Introduction
Chronic traumatic encephalopathy (CTE) is a progressive neurodegenerative disease associated with a prolonged history of repetitive head impacts (RHI) including concussive and subconcussive hits [
1,
2]. Athletes participating in sports such as American football, hockey, boxing, soccer, and rugby, as well as military personnel, may be at risk due to years of exposure to head impacts [
1,
2]. The clinical features of CTE typically manifest years or decades after exposure to RHI and consist of impairments in mood, behavior, cognition, and motor functioning [
3]. Recently, consensus criteria have established CTE as a distinct tauopathy defined by abnormally phosphorylated tau (ptau) accumulation within neurons, astrocytes, and cell processes in an irregular and patchy distribution that is perivascular and concentrated within the depths of sulci [
4]. The earliest pathological changes of CTE are often observed within the sulcal depths of the dorsolateral frontal cortex. In later, more severe stages, ptau pathology is present within neighboring cortical regions and within medial temporal lobe structures such as the hippocampus [
4,
5].
Growing evidence suggests that cumulative RHI exposure is associated with the development and increased severity of CTE [
3,
5‐
8]. We have previously shown in a heterogeneous cohort of deceased contact sport athletes and military personnel that the number of years of exposure to RHI significantly predicts increased CTE stage as defined by the extent of ptau pathology [
5,
9]. Furthermore, a cumulative head impact index based on position, level of play, and impact frequencies from helmet accelerometer studies in American football players was associated with later-life cognitive and neurobehavioral impairment [
8]. In contrast, the reported number of concussions was not significantly correlated with CTE stage [
5,
9] and was less predictive than the cumulative head impact index of cognitive and neurobehavioral impairment [
8]. Furthermore, 16 % of individuals diagnosed with CTE had no reported history of concussions suggesting subconcussive hits are sufficient for the development of the disease [
10]. However, to date, an association between years of contact sports play and a quantitative measure of ptau pathology in the brain has not been shown, and the underlying factors linking RHI exposure to the development of CTE are unknown.
Although mild traumatic brain injury has been associated with multifocal axonal loss and microglial activation [
1], the factors that initiate the development of ptau pathology are unknown. While the neuroinflammatory response to mild acute insults is usually short-lived [
11,
12], RHI may lead to chronic neuroinflammation that induces a self-perpetuating inflammatory cycle with longstanding activation of microglia, including sustained release of inflammatory mediators [
13]. Recent evidence suggests that neuroinflammatory cytokines and reactive microglia exacerbate tau pathology and contribute to the spreading of ptau in rodent models of Alzheimer disease and other tauopathies, suggesting a potential link between traumatic brain injury and CTE [
14,
15]. However, the neuroinflammatory state years after a period of RHI is unknown.
Here, we test the hypothesis that increased neuroinflammation, defined by an increased astrocyte and microglial number and an upregulation of the inflammatory/phagocytic marker CD68, is associated with a history of longer exposure to RHI, increased CTE ptau pathology, and increased risk of developing dementia in a cohort of American football players and non-exposed control subjects.
Discussion
We show that increased neuroinflammation, as evidenced by increased CD68 cell density and enhanced microglia reactive morphology, was associated with more severe AT8 immunopositive ptau pathology in the DLF cortex of subjects neuropathologically diagnosed with CTE. Furthermore, duration of RHI exposure, as defined by the years of football played, predicted greater density of CD68 positive inflammatory microglia in American football players with and without CTE pathology. A simultaneous equations regression model demonstrated that exposure to RHI had a significant effect on ptau pathology that was both direct and indirect (via increased CD68 cell density). Last, we found that increased neuroinflammation was related to the risk of a subject being diagnosed with dementia and that the relationship was mediated through ptau, independent of age. Altogether, this suggests that increased neuroinflammation as a consequence of prolonged RHI exposure may play an important role in the development of ptau pathology in CTE. To our knowledge this is the first data directly linking neuroinflammation to the development of ptau pathology in CTE.
Although moderate/severe TBI has been observed to result in long lived neuroinflammation [
20], even mild head impacts may lead to brain injury, including multifocal traumatic axonal disruption, that is concentrated in regions such as the white matter-gray matter junction, around blood vessels, and at the depth of the cortical sulci [
12,
21‐
23]. Multifocal traumatic axonal injury may be an early trigger for neuroinflammation [
24], with microglial activation and initiation of an immune response that serves to repair or limit the damage [
25]. The interval between injuries also impacts the long-term CNS response [
26]. After isolated or infrequent mild TBI the induced neuroinflammatory response may dissipate as the CNS tissue is repaired. However, repetitive injuries, particularly RHI that occur during a short interval that do not allow for complete recovery, may lead to a persistent proinflammatory state. Furthermore, neuronal death resulting from mechanical injuries has the potential to result in cell lysis and exposure of intracellular components into the extracellular environment. Intracellular proteins like HMGB1, ATP, and other damage associated molecular patterns (DAMPS) can trigger a rapid innate immune response and induce chronic neuroinflammation [
27].
Numerous studies have demonstrated that neuroinflammation occurs acutely following TBI [
11,
28‐
30]. In addition, mouse models that mimic the repetitive injury observed in humans have shown that glial changes and neuroinflammation precede ptau pathology [
31,
32]. We also observed elevated CD68 positive cell density in RHI-exposed subjects in the absence of any ptau pathology (Fig.
1c). These individuals were younger and the duration of exposure to RHI was less than the CTE group; thus, they may represent individuals in a prodromal state.
The simultaneous equations regression analysis is a form of structural equation modeling that incorporates feedback loops [
33] and is therefore well suited to model the effects of interacting pathologies. In a cohort of subjects with and without exposure to RHI, we found that RHI exposure had a cumulative effect on the extent of ptau pathology that was both direct and mediated through increased CD68 cell density (Table
2, Fig.
2). An activated microglial-mediated increase in ptau is consistent with previous work demonstrating that neuroinflammation increased the activity of GSK3β and p38MAPK, both of which are involved in tau hyperphosphorylation [
14] and promote ptau pathology [
34]. The simultaneous equations regression model also demonstrated that both CD68 and ptau had a significant effect on each other (Fig.
2). This is in agreement with previous studies in mouse models that have shown an inflammatory mediated increase in ptau as well as tau induced neuroinflammation [
35]. Our data suggest that a similar feedback mechanism may be present in human CTE. Thus, down regulation of chronic microglia activation might represent an important therapeutic target for CTE. Furthermore, there was a significant direct effect of RHI exposure on increased ptau pathology independent of CD68. The mechanisms underlying this increase in ptau following RHI are unknown, but might involve calcium influx and activation of kinases which are known to occur after mild traumatic brain injury [
26]. In addition to its association with RHI exposure and ptau pathology, CD68 cell density in the DLF appears to plays an indirect role in the development of dementia in CTE, mediated through ptau pathology. A binary logistic regression demonstrated dementia was significantly predicted by CD68 independent of age, but that association was eliminated when a measure of ptau pathology was included in the model. Our data suggest that increased activated microglia occur early following RHI, followed by a later, but steeper increase in ptau pathology, and finally cumulating in dementia in some cases (Fig.
3). In support of this, cumulative RHI exposure has recently been associated with cognitive impairment later in life [
8] and TSPO PET ligands for activated microglia have shown elevated activity in retired NFL players at risk for CTE [
36]. Activated microglia and persistent neuroinflammation are one mechanism by which RHI exposure may drive the development and spread of ptau pathology and lead to dementia in CTE.
There are several limitations to our study. There is selection bias in an autopsy-based study of individuals whose brains are donated by the family, and the subjects may not represent the population as a whole. Additionally, clinical and RHI exposure histories are obtained retrospectively and are subject to bias. Future longitudinal, prospective studies will be needed to verify these results. For the simultaneous equations regression analysis, CD68 and AT8 densities were treated as both outcomes and predictors to avoid the “endogeneity” problem (i.e., (1) a linear regression model that has independent variables as predictors would give invalid inference due to the significant correlation between the predictor and the error term, and (2) their coefficients would be biased) [
37]. To address the inference and bias problems caused by endogeneity, we used simultaneous equation models, which are widely used in econometrics [
33]. For cell quantitation we used the Aperio nuclear algorithm to determine the number of stained cells. In contrast to design based stereological methods, the nuclear algorithm lacks 3-dimensional volume measurements and does not provide an estimate of the total number of microglia [
18]. However, we obtained similar results when using the positive pixel algorithm. Finally, while CD68 is typically associated with a more inflammatory, phagocytic microglia phenotype [
38], glial activation is too complex and dynamic to be completely recapitulated by one marker. Future studies will be needed to explore additional markers and inflammatory cytokines to better define the neuroinflammatory phenotype in CTE.
Acknowledgments
This work was supported by the Department of Veterans Affairs, Veterans Health Administration, Clinical Sciences Research and Development Merit Award (I01-CX001038); Alzheimer’s Association (NIRG-305779); National Institute of Aging (1RF1AG054156-01); Veterans Affairs Biorepository (CSP 501); Translational Research Center for Traumatic Brain Injury and Stress Disorders (TRACTS) Veterans Affairs Rehabilitation Research and Development Traumatic Brain Injury Center of Excellence (B6796-C); National Institute of Neurological Disorders and Stroke, National Institute of Biomedical Imaging and Bioengineering (U01NS086659-01); National Institute of Aging Boston University AD Center (P30AG13846; supplement 0572063345–5); Department of Defense Peer Reviewed Alzheimer’s Research Program (DoD-PRARP #13267017); National Heart, Lung, and Blood Institute, Framingham Heart Study (NHLBI/NIH #N01-HC-25195 and R01AG08122); NINDS (1F32NS096803-01); Concussion Legacy Foundation. This work was also supported by unrestricted gifts from the Andlinger Foundation and WWE. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. We gratefully acknowledge the use of resources and facilities at the Edith Nourse Rogers Memorial Veterans Hospital (Bedford, MA) as well as all the individuals whose participation and contributions made this work possible.