Graphical abstract
graphic file with name fx1.jpg
[91]Open in a new tab
Highlights
* •
17q21.31 structural haplotypes include a 1 Mb inversion (H) and
duplications (α, β, γ)
* •
H1β1γ1 was associated with CTE neuropathological and clinical
endophenotypes
* •
Among football players, H1β1γ1 effect was similar to effect of
playing 9.6 years
* •
In DGE analysis, H1β1γ1 showed up-regulation of immune-related
genes and pathways
__________________________________________________________________
Han et al. identify H1β1γ1, a structural haplotype in the 17q21.31
region that contains MAPT, as a genetic risk factor for CTE
neuropathological and clinical endophenotypes. H1β1γ1 DGE analyses in
the dorsolateral frontal cortex implicate cis-acting genes and
inflammatory pathways. H1β1γ1 may help explain CTE risk beyond
repetitive head impact exposure.
Introduction
Chronic traumatic encephalopathy (CTE) is a neurodegenerative disease
associated with repetitive head impacts (RHIs).[92]^1 The diagnosis of
CTE is confirmed only by postmortem neuropathological examination and
is characterized by hyperphosphorylated tau (ptau) located around blood
vessels, usually at the depths of the sulci.[93]^2 CTE has been
previously identified in brain donors who had participated in contact
sports, such as professional football, boxing, and soccer, and in
veterans with RHI exposure.[94]^3^,[95]^4 However, not all individuals
who participate in contact sports may go on to develop CTE, and among
individuals with CTE and similar levels of RHI exposure, there is a
marked variation in disease severity,[96]^5^,[97]^6 suggesting that
factors beyond RHI may also play a role. This was recently demonstrated
for the apolipoprotein E ε4 (APOE ε4) allele, the genetic locus most
investigated for outcomes following traumatic brain injury and the
largest genetic risk factor for late-onset Alzheimer’s disease (AD).
Among brain donors with RHI exposure, APOE ε4 was significantly
associated with CTE stage and quantitative and semi-quantitative
measures of ptau pathology.[98]^7
Microtubule-associated protein tau (MAPT), a gene located at 17q21.31,
encodes the tau protein, which aggregates in neurons as neurofibrillary
tangles (NFTs) in multiple neurodegenerative diseases,[99]^8 including
CTE. Two highly divergent extended haplotypes, referred to as H1 and
H2, are defined by a large ancestral inversion in the 17q21.31 region
that includes MAPT. The H1 haplotype has been associated with primary
tauopathies including progressive supranuclear palsy (PSP)[100]^9 and
corticobasal degeneration,[101]^10 as well as other neurodegenerative
diseases, including Lewy body dementia,[102]^11 AD,[103]^12 and
Parkinson’s disease.[104]^13 The inverted H2 haplotype is rare in
African ancestry but seems to be positively selected in European
ancestry.[105]^14 In addition to the inversion, there are distinct
duplications, making this region particularly complex. Boettger et al.
identified nine segregating structural variants, defined by the H1/H2
inversion and three distinct duplications, α, β, and γ.[106]^15 They
also showed that these structural variants could be identified by
imputation from single-nucleotide polymorphisms (SNPs). Here, we
leverage array SNP data across the 17q21.31 region and the Boettger et
al. reference panel to impute these structural variants and test their
association with CTE and related, richly characterized endophenotypes
from the world’s largest collection of brain donors with RHI exposure
and CTE.
Results
Sample and structural haplotype descriptives
[107]Figure 1 shows a schematic overview of the methods. The primary
sample included 447 donors of European ancestry, with median age of 71
(interquartile range [IQR] = 61–80) years, all of whom had a history of
exposure to RHI from contact sports or military service. A flowchart
describing the included and excluded donors is presented in [108]Figure
S1. [109]Table 1 shows demographic, RHI-related, clinical, and
neuropathological characteristics of donors stratified by CTE stage. A
total of 305 had neuropathologically confirmed CTE (68.2%). Of these,
38 (12.5%) had stage I CTE, 41 (13.4%) had stage II CTE, 108 (35.4%)
had stage III CTE, and 118 (38.7%) had stage IV CTE. [110]Table S1
shows missing values for neuropathological outcomes (semi-quantitative
and quantitative tau measures) that were imputed. [111]Tables S2A and
S2B show the imputation quality scores corresponding to the H1/H2
inversion and copy-number variation (CNV) markers for batches 1 and 2.
[112]Table 2 shows each haplotype frequency stratified by CTE
diagnosis. Only 8 of 9 structural haplotypes were observed. H2α2γ1,
previously found to have a 0.01 frequency in European ancestry,[113]^15
was not observed.
Figure 1.
[114]Figure 1
[115]Open in a new tab
Methodological schematic
Abbreviations: CTE, chronic traumatic encephalopathy; DLFL,
dorsolateral frontal lobe; Hx, history; MRI, magnetic resonance
imaging; PC, principal components of population substructure; PET,
positron emission tomography; SNP, single nucleotide polymorphism.
Table 1.
Sample descriptives, stratified by CTE status and stage
Characteristic All (n = 447) No CTE (n = 142) All CTE (n = 305) CTE
stage I (n = 38) CTE stage II (n = 41) CTE stage III (n = 108) CTE
stage IV (n = 118)
Demographic and clinical
__________________________________________________________________
Male sex 438 (98) 134 (94) 304 (99) 38 (100) 41 (100) 108 (100) 117
(99)
Median age at death (IQR) 71 66 74 57.5 66 75.5 77
IQR 61–80 54–76 64–81 40–71 49–74 64–83 71–83
Range 20–99 21–99 20–98 20–89 25–89 32–97 46–98
__________________________________________________________________
Cause of death
__________________________________________________________________
Neurodegenerative 206 (46) 58 (41) 148 (49) 8 (21) 9 (22) 42 (39) 89
(75)
Cardiovascular disease 68 (15) 21 (15) 47 (15) 9 (24) 9 (22) 21 (19) 8
(7)
Suicide 30 (7) 17 (12) 13 (4) 6 (16) 4 (10) 3 (3) 0
Cancer 33 (7) 9 (6) 24 (8) 2 (5) 4 (10) 13 (12) 5 (4)
Motor neuron disease 16 (4) 4 (3) 12 (4) 1 (3) 1 (2) 8 (7) 2 (2)
Accidental overdose 10 (2) 5 (4) 5 (2) 0 3 (7) 2 (2) 0
Injury 8 (2) 2 (1) 6 (2) 3 (8) 0 0 3 (3)
Other 60 (13) 19 (13) 41 (13) 8 (21) 8 (20) 16 (15) 9 (8)
Unknown 16 (4) 7 (5) 9 (3) 1 (3) 3 (7) 3 (3) 2 (2)
Dementia diagnosed by consensus 286 (64) 78 (55) 208 (68) 13 (34) 14
(34) 73 (68) 108 (92)
Dementia diagnosed in life 262 (59) 76 (55) 186 (61) 8 (21) 10 (24) 61
(57) 107 (91)
__________________________________________________________________
RHI-related
__________________________________________________________________
Contact sport play history 424 (95) 123 (87) 301 (99) 36 (95) 41 (100)
106 (98) 118 (100)
Football 394 (88) 110 (77) 284 (93) 34 (89) 37 (90) 101 (93) 112 (95)
Professional 163 (36) 12 (9) 151 (50) 8 (18) 15 (37) 60 (56) 69 (59)
College/semi-professional 168 (38) 54 (38) 114 (37) 16 (42) 19 (46) 39
(36) 40 (34)
High school/youth 63 (14) 44 (31) 19 (6) 11 (29) 3 (7) 2 (2) 3 (3)
Mean duration of football play in years (SD) 12.1 (5.6) 8.5 (4.6) 13.5
(5.3) 9.1 (3.8) 11.6 (4.1) 13.6 (4.4) 15.2 (5.9)
Ice hockey 24 (5) 8 (6) 16 (5) 4 (11) 3 (7) 7 (7) 2 (2)
Soccer 31 (7) 15 (11) 16 (5) 9 (24) 2 (5) 1 (1) 4 (3)
Amateur wrestling 18 (4) 8 (6) 10 (3) 2 (5) 3 (7) 4 (4) 1 (1)
Boxing 18 (4) 5 (4) 13 (4) 1 (3) 1 (2) 6 (6) 5 (4)
Rugby 8 (2) 2 (1) 6 (2) 2 (5) 3 (7) 1 (1) 0
Military veterans 155 (35) 51 (36) 104 (34) 14 (37) 12 (29) 35 (32) 43
(36)
Combat 34 (8) 18 (13) 16 (5) 4 (11) 3 (7) 6 (6) 3 (3)
__________________________________________________________________
Pathology
__________________________________________________________________
Mean log-transformed %AT8-positive cells per unit area (SD) −1.3 (1.1)
−1.8 (1.1) −1.1 (0.9) −2.2 (0.4) −1.8 (0.6) −1.1 (0.7) −0.4 (0.6)
AD pathology 124 (28) 42 (30) 82 (27) 3 (8) 2 (5) 22 (20) 55 (47)
Mean CERAD neuritic plaque score (SD) 0.8 (1.0) 0.8 (1.1) 0.8 (0.9) 0.2
(0.7) 0.3 (0.5) 0.7 (0.9) 1.2 (0.9)
Mean Braak NFT stage (SD) 2.9 (2.0) 2.3 (2.4) 3.2 (1.8) 1.34 (1.6) 1.7
(1.4) 3.4 (1.3) 4.2 (1.4)
Lewy body pathology 92 (21) 24 (17) 68 (22) 4 (11) 4 (10) 27 (25) 33
(28)
Brainstem predominant 36 (8) 8 (6) 28 (9) 0 0 11 (10) 17 (14)
Limbic/neocortical predominant 56 (13) 16 (11) 40 (13) 4 (11) 4 (10) 16
(15) 16 (14)
FTLD-tau 26 (6) 12 (9) 14 (5) 1 (3) 0 3 (3) 10 (9)
FTLD-TDP43 20 (5) 3 (2) 17 (6) 1 (3) 3 (7) 1 (1) 12 (10)
[116]Open in a new tab
Percentages are based on non-missing data.
Abbreviations: AD, Alzheimer’s disease; CERAD, Consortium to Establish
a Registry for Alzheimer’s Disease; CTE, chronic traumatic
encephalopathy; FTLD, frontotemporal lobar degeneration.
Table 2.
Haplotype frequency by CTE diagnosis
CTE diagnosis n H2 H1β1γ1 H1β1γ2 H1β1γ3 H1β1γ4 H1β2γ1 H1β3γ1 H2
[MATH: α :MATH]
1γ2 H2
[MATH: α :MATH]
2γ2
All 447 0.17 0.39 0.11 0.05 0.001 0.28 0.006 0.13 0.04
CTE 305 0.18 0.41 0.10 0.04 0 0.26 0.005 0.15 0.04
No CTE 142 0.14 0.34 0.12 0.05 0.004 0.34 0.007 0.10 0.05
[117]Open in a new tab
Abbreviation: CTE, chronic traumatic encephalopathy.
H1β1γ1 associations with CTE neuropathological and clinical endophenotypes in
the UNITE Brain Bank
No significant associations were identified between these eight
structural haplotypes and CTE diagnosis. However, there were
significant associations between H1β1γ1 and various CTE endophenotypes
that capture measures of CTE severity. The overall H1β1γ1 allele
frequency was 0.39 (0.41 for individuals with neuropathologically
confirmed CTE; 0.34 for those without CTE). Among the global measures,
history of dementia and quantitative tau burden in dorsolateral frontal
cortex (DLFC) were significantly associated with H1β1γ1after correction
for multiple testing ([118]Table 3). Each additional copy of H1β1γ1
corresponded to a 1.90-fold increase in odds of dementia (95%
confidence interval [CI], 1.27–2.86; p[adj] = 0.007) and a 0.45-unit
elevation in log-transformed AT8+ cell count per mm^2 (95% CI,
0.14–0.75; p[adj] = 0.015). H1β1γ1 dosage was non-significantly
associated with CTE stage. Each additional copy of H1β1γ1 corresponded
to a 1.35X odds of increasing 1 stage (95% CI, 1.04–1.77). Among the
regional measures, there were significant associations after correction
for multiple testing with semi-quantitative tau burden in the amygdala,
entorhinal cortex, inferior parietal cortex, DLFC, and superior
temporal cortex, with odds ratios (ORs) ranging from 1.50 to 1.66
([119]Figure 2; [120]Table S3).
Table 3.
Estimated associations of H1β1γ1 status with CTE diagnosis, CTE stage,
dementia diagnosis, and quantitative tau burden in the dorsolateral
frontal cortex
Outcome OR (95% CI) Unadjusted p value Permutation adjusted p value
CTE diagnosis[121]^a 1.31 (0.94–1.83) 0.11 0.30
CTE stage[122]^b 1.35 (1.04–1.77) 0.03 0.08
Dementia[123]^c 1.90 (1.27–2.86) 0.002 0.007
__________________________________________________________________
Outcome beta (95% CI) Unadjustedpvalue Permutation adjustedpvalue
__________________________________________________________________
Quantitative tau burden in dorsolateral frontal cortex[124]^d 0.45
(0.14–0.75) 0.005 0.02
[125]Open in a new tab
All analyses included European ancestry, used an additive model for
H1β1γ1, and were adjusted for age and 10 principal components of
population substructure. Abbreviations: CI, confidence interval; CTE,
chronic traumatic encephalopathy; OR, odds ratio.
^a
OR is the odds of having a CTE diagnosis for each additional H1β1γ1
copy.
^b
OR is the odds of increasing 1 stage (scale of 0–4) for each additional
H1β1γ1 copy.
^c
OR is the odds of having dementia for each additional H1β1γ1 copy.
^d
Beta value is the increase in log tau+ cells/mm^2 in the dorsolateral
frontal cortex for each additional H1β1γ1 copy.
Figure 2.
[126]Figure 2
[127]Open in a new tab
Brain heatmap of estimated associations of H1β1γ1 status with
semi-quantitative tau burden in brain regions commonly affected in
chronic traumatic encephalopathy
OR is the odds of increasing 1 level (scale of 0–3) for each additional
H1β1γ1 copy. The H1β1γ1 structural haplotype was found to be
significantly associated with semi-quantitative tau burden in the
amygdala (AMY) (OR = 1.52; p[adj]= 0.025), entorhinal cortex (EC) (OR =
1.50; p[adj]= 0.047), inferior parietal cortex (IP) (OR = 1.50; p[adj]=
0.039), dorsolateral frontal cortex (DLFC) (OR = 1.47; p[adj]= 0.045),
and superior temporal cortex (ST) (OR = 1.66; p[adj]= 0.002).
When sensitivity analyses for the same outcomes were conducted
exclusively among American football players, including duration of play
in years as an additional covariate, the magnitudes of association for
H1β1γ1 remained similar ([128]Tables 4 and [129]S4). For quantitative
tau burden in the DLFC, having a copy of the H1β1γ1 allele conferred a
similar risk of pathological tau accumulation to playing 9.6 years of
football. We did not observe a significant H1β1γ1 × duration of play
interaction across any outcome.
Table 4.
Identified pathways based on differentially expressed genes for H1β1γ1
Pathway Total genes in pathway Significantly differently expressed
genes in pathway Fold enrichment[130]^a FDR-adjusted p value
Up-regulated pathways
__________________________________________________________________
Reactome database
Cytokine signaling in immune system 798 659 1.07 3.81E−13
Extracellular matrix organization 321 263 1.06 6.43E−13
Interferon alpha/beta signaling 80 61 0.990 1.20E−12
Interleukin-4 and interleukin-13 signaling 111 89 1.04 1.18E−08
Signaling by interleukins 458 390 1.11 3.30E−08
PANTHER database
Apoptosis signaling pathway 117 114 1.26 2.14E−04
Integrin signaling pathway 192 179 1.21 4.70E−04
Angiogenesis 169 164 1.26 1.25E−02
p53 pathway 88 82 1.21 1.37E−02
TGF-β signaling pathway 100 91 1.18 2.20E−02
__________________________________________________________________
Downregulated pathways
__________________________________________________________________
Reactome database
Neuronal system 419 369 1.14 <1E−20
Transmission across chemical synapses 234 234 1.12 <1E−20
Neurotransmitter receptors and postsynaptic signal transmission 208
178 1.11 3.67E−12
Voltage-gated potassium channels 43 40 1.21 4.48E−10
Protein-protein interactions at synapses 91 86 1.23 3.47E−08
PANTHER database
Synaptic vesicle trafficking 151 29 0.147 9.83E−06
Ionotropic glutamate receptor pathway 357 47 0.171 4.34E−04
Metabotropic glutamate receptor group III pathway 523 65 0.161
1.65E−03
Heterotrimeric G protein signaling pathway-Gq alpha and Go
alpha-mediated pathway 781 114 0.189 4.44E−03
Metabotropic glutamate receptor group I pathway 179 22 0.159 7.30E−03
[131]Open in a new tab
The table shows the top 5 up- and downregulated pathways for each
database based on p value. Note that for downregulated pathways, the
number of differentially expressed genes listed for the Reactome
database represents the number of downregulated differentially
expressed genes, but for the PANTHER database, it represents the number
of up-regulated differentially expressed genes. Thus, the fold
enrichment is high for the Reactome database but low for the PANTHER
database.
Abbreviations: FDR, false discovery rate; PANTHER, protein analysis
through evolutionary relationships.
^a
Indicates how much more often genes from a pathway were differentially
expressed compared with what would be expected by chance.
[132]Tables S6 and [133]S7 show H1β1γ1 post hoc analyses limited to all
donors with CTE and limited to football-playing donors with CTE,
respectively. Compared with effect sizes for analyses of the full
sample, effect sizes for analyses limited to those with CTE were
varied. Effect sizes were generally smaller among all donors with CTE
but were similar among football players with CTE.
In H1β1γ1 post hoc analyses evaluating additional clinical outcomes
among brain donors, H1β1γ1 was non-significantly associated with
Functional Assessment Questionnaire (FAQ) score (β = 1.58, 95% CI,
0.10–3.05) and non-significantly associated with parkinsonism (OR =
1.40, 95% CI, 0.99–1.97) ([134]Table S8).
H1β1γ1 differential gene expression in the DLFC
In the differential gene expression (DGE) analysis for H1β1γ1, using
expression data from the DLFC from a subset of brain donors (n = 167;
see [135]Table S9 for descriptive statistics), 8 genes were
significantly differentially up-regulated with a log fold change (LFC)
> 0.4, and 7 genes were significantly differentially downregulated with
an LFC < −0.4. [136]Figure 3 shows a volcano plot of the results.
Participants harboring the H1β1γ1 allele exhibited a pronounced
up-regulation of genes linked to immunity, such as CXCR1,[137]^16
CHI3L1,[138]^17 and SERPINA3.[139]^18 For the up-regulated genes, the
H1β1γ1 allele had a trans-acting effect, with top hits distal to
17q21.31. Reactome analysis revealed networks implicating the immune
system, including cytokine, interferon alpha/beta, and interleukin
signaling. Protein analysis through evolutionary relationships
(PANTHER) protein analysis implicated the immune system (integrin and
transforming growth factor β [TGF-β] signaling), as well as other
pathways linked to neurodegeneration, including cell death (apoptosis,
p53) and blood-brain barrier integrity (angiogenesis, integrin
signaling) ([140]Table 4). For the downregulated genes, the H1β1γ1
allele had a cis-acting effect on ARL17B and LRRC37A2, located in the
17q21.31 region. ARL17B is part of the overlapping component of the β
and γ duplication, whereas LRRC37A2 is part of the second γ
duplication.[141]^19 MAPT was not significantly differentially
expressed. H1β1γ1 allele also had a trans-acting effect to downregulate
genes linked to synaptic function, such as CRH[142]^20 and VGF.[143]^21
Reactome and PANTHER analyses consistently implicated downregulation of
neuronal, synaptic, and neurotransmitter-related pathways ([144]Table
4).
Figure 3.
[145]Figure 3
[146]Open in a new tab
Volcano plot of differentially expressed genes for H1β1γ1
Abbreviations: FDR, false discovery rate; LogFC, log fold change.
^aAnalyses included 167 donors of European ancestry, used an additive
model for H1β1γ1, and were adjusted for age at death and RNA integrity
number (RIN).
H1β1γ1 associations with CTE clinical endophenotypes in the DIAGNOSE CTE
study
To replicate/expand on findings from Understanding Neurological Injury
and Traumatic Encephalopathy (UNITE), we carried out additional
analyses in living former professional and college American football
players from the Diagnostics, Imaging, and Genetics Network for the
Objective Study and Evaluation of Chronic Traumatic Encephalopathy
(DIAGNOSE CTE) study.[147]^22 Of the 180 former football players, 115
had available genotype data and were of European ancestry. We examined
10 outcomes selected based on associations identified in the
neuropathology sample and their known clinical correlates. H1β1γ1 was
non-significantly associated with worse cognitive performance on the
executive/speed factor (β = −0.316; 95% CI, −0.588 to −0.044) and with
greater flortaucipir positron emission tomography (PET) standardized
uptake value ratio (SUVR) in the amygdala (β = 0.030; 95% CI, 0–0.60)
([148]Table S10).
Discussion
Among 447 brain donors of European ancestry with RHI exposure from
contact sports or military service and deeply characterized CTE-related
endophenotypes, we imputed structural haplotypes across the 17q21.31
region. After correction for multiple testing, the H1β1γ1 haplotype was
significantly associated with a quantitative measure of ptau burden in
the DLFC and dementia and semi-quantitative measures of ptau pathology
across the cerebral cortex, amygdala, and entorhinal cortex.
Association sizes were largest for the superior temporal cortex (1.66
odds of increasing 1 level with each additional allele, p[adj] = 0.002)
and for dementia (1.9 odds of having dementia with each additional
allele, p[adj] = 0.007). Although the association with CTE stage did
not survive permutation correction for multiple testing, the
association demonstrated a trend (1.35 odds of increasing 1 stage with
each additional allele, p[adj] = 0.08). We did not observe a
significant association with CTE diagnosis. DGE analysis in the DLFC
for H1β1γ1 haplotype demonstrated up-regulated genes and pathways
related to neuroinflammation and downregulated cis (ARL17B and
LRRC37A2) and trans genes and pathways related to neuronal, synaptic,
and neurotransmitter function. Among a living sample of former college
and professional American football players, the H1β1γ1 haplotype was
non-significantly associated with worse cognitive performance on the
executive/speed factor and with greater flortaucipir PET SUVR in the
amygdala.
Human 17q21.31 is a structurally complex region containing a
megabase-long inversion polymorphism and several CNVs.[149]^14 The
inversion polymorphism exists either as direct (H1) or inverted (H2)
haplotypes.[150]^13^,[151]^23 Within this region is the MAPT gene that
codes for the tau protein that accumulates in CTE and other
tauopathies. Although clear associations have been identified between
the H1 haplotype and primary tauopathies, such as PSP, we did not
observe associations between H1 and CTE. Previous
studies[152]^9^,[153]^12^,[154]^24^,[155]^25 aiming to identify the
association between subhaplotypes on the H1 and H2 backbones and
neurodegenerative diseases have produced mixed results, possibly due to
the use of just a handful of SNPs to tag exceedingly complex variation
defined by multiple CNVs. Recently, investigators have taken a more
sophisticated approach to this region. Wang et al. leveraged
whole-genome sequencing data to show that the number of γ duplications
was associated with an increased risk of PSP.[156]^26 A recent review
article proposed an alternative naming convention in the region to
integrate existing CNV and SNP-based nomenclature, with H1β1γ1 being
renamed as H1.1_b-z.[157]^27 In the current study, we leveraged array
SNP data across the 17q21.31 region and the Boettger et al. reference
panel to impute and examine structural haplotypes defined by the H1/H2
inversion and CNVs. Similar to using whole-genome sequence data, our
imputation strategy is a more robust approach than using a small number
of SNPs to understand the genetic architecture of this complicated
region. As we used the Boettger et al. reference panel, we maintained
their nomenclature. In the future, this imputation approach that allows
investigation in large numbers of samples without the need for
long-read sequencing could be applied quickly to other tauopathy
datasets, including AD datasets.
H1β1γ1 was associated with semi-quantitative tau burden in the
amygdala, entorhinal cortex, inferior parietal cortex, DLFC, and super
temporal cortex. The cortical regions and the entorhinal cortex are
affected early in CTE (stages I and II), while the amygdala is affected
later in the disease course (stages III and IV).[158]^28 H1β1γ1 status
also was associated with dementia, functional impairment on the FAQ,
and parkinsonism, all of which typically occur late in the disease
course and have been clearly linked to CTE tau
pathology.[159]^29^,[160]^30^,[161]^31 Our findings suggest that H1β1γ1
likely plays a role both early and late in the disease course of CTE.
Dementia may be predominantly driven by burden of tau pathology in the
cortex, as has been shown in AD.[162]^32 Notably, we did not find
H1β1γ1 associations with hippocampal tau pathology, which is important
for CTE staging by the McKee criteria. Hippocampal tau pathology is a
component of stage III and IV disease. This may explain why we only
found a non-significant association with CTE stage, which is defined by
the location of ptau pathology and, to a lesser extent, the burden of
ptau pathology. Among late-stage CTE, we recently described CTE
subtypes defined by whether there is “cortical sparing” of tau
pathology relative to the hippocampus. Given the disproportionate
effect of H1β1γ1 on cortical tau compared with hippocampal tau, H1β1γ1
may contribute to this subtyping, which is an area of future
work.[163]^33
Among football players, comparison of the relationships between the
various outcomes of interest, H1β1γ1 status, and duration of football
play may provide additional insight into CTE initiation and
progression. As we have shown previously, the duration of play was
strongly associated with CTE diagnosis and with CTE stage, which is
defined by the location of ptau pathology and, to a lesser extent, the
burden of ptau pathology.[164]^1^,[165]^30 Conversely, H1β1γ1 was not
associated with CTE diagnosis and was only non-significantly associated
with CTE stage, but had a substantial association with dementia and
quantitative tau burden in the DLFC, with each copy of H1β1γ1
conferring a similar risk of pathological tau burden to more than 9
years of football play. Taken together, these findings suggest that the
duration of play may drive disease initiation and have a role in
disease progression, whereas H1β1γ1 may be particularly important for
disease progression and severity. This finding is similar to what we
have previously observed for APOE e4 and TMEM106B, which were not
associated with CTE diagnosis but were associated with several measures
reflecting CTE severity.[166]^7^,[167]^34 In the current manuscript,
H1β1γ1 had similar effects on ptau-related outcomes in analyses limited
to those with CTE as opposed to all donors with RHI exposure,
suggesting that its effect may not be CTE specific. The combination of
these observations provides a growing understanding of the landscape
for neurodegenerative disease risk after RHI exposure and may offer a
future roadmap for a personalized medicine approach.
In DGE analyses, H1β1γ1 dosage corresponded to decreased expression of
LRRC37A2 and ARL17B, both of which are located in the 17q21.31 region.
More specifically, they are members of highly conserved gene families
located on the 5′ end of the γ duplication where there is also overlap
with the β duplication. Copy number increases the expression of
full-length protein isoforms from these gene families.[168]^27 Genetic
variants in both genes have been associated with circulating plasma
total tau levels.[169]^35 The decreased expression in cis genes for
H1β1γ1, yet increased tau pathology and associated dementia diagnosis,
may be related to an aberrant feedback loop, in which H1β1γ1 dosage is
related to increased cortical tau, which may then downregulate genes
located within the 17q21.31 region.[170]^19
In DGE analyses, H1β1γ1 dosage corresponded to increased expression of
immunity-related genes and pathways. Up-regulated genes included
CXCR1,[171]^16 CHI3L1,[172]^17 and SERPINA3,[173]^18 which are related
to the innate immune system, and Reactome and PANTHER analyses
identified multiple up-regulated immunity-related pathways including
cytokine, interferon alpha/beta, interleukin, integrin, and TGF-β
signaling. Mild brain injury may trigger multifocal traumatic axonal
injury,[174]^36 initiating neuroinflammation,[175]^37 which then
initiates an immune response with microglial activation to repair or
limit damage.[176]^38 Although this neuroinflammatory response may
dissipate between injuries, repetitive injuries that occur within a
short interval of time may prevent complete recovery and result in a
persistent proinflammatory state. Ongoing inflammation may then drive
the development and spread of CTE ptau pathology, which further induces
inflammation via a bidirectional pathway.[177]^39 Given H1β1γ1’s
association with immune pathways, it may further amplify the
CTE-related inflammatory response, driving disease progression.
To extend the findings from the neuropathological analyses, we
conducted additional analyses of H1β1γ1 in the living DIAGNOSE
study.[178]^22 Participants in the study were thoroughly characterized
former college and professional American football players. H1β1γ1 was
non-significantly associated with worse cognitive performance on the
executive/speed factor. Executive dysfunction is a core component of
the 2021 Traumatic Encephalopathy Syndrome Criteria, proposed to
diagnose CTE in life. Executive function tends to localize to the
frontal cortex, a region that is affected early and severely in CTE and
which was implicated in the H1β1γ1 neuropathological analyses. Further,
we recently showed that frontal cortex ptau pathology was the primary
driver of cognitive impairment, including executive dysfunction, in
CTE.[179]^31 H1β1γ1 was also non-significantly associated with greater
flortaucipir PET SUVR in the amygdala. The amygdala was implicated in
the H1β1γ1 neuropathological analyses, is greatly affected in
high-stage CTE, and has been implicated in cognitive performance in
CTE. More robust associations between H1β1γ1 and clinical outcomes in
the DIAGNOSE study may not have been observed due to the small sample
size (only 77 participants had at least one H1β1γ1 allele), the fact
that the phenotypes studied were different than the outcomes in the
neuropathological analyses, and the fact that DIAGNOSE data were
cross-sectional, and therefore may not have accurately represented
outcomes experienced near the end of life, as with the pathology data.
Future studies of the H1β1γ1 haplotype in other living samples with RHI
exposure would be valuable.
Strengths of this study include imputation of the structural haplotypes
of 17q21.31 and identification of an association with CTE
endophenotypes. Brain donors for this study were carefully
characterized, further increasing power to detect genetic associations.
Additionally, all donors, both individuals with and without CTE, had
RHI exposure, putting them at risk for CTE.
This study investigated how structural variants of 17q21.31 may be
associated with CTE-related pathological and clinical outcomes. The
H1β1γ1 haplotype may help explain CTE heterogeneity among those with
similar RHI exposure. Understanding the genetic underpinnings of CTE
pathology may provide insights into disease mechanism and offers a
precision medicine approach to harm reduction, including guiding
decisions regarding contact sport play and providing a target of
therapies.
Limitations of the study
The study also has limitations. Although we made use of the largest
available sample of brain donors with CTE, the sample was still small
by genetic standards. Sample size was also a concern for the living
DIAGNOSE sample. The study only included individuals of European
ancestry because this was the only ancestry with a large enough sample
in the UNITE Brain Bank to make reasonable inferences. The genetic
architecture of the 17q21.31 region differs markedly by
ancestry,[180]^27 limiting a combined analysis across ancestries. The
small number of females in the UNITE Brain Bank was included in
analyses, but clarity regarding whether findings generalize to females
will require larger sample sizes. Efforts are currently underway to
recruit and obtain genome-wide genotyping on additional donors.
Additionally, brain donors were not followed during life, and clinical
and RHI exposure came largely from retrospective informant report,
making recall bias a possibility. CTE pathology and dementia may be
independently associated with brain bank selection, introducing
potential selection bias. Larger sample sizes and additional
information about the general population of individuals exposed to RHI
will be valuable for future analyses to account for selection bias.
Resource availability
Lead contact
Any additional information required to reanalyze the data reported in
this paper is available from the lead contact upon request, Dr. Jesse
Mez (jessemez@bu.edu).
Materials availability
This study did not generate new unique reagents.
Data and code availability
* •
Raw genetic data, including SNPs in the 17q21.31 region, are
available at
[181]https://osf.io/qj5rf/?view_only=9b56c078159d4ae69c5b2e7409196e
f8. The 17q21.31 reference panel and instructions on its use can be
found at [182]https://github.com/freeseek/impute17q21. Deidentified
UNITE and DIAGNOSE CTE data are available in the Federal
Interagency Traumatic Brain Injury Research (FITBIR) repository,
[183]https://fitbir.nih.gov. Data are also available through a data
sharing portal for UNITE,
[184]https://www.bu.edu/alzresearch/information-for-investigators/
and the DIAGNOSE CTE Research Project, [185]http://diagnosecte.com.
Please note that participants and donors may have identifiable
information related to their elite contact sport play (e.g., sport,
position, years of play, and age), and these data may be removed to
maintain confidentiality.
* •
All original code has been deposited at
[186]https://osf.io/qj5rf/?view_only=9b56c078159d4ae69c5b2e7409196e
f8 and is publicly available as of the date of publication. Full
results are available at
[187]https://osf.io/qj5rf/?view_only=9b56c078159d4ae69c5b2e7409196e
f8.
* •
Any additional information required to reanalyze the data reported
in this paper is available from the [188]lead contact upon request.
Acknowledgments