Neurodegeneration in the cortical sulcus is a feature of chronic traumatic encephalopathy and associated with repetitive head impacts Authors: Raymond Nicks, Arsal Shah, Spiro Anthony Stathas, Daniel Kirsch, Sarah M Horowitz, Nicole Saltiel, Samantha M Calderazzo, Morgane L M D Butler, Kerry A Cormier, Nurgul Aytan, Fatima Tu-Zahra, Rebecca Mathias, Farwa Faheem, Suzie Marcus, Elizabeth Spurlock, Lucas Fishbein, Camille D Esnault, Alexandra Boden, Grace Rosen, Weiming Xia, Sarah Daley, Gaoyuan Meng, Brett R Martin, Daniel H Daneshvar, Christopher J Nowinski, Michael L Alosco, Jesse Mez, Yorghos Tripodis, Bertrand R Huber, Victor E Alvarez, Jonathan D Cherry, Ann C McKee, Thor D Stein Published: 2024 ABSTRACT AbstractNeurodegeneration is a seminal feature of many neurological disorders. Chronic traumatic encephalopathy (CTE) is caused by repetitive head impacts (RHI) and is characterized by sulcal tau pathology. However, quantitative assessments of regional neurodegeneration in CTE have not been described. In this study, we quantified three key neurodegenerative measures, including cortical thickness, neuronal density, and synaptic proteins, in contact sport athletes (n = 185) and non-athlete controls (n = 52) within the sulcal depth, middle, and gyral crest of the dorsolateral frontal cortex. Cortical thickness and neuronal density were decreased within the sulcus in CTE compared to controls (p’s < 0.05). Measurements of synaptic proteins within the gyral crest showed a reduction of α-synuclein with CTE stage (p = 0.002) and variable changes in PSD-95 density. After adjusting for age, multiple linear regression models demonstrated a strong association between the duration of contact sports play and cortical thinning (p = 0.001) and neuronal loss (p = 0.032) within the sulcus. Additional regression models, adjusted for tau pathology, suggest that within the sulcus, the duration of play was associated with neuronal loss predominantly through tau pathology. In contrast, the association of duration of play with cortical thinning was minimally impacted by tau pathology. Overall, CTE is associated with cortical atrophy and a predominant sulcal neurodegeneration. Furthermore, the duration of contact sports play is associated with measures of neurodegeneration that are more severe in the cortical sulcus and may occur through tau-dependent and independent mechanisms.Supplementary InformationThe online version contains supplementary material available at 10.1007/s00401-024-02833-8.Keywords: Chronic traumatic encephalopathy, Repetitive head impacts, Contact sports, Neurodegeneration, Cortical thinning, Neuronal loss, Synaptic loss, Cortical sulcus, Tau pathology Introduction Chronic traumatic encephalopathy (CTE) is a progressive neurodegenerative disease caused by repetitive head impacts (RHI), typically sustained through playing contact or collision sports, including American football, ice hockey, boxing, soccer, or rugby [10, 11, 50, 52, 53]. Clinical symptoms of CTE include impairments in cognition, behavior, mood, and motor functioning and typically manifest years or decades after exposure to RHI [56, 58, 70]. Gross neuropathological changes in CTE include prominent frontal lobe atrophy as well as temporal lobe, thalamic, and mammillary body atrophy in higher stage disease [47, 50, 53]. CTE is defined neuropathologically by the perivascular accumulation of abnormally hyperphosphorylated tau in neurons and occasionally in astrocytes with a predilection for the depths of cortical sulci [4, 10, 48]. Computational modeling of head impacts demonstrates that strain forces concentrate at convexities of the brain, such as sulcal depths, and blood vessels, potentially explaining the pattern of pathology observed in CTE [27, 30, 83]. Neurodegeneration, including cortical thinning and synaptic and neuronal loss, is a prominent feature of tauopathies. For instance, in Alzheimer’s disease (AD), numerous studies have demonstrated cortical thinning [25, 32, 62], neuronal loss [21], and synaptic loss [23, 24, 41, 44, 64, 65] within the medial temporal lobe and neocortical regions [9, 22, 78]. Measures of synaptic proteins are also decreased in frontotemporal lobar degenerations [12, 20]. However, quantitative assessments of neurodegeneration in CTE are lacking and have not considered the regional involvement of the cortical sulcus. Here, we quantify neurodegeneration in CTE by measuring changes in gray matter cortical thickness, neuronal density, and synaptic protein levels. We hypothesized that greater neurodegeneration will be present at the depths of the sulcus and with increasing CTE stage. Using the duration of contact sports play as a proxy for cumulative RHI exposure, we also test the hypothesis that years of play would predict increased neurodegeneration in CTE. Materials and methods Participants Participants were selected from two different brain donor groups, including the Understanding Neurologic Injury and Traumatic Encephalopathy (UNITE) study that includes those with a history of RHI through contact/collision sports exposure such as football, ice hockey, boxing, soccer, rugby, and martial arts [57] and the Framingham Heart Study (FHS), a longitudinal and community-based study. Neuropathological processing included comprehensive screening for neurodegenerative conditions following procedures established for the UNITE brain bank [57, 76]. Participants with the following comorbidities were excluded: intermediate or high Alzheimer’s disease by NIA-Reagan criteria, motor neuron disease (MND), frontotemporal lobar degeneration (both FTLD-TDP and FTLD-tau, including corticobasal degeneration [CBD] and progressive supranuclear palsy [PSP]), neocortical Lewy body disease (LBD), and primary brain neoplasms. Age-related pathologies, including primary age-related tauopathy (PART) and limbic-predominant TDP-43 inclusions, were not excluded in order to increase the generalizability of the study. Of the 1109 participants from UNITE and FHS who had tissue available prior to February 24, 2020, 316 participants met inclusion criteria and had available fresh frozen dorsolateral frontal cortex tissue. Since 97% of the participants who had repetitive head impacts were male, we did not include 79 female controls, leaving our final dataset with n = 237 participants. These participants were broken down into four groups: control (n = 52), RHI (n = 48), Low CTE (n = 49), and High CTE (n = 88). All brains were fully assessed for gross and microscopic pathology by neuropathologists (AM, TS, VA, BH) blinded to the clinical evaluation. The brains were hemisected, and pathological assessment was performed on the fixed hemisphere, while the opposing hemisphere was frozen for biochemical protein quantification. Atrophy patterns were graded with a four-point scale including none (0), mild (1), moderate (2), and severe (3) atrophy using criteria from the National Alzheimer’s Coordinating Center (NACC) neuropathology assessment. Subcortical atrophy was determined as present or absent, and the presence of a thalamic notch was defined as medial thalamic atrophy that results in a sharp curvature of the thalamus [51]. Such gross pathology measures have been shown to correlate with clinical symptoms [80]. CTE was diagnosed based on consensus criteria [10, 48]. Staging was based on regional p-tau involvement according to the McKee staging system since this has been shown to correlate with duration of play and clinical symptoms [4, 53]. Stages were then dichotomized into low (I and II) and high (III and IV) stages, which shows good agreement with low and high stages using consensus criteria [4, 53]. Participants with no evidence of CTE and without RHI exposure, were labeled the “control” group. Participants who had no evidence of CTE but had RHI exposure were labeled the “RHI” group. Participants diagnosed with CTE Stage I or II were grouped as “Low CTE.” Lastly, participants diagnosed with CTE Stage III or IV were grouped as “High CTE.” There were no cases of CTE without a history of RHI. Next of kin and other informants of the brain donors in both UNITE and FHS completed the Boston University Repetitive Head Impact Exposure Assessment (BU-RHIEA) questionnaire to determine RHI status and duration of contact sports play [14]. The duration of contact sports play was used as a proxy for cumulative RHI exposure as previously described [55]. Total years of all contact sports participation, including American football, was determined from next-of-kin after death and by checking with an online database for professional players [55] in the UNITE and FHS study groups. For participants who played multiple sports, the primary sport was determined by the highest level of play (e.g., professional, semi-professional, college, high school, and youth). If a participant played both sports at the same level, the sport with the longer duration was determined as their primary sport. Next-of-kin provided written consent for research participation and brain donation. Institutional review boards of the Boston University Medical Center and the Bedford Veteran’s Affairs (VA) Healthcare System approved all study protocols. Histochemistry and immunohistochemistry All brain tissue was processed identically by fixation in periodate-lysine-paraformaldehyde and stored at 4 °C. Multiple tissue blocks were taken, including from the dorsolateral frontal cortex (DLFC) perpendicular to the superior frontal sulcus, embedded in paraffin, and cut at 10 μm. The sections of paraffin-embedded tissue from the DLFC were stained for hyperphosphorylated tau (AT8; Invitrogen MN1020; 1:1000), NeuN (BioLegend; 1:750), and luxol fast blue, and hematoxylin and eosin (LHE), using previously described methods [49]. Stained slides were scanned at 20 × magnification with a Leica Aperio Scanscope (Leica Biosystems, Richmond, IL). Slides were examined using the Aperio eSlide Manager (Leica Biosystems) and the Aperio ImageScope (Leica Biosystems) software. Gray matter cortical thickness was measured by using the ImageScope ruler tool on LHE stained sections. Cortical thickness was calculated by measuring three lines orthogonal to the pial surface and then computing the average. The average gray matter cortical thickness was calculated at the depth of the cortical sulcus (defined as the bottom third of two connecting gyri), in the middle (defined as the middle third of two connecting gyri), and at the gyral crest (defined as the top third of two connecting gyri). In the sulcus and middle, sample sizes were 40 for control, 31 for RHI, 34 for Low CTE, and 54 for High CTE due to missing data. In the crest, sample sizes were 39 for control, 31 for RHI, 34 for Low CTE, and 54 for High CTE. The reasons for missingness included poor tissue quality or incomplete brain specimens that excluded the area of analysis. Group sizes were based on available tissue and not reduced to match between analyses to prevent loss of statistical power. Neuronal density was quantified by using ImageScope (Leica Biosystems). The gray matter was highlighted from the pia to the boundary between the white and gray matter. NeuN has previously been used as a reliable tool for the quantitative study of neuronal morphometry in postmortem human brain tissue [31]. Therefore, neuronal density was quantified by Leica’s image analysis and automated counting software (Aperio nuclear algorithm, Version 9, Leica Biosystems) that identified cells positively labeled for NeuN within the gray matter. The total number of NeuN cells was divided by the area of analysis to determine the NeuN cell density. NeuN density was measured in the sulcus, middle, and crest. In the sulcus, sample sizes were reduced to 26 for control, 20 for RHI, 19 for Low CTE, and 36 for High CTE. In the middle and crest, sample sizes were reduced to 26 for control, 19 for RHI, 15 for Low CTE, and 30 for High CTE. Neuronal density quantification was not performed if the stain failed due to excessive fixation, there was poor tissue quality, the tissue source was exhausted, or the region for analysis was absent. NeuN staining is sensitive to fixation and, therefore, the sample size was reduced compared to other measures. For tau density, Leica’s image analysis and automated counting software (Aperio nuclear algorithm, Version 9, Leica Biosystems) was calibrated for shape, size, and staining intensity to detect AT8-immunoreactive neurofibrillary tangles (NFTs) within the gray matter. Counts were normalized to the area measured and are presented as density within the analyzed region. For both NeuN and NFT counts, all identified objects were manually checked and any flagged cases as well as a random subset were further inspected by a neuropathologist (TDS) or trained morphologist (JDC) to validate that the algorithm was correctly identifying either NeuN+ cells or AT8+ neuronal tangles. Mis-identified object were re-classified and analyzed again. Astrocytic tau tangles were generally smaller, lacked a pyramidal shape, and were largely excluded from the analysis. Quantitative immunoassay measurements of α-synuclein and PSD-95 Frozen tissue from the gyral crest of the dorsolateral prefrontal cortex was weighed and placed on dry ice. Brain tissue was homogenized in a 5:1 volume of freshly prepared, ice-cold 5 M Guanidine Hydrochloride in Tris-buffered saline (20 mM Tris–HCl, 150 mM NaCl, pH 7.4), which contained 1:100 Halt protease inhibitor cocktail (Thermo Fischer Scientific, Waltham, MA) and 1:100 Phosphatase inhibitor cocktail 2 & 3 (Sigma-Aldrich, St. Louis, MO) as previously reported [68, 69]. The homogenate was then shaken (regular rocker) overnight at room temperature. The lysate was diluted with 1% Blocker A (Meso Scale Discovery (MSD), Rockville, Maryland, #R93BA-4) in wash buffer according to specific immunoassays: 1:4000 for α-synuclein (MSD #K151WKK-2) and 1:3000 for postsynaptic density protein-95 (PSD-95, MSD #K250QND). Samples were centrifuged at 17,000 g and 4 °C for 15 min. The supernatant was subsequently applied to the immunoassays, and the original homogenate was aliquoted and stored at − 80 °C. Relative PSD-95 units were calculated based on a standard curve for a reference brain lysate. Standards with known concentrations were used for α-synuclein, and all standards and samples were run in duplicate. Measurements were made using the multi-detection SPECTOR 2400 Imager (MSD). The sample size for α-synuclein was 48 for control, 45 for RHI, 42 for Low CTE, and 80 for High CTE. The sample size for PSD-95 was 49 for control, 46 for RHI, 45 for Low CTE, and 80 for High CTE. Sample sizes were reduced due to missing data or exhausted tissue sources. Statistical methodology Statistical analysis was performed using SPSS 27.0 (IBM Corp) and Prism v9 (GraphPad Software). The means were compared between all four groups (control, RHI, Low CTE, High CTE) using the Kruskal–Wallis test, and values were adjusted by Bonferroni correction for multiple tests for continuous and ordinal variables. The chi-square test was used for proportions to evaluate dichotomous variables. An analysis of covariance (ANCOVA) was used to compare differences between the pathology groups for cortical thickness, neuronal density, and synaptic protein density measurements, adjusting for age at death and postmortem interval (PMI). Gray matter cortical thickness measurements, as well as α-synuclein and PSD-95 concentrations, underwent rank-based normalization for regression analyses [71]. Multiple linear regression analyses were used to evaluate associations between the dependent variables (cortical thickness, neuronal density, and synaptic protein density) and predictors, total years of contact sports play and tau pathology (AT8), adjusting for age at death and PMI. For a final model, we employed a mediation analysis technique [34] to examine the direct and indirect effects of the independent variables (RHI and age) on the dependent variables NeuN density and tau pathology (AT8) within the frontal sulcus. Parameter estimates were used to quantify the strength and significance of the mediated pathway on the total sample of participants with a history of repetitive head impacts. Statistical significance throughout was set as 0.05. Results Demographic differences Based on RHI exposure history and the presence of CTE, participants were grouped as control, RHI, Low CTE, or High CTE (Table 1). There was no significant difference in race. Age at death and postmortem interval (PMI) differed significantly between groups, and these variables were controlled for in subsequent analyses. Those with a history of RHI were also significantly more likely to have had a traumatic brain injury (TBI) with a loss of consciousness (LOC) compared to controls (Table 1). In the control group, six participants reported a history of TBI with LOC resulting from diverse causes, including significant falls, motor vehicle accidents, or military blast exposure. In contrast, nearly all of those with a history of RHI sustained a TBI with LOC from contact sports (> 90%). Table 1Demographic and contact sports exposure of pathological groupsCharacteristicControl (%) (n = 52)RHI (%)(n = 48)Low CTE (%)(n = 49)High CTE (%) (n = 88)p-valueCohortUNITE0100% (48)100% (49)100% (88)FHS100% (52)000Race0.508White42.3% (22)87.5% (42)83.7% (41)88.6% (78)Black/African American010.4% (5)12.3% (6)11.4% (10)Other02.1% (1)2% (1)0Unknown57.7% (30)02% (1)0Age at death, years82.4 (1.53)52.8 (3.26)55.0 (2.49)73.5 (1.34) < 0.001*abcefRange53–9614–8825–8432–97PMI, hours24.6 (3.6)43.8 (3.5)42.9 (3.2)34.3 (2.2) < 0.001*abcefDuration of contact sports play, years010.7 (1.0)15.1 (1.3)17.0 (0.7) < 0.001*abcdeMain sport played, % (n)American Football085.4% (41)86% (42)94.3% (83)Ice Hockey02.1% (1)2% (1)3.4% (3)Boxing02.1% (1)2% (1)2.3% (2)Otherg010.4% (5)10% (5)0History of TBITBI with LOC % (n)17.1% (6)56.8% (25)70.4% (31)65.8% (52) < 0.001*abcData are presented as mean (SEM) for continuous variables and % (n) for categorical variables. *p < 0.05, Analysis of Variance with post hoc least significant difference statistical testing p < 0.05 as follows:aRHI vs. controlbLow CTE vs. controlcHigh CTE vs. controldLow CTE vs. RHIeHigh CTE vs. RHIfHigh CTE vs. Low CTEgIncludes soccer, rugby, amateur wrestling, martial arts, and karateFor PMI, N = 220; main sport played, N = 185; duration of contact sports play, N = 235; TBI with LOC, N = 202CTE chronic traumatic encephalopathy, FHS framingham heart study, LOC loss of consciousness, PMI post-mortem interval, RHI repetitive head impacts, SEM standard error of mean, TBI traumatic brain injury, UNITE understanding neurologic injury and traumatic encephalopathy Demographic and contact sports exposure of pathological groups Data are presented as mean (SEM) for continuous variables and % (n) for categorical variables. *p < 0.05, Analysis of Variance with post hoc least significant difference statistical testing p < 0.05 as follows: aRHI vs. control bLow CTE vs. control cHigh CTE vs. control dLow CTE vs. RHI eHigh CTE vs. RHI fHigh CTE vs. Low CTE gIncludes soccer, rugby, amateur wrestling, martial arts, and karate For PMI, N = 220; main sport played, N = 185; duration of contact sports play, N = 235; TBI with LOC, N = 202 CTE chronic traumatic encephalopathy, FHS framingham heart study, LOC loss of consciousness, PMI post-mortem interval, RHI repetitive head impacts, SEM standard error of mean, TBI traumatic brain injury, UNITE understanding neurologic injury and traumatic encephalopathy Gross anatomical changes Atrophy patterns were significantly different between groups (Table 2). High CTE demonstrated greater atrophy across all cortical and hippocampal regions compared to control, RHI, and Low CTE groups (p’s < 0.05). Atrophy scores were greatest in the frontal lobe and hippocampus in High CTE. There were no significant differences in lobar atrophy scores between control, RHI, and Low CTE groups. In addition, High CTE had a greater frequency of atrophy noted in the hypothalamus and mammillary body as well as a greater proportion of cases with a thalamic notch (indicative of thalamic atrophy) compared to the other groups (Table 2; p’s < 0.05). Table 2Gross pathological changesCharacteristicControl (n = 52)RHI(n = 48)Low CTE(n = 49)High CTE(n = 88)p-valueAtrophy, 0–3, mean (SEM)Frontal lobe0.73 (0.117)0.72 (0.134)0.69 (0.110)1.83 (0.097) < 0.001*abcTemporal lobe0.42 (0.088)0.50 (0.119)0.50 (0.103)1.65 (0.098) < 0.001*abcParietal lobe0.25 (0.061)0.35 (0.099)0.31 (0.090)1.13 (0.102) < 0.001*abcOccipital lobe0.13 (0.048)0.13 (0.067)0.13 (0.057)0.58 (0.094) < 0.001*abcTotal cortical1.29 (0.221)0.65 (0.125)0.60 (0.106)1.71 (0.096) < 0.001*acHippocampus0.79 (0.187)0.48 (0.103)0.62 (0.106)1.87 (0.104) < 0.001*abcSubcortical atrophy, % present (present/absent)Hypothalamus0.0% (0/13)17.1% (7/34)21.7% (10/36)62.4% (53/32) < 0.001*abcMamillary body0.0% (0/13)16.7% (7/35)23.9% (11/35)61.0% (50/32) < 0.001*abcThalamic notch7.7% (1/12)11.9% (5/37)25.0% (11/33)47.7% (41/45) < 0.001*abSeptum pellucidum abnormalities, % present (present/absent)Septum cavum10.0% (1/9)16.7% (6/30)33.3% (11/22)45.6% (26/31)0.012*cSeptal fenestrations0.0% (0/10)0.0% (0/36)21.2% (7/26)47.2% (25/28) < 0.001*cData are presented as mean (SEM) for continuous variables and % (n) for categorical variables. *p < 0.05, Analysis of Variance with post hoc least significant difference statistical testing p < 0.05 as follows:aHigh CTE vs. controlbHigh CTE vs. RHIcHigh CTE vs. low CTEFor Septum cavum, N = 136; Septal fenestrations, N = 136; Thalamic notch, N = 185; Hypothalamus, N = 185; Mamillary bodies, N = 181; Total cortical, N = 194CTE chronic traumatic encephalopathy, RHI repetitive head impacts, SEM standard error of mean Gross pathological changes Data are presented as mean (SEM) for continuous variables and % (n) for categorical variables. *p < 0.05, Analysis of Variance with post hoc least significant difference statistical testing p < 0.05 as follows: aHigh CTE vs. control bHigh CTE vs. RHI cHigh CTE vs. low CTE For Septum cavum, N = 136; Septal fenestrations, N = 136; Thalamic notch, N = 185; Hypothalamus, N = 185; Mamillary bodies, N = 181; Total cortical, N = 194 CTE chronic traumatic encephalopathy, RHI repetitive head impacts, SEM standard error of mean Abnormalities of the septum pellucidum were also frequent in CTE. Notably, High CTE had a higher frequency of a cavum septum pellicidum than the RHI group (p < 0.05). Both Low CTE and High CTE were more likely to have septum fenestrations compared to control and RHI groups (p’s < 0.05). Associations between RHI exposure and gray matter cortical thickness We examined the burden of tau pathology across multiple cortical regions. The tau pathology burden increased with RHI and CTE across regions in the frontal, temporal, and parietal lobes with the greatest burden in the middle frontal region (Supplementary Table 1, online resource). We, therefore, focused on the dorsolateral middle frontal cortex for the remainder of the study. Cortical thickness measurements were conducted along the superior frontal sulcus with the dorsolateral frontal cortex (DLFC) divided into three regions of interest: sulcus, middle, and crest (Fig. 1a). After adjusting for age and PMI, the greatest differences between groups were seen within the DLFC sulcus (Fig. 2a, p = 0.008), with reduced cortical thickness in both Low CTE and High CTE compared to the RHI group (p’s < 0.05) as well as reduced cortical thickness in High CTE compared to the control group (p = 0.015). No significant differences were found among the groups in the middle third of the gyrus (Fig. 2b). In the gyral crest, the RHI group showed increased cortical thickness compared to the control group (p = 0.021, Fig. 2c). Comparison of non-age adjusted cortical thickness also showed significant decreases in the sulcus within High CTE compared to control (p = 0.022) and RHI (p = 0.004) groups (Supplementary Figure a, online resource). To examine differences within individuals, we calculated the ratio of the cortical thickness at the sulcus to the thickness in the gyral crest and found that RHI (p = 0.004), Low CTE (p = 0.001), and High CTE (p < 0.001) groups all had lower sulcus/crest ratios than the control group (Supplementary Figure b, online resource). Representative images of NeuN staining in CTE. a Cortical thickness was calculated by measuring three lines orthogonal to the pial surface and then computing the average. Analysis was based on three areas: crest, middle, and sulcus. b, c The number of NeuN+ cells in the sulcus of the dorsolateral frontal cortex was greater in control (b) compared to High CTE (c). Scale bars: a 5 mm; b, c 200 μm Cortical thickness measurements by pathology group in dorsolateral frontal gray matter in a sulcus, b middle, and c crest. a Cortical thickness within the sulcus was significantly different between groups, as shown by ANCOVA (p = 0.008). Post hoc pairwise comparisons showed that High CTE was significantly less than control (p = 0.015) and RHI (p = 0.04), and Low CTE was reduced compared to the RHI group (p = 0.039). b Within the middle of the gyrus, there were no significant differences between groups. c At the gyral crest, the cortical thickness was significantly different between groups, as shown by ANCOVA (p = 0.01). Post hoc pairwise comparison showed that the RHI group was significantly greater than control (p = 0.021). *p < 0.05, analysis of covariance adjusting for age at death and PMI Multiple linear regression analyses were run in the sulcus, middle, and crest to determine potential associations between duration of contact sports play and cortical thickness, adjusting for age at death and PMI (Table 3a). There was a significant negative association between the duration of contact sports play and cortical thickness in the sulcus (β = −0.254, p = 0.001), but not in the middle third of the gyrus or in the gyral crest. Because duration of play is associated with tau pathology, which may partially mediate cortical thinning, a second multiple linear regression tested associations between cortical thickness and duration of contact sports play while adjusting for tau pathology (Table 3b). This analysis demonstrated that both years of contact sports play and tau pathology (AT8 density at the sulcus) were significant predictors for reduced cortical thickness in the sulcus. To determine if the primary sport of participation had an effect, we employed a sensitivity analysis consisting of participants who had a history of American football play. The results were similar and showed a significant association between the duration of football participation and reduced cortical thickness in the sulcus (β = −0.200, p = 0.012). An additional regression was run adjusting for a history of TBI with a loss of consciousness, and the results were similar without a significant contribution from TBI (duration of play: β = −0.248, p = 0.005; TBI: β = 0.026, p = 0.756). Within a subset of CTE cases (N = 22), 13 (59%) had the CTE pathognomonic perivascular tau pathology lesion within the dorsolateral frontal cortex section examined. Table 3Multiple linear regression analyses modelling cortical thickness with duration of contact sports play in the sulcus, middle, and crest of dorsolateral frontal cortexa) Associations with cortical thickness in the sulcus, middle, and crestPredictorSulcusMiddleCrestβp-valueβp-valueβp-valueDuration of contact sports play (years)−0.2540.0010.0740.3470.0810.296b) Associations with cortical thickness in the sulcus, adjusting for tau pathologyPredictorSulcusβp-valueDuration of contact sports play (years)−0.2270.013Tau pathology (AT8) sulcus−0.304 < 0.001a) Multiple linear regressions were run in the sulcus, middle, and crest to determine associations between cortical thickness (dependent variable) and total years of playing contact sports. b) A second multiple linear regression model was run in the sulcus to determine associations between cortical thickness and total years of playing contacts sports, adjusting for tau pathology. The cortical thickness variables displayed for the sulcus, middle, and crest underwent rank-based normalization, and both models were adjusted for age at death and post-mortem interval. β indicates standardized beta value. Significant associations are shown in bold Multiple linear regression analyses modelling cortical thickness with duration of contact sports play in the sulcus, middle, and crest of dorsolateral frontal cortex a) Multiple linear regressions were run in the sulcus, middle, and crest to determine associations between cortical thickness (dependent variable) and total years of playing contact sports. b) A second multiple linear regression model was run in the sulcus to determine associations between cortical thickness and total years of playing contacts sports, adjusting for tau pathology. The cortical thickness variables displayed for the sulcus, middle, and crest underwent rank-based normalization, and both models were adjusted for age at death and post-mortem interval. β indicates standardized beta value. Significant associations are shown in bold Associations between RHI exposure and neuronal density In order to examine whether the reduction in cortical thickness might be partially due to decreased neuronal density, we examined associations with NeuN+ cellular density in the sulcus, middle, and gyral crest. Within the sulcus, there was significantly lower neuronal density in both Low CTE and High CTE groups compared to the control group (p’s < 0.05, Fig. 3a). In both the middle and gyral crest regions, no significant differences in neuronal density were observed among the groups (Fig. 3b and Fig. 3c). Immunohistochemical staining for NeuN illustrates the decreased neuronal density within the sulcus in High CTE (Fig. 1c) compared to control (Fig. 1b). Neuronal density by pathology group in dorsolateral frontal gray matter in a sulcus, b middle, and c crest. a Neuronal density within the sulcus was significantly different between groups, as shown by ANCOVA (p = 0.05). Post hoc pairwise comparisons showed that neuronal density was reduced compared to the control group in both Low CTE (p = 0.03) and High CTE (p = 0.014). b Within the middle of the gyrus there were no significant differences between groups. c At the gyral crest, there were no significant differences between groups. *p < 0.05, analysis of covariance adjusting for age at death and PMI Multiple linear regression analyses tested for associations between duration of contact sports play and neuronal density, adjusting for age at death and PMI (Table 4). There was a significant negative association between the duration of contact sports play and neuronal density in the sulcus (β = −0.231, p = 0.032), but the association with neuronal density in the middle third and gyral crest were not significant (Table 4a). In order to test whether tau pathology might partially underlie the decreased neuronal density within the sulcus, a second multiple linear regression model included levels of AT8+ tau pathology within the sulcus. Sulcal tau pathology, but not duration of play, significantly predicted decreased neuronal density (β = −0.344, p = 0.007, Table 4b), suggesting that duration of play leads to decreased neuronal density through tau pathology within the sulcal depths. A sensitivity analysis restricted to football players also showed an association between the duration of football participation and decreased neuronal density in the sulcus (β = −0.229, p = 0.035). Table 4Multiple linear regression analyses modelling neuronal density (NeuN/mm2) with duration of contact sports play in the sulcus, middle, and crest of dorsolateral prefrontal cortexa) Associations with neuronal density in the sulcus, middle, and crestPredictorSulcusMiddleCrestβp-valueβp-valueβp-valueDuration of contact sports play (years)−0.2310.032−0.1950.064−0.1300.287b) Associations with neuronal density in the sulcus, adjusting for tau pathologyPredictorSulcusβp-valueDuration of contact sports play (years)−0.1740.219Tau pathology (AT8) sulcus−0.3440.007a) Multiple linear regressions were run in the sulcus, middle, and crest to determine associations between neuronal density and total years of playing contact sports. b) A second multiple linear regression model was run in the sulcus to determine associations between neuronal density and total years of playing contact sports, adjusting for tau pathology. Both models were adjusted for age at death and post-mortem interval. β indicates standardized beta value. Significant associations are shown in bold Multiple linear regression analyses modelling neuronal density (NeuN/mm2) with duration of contact sports play in the sulcus, middle, and crest of dorsolateral prefrontal cortex a) Multiple linear regressions were run in the sulcus, middle, and crest to determine associations between neuronal density and total years of playing contact sports. b) A second multiple linear regression model was run in the sulcus to determine associations between neuronal density and total years of playing contact sports, adjusting for tau pathology. Both models were adjusted for age at death and post-mortem interval. β indicates standardized beta value. Significant associations are shown in bold Although participants with significant other pathologies including AD or FTLD were excluded from this study, low level pathologies might contribute to neurodegeneration. Therefore, sensitivity analyses were performed for additional pathologies that might contribute to cortical thinning and neuronal loss, including neuritic beta-amyloid plaques, AD-type tau pathology, and TDP-43 inclusions. Separate multiple linear regressions showed that together with tau pathology in the model neither neuritic beta-amyloid plaque score nor Braak stage significantly contributed to cortical thickness or neuronal density at the cortical sulcus. The presence of TDP-43 was significantly associated with cortical thickness (β = −0.353, p = 0.002), but not neuronal density. CTE-associated loss of synaptic proteins Synaptic protein levels were quantified by using immunoassays to assess the concentration of α-synuclein (presynaptic) and PSD-95 (postsynaptic) in the gyral crest of the DLFC gray matter tissue. Levels of α-synuclein have been shown to decrease in AD and to decrease with the duration of dementia [75]. PSD-95 served as a postsynaptic marker [33, 36]. RHI, Low CTE, and High CTE groups all showed significantly lower α-synuclein density compared to the control group (p’s < 0.01, Fig. 4a). There was significantly decreased PSD-95 density in Low CTE compared to the control group (p = 0.009), but paradoxically increased PSD-95 density in High CTE compared to Low CTE (p = 0.039, Fig. 4b). Previous studies have shown increased variability and altered distributions in dendritic/spine measures [77], and a dynamic damage and repair process might explain the increased PSD-95 density in High CTE. In order to further test whether variation in synaptic proteins are higher with RHI and CTE, we calculated the coefficient of variation (CV = [standard deviation/mean] × 100) for both α-synuclein and PSD-95 protein densities. For α-synuclein, CVs were slightly higher in RHI (49.9%), Low CTE (52.3%), and High CTE (48.3%) than in the control (45.7%) group. Similarly, for PSD-95, CVs were higher in the RHI (54.2%), Low CTE (50.1%), and High CTE (57.5%) groups than in the control (47.7%) group. α-synuclein and PSD-95 concentrations as measures of synaptic density in dorsolateral frontal gray matter tissue. a α-synuclein levels were significantly different between groups, as shown by ANCOVA (p = 0.028). Post hoc pairwise comparisons showed that RHI, Low CTE, and High CTE group all had significantly less α-synuclein than the control group (p < 0.01). b PSD-95 levels were significantly different between groups, as shown by ANCOVA (p < 0.001). Post-hoc pairwise comparisons showed that the Low CTE group had less PSD-95 than the control group (p = 0.009) and the High CTE group was increased as compared to Low CTE (p = 0.039). *p < 0.05 and **p < 0.01, analysis of covariance adjusting for age at death and PMI Pathway analysis of the effects of RHI on neuronal density within the sulcus For a final model, we employed a mediation analysis, with neuronal density and tau pathology (AT8) in the sulcus as the dependent variables. We then tested the direct and indirect effects of age and duration of play. Both age (β = 0.367, p < 0.001) and duration of play (β = 0.424, p < 0.001) had a direct effect on tau pathology (Supplementary Table 2). Tau pathology was negatively associated with neuronal density (β = −0.335, p < 0.001). Overall, this model suggests that the duration of play acts through increased tau pathology to decrease neuronal density within the sulcus (Fig. 5). Schematic representation of mediation analysis modeling associations between repetitive head impacts, age of death, tau pathology, and neuronal density in the sulcus. Black arrows represent significant positive associations, and the red arrow represents significant negative association (p’s < 0.001). Standardized β’s are shown above arrows. Created in BioRender. Stein, T. (2024) BioRender.com/h62i404 Discussion Chronic traumatic encephalopathy can be a patchy and focal disease, and its association with gross and microscopic measures of neurodegeneration has not been systematically explored. Here, we show that frontal, hippocampal, hypothalamic, mammillary body, and thalamic atrophy are prominent in high stage CTE. Furthermore, both RHI exposure and increasing CTE stage are associated with neurodegeneration in the dorsolateral frontal cortex, as evidenced by cortical thinning and decreased neuronal density predominantly within the sulcus, as well as synaptic changes in the gyral crest. The cortical sulcus appears to be uniquely vulnerable to RHI. Computational and physical models that incorporate the gyral architecture of the human brain demonstrate increased strain at the convexities and sulcal depths [27, 30] as well as the formation of cavitation vapor bubbles with the greatest strain at the depths of sulci [37]. The tau pathology in CTE is primarily neuronal and concentrated at the cortical sulcal depths [15]. Furthermore, recent single nuclear RNA sequencing of contact sport athletes with and without CTE shows neuronal loss within the superficial layers at the sulcal depths involving excitatory layer 2/3 CUX2/LAMP5 neurons [16]. Utilizing a larger group of contact/collision sports athletes with and without CTE, we similarly find a sulcal specific pattern of neurodegeneration involving pathways that are both dependent and independent of tau pathology. Previous MRI studies have assessed different patterns of cortical thinning in AD and frontotemporal dementia (FTD), revealing distinct patterns of regional involvement in AD compared to the frontal and temporal lobes in FTD [29]. Du et al. demonstrated that MRI-based cortical thinning patterns can differentiate between neurodegenerative disease groups [29]. Spotorno et al. investigated cortical thickness in AD and LBD pathology with a direct correlation to postmortem tau burden [66]. Additionally, Mak et al. used PET scans to demonstrate that cortical thinning is associated with tau pathology [43]. Measurement of cortical thickness has proven valuable in detecting cognitive impairment, with evidence that suggests that gray matter loss may underlie cognitive disease [35, 82]. Cortical thinning in youths who had sustained a TBI has also been demonstrated in previous MRI studies [46, 54]. Koerte et al. observed greater cortical thinning in a small group of former professional soccer players compared to non-contact sports athletes, suggesting a potential role of RHI in cortical thinning and cognitive decline [38]. More recently, Albaugh et al. found reduced cortical thickness in the frontal, parietal, and temporal cortices in RHI-exposed ice hockey players, which was associated with post-concussive symptoms [2]. Doughty et al. demonstrated significant cortical thinning, microstructural anomalies, and functional abnormalities in former professional football players that could be distinguished decades after their retirement [28]. Our results support and extend previous MRI findings in patients with CTE [1, 5–7, 74] by demonstrating cortical thinning within the sulcus of the dorsolateral frontal cortex. Future research should focus on distinguishing cortical thickness patterns among neurodegenerative diseases, including CTE, to explore whether in vivo measures can effectively differentiate CTE from other neuropathologies. Our research builds upon previous studies that have established that AD is characterized by synaptic loss and alterations in synaptic morphology, which may represent compensatory responses to neurodegeneration [23, 24, 42, 45, 61, 64, 65]. Synapse loss, indicating synaptic dysfunction, is considered a reliable marker of cognitive decline in both postmortem and biopsied Alzheimer’s disease brains [23, 64, 65, 72, 79]. Understanding dendritic alterations in neurodegenerative diseases like CTE may provide insights into the underlying mechanisms of cognitive impairment associated with repetitive head injuries. Warling et al. quantified dendrites and dendritic spines of supragranular pyramidal neurons within the frontal and occipital lobes of brain donors with CTE and found increased variability with an overall decrease in dendritic extent in CTE [77]. Furthermore, Braun et al. demonstrated that the mechanical stretching of neurons induces the mislocalization of tau to the dendritic spines, which contributes to synaptic dysfunction [13]. Thus, tau phosphorylation might lead to functional deficits in synaptic function following traumatic brain injuries. Chapman et al. found impaired and reversible synaptic plasticity following high-frequency head impacts in mice [18]. Our previous studies on gene expression networks have shown that genes involved in synaptic pathways are variably altered in Low CTE versus High CTE [39]. This suggests that different processes may be involved depending on the disease progression. In addition, we found sulcal specific network changes in those with a history of RHI and in low stage CTE [19]. Here, we show variation in the levels of α-synuclein and PSD-95 in those with a history of RHI and with CTE, suggesting a continual process of damage and repair that may contribute to disease progression and may have functional consequences. There is likely a dynamic process of synaptic damage and regrowth that underlies resilience and vulnerability to disease and that affects certain regions more than others. Future studies using spatial techniques and super-resolution microscopy should further examine regional vulnerability and the processes that might mediate structural changes in synapses, such as astrocyte and microglia ingestion of synapses, as demonstrated in AD [17, 73]. Finally, cortical thinning has been associated with cognitive decline in Alzheimer’s disease [23–26, 60, 72, 81]. In the context of CTE, cognitive impairments have similarly been observed, with tau pathology in the frontal cortex contributing significantly to cognitive, functional, and neuropsychiatric symptoms [3, 8]. However, tau pathology alone does not fully account for the extent of cognitive decline seen in CTE patients. Our findings suggest that cortical thinning could play an important role in this cognitive decline. In addition, RHI has been associated with a variety of neuropathologies, many of which contribute to functional impairment [63]. Of these, TDP-43 pathology is particularly frequent in CTE [59] and one of the biggest contributors to cognitive impairment [63]. Consistent with this we found in a sensitivity analysis that the presence of TDP-43 inclusions was associated with decreased cortical thickness. While we did not include measures of cognitive decline in our current model due to limitations in statistical power, future research should explore the potential of cortical thinning together with comorbid pathologies as predictive markers for cognitive decline in CTE. Additionally, future studies should include comparisons with other neurodegenerative diseases such as AD and examine markers of cell death and the susceptibility of various cell types, including interneurons. Such studies could provide valuable insights into the multifaceted nature of cognitive impairment following RHI and in CTE and may help identify new diagnostic and therapeutic targets. Limitations There are several limitations to the present study. Brain donors were largely recruited via self-selection or next-of-kin referral, which introduces autopsy-based selection bias that may hinder generalizability. However, inverse probability weighting recently demonstrated that study selection did not significantly affect the relationship between RHI and CTE pathology [40, 55]. The average age of the FHS controls was much older than that of the UNITE brain bank, and PMI was significantly higher in UNITE, given the national catchment area. Although we adjusted for both age and PMI in our analyses, these factors may still contribute to differences. The vast majority of the groups are Caucasian men, further limiting generalizability. Future studies within larger community-based aging cohorts with RHI history, as well as prospective studies, will be necessary to confirm and expand these findings. Finally, synaptic measures were taken from the gyral crest due to the difficulty of sulcal tissue dissections. Future studies, such as super-resolution microscopy, may be necessary to examine alterations in the sulcus [67]. Conclusions In those with CTE as the only significant pathology, there is marked cortical atrophy, cortical thinning and decreased neuronal density within the sulcal depth, and altered synaptic protein levels. Contact sports exposure duration was associated with cortical thinning and neuronal loss through tau-dependent and independent mechanisms. These findings may have implications for future MRI and other biomarker studies, where cortical thinning may be useful as a biomarker to detect differential thinning in participants suspected of having CTE. They may also inform assessments of regional-based neurodegeneration following RHI and in CTE. Supplementary Information Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 79 KB) Acknowledgements This work was supported by the United States (US) Department of Veterans Affairs, Veterans Health Administration, Clinical Sciences Research and Development Merit Award (I01-CX001038); Alzheimer’s Association (NIRG-305779, NIRG-362697); National Institute of Aging (RF1AG054156, R56AG057768, RF1AG057768, K23AG046377, U19AG068753, AG08122, AG054076); National Institute of Neurological Disorders and Stroke (U54NS115266, U01NS086659, K23NS102399, F31NS127449), National Institute of Aging Boston University AD Center (P30AG13846; supplement 05720633455; P30AG072978); National Heart, Lung and Blood Institute (75N92019D00031 and HHSN2682015000011); Department of Defense Peer Reviewed Alzheimer’s Research Program (PRARP #13267017); and the Concussion Legacy Foundation. This work was also supported by unrestricted gifts from the Andlinger Foundation and WWE. 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. Funding This work was supported by U.S. Department of Veterans Affairs, I01BX005933, Thor D Stein, I01BX005161, Thor D Stein, Alzheimer’s Association, NIRG-305779, NIRG-362697, National Institute on Aging, RF1AG054156, Thor D Stein, R56AG057768, Thor D Stein, RF1AG057768, Thor D Stein, K23AG046377, U19AG068753, AG054076, National Institute of Aging, AG08122, National Institute of Neurological Disorders and Stroke, U54NS115266, Ann McKee, U01NS086659, Ann McKee, K23NS102399, F31NS127449, Daniel Kirsch, National Institute of Aging Boston University AD Center, P30AG13846, Ann McKee, National Heart, Lung, and Blood Institute, 75N92019D00031, HHSN2682015000011, Peer Reviewed Alzheimer’s Research Program, PRARP #13267017. Data availability Data used in this study are available from the Boston University Alzheimer’s Disease Center (https://www.bu.edu/alzresearch/information-for-investigators/) and from the authors by request. Declarations Conflict of interest Outside of the submitted work, WX received research support from the National Institutes of Health, Veterans Health Administration, Biomedical Laboratory Research and Development, and Veterans Health Administration, Clinical Sciences Research and Development Merit Awards. He reports being principal investigator and co-investigator on clinical trials, and has a patent pending regarding the diagnosis of AD using machine learning. JDC received grant support from the Department of Veterans Affairs Career Development Award and National Institute of Aging Boston University AD Center. YT reports receiving grant support to the Boston University School of Public Health. ACM reports grant support for other works from the NINDS/NIA, NIA, and VA, and honoraria from the University of Massachusetts, Montefiore Medical Center, Korean Dementia Society, and Texas Neurological Society. The remaining authors report no relevant conflicts of interest. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Raymond Nicks and Arsal Shah are co-first authors. References Associated Data Supplementary Materials Supplementary file1 (DOCX 79 KB) Data Availability Statement Data used in this study are available from the Boston University Alzheimer’s Disease Center (https://www.bu.edu/alzresearch/information-for-investigators/) and from the authors by request.