeling of the average mean diffusivity–VOI(1–3) as a function of the number of blast exposures (log10) accounting for PTSD symptoms and PTSD-related factors (CAPS, PHQ-9, AUDIT-C) still resulted in a statistically significant correlation between log10 number of blast exposures and mean diffusivity (P ≤ 0.012), thus further strengthening the conclusion that increasing numbers of blast exposures may be associated with reduced mean diffusivity in white matter tracts of the inferior cerebellum.

Number of blast-related mild TBIs correlates with symptoms of impaired sensorimotor integration in combat veterans

Sensorimotor integration is a critical function of the cerebellum. To address whether symptoms related to impaired sensorimotor integration correlate with an increase in the number of blast exposures, we correlated symptoms of dizziness, loss of balance, and poor coordination (items of the NSI) in 41 blast-exposed veterans, with the number of blast-related mild TBIs (ranging from 1 to >100). Symptoms were rated by participants as none to severe (0 to 4). Table 3 shows that the number of blast exposures significantly correlated with sensorimotor symptoms of dizziness, loss of balance, and poor coordination (Spearman correlation: P ≤ 0.016, P ≤ 0.005, and P ≤ 0.019, respectively), as well as total symptoms (Spearman correlation: P ≤ 0.004). In addition, the association between total symptoms and the number of blast exposures remained statistically significant after adjusting for PTSD symptoms, depression, and alcohol use by multivariate regression analysis (P ≤ 0.02), thus arguing that these impairments are not likely to be the result of comorbid PTSD, depression, or alcohol use.

Table 3. Correlation between number (log10) blast exposures and symptoms related to sensorimotor integration (n = 41).

D, dizziness; LB, loss of balance; PC, poor coordination.

View this table:


Numerous reports indicate that blast-exposed individuals exhibit abnormalities in the brain after imaging with FDG-PET, DTI, DTI tractography, or macromolecular proton fraction mapping (35, 26, 31, 46, 47). This raises concerns that repetitive blast exposure may set in motion latent pathological processes that place these individuals at risk for developing neurodegenerative disorders in mid-to-late life, as has been argued for repetitive impact-related mild TBIs (4853).

A number of obstacles currently limit our understanding of the significance of aberrant neuroimaging findings in veterans with blast-related mild TBIs. First, with one exception (28), most of the subjects studied experienced multiple mild TBIs. Second, it can be challenging to assess the relative contributions of primary blast overpressure and secondary/tertiary (impact) injury to overall TBI symptoms and persisting neuroimaging abnormalities. Third, little is known regarding early-stage pathology among individuals who currently have only subtle cognitive symptoms yet nonetheless have abnormal neuroimaging findings (5).

To begin closing these important knowledge gaps, we studied the effects of increasing numbers of blast exposures on neuronal activity measured by FDG-PET. We also carried out DTI tractography to assess the structural integrity of tracts in the inferior cerebellum. We focused on the impact of repetitive blast in the cerebellum to facilitate potential translational correspondence between the neuropathological findings in blast-exposed mice and findings in blast-exposed veterans.

That unbiased whole-brain voxel-wise analyses demonstrated a marked dose-response relationship between increasing numbers of blast-related mild TBIs and decreased metabolic activity specifically in the cerebellum argues that the cerebellum may be particularly vulnerable to graded increases in the number of blast exposures. Our findings in the repetitive blast-exposed mice support this notion. We found that an increase in the number of blasts produced graded increases in Purkinje cell loss (Fig. 3). Although Purkinje cell loss was statistically significant, any specific injured neuron was surrounded by neighboring Purkinje cells that appeared, at least initially, to be intact. More extensive subclusters of empty baskets became apparent 30 days after mice received three blast exposures. Although the absolute number of lost Purkinje cells was small compared to the number of remaining morphologically intact cells, the extensive synaptic arbors supported by each Purkinje cell might be expected to amplify the functional consequences of individual Purkinje cell loss. In keeping with this, we observed slowly developing blast-induced reductions in the synaptic scaffolding protein, PSD-95, at least 4 months after blast exposure, further supporting that synapses in the cerebellum are lost or become chronically structurally abnormal.

We found that the ventral lobules (7 to 10) of the murine cerebellum are more vulnerable to blast-related injury compared to other regions of the cerebellum. This highly specific pattern of neuron loss corresponded with the overall pattern of BBB dysfunction and altered Purkinje cell plasma membrane integrity occurring within 1 to 4 hours of blast exposure. This suggested that early-occurring BBB disturbances may play a key role in subsequent chronic CNS injury. This emphasizes the importance of developing early treatment strategies to limit the pathophysiological consequences of even limited BBB injury to prevent long-term neuron loss.

The negative dose-response relationship between mild TBIs from blast exposure and metabolic activity assessed by FDG-PET was most pronounced in the inferior aspect of the cerebellum in blast-exposed veterans. Nonetheless, these correlations may also include contributions from more superior cerebellar domains, which may not be distinguishable due to the inherently lower spatial resolution and lack of neuroanatomical discrimination of FDG-PET imaging in human brains, particularly in inferior brain regions where slight co-registration mismatches may reduce localization specificity.

In landmark neuropathological studies, Corsellis and colleagues mapped Purkinje cell loss (54) in former boxers and found an anatomical pattern of loss that was strikingly similar to our findings in blast-exposed mice. Specifically, they found the most prominent Purkinje cell loss in the inferior/ventral aspects and, to a lesser extent, limited losses in more superior domains of the cerebellum (54). Purkinje cell damage was one of the most prominent neuropathological findings in nonhuman primates after exposure to explosive blasts that caused behavioral changes similar to humans with mild TBIs (55).

The neuroanatomical correspondence between blast-exposed mice and veterans with repetitive blast-related mild TBIs or former boxers with impact-related mild TBIs raises interesting questions about the mechanisms by which blast injures the cerebellum. It has been argued that the anatomical structure of the cerebellum may render it vulnerable to injury from shear stress or rapid acceleration/deceleration impacts (26, 5659), particularly impacts of the inferior cerebellar lobules against the posterior fossa and/or herniation of the cerebellar tonsils through the magnum foramen itself (54, 60). In the mouse model of blast-induced mild TBI used in this study, head movements were attenuated (9), consistent with previous studies using comparably restrained movement models of blast that produce neuropathology and behavioral deficits (7, 1113, 19, 39, 6167). The blast parameters we used were comparable to those estimated to be sufficient to produce brain injury in humans without substantial acceleration/deceleration (8). Nonetheless, it is not possible to rule out acceleration/deceleration as a contributing injury mechanism.

Factors other than biomechanical insults may also contribute to blast-related cerebellar injury. The high basal metabolic activity of Purkinje cells may influence their susceptibility to injury (68). This may partially explain their vulnerability to injury by ischemia and hypoxia (69). The microvasculature of the Purkinje cell layer renders it vulnerable to the development of ischemia during episodes of hypoperfusion (70), and vasospasm is uniquely associated with blast-related versus impact-related TBI (71). Thus, these factors also may contribute to cerebellar vulnerability to blast-related mild TBIs. In addition, unrelated to displacement or shear stress, the cerebellum is predisposed to CNS insults that involve susceptibility to BBB breakdown in the experimental autoimmune encephalomyelitis mouse model (72). Also, a large number of human genetic diseases target the cerebellum (73). Thus, it would be premature to conclude that any one factor governs susceptibility of the cerebellum to blast-related mild TBI. It is more likely that complex interactions among multiple factors underlie the correspondence we have found between blast-exposed mice and veterans with blast-related mild TBI, and that have previously been found in former boxers (54).

In addition to Purkinje cell and synaptic disturbances, we also observed increased cerebellar expression of APP, axonal varicosities in cerebellar white matter tracts, and persistent reactive gliosis in white matter tracts in proximity to the deep cerebellar nuclei. In keeping with previous findings in both humans and blast-exposed animals (26, 39, 55, 66, 74), these data suggest that deep cerebellar white matter tracts in blast-exposed veterans may also exhibit graded vulnerability to increasing numbers of blast-related mild TBIs. The DTI tractography results (Fig. 8) were consistent with this notion, as we observed a statistically significant negative correlation between blast number and mean diffusivity in these tracts.

In this regard, it is often reported that white matter damage, typically exemplified by distinct myelin disruption and axonal loss, is associated with increased, rather than decreased, mean diffusivity, because the loss of overall white matter integrity allows for more freely moving water molecules in all directions (7577). However, a growing body of data support that different types of white matter tract pathology can result in complex and sometimes opposing changes in mean diffusivity. For example, inflammatory processes such as reactive gliosis and microgliomatosis accompanied by increased cellularity can restrict water diffusivity (78). In some instances, a CNS insult results in an initial decrease in mean diffusivity or the apparent diffusion coefficient of water (ADCw), followed by increased diffusivity (79). Similar observations have recently been reported among individuals with presenilin-1 (PS1) mutations (45). Compared to normal controls, mean diffusivity was reduced in PS1 mutation carriers before onset of symptoms. However, mean diffusivity was increased in symptomatic Alzheimer’s disease (AD) patients with PS1 mutations compared to controls, thereby raising the possibility that the initial mean diffusivity decrease may reflect early-occurring glial activation or cellular swelling preceding overt AD-related pathology (45). Additional support for the idea that reactive gliosis attending neurodegenerative disease processes could be associated with decreased mean diffusivity comes from recent findings in patients with sporadic Creutzfeldt-Jakob disease (CJD) in whom DTI analyses revealed widespread decreased mean diffusivity in white matter tracts (80, 81). In addition, among a limited number of CJD cases in which it was possible to relate the mean diffusivity findings to histopathological examinations, reactive astrocytic gliosis and microglial activation appeared to be roughly associated with the degree of reduction in mean diffusivity (80).

These reports raise the possibility that our finding of lower mean diffusivity in deep cerebellar white matter in veterans with the largest number of blast exposures may indicate chronic subacute reactive gliosis and microgliomatosis. Nonetheless, in the absence of more direct neuropathological evidence and with the lack of correlated changes in fractional anisotropy, this possibility remains to be confirmed. Alternatively, the relationship between mean diffusivity and the number of blast exposures could be due to other factors that include impaired axonal transport, cellular swelling, or changes in crossing fiber anatomy (78, 80). It also remains unclear whether or not such underlying white matter DTI abnormalities in these blast-related mild TBI subjects are an early indication of latent disease processes that may later give rise to neurological symptoms.

Our study has limitations with respect to estimates of the number of blast exposures, which relied on retrospective recollection facilitated by a semistructured interview. However, unbiased whole-brain voxel-wise analysis yielded a correlation between the number of recalled blast exposures and glucose hypometabolism in the cerebellum. The number of blast exposures also correlated with self-reported sensorimotor symptoms. Another important limitation is that the possible link between decreased mean diffusivity on diffusion tensor images in the blast-exposed veterans and mild, chronic white matter neuroinflammation-related processes is only suggestive. Nonetheless, this possibility is supported by recent PET translocator protein imaging data in former National Football League players, suggesting that neuroinflammation is associated with repetitive impact mild TBI (82). Despite the cell loss seen in the cerebellum of blast-exposed mice, which is indicative of lasting CNS injury (Fig. 3), the microglial and astrocytic pathology we observed 30 days after repetitive blast exposure in the mice requires further study.

The deep cerebellar white matter tracts imaged here subserve functional networks connecting the cerebellum with brain regions that are critical to both cognitive and emotional function, including the prefrontal cortex and limbic regions (33, 83, 84). Thus, it is possible that the cerebellar white matter and Purkinje cell abnormalities observed in our study could disrupt cerebellar-cortical connections and thereby contribute to the chronic behavioral and postconcussive symptoms that are frequently observed in this population, as well as impaired sensorimotor integration. Continued longitudinal investigation of the veterans in this study is currently under way in an effort to address this important question.


Study design

The overall aim of this study was to investigate the neurological consequences of repetitive mild TBI in the cerebellum and to test the correlation between cerebellar injury in a mouse model of repetitive blast-induced mild TBI and neuroimaging findings in veterans with multiple blast-related mild TBI. A total of 41 veterans reporting from 1 to greater than 100 blast-related mild TBI events were studied. All participants were male. Females were eligible for study inclusion, but no females with blast-related mild TBI were enrolled. Inclusion criteria for all veteran participants included documented hazardous duty in Iraq and/or Afghanistan with the U.S. Armed Forces during Operation Iraqi Freedom and/or Operation Enduring Freedom. Veterans with mild TBI must have had exposure to at least one blast with acute symptoms that meet VA/Department of Defense/American Congress of Rehabilitation Medicine criteria for mild TBI. Exclusion criteria for all participants included moderate-severe TBI, seizure disorder, insulin-dependent diabetes, current DSM-IV (Diagnostic and Statistical Manual of Mental Disorders, 4th Edition) diagnosis of alcohol abuse or other substance abuse, schizophrenia or other psychotic disorders, bipolar disorder, or dementia, and taking medications likely to affect cognitive performance such as opiates, benzodiazepines, and sedating antihistamines. Veterans with retained shrapnel or metal fragments or foreign objects in the eyes, skin, or body that would contraindicate a brain MRI scan were excluded from MRI. All participants who had undergone FDG-PET scanning were included in the reported FDG-PET analysis. A subset of these underwent MRI scanning; all of those with MRI scans of sufficient quality for DTI tractography were included in this study. Sensorimotor-related NSI symptoms for all participants in this report are included.

A total of 175 male wild-type C57BL/6 mice ranging in age from 3 to 4 months were studied. Animals were randomly assigned to experimental (blast) and control (sham) groups. Experimental group sample sizes were based on previous pathological findings (9). For Western blots, entire flash-frozen cerebellum samples were analyzed and reported. Fixed floating tissue sections from each brain were collected into buffer holding tubes for each animal. For each experiment, fixed sections from each brain were randomly pulled from the holding tubes; subjected to immunostaining, mounting, and confocal microscopy; and analyzed under identical conditions for both blast-exposed versus sham control samples with subsequent data reported, including sections with missing lobules. Except where indicated, analyses were performed nonblinded using identical experimental procedures for data/image collection and with image adjustments limited only to linear contrast and brightness adjustments applied identically to data from blast- and sham-treated animals in each experiment.

Human subjects

These studies were approved by the VA Puget Sound Health Care System Human Subjects Committee. Participants provided written informed consent before study. The study conformed to institutional regulatory guidelines and principles of human subject protection in the Declaration of Helsinki. Veterans with mild TBI were characterized by physical and neurological examination. Behavioral assessments were made for PTSD using CAPS (85), for depression using PHQ-9 (86), for alcohol consumption using AUDIT-C (87), and for potential postconcussive symptoms using NSI. A lifetime history of both blast-related and impact-related mild TBI was obtained using a semistructured interview by two expert clinicians [described in detail elsewhere (5)]. FDG-PET and DTI were performed within 3 months of clinical evaluation. Participants with retained shrapnel did not undergo DTI.

Animal subjects

This study used wild-type C57BL/6 male, 3- to 4-month-old mice (The Jackson Laboratory) that were group-housed with access to food and water ad libitum. All animal experiments were conducted in accordance with Association for Assessment and Accreditation of Laboratory Animal Care guidelines and were approved by the VA Puget Sound Institutional Animal Care and Use Committee.

Statistical analyses

Standard ANOVAs or t tests were used (two-tailed). Where noted, following a statistically significant overall ANOVA, Helmert tests were used to carry out a priori, planned comparisons (88, 89). Rank-order Spearman correlations (two-tailed P values) were performed using log10-transformed number of blast exposures to linearize blast number. Subsequent multivariate linear regression analyses reported correlations (two-tailed) adjusted for CAPS, PHQ-9, and AUDIT-C scores. Using established methods (90), NEUROSTAT performed an r-to-z transform, with statistical significance of resultant z-score values evaluated using random Gaussian fields and a Euler characteristic algorithm (91) to control for multiple comparisons and maintain a type I error rate of P < 0.05 (corresponding to z > 4.0). All reported P values denote two-tailed critical values defined as P ≤ 0.05. Statistical likelihood of correctly identifying the conditions for treatment pairs (sham versus blast) was performed using a test for one proportion. Statistical analyses were done using SPSS software (IBM) and R version 3.2.2 (R Core Team).



Tabulated data for Figs. 1B, 2E, 3 (C and D), 4 (A and B), 5 (B and C), 6D, 7 (B, D, H, and I), and 8.

References (92101)


  1. Funding: This work was supported by the Department of Veterans Affairs Office of Research and Development Medical Research Service (D.G.C. and B.C.K.); VA Rehabilitation Research & Development Service (E.R.P.); University of Washington Friends of Alzheimer’s Research (D.G.C. and E.R.P.); University of Washington Royalty Research Fund (D.G.C.); Northwest Network Mental Illness Research, Education and Clinical Center (J.S.M., B.R.H., K.F.P., E.C.P., M.A.R., and E.R.P.); Office of Academic Affiliations Advanced Fellowship Program Mental Illness Research and Treatment, Department of Veterans Affairs (B.R.H.); NIH T32 AG000258 (J.S.M.); and VA Clinical Science R&D Career Development Award Program #CX00516 (K.F.P.). Author contributions: J.S.M. and B.R.H. conducted blast procedures, biochemistry experiments, and microscopy. K.D.M. conducted biochemistry experiments. J.S.M., B.R.H., D.J.C., T.L.R., S.M., G.L., E.C.P., M.A.R., E.R.P., and D.G.C. contributed to study design. J.S.M., D.J.C., E.R.P., G.L., E.R.P., M.A.R., and D.G.C. performed data analysis and data interpretation. J.S.M., D.J.C., T.L.R., K.F.P., G.L., B.C.K., E.C.P., M.A.R., E.R.P., and D.G.C. drafted the manuscript and/or edited it critically for important intellectual content. Competing interests: J.S.M. is chief operations officer of Neurogenix Pharmaceuticals. The other authors declare no competing interests.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article