Research ArticleTraumatic Brain Injury

In vivo detection of cerebral tau pathology in long-term survivors of traumatic brain injury

See allHide authors and affiliations

Science Translational Medicine  04 Sep 2019:
Vol. 11, Issue 508, eaaw1993
DOI: 10.1126/scitranslmed.aaw1993

Imaging long-lasting deposition

Positron emission tomography (PET) using the radioligand for tau protein flortaucipir is a recently developed noninvasive method for measuring tau protein deposition. Although traumatic brain injury (TBI) has been associated with increased tau in postmortem samples, the long-term effects of a single TBI on tau deposition and its relationship with brain damage is unclear. Now, Gorgoraptis et al. reported increased flortaucipir signal (index of increased tau deposition) in long-term TBI survivors compared to healthy controls. Cerebrospinal fluid biomarkers of neurodegeneration and white matter damage correlated with flortaucipir signal, suggesting that flortaucipir PET imaging might be useful for diagnosis and prognostication of neuronal damage after TBI.

Abstract

Traumatic brain injury (TBI) can trigger progressive neurodegeneration, with tau pathology seen years after a single moderate-severe TBI. Identifying this type of posttraumatic pathology in vivo might help to understand the role of tau pathology in TBI pathophysiology. We used flortaucipir positron emission tomography (PET) to investigate whether tau pathology is present many years after a single TBI in humans. We examined PET data in relation to markers of neurodegeneration in the cerebrospinal fluid (CSF), structural magnetic resonance imaging measures, and cognitive performance. Cerebral flortaucipir binding was variable, with many participants with TBI showing increases in cortical and white matter regions. At the group level, flortaucipir binding was increased in the right occipital cortex in TBI when compared to healthy controls. Flortaucipir binding was associated with increased total tau, phosphorylated tau, and ubiquitin carboxyl-terminal hydrolase L1 CSF concentrations, as well as with reduced fractional anisotropy and white matter tissue density in TBI. Apolipoprotein E (APOE) ε4 genotype affected the relationship between flortaucipir binding and time since injury, CSF β amyloid 1–42 (Aβ42) concentration, white matter tissue density, and longitudinal Mini-Mental State Examination scores in TBI. The results demonstrate that tau PET is a promising approach to investigating progressive neurodegeneration associated with tauopathy after TBI.

INTRODUCTION

Traumatic brain injury (TBI) can lead to chronic neurodegeneration and dementia in later life (1, 2). Deposition of hyperphosphorylated protein tau neurofibrillary tangles is a pathological hallmark of this neurodegenerative process (3, 4). Since the first observations in postmortem examinations of brain of boxers (5), the pathological features and clinical correlations of tau neurofibrillary tangle deposition in chronic traumatic encephalopathy (CTE) after repetitive TBI have been increasingly well characterized (3, 6). However, tau deposition also occurs after a single TBI, after which abundant and widely distributed neurofibrillary tangles have been found postmortem in about one-third of patients with TBI (7). As in Alzheimer’s disease (AD), all six isoforms of tau (including both 3- and 4-repeat isoforms) are observed in post-TBI tauopathy (3, 4). However, the distribution of tau pathology after TBI follows a pattern distinct from that observed in AD, concentrating primarily in the depths of sulci and at points of geometric inflection in the cerebral neocortex, whereas the medial temporal lobe is relatively spared in early disease (7, 3). Although the pathogenic role of TBI as a trigger for tau aggregation is not fully understood, traumatic axonal injury (TAI) appears to lead directly to tau hyperphosphorylation (8, 9).

Flortaucipir ([18F]AV-1451 and [18F]T807), a recently developed positron emission tomography (PET) radioligand for tau, enables the examination of tau pathology in vivo (10, 11). Flortaucipir demonstrated potent and specific nondisplaceable binding to tau neurofibrillary tangles in postmortem human brain tissue in AD (1214). Flortaucipir is relatively selective for tau, with no substantial binding to β amyloid (Aβ), α-synuclein, or TAR DNA binding protein 43 (TDP-43) in postmortem brain tissue (12), although it also binds off-target to monoamine oxidases (15) as well as neuromelanin- and melanin-containing cells (12, 16). In vivo, flortaucipir binding is increased in patients with AD with a regional pattern in keeping with the clinical phenotype (1719), cognitive profile (17, 18), and estimated Braak and Braak staging (20). In tau mutation carriers (MAPT gene) and in AD, the distribution of in vivo flortaucipir binding and that of postmortem tau pathology are strongly concordant (19, 21). Furthermore, in both typical AD and posterior cortical atrophy, flortaucipir binding follows the pattern of regional atrophy as quantified by magnetic resonance imaging (MRI) (22, 23).

A recent report on 26 former National Football League players (24) and a previous single case study (25) showed increased cerebral flortaucipir binding after exposure to repetitive TBI. However, the role of flortaucipir in quantifying tau pathology and its distribution in long-term survivors of a single TBI have not been studied. The relationships of tau pathology to clinical outcome and biomarkers of neurodegeneration are also unclear. Of particular interest is the relationship between tau pathology and white matter microstructural changes related to TAI, because these will help to clarify the pathophysiological relationship between the initial injury and progressive neurodegeneration.

Here, we used flortaucipir PET to study the distribution of tau pathology in individuals at least 18 years after a single moderate-severe TBI and in healthy controls. TBI participants were primarily recruited from a patient cohort under follow-up at the University of Glasgow, who have previously been followed up longitudinally in terms of cognitive and disability outcomes (26, 27). We hypothesized that (i) flortaucipir binding would be increased in a proportion of participants with TBI many years after their injury; (ii) flortaucipir binding would correlate with cerebrospinal fluid (CSF) and blood biomarkers of neuronal damage and neurodegeneration, including tau; (iii) flortaucipir binding would be associated with the extent and distribution of diffuse axonal injury, quantified by diffusion MRI and voxel-based morphometry (VBM); and (iv) flortaucipir binding would be associated with poor long-term cognitive outcomes and disability.

RESULTS

Participant demographic and clinical characteristics

Twenty-one participants [7 females and 14 males; median age, 49 years (range, 29 to 72 years)] with a history of a single moderate-severe TBI were included in the study. Nineteen participants with TBI were recruited from the Institute of Health and Wellbeing, Head Injury Research Group, University of Glasgow, UK, and the remaining two (participants P9 and P17 in Table 1) were recruited from the specialist TBI clinic at Imperial College Healthcare NHS Trust, London, UK. Participants were examined at a median time of 32 years (range, 18 to 51 years) after their injury. TBI was caused by road traffic accidents in 18 participants (86%), assault in 2 participants (10%), and fall from a height in 1 participant (5%). Glasgow Outcome Scale–Extended (GOS-E) median score was 6 (range, 4 to 8) at the time of assessment. In 15 TBI participants (71%), longitudinal clinical data were also available (27), including GOS-E and Mini-Mental State Examination (MMSE) scores obtained at a median time point of 16 years (range, 9 to 19 years) after the injury and 17 years (range, 15 to 19 years) before the current clinical and imaging assessment. The demographic and clinical characteristics and apolipoprotein E (APOE) genotype of individual TBI participants are presented in Table 1. Focal lesions were delineated on MRI as specified in the Supplementary Materials. Nineteen TBI participants (90%) had focal lesions apparent on MRI with a median volume of 1634 voxels (range, 12 to 23,118 voxels). A cerebellar lesion was present on structural MRI in one participant with TBI (P7). There were no cerebellar abnormalities on structural MRI in any other participants (fig. S1).

Table 1 Clinical characteristics of individual TBI participants.

Participant designations correspond to those in Figs. 1 and 2 and fig. S1. APOE, apolipoprotein E genotype; GOS-E, Glasgow Outcome Scale–Extended; MMSE, Mini-Mental State Examination; NA, not available; RTA, road traffic accident; TBI, traumatic brain injury.

View this table:

Eleven healthy controls [five females and six males; median age, 57 years (range, 29 to 72 years)] of similar educational background and estimated premorbid intelligence to the TBI participants were also examined. Eight healthy control participants were recruited from the Institute of Health and Wellbeing, Head Injury Research Group, University of Glasgow, UK, and three healthy controls were recruited from the National Institute for Health Research (NIHR) Imperial Clinical Research Facility, London, UK. The demographic characteristics of the TBI and control groups are presented in Table 2.

Table 2 Demographic and neuropsychological comparisons between TBI and healthy control groups, as well as between disabled TBI and good recovery TBI subgroups.

Independent sample Mann-Whitney-Wilcoxon test W and P values (P < 0.05 in bold) for comparisons between groups and subgroups. BIS, Barratt Impulsivity Scale; BVMT, Brief Visuospatial Memory Test; CRT, choice reaction time; PT, People’s Test; FrSBe, Frontal Systems Behavior; HADS, Hospital Anxiety and Depression Score; HVLT, Hopkins Verbal Learning Test; LARS, Lille Apathy Rating Scale; RDI, Recognition Discrimination Index; RT, reaction time; WASI, Wechsler Abbreviated Scale for Intelligence; WTAR, Wechsler Test of Adult Reading.

View this table:

We divided TBI participants into a good recovery and a disabled subgroup based on their GOS-E scores (disabled subgroup, GOS-E ≤ 6; good recovery subgroup, GOS-E > 6). Twelve participants [five females and seven males; median age, 48.5 years (range, 39 to 72 years)] with a median GOS-E of 5 (range, 4 to 6) were included in the disabled subgroup, and nine participants with TBI [two females and seven males; median age, 54 years (range, 29 to 65 years)] with a median GOS-E of 8 (range, 7 to 8) were included in the good recovery subgroup. Longitudinal clinical data were available for nine participants with TBI (75%) in the disabled group and six participants with TBI (67%) in the good recovery group. The demographic characteristics of these subgroups are presented in Table 2. There was no significant difference in age or years of education when each of the subgroups was compared to healthy controls (disabled TBI versus healthy controls: age, independent t test t = −0.91, P = 0.37; education, Mann-Whitney-Wilcoxon W = 60, P = 0.73; good recovery TBI versus healthy controls: age, t = −0.93, P = 0.36; education, W = 64.5, P = 0.26). Estimated premorbid intelligence [Wechsler Test of Adult Reading (WTAR)] was lower in the disabled TBI subgroup when compared to healthy controls (W = 23, P = 0.009) but not significantly different between good recovery TBI and healthy controls (W = 53.5, P = 0.79). WTAR score did not correlate with age at injury in either the disabled TBI subgroup or the TBI group overall (Spearman ρ = −0.04, P = 0.89 and ρ = 0.18, P = 0.43, respectively).

Neuropsychological performance

When compared to healthy controls, participants with TBI manifested impairments on multiple cognitive domains including processing speed, executive function, motivation, inhibition, and verbal and visual memory (Table 2), in a neuropsychological profile typical of TBI. On most neuropsychological measures, these differences were driven by the disabled TBI subgroup (fig. S2). The good recovery TBI subgroup had similar outcomes as healthy controls on most measures, with the exception of the Frontal Systems Behavior (FrSBe) executive and apathy scores, on which both good recovery and disabled TBI subgroups were impaired when compared to healthy controls (good recovery TBI: FrSBe-E, W = 74, P = 0.003; FrSBe-A, W = 63.5, P = 0.046; disabled TBI: FrSBe-E, W = 107.5, P < 0.001; FrSBe-A, W = 94, P = 0.005). The good recovery TBI subgroup performed better than the disabled subgroup on several cognitive domains including logical reasoning, processing speed, executive function, motivation, inhibition, verbal and visual memory, and mood assessment (Table 2).

As shown in Table 2, there was evidence of cognitive decline on longitudinal assessment (average, −1.3 or −0.073 MMSE points per year) in the disabled TBI subgroup, whereas the good outcome TBI subgroup showed improvement of cognitive performance over time (average, 1 or 0.058 MMSE points per year), with longitudinal change being significantly different between the TBI subgroups (W = 7, P = 0.041; Table 2).

Flortaucipir binding in individual participants with TBI

To assess flortaucipir binding and its distribution in the brain in individual participants with TBI, voxel-wise z-scores of Montreal Neurological Institute (MNI)–registered flortaucipir nondisplaceable binding potential (BPND) images were derived for each patient versus the healthy control group. The resulting z-score maps describe the distribution of flortaucipir in individual participants with TBI relative to variations in binding seen in healthy participants. As illustrated in Fig. 1, several participants with TBI had increased flortaucipir binding in a patchy cortical and subcortical distribution, involving both gray and white matter (Fig. 1, P1 to P8 and P10). In contrast, other participants with TBI had similar flortaucipir BPND to that of healthy controls (Fig. 1, P17 to P22).

Fig. 1 Flortaucipir BPND z-score maps for each TBI individual are compared voxel-wise to healthy controls.

TBI participants are presented in descending order of number of voxels with flortaucipir BPND z > 1.645 (see Fig. 2). Axial images are displayed in radiological convention at MNI coordinate 55. Patchy cortical and subcortical increase in tracer uptake is observed in some TBI participants, most consistently in the lateral occipital cortex (P1 to P8 and P10), whereas others show similar BPND values as controls (P17 to P21). The corresponding clinical characteristics are presented in Table 1. Participant designations correspond to those in Table 1, Fig. 2, and fig. S1.

The number of voxels with high BPND (z > 1.645; corresponding to one-tailed P < 0.05) was used to measure the spatial extent of increased flortaucipir signal in individual TBI participants (Fig. 2). Participants with TBI had a median of 841 voxels above threshold (range, 0 to 15,809). For comparison, voxel-wise z-scores were also derived for each healthy control versus the rest of the control group. In controls, the median number of voxels above threshold was 4 (range, 0 to 253; Fig. 2).

Fig. 2 The spatial extent of increased flortaucipir binding in TBI is expressed as the number of voxels with flortaucipir BPND z > 1.645 for each patient with TBI compared to the healthy control group.

The dashed line represents the number of voxels above threshold in the healthy control participant with the maximum number of voxels above that threshold (compared to the rest of the controls). Participant designations correspond to those in Table 1, Fig. 1, and fig. S1. The inset figure shows the number of voxels above threshold (in logarithmic scale) in TBI and healthy controls.

Eight participants with TBI (38%) had >2000 voxels (equivalent to 16 cm3 of brain volume) above threshold, indicating spatially extensive flortaucipir signal increase, whereas further seven participants (33%) showed increased signal of more spatially limited extent (254 to 1999 voxels; Fig. 2). The remaining six participants (29%) had less than 253 voxels above threshold, falling within the range of the healthy control group. There was no difference in flortaucipir spatial extent when comparing between disabled and good functional outcome subgroups (Mann-Whitney-Wilcoxon W = 68.5, P = 0.32).

Reduced flortaucipir binding within focal lesions in TBI

Flortaucipir binding within focal lesions was significantly reduced when compared to nonlesioned cerebral areas (Wilcoxon signed-rank V = 22, P = 0.002; fig. S3). Areas of markedly reduced flortaucipir uptake (in black) in Fig. 1 correspond to focal lesions in individual TBI participants (cf. fig. S1). Focal lesion size (number of voxels) did not correlate with average flortaucipir BPND within the nonlesioned gray or white matter (Spearman ρ = −0.37, P = 0.10 and ρ = −0.15, P = 0.51, respectively; fig. S3) or with flortaucipir spatial extent (number of voxels above threshold; ρ = −0.06, P = 0.78; fig. S3).

Increased flortaucipir binding in the TBI patient group versus healthy controls

Flortaucipir BPND was next compared between TBI and healthy control groups, in a nonparametric voxel-wise analysis, excluding lesioned areas (fig. S1) and the striatum where off-target binding is observed with this tracer. Flortaucipir binding was significantly higher in TBI participants than in healthy controls in the right lateral occipital cortex (P < 0.05, adjusted for multiple comparisons; Fig. 3, inset). When each of the TBI subgroups was independently compared to healthy controls, significantly increased flortaucipir binding was found in both TBI subgroups, in the same right lateral occipital area (P < 0.05, adjusted for multiple comparisons; fig. S4 and cf. Fig. 3). There was no significant difference in flortaucipir binding when comparing between the disabled (GOS-E ≤ 6) and good outcome (GOS-E > 6) TBI subgroups (fig. S4). There were no areas of significantly decreased binding in the TBI group or subgroups compared to healthy controls (all P > 0.05; fig. S4).

Fig. 3 Flortaucipir BPND is increased in TBI compared to healthy controls.

Voxels with increased flortaucipir uptake in TBI at P < 0.05, not adjusted for multiple comparisons, are shown on axial slices for illustration. Flortaucipir binding was increased in TBI in a small cluster of voxels in the right parietal cortex (inset, P < 0.05, corrected for multiple comparisons).

In addition, we examined flortaucipir BPND within a medial temporal lobe region of interest (ROI) including the entorhinal, perirhinal, and parahippocampal cortices and the hippocampus, a region typically affected by tau accumulation in AD. There was no significant difference in flortaucipir binding within this ROI between the TBI group, or either of the disabled or good recovery TBI subgroups, and healthy controls (all TBI versus healthy controls: Mann-Whitney-Wilcoxon W = 108, P = 0.78; disabled TBI versus healthy controls: W = 57, P = 0.61; good recovery TBI versus healthy controls: W = 51, P = 0.94; fig. S5).

Correlations between flortaucipir and CSF total tau, phosphorylated tau, and ubiquitin C-terminal hydrolase L1 in TBI

Next, CSF and plasma biomarkers of neurodegeneration and neuronal damage, including total tau (T-tau), tau phosphorylated at amino acid 181 (P-tau), ubiquitin carboxy-terminal hydrolase L1 (UCH-L1), Aβ42, neurofilament light (NFL), glial fibrillary acidic protein (GFAP), and protein S100 were assessed in TBI and healthy controls, and their relationship with flortaucipir binding was explored. CSF data were available from 19 participants with TBI and 10 healthy controls.

In TBI, CSF T-tau was significantly correlated with flortaucipir BPND in the cerebral cortical gray matter (Spearman ρ = 0.53, P = 0.022; Fig. 4A) but not in the white matter (ρ = 0.44, P = 0.059; fig. S6). Conversely, CSF P-tau correlated with flortaucipir BPND in the cerebral white matter (ρ = 0.52, P = 0.024; Fig. 4B) but not in the gray matter (ρ = 0.32, P = 0.18; fig. S6). There was no correlation between CSF T-tau or P-tau and flortaucipir spatial extent (T-tau: ρ = 0.21, P = 0.38; P-tau: ρ = 0.29, P = 0.24; fig. S6). CSF tau in healthy controls did not correlate with flortaucipir binding in the cerebral gray matter (T-tau: ρ = 0.15, P = 0.68; P-tau: ρ = −0.21, P = 0.56; fig. S6) or white matter (T-tau: ρ = −0.09, P = 0.81; P-tau: ρ = 0, P = 1; fig. S6). Plasma T-tau concentration did not correlate with flortaucipir binding in either TBI or healthy controls (fig. S6). There were no differences in T-tau or P-tau concentration in CSF or plasma between TBI and healthy control groups (CSF T-tau: W = 93, P = 0.95; CSF P-tau: W = 67, P = 0.21; plasma T-tau: W = 118, P = 0.76; fig. S6).

Fig. 4 Flortaucipir binding is associated with CSF biomarkers in TBI.

CSF total tau (T-tau) (A), phosphorylated (P-tau) (B), and UCH-L1 (C) concentrations in the CSF are positively correlated with flortaucipir BPND (normalized against the healthy control group) in the cerebral gray matter (GM; T-tau and UCH-L1) and in the cerebral white matter (WM; P-tau) in TBI participants.

UCH-L1 in the CSF in TBI participants also positively correlated with flortaucipir BPND in the gray matter (Spearman ρ = 0.52, P = 0.023; Fig. 4C) but not in the white matter (ρ = 0.42, P = 0.073; fig. S7). There was no association between plasma UCH-L1 and flortaucipir binding in TBI (fig. S7). In healthy controls, there were no associations between flortaucipir binding and UCH-L1 in the CSF or plasma (fig. S7). There was no difference between TBI and control groups in UCH-L1 concentration (fig. S7). There were also no associations between flortaucipir BPND and Aβ42, NFL, GFAP, or S100 in TBI or healthy controls and no difference between TBI and control groups in the CSF or plasma (fig. S7).

Flortaucipir spatial extent and white matter damage in TBI

Next, we investigated the relationship between flortaucipir binding and structural brain injury produced by TBI (28). White matter integrity measured by fractional anisotropy (FA) was significantly reduced in TBI when compared to healthy controls (Mann-Whitney-Wilcoxon W = 44, P = 0.004; Fig. 5A), in keeping with the presence of TAI in TBI. As expected in TBI (29), reduced FA correlated with worse performance on multiple cognitive measures (table S1). Tissue density measured by VBM was also significantly reduced in the nonlesioned white matter in TBI (W = 61, P = 0.031; Fig. 5B). Tissue density in the nonlesioned gray matter was not significantly different between TBI and healthy control groups (W = 113, P = 0.94; Fig. 5C).

Fig. 5 White matter microstructural changes in TBI are associated with flortaucipir spatial extent.

(A) Cerebral white matter fractional anisotropy (FA) and (B) voxel-based morphometry (VBM)–derived white matter tissue density are reduced in TBI when compared to healthy controls (HC). This was not observed with gray matter tissue density (C). Flortaucipir spatial extent (number of voxels above threshold) in TBI is associated with reduced white matter FA (D) and white matter tissue density (E) but not with gray matter tissue density (F). **P < 0.005, *P < 0.05.

The spatial extent of flortaucipir binding (number of voxels with BPND z > 1.645) correlated with the degree of white matter damage seen in TBI participants. The number of voxels showing increased flortaucipir binding was negatively correlated with both average FA (Spearman ρ = −0.48, P = 0.027; Fig. 5D) and average white matter density (ρ = −0.46, P = 0.037; Fig. 5E), in keeping with a relationship between cerebral tau pathology with microstructural white matter disruption. There was no significant correlation between flortaucipir spatial extent and gray matter tissue density (ρ = −0.31, P = 0.17; Fig. 5F).

White matter damage in areas of increased flortaucipir binding

We next examined whether greater white matter pathology was seen in areas of increased flortaucipir binding. Because the localization of increased tau varies from patient to patient, we defined areas of high and normal flortaucipir signal (BPND z > or < 1.645) in each patient. Eight TBI participants who did not show high flortaucipir signal within the white matter were excluded from these analyses. Significantly lower FA (Wilcoxon signed-rank V = 19, P = 0.018; Fig. 6, A and B) and white matter density (V = 17, P = 0.025; Fig. 6, C and D) were seen in areas of high flortaucipir binding compared to areas where binding was not increased, demonstrating relatively high flortaucipir binding in areas of posttraumatic white matter damage. The corresponding VBM analysis for gray matter density did not reveal differences in gray matter density between areas of high and those of normal flortaucipir binding (V = 97, P = 0.95; fig. S8). There was considerable between-subject variability in these effects (Fig. 6, B and D, and fig. S8B).

Fig. 6 Increased white matter flortaucipir BPND colocalizes with white matter microstructural changes in TBI.

Skeletonized FA (A) and white matter density (C), standardized against controls, were compared within participant between areas of increased flortaucipir binding (in red) and areas where binding was not increased (in blue). A representative axial slice for each patient shows the spatial distribution of the white matter areas compared, and images are presented in the order of greatest to smallest effect size (greatest decrease in FA or white matter density z-score in high flortaucipir areas shown top left). Both FA z-scores (B) and white matter density z-scores (D) are lower within areas of increased flortaucipir binding, defined as BPND z > 1.645.

Cortical flortaucipir binding and white matter microstructural damage

Next, we examined whether flortaucipir BPND within the right lateral occipital cortical area where flortaucipir signal was increased in TBI versus controls (Fig. 3) was associated with white matter tract FA. Increased right lateral occipital flortaucipir binding was associated with reduced FA in remote white matter regions, including association, commissural, and projection tracts (Fig. 7, A and B, and table S2). Correlations were observed in the genu and body of the corpus callosum, as well as in several association tracts within the ipsilateral (right) hemisphere, including the cingulum bundle, inferior longitudinal fasciculus, uncinate fasciculus, and anterior thalamic radiation (Fig. 7B and table S2), but not in the contralateral hemisphere (fig. S9 and table S2). Higher cortical flortaucipir BPND was associated with reduced tissue density in remote white matter regions including the corpus callosum and right prefrontal white matter (Fig. 7C). The same analysis for gray matter density did not show an association between flortaucipir BPND within the lateral occipital cluster and cerebral cortical gray matter density (P > 0.05, corrected for multiple comparisons).

Fig. 7 Increased cortical flortaucipir binding in the right lateral occipital area is associated with microstructural white matter damage in TBI.

(A) Increased flortaucipir BPND in the right lateral occipital cortex in TBI participants was associated with reduced diffusion MRI–derived FA in remote white matter tracts (top), partly overlapping with white matter tracts of reduced FA in TBI when compared to controls (bottom). (B) Flortaucipir BPND in the right lateral occipital cortex in TBI participants correlated negatively with average skeletonized FA within the genu and body of the corpus callosum as well as within the cingulum bundle, inferior longitudinal fasciculus (ILF), anterior thalamic radiation (ATR), and uncinate fasciculus in the right hemisphere. Correlation coefficients and P values are presented in table S2. (C) Top: Increased flortaucipir BPND in the right lateral occipital cortex in TBI participants is associated with reduced VBM-derived white matter density in the corpus callosum (cal) and right prefrontal area (PF). This distribution partly overlaps with areas of white matter atrophy in TBI when compared to controls (bottom). Results are presented in radiological convention; color maps represent P < 0.05, adjusted for multiple comparisons.

Flortaucipir binding and clinical measures in TBI

There were no correlations of average flortaucipir BPND in the cerebral gray or white matter or the spatial extent of high flortaucipir signal in TBI participants (number of voxels above BPND z > 1.645) with age, time since injury, disability (GOS-E), or any neuropsychological measures in the TBI participants (all P > 0.05, false discovery rate (FDR)–adjusted; table S3).

APOE ε4 genotype interactions with flortaucipir binding and time since injury, CSF Aβ42, white matter density, and MMSE scores in TBI

There were no overall differences between participants with at least one APOE ε4 allele and those without an APOE ε4 allele in either cerebral flortaucipir BPND or in flortaucipir spatial extent in the TBI group (Mann-Whitney-Wilcoxon W = 41, P = 0.46 and W = 44, P = 0.60, respectively; fig. S10). In further exploratory analyses, we examined whether APOE carrier status (presence of at least one ε4 allele) influenced the relationship between flortaucipir measures (average cerebral BPND and spatial extent) and age, time since injury, CSF biomarkers, white matter MRI measures, and neuropsychological performance in TBI. Interactions between APOE status and each of the above variables on flortaucipir measures were examined using linear regression in the TBI group. Interaction plots are shown in fig. S11.

There was an interaction between APOE genotype and time since injury on cerebral flortaucipir BPND (F1,17 = 15.11, P = 0.001) but not on flortaucipir spatial extent. Flortaucipir BPND increased with time elapsed since the TBI in APOE ε4 carriers but not in ε4 noncarriers (fig. S11). Age did not interact with APOE genotype on average cerebral flortaucipir BPND or on flortaucipir spatial extent (fig. S11).

There was an interaction between APOE genotype and CSF Aβ42 on flortaucipir BPND (F1,15 = 6.15, P = 0.026) but not on flortaucipir spatial extent in TBI (fig. S11). Higher cerebral flortaucipir binding was associated with lower Aβ42 concentration in CSF in APOE ε4 carriers but not in ε4 noncarriers (fig. S11). There were no interactions on flortaucipir measures between APOE ε4 carrier status and T-tau, P-tau, UCH-L1, NFL, GFAP, or S100 CSF concentrations (fig. S11).

The effect of white matter tissue density (derived from VBM) on flortaucipir spatial extent was influenced by APOE genotype (F1,17 = 6.90, P = 0.018), with ε4 noncarriers showing increased flortaucipir binding with reducing white matter tissue density, which was not observed in APOE ε4 carriers (fig. S11). There were no interactions between white matter tissue density and APOE genotype on average cerebral flortaucipir BPND and no interactions between gray matter tissue density or FA and APOE genotype on flortaucipir measures (fig. S11).

APOE genotype influenced the relationship of longitudinal MMSE scores and flortaucipir binding, with decreasing MMSE scores over time associated with higher cerebral flortaucipir BPND in ε4 noncarriers but not in ε4 carriers (F1,10 = 40.76, P < 0.0001; fig. S11). Current MMSE scores also interacted with APOE ε4 carrier status on flortaucipir spatial extent, with current MMSE scores being lower with increasing spatial extent in ε4 noncarriers but not in ε4 carriers (F1,17 = 6.14, P = 0.024; fig. S11).

Similar off-target binding in TBI and healthy controls

In keeping with the known off-target binding of flortaucipir to neuromelanin (12), flortaucipir BPND was increased in the striatum both in TBI participants and in healthy controls when compared to the rest of the cerebral white and gray matter, excluding focal lesions (TBI: Wilcoxon signed-rank V = 214, P = 0.0002; controls: V = 66, P = 0.001; fig. S12). However, flortaucipir BPND in the striatum and in the choroid plexus (16) was not different between TBI participants and healthy controls (fig. S12).

DISCUSSION

Hyperphosphorylated tau deposition is a key pathological feature of neurodegeneration triggered by TBI (3, 4). Patients are at increased risk of AD and CTE after TBI, both characterized by the accumulation of tau pathology (1, 30, 31). PET ligands have recently been developed that specifically bind hyperphosphorylated tau, enabling the detection of tau pathology in vivo (32). We used flortaucipir PET to investigate tau pathology in a well-characterized cohort of individuals who were examined many years after a single moderate-severe TBI (26, 27). As predicted, there was considerable variability in the extent of flortaucipir binding across TBI participants. Broadly, a third of TBI participants showed extensive increases in cerebral flortaucipir binding, a third showed more limited increases, and a third showed no abnormality. The proportion of individuals with increased flortaucipir binding is much higher than would be expected in the general population in this age group (33). At the group level, TBI participants showed elevated flortaucipir binding when compared to age- and education-matched healthy controls.

Flortaucipir binds to neurofibrillary tangles containing abnormal tau in postmortem human brain tissue in AD (1214). A lack of specificity has been a critical issue with previous PET tau ligands (34), but flortaucipir is more selective and does not show substantial Aβ, α-synuclein, or TDP-43 binding in postmortem brain tissue (12). However, off-target binding to monoamine oxidases and neuromelanin- and melanin-containing cells is a known limitation (12, 15, 16). Flortaucipir has been validated for the study of tau pathology in patients with AD, showing binding that correlates with clinical phenotype (1719), cognitive profile (17, 18), and estimated Braak and Braak staging (20). Tau pathology seen after TBI appears to be similar to AD, with all six isoforms of tau (including both 3- and 4-repeat isoforms) observed (3, 4). As a result, flortaucipir is expected to bind tau pathology seen after TBI in a similar way to AD, albeit with a distinct spatial pattern (3).

The pattern of flortaucipir binding potentially provides diagnostic information about the type of posttraumatic neurodegeneration. Tau pathology seen after TBI often has a distribution distinct from AD. In cases of CTE, tau pathology is concentrated in depths of cortical sulci (3, 6) and can be widely distributed in cortical areas after single TBI (7). In our study, some patients showed extensive flortaucipir binding in cortical areas and in parts of the white matter. This spatial pattern would be atypical for AD (35) but is more in keeping with CTE (3, 4). It is also compatible with the spatially extensive tau neurofibrillary tangle pathology seen in about one-third of patients with chronic TBI (7).

The relationship between tau pathology and flortaucipir binding is supported by our CSF results. CSF tau is raised in incipient AD (36) and other tauopathies (37) and provides a diagnostic biomarker for AD (38). CSF tau increases in the days after TBI, generally falling to normal concentration 8 to 12 weeks after the injury (39, 40). To our knowledge, our study is the first to examine CSF tau several years after moderate-severe TBI. CSF T-tau and P-tau were not increased across the TBI group, which might be due to the high variability within the TBI group. CSF T-tau and P-tau concentrations correlated positively with flortaucipir binding in the cerebral cortex and white matter, respectively, supporting the conclusion that increased flortaucipir binding is indicative of the presence of tau pathology. We also observed a positive correlation between CSF UCH-L1 concentration and cortical flortaucipir uptake. In the acute phase, UCH-L1 may indicate the presence of neuronal injury after TBI (41). It is abundant in cerebral neurons and is an important component of the ubiquitin-proteasome system, (42). Dysfunction of UCH-L1 is implicated in a number of neurodegenerative diseases (43, 44), and UCH-L1 is found within misfolded protein aggregates, including neurofibrillary tangles (45). CSF UCH-L1 is elevated in patients with AD (46), and our results suggest that the link between TBI and dysfunction in the ubiquitin-proteasome system warrants further investigation.

We also showed a relationship between flortaucipir binding and TAI. This is in keeping with a causative role for TAI in the pathophysiology of posttraumatic tau pathology. Mechanical forces exerted at the time of head injury are thought to disrupt axonal organization, producing damage to microtubule structure and associated axonal tau (4). This damage may lead to hyperphosphorylation of tau, misfolding, and neurofibrillary tangle formation, which eventually causes neurodegeneration (8, 9, 47). Mechanical forces are maximal in points of geometric inflection such as the base of cortical sulci (48), where tau pathology is seen in CTE.

In AD, increased flortaucipir binding is greater in cortical regions that show atrophy (22, 23). We observed a different relationship after TBI. Flortaucipir binding was increased in white matter areas that also showed reduced FA and atrophy. This suggests that tau pathology was more marked in areas of TAI. In addition, cortical flortaucipir binding also correlated with the extent of TAI within the white matter. For example, flortaucipir binding in the right lateral occipital cortex, the most consistent area of increase, was correlated with reduced FA in a tract directly connected to that cortical area, the inferior longitudinal fasciculus. This relationship might be explained by slow Wallerian degeneration that follows the initial injury and leads to the accumulation of tau pathology in connected cortical regions (49). Alternatively, prion-like spread of tau pathology might over time result in the cortical accumulation of tau pathology in regions that are connected to white matter regions initially damaged by TAI (50, 51). We have previously reported a similar relationship between cortical amyloid pathology measured using [11C]-PiB PET and the degree of TAI in connected tracts (52), suggesting that structural connectivity may influence the neuroanatomical distribution of posttraumatic tau and amyloid pathology in similar ways.

A recent study on former National Football League players, exposed to repetitive TBI and manifesting cognitive, mood, and behavioral symptoms, also found raised cerebral flortaucipir binding, which increased with years of exposure to repetitive TBI (24). With our results, this supports the hypothesis that the risk for neurodegeneration from TBI may be dose dependent (1), produced either by a single severe TBI or repeated exposure to more minor TBI.

Genetic factors may partially explain the variability in the neurodegenerative trajectory after TBI (53, 54). APOE ε4, the strongest risk factor for sporadic AD, is associated with increased Aβ and tau pathology (55), and it is also the strongest genetic predictor of adverse outcome after TBI (56). APOE ε4 may act synergistically with TBI to elevate the risk of neurodegeneration (30, 57). Therefore, in exploratory analyses, we examined how APOE ε4 carrier status might influence flortaucipir binding. Flortaucipir binding increased with time since TBI in APOE ε4 carriers and was associated with lower Aβ42 concentrations in the CSF (indicating higher Aβ42 concentrations in the brain) in ε4 carriers. This supports an influence of APOE ε4 on the risk of neurodegeneration after TBI. In keeping with these findings, Aβ plaque pathology after TBI is more prevalent in individuals with at least one APOE ε4 allele (57). Results from animal models also support the association between APOE ε4 genotype and Aβ pathology after TBI (58).

Our study has a number of limitations. The relatively small sample size may have limited our ability to detect associations between flortaucipir binding and clinical outcomes. Hence, the absence of a clinical effect should be interpreted with caution, because our study is unlikely to have been adequately powered for these outcomes. In AD, flortaucipir uptake was associated with poorer performance in various cognitive domains in regionally specific patterns (17). However, the situation is likely to be more complex after TBI because the influence of the initial injury on cognitive state will confound progressive neurodegenerative effects, resulting in the need for larger sample sizes. Estimated premorbid intelligence was lower in the disabled TBI subgroup when compared to healthy controls. However, vocabulary development may be adversely affected by TBI during development (59). Hence, it is possible that performance on WTAR was directly reduced by the TBI, especially in individuals injured during childhood. Flortaucipir binds off-target to monoamine oxidases and neuromelanin, which is particularly apparent in the striatum and choroid plexus (12, 13, 15, 16). This off-target flortaucipir binding may have limited our ability to detect underlying tau pathology in these regions. In addition, a contribution from non–tau pathology cannot be excluded with certainty. Flortaucipir binding increases have been reported in semantic variant primary progressive aphasia, where the underlying pathology is TDP-43 (60), although flortaucipir binding to TDP-43 has not been demonstrated postmortem (12). Off-target binding has also been reported in areas of acute parenchymal and subarachnoid hemorrhage but not in areas showing superficial siderosis due to chronic hemorrhage (12). This is unlikely to be a major confound in our study because there were no acute hemorrhagic lesions in our participants and binding was reduced within focal lesions. Flortaucipir signal is reduced at the site of focal lesions. Therefore, the presence of a focal lesion in one or more TBI participants within a given brain area would be expected to reduce our sensitivity in detecting flortaucipir group effects within that area. To minimize the impact of lesions in our analysis, we took into account the location of focal lesions as regressors in voxel-wise group comparisons. Last, brain atrophy can affect on PET results, but we used stringent measures to minimize the effects of atrophy during preprocessing. Moreover, the relationship between increased flortaucipir binding and reduced white matter tissue density is unlikely to be artifactual, because nonspecific flortaucipir binding is expected to be reduced within areas of reduced tissue density.

In summary, we show that flortaucipir binding is increased many years after a single moderate-severe TBI. Increased binding was associated with CSF indicators of neurodegeneration, including T-tau and P-tau, as well as the presence of TAI that may provide the initial trigger to its accumulation. The ability to detect tau pathology in vivo after TBI has major potential implications for diagnosis and prognostication of clinical outcomes after TBI. It is also likely to assist in patient selection and stratification for future treatment trials targeting tau.

MATERIALS AND METHODS

Study design

This study was designed to assess whether flortaucipir PET provides evidence of tau pathology in individuals several years after a single moderate-severe TBI when compared to healthy controls. We also examined the relationship of flortaucipir signal with CSF and blood biomarkers, MRI measures of axonal injury, and cognitive and disability outcomes. Imaging assessment and biological sampling were cross sectional, and longitudinal cognitive and disability measures were available in a subset of individuals with TBI. The sample size was calculated to provide 80% statistical power with a type 1 error rate of 0.05 to detect an increase in flortaucipir binding in TBI compared to controls, based on effect sizes extrapolated from previous flortaucipir data on AD (11) and from postmortem data showing tau pathology in about a third of individuals after moderate-severe TBI (7). Investigators performing the PET modeling, MRI processing, and lesion delineation were blinded as to clinical outcomes and biomarker data. Randomization was not applicable in this observational study.

TBI participants were primarily recruited from two TBI cohorts under follow-up at the Institute of Health and Wellbeing, Head Injury Research Group, University of Glasgow, UK. The original cohorts included all individuals acutely admitted with TBI to the Southern General Hospital, Glasgow between 1968 and 1985 (27) and between 1996 and 1999 (26). These individuals have been followed up longitudinally in terms of cognitive, well-being, and disability outcomes (26, 27). Members of these follow-up cohorts fulfilling the inclusion criteria for the current study were invited to participate with the assent of their general practitioner (primary care physician). Additional TBI participants fulfilling the inclusion criteria were recruited from the specialist multidisciplinary TBI clinic at Imperial College Healthcare NHS Trust, London, UK. Individuals were recruited into the TBI group with the following inclusion criteria: (i) a history of a single moderate-severe TBI [according to Mayo classification (61)], (ii) age over 18 years, (iii) capacity to provide written informed consent, (iv) no prior neurological or psychiatric illness, (v) no contraindication to PET or prior radiation exposure that, when combined with the dose from the present study, would exceed 10 mSv in addition to the natural background radiation in the previous 3 years, (vi) no contraindication to MRI, and (vii) no medication use or allergies that may compromise participant safety or interfere with study procedures. TBI participants were divided into two subgroups based on functional outcome. Poor functional outcome was defined as a GOS-E score of 6 or less (moderate-severe disability) at the time of assessment and those with GOS-E greater than 6 as good outcome. Healthy volunteers of similar age and socioeconomic background as the TBI participants were also recruited. The study was approved by the Westminster Research Ethics Committee and the Administration of Radioactive Substances Advisory Committee. All participants gave written informed consent.

Procedures

All participants were administered an intravenous bolus of flortaucipir (also known as [18F]AV-1451 and [18F]T807; average dose, 250 megabecquerels), and dynamic PET scans were acquired over 90 min. Flortaucipir was supplied by Avid Radiopharmaceuticals, a wholly owned subsidiary of Eli Lilly and Company, and PET image acquisition was carried out at Imanova Centre for Imaging Sciences (currently Invicro London). Flortaucipir PET data were analyzed with the Molecular Imaging and Kinetic Analysis Toolbox (www.miakat.org) using the simplified reference tissue model (62, 63) with cerebellar gray matter as the reference tissue, based on postmortem studies in CTE that demonstrated relative sparing from tau pathology in that region (3, 64). All participants also underwent structural 3-T MRI, including volumetric T1 and diffusion tensor imaging (DTI). Details on PET and MRI acquisition and analyses are provided in the Supplementary Materials.

TBI and control participants took part in neuropsychological assessment and had venous blood and CSF sampling. Clinical assessment and blood and CSF sample acquisition were carried out at the NIHR/Wellcome Trust Imperial Clinical Research Facility. Details on neuropsychological assessments and biological sample analyses are provided in the Supplementary Materials.

Statistical analyses

Statistical analyses were carried out using R statistical software (www.r-project.org) (65), unless specified otherwise. Voxel-wise differences in BPND between TBI participants and healthy controls, between each of the disabled TBI and good recovery TBI subgroups and healthy controls, and between these subgroups were assessed using permutation tests in the FMRIB Software Library (FSL) (66). The same procedure was used to assess voxel-wise differences between TBI participants and healthy controls in (i) VBM-derived white matter density, (ii) gray matter density, and (iii) diffusion MRI–derived skeletonized FA. Voxels corresponding to structural lesions were excluded from the analyses using participants’ lesion maps as individual voxel-wise repressors. All cerebral gray and white matter voxels were included except those corresponding to the striatum, where there is known off-target binding (12). One thousand permutations were used, and results were cluster-corrected for multiple comparisons using threshold-free cluster enhancement (TFCE) and a family-wise error rate of 0.05. Localization of voxel clusters was reported on the basis of the Harvard-Oxford probabilistic atlas (67) and the Johns Hopkins University (JHU) DTI-based white matter atlas (68).

To investigate whether white matter tau pathology shared common localization with TAI and/or white matter atrophy in TBI, we compared skeletonized FA and, separately, average white matter density (z-scores) within areas of high flortaucipir signal (BPND z > 1.645) versus white matter areas where binding was not increased (BPND z < 1.645) using (paired) Wilcoxon signed-rank tests. Any voxels corresponding to focal lesions were excluded. Only TBI participants with 15 or more white matter voxels with increased flortaucipir binding within the regions of interest were included.

To examine the localization of white matter microstructural damage predicted by increased flortaucipir cortical signal in TBI participants, standardized BPND within the right parietal cortical area of flortaucipir increase in TBI was used as a regressor in a voxel-wise general linear model (GLM) of skeletonized FA and, separately, in a voxel-wise GLM of white matter density. Statistical significance was tested using permutation tests in FSL (66). As previously, voxels corresponding to lesions were excluded for each individual, 1000 permutations were used and results were obtained using TFCE, corrected for multiple comparisons using a family-wise error rate of 0.05. Localization of voxel clusters was reported on the basis of the JHU DTI-based white matter atlas (68). Post hoc correlations were examined using Spearman’s rank correlation between standardized BPND within the right parietal cortical area of flortaucipir increase in TBI participants and skeletonized FA in the following tracts: right and left superior longitudinal, inferior longitudinal, inferior fronto-occipital fasciculi and cingulum bundles, corpus callosum (genu, body, and splenium), as well as in the right and left corticospinal tracts as a control.

Correlations between clinical/neuropsychological measures and PET markers in TBI participants were examined using Spearman’s rank correlation with FDR correction for multiple comparisons. Interactions between APOE genotype (presence of at least one ε4 allele) and age, time since injury, CSF biomarkers, white matter MRI measures, and neuropsychological performance on cerebral flortaucipir BPND and flortaucipir spatial extent were examined in the TBI group using linear regression in each case to assess the statistical significance of the interaction term and interaction plots to determine the direction of the effect in ε4 carriers versus noncarriers. Assumptions of multiple linear regression analysis were tested as follows: Analysis of standard residuals was carried out to identify outliers; variance inflation factor was used to test for multicollinearity; Durbin-Watson test was conducted to test for independence of errors; and approximate normal distribution of errors, homoscedasticity, and linearity were assessed by plotting regression-standardized residuals.

SUPPLEMENTARY MATERIALS

stm.sciencemag.org/cgi/content/full/11/508/eaaw1993/DC1

Materials and Methods

Fig. S1. Focal lesions in TBI participants.

Fig. S2. Individual data points as well as means and SDs for demographic and neuropsychological results in the healthy control group, disabled TBI subgroup, and good recovery TBI subgroup.

Fig. S3. Flortaucipir binding and focal lesions.

Fig. S4. Voxel-wise comparisons between subgroups in flortaucipir binding.

Fig. S5. Flortaucipir medial temporal lobe BPND in the healthy control, disabled TBI, and good recovery TBI groups.

Fig. S6. Relationships of CSF T-tau and P-tau and plasma T-tau concentrations with flortaucipir gray matter BPND, white matter BPND, and spatial extent in TBI and healthy controls.

Fig. S7. Relationships of CSF and plasma UCH-L1, NFL, GFAP, S100, and CSF Aβ42 with cerebral gray matter and white matter flortaucipir BPND in TBI and healthy controls.

Fig. S8. Gray matter density within areas of increased flortaucipir BPND in TBI.

Fig. S9. Flortaucipir right parietal cluster (TBI > control) BPND (z-score compared to healthy controls) and FA in white matter tracts in TBI.

Fig. S10. No difference in flortaucipir BPND (z-score compared to healthy controls) or spatial extent (number of voxels above threshold) in TBI when comparing individuals with at least one APOE ε4 allele and those not carrying an ε4 allele.

Fig. S11. Interactions between APOE genotype and each of time since injury, age, CSF T-tau, P-tau, Aβ42, NFL, GFAP, S100 concentrations, white matter and gray matter densities, and MMSE scores (longitudinal change and current) on flortaucipir BPND [z-score compared to controls and spatial extent (N voxels above threshold)] in TBI.

Fig. S12. Off-target flortaucipir binding in the striatum and choroid plexus compared to other cerebral areas in healthy controls and TBI.

Fig. S13. Flortaucipir reference region (cerebellar gray matter) time activity curves.

Table S1. Correlations between neuropsychological measures and average white matter tract FA in TBI.

Table S2. Correlations between flortaucipir right parietal cluster (TBI > control) BPND (z-score compared to healthy controls) and FA in white matter tracts in TBI participants.

Table S3. Correlations between clinical measures and each of flortaucipir right parietal cluster (TBI > control) BPND, whole cerebral cortex BPND (z-score compared to the healthy control group), white matter BPND (z-score compared to the healthy control group), and spatial extent (N of voxels with BPND z-score > 1.645 of healthy controls) in TBI participants.

References (6982)

REFERENCES AND NOTES

Funding: This project was funded by the Medical Research Council (MRC) UK grant number MR/L022141/1 (to D.J.S.). D.J.S. was supported by an NIHR Research Professorship (NIHR-RP-011-048). This work was supported by the UK Dementia Research Institute. Infrastructure was supported by the National Institute of Health Research (NIHR) Imperial Biomedical Research Centre. Author contributions: N.G.: Acquisition, analysis, and interpretation of data, drafting and critical revision of the article, and final approval of the version to be published. L.M.L.: Acquisition, analysis and interpretation of data, critical revision of the article, and final approval of the version to be published. A.W.: Analysis and interpretation of data, drafting and critical revision of the article, and final approval of the version to be published. K.A.Z.: Acquisition of data, critical revision of the article, and final approval of the version to be published. L.M.M.: Acquisition of data, critical revision of the article, and final approval of the version to be published. C.M.: Acquisition of data, critical revision of the article, and final approval of the version to be published. E.R.: Acquisition of data, critical revision of the article, and final approval of the version to be published. A.H.: Analysis of data, critical revision of the article, and final approval of the version to be published. H.Z.: Design and interpretation of data, critical revision of the article, and final approval of the version to be published. J.P.: Design and interpretation of data, critical revision of the article, and final approval of the version to be published. P.M.M.: Study design, interpretation of data, critical revision of the article, and final approval of the version to be published. R.N.G.: Study design, interpretation of data, critical revision of the article, and final approval of the version to be published. T.M.M.: Study design, interpretation of data, critical revision of the article, and final approval of the version to be published. D.J.S.: Conception and design of the study, interpretation of data, drafting and critical revision of the article, and final approval of the version to be published. Competing interests: None of the authors declare any competing interests related to the current study. H.Z. has served at scientific advisory boards for Roche Diagnostics, Wave, Samumed, and CogRx, has lectured for AlzeCure, and is a cofounder of Brain Biomarker Solutions in Gothenburg AB, a GU Ventures–based platform company at the University of Gothenburg. P.M.M. acknowledges personal and research support from the Edmond J. Safra and Lily Safra Foundation, an NIHR Senior Investigator Award, and the UK Dementia Research Institute. P.M.M. is also reimbursed for service on a Scientific Advisory Board to Ipsen Biopharmaceuticals; has received consultancy fees from Roche, Adelphi Communications, Celgene, and Biogen OrbiMed; has received honoraria or speakers’ fees from Novartis and Biogen; and has received research or educational funds from Biogen, Novartis, and GlaxoSmithKline. R.N.G. is a consultant for AbbVie, Biogen, and Cerveau. N.G., L.M.L., A.W., K.A.Z., L.M.M., C.M., E.R., A.H., J.P., T.M.M., and D.J.S. have no disclosures. Data and materials availability: All data associated with this study are present in the paper or the Supplementary Materials.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article