Putative risk alleles for LATE‐NC with hippocampal sclerosis in population‐representative autopsy cohorts

Abstract Limbic‐predominant age‐related TAR‐DNA‐binding protein‐43 (TDP‐43) encephalopathy with hippocampal sclerosis pathology (LATE‐NC + HS) is a neurodegenerative disorder characterized by severe hippocampal CA1 neuron loss and TDP‐43‐pathology, leading to cognitive dysfunction and dementia. Polymorphisms in GRN, TMEM106B and ABCC9 are proposed as LATE‐NC + HS risk factors in brain bank collections. To replicate these results in independent population‐representative cohorts, hippocampal sections from brains donated to three such studies (Cambridge City over 75‐Cohort [CC75C], Cognitive Function and Ageing Study [CFAS], and Vantaa 85+ Study) were stained with hematoxylin–eosin (n = 744) and anti‐pTDP‐43 (n = 713), and evaluated for LATE‐NC + HS and TDP‐43 pathology. Single nucleotide polymorphism genotypes in GRN rs5848, TMEM106B rs1990622 and ABCC9 rs704178 were determined. LATE‐NC + HS (n = 58) was significantly associated with the GRN rs5848 genotype (χ2(2) = 20.61, P < 0.001) and T‐allele (χ2(1) = 21.04, P < 0.001), and TMEM106B rs1990622 genotype (Fisher's exact test, P < 0.001) and A‐allele (χ2(1) = 25.75, P < 0.001). No differences in ABCC9 rs704178 genotype or allele frequency were found between LATE‐NC + HS and non‐LATE‐NC + HS neuropathology cases. Dentate gyrus TDP‐43 pathology associated with GRN and TMEM106B variations, but the association with TMEM106B nullified when LATE‐NC + HS cases were excluded. Our results indicate that GRN and TMEM106B are associated with severe loss of CA1 neurons in the aging brain, while ABCC9 was not confirmed as a genetic risk factor for LATE‐NC + HS. The association between TMEM106B and LATE‐NC + HS may be independent of dentate TDP‐43 pathology.


INTRODUCTION
Hippocampal sclerosis in old age is a dementing disorder characterized by severe loss of pyramidal neurons and gliosis in the hippocampal CA1 area. In addition to the anatomically defined neuron loss, hippocampal sclerosis is frequently associated with aggregations of transactive response DNA binding protein 43kDa (TDP-43) (15). Recently, neuropathological changes of limbic-predominant age-related TDP-43 encephalopathy (LATE) were recognized as disease entity (13). Hippocampal sclerosis in old age with TDP-43 pathology is thus a subgroup of LATE-NC (LATE-NC + HS).
In three European population based studies of older people with brain donation, two from the UK and one from Finland, 79-100% LATE-NC + HS cases were found to have TDP-43 pathology in the hippocampal dentate gyrus (6,10). TDP-43 pathology is also linked to hippocampal neuron loss of lesser extent than LATE-NC + HS, indicating that TDP-43 pathology without severe CA1 neuron loss may represent a precursor state for LATE-NC + HS (6). The exact etiological pathophysiology of LATE-NC + HS is still unclear, but polymorphisms in the progranulin-encoding gene GRN (rs5848) (3), transmembrane protein 106B-encoding gene TMEM106B (rs1990622) (16), and sulfonylurea receptor 2-encoding ABCC9 (rs704178, in near-perfect linkage disequilibrium with rs704180) (14) have emerged as potential risk factors within US brain bank-based genome-wide association studies (GWAS).
GRN rs5848 TT genotype is associated with lower progranulin levels and has been linked with frontotemporal lobar degeneration with TDP-43 positive inclusions (FTLD-TDP) without GRN mutation (19). Progranulin has direct neurotrophic and inflammatory response-modulating functions, and is found to play a role in TDP-43 processing (22). TMEM106B is reported to modulate progranulin levels (4). ABCC9 has not been associated with FTLD-TDP, TDP-43 accumulation or other neurodegenerative diseases than LATE-NC + HS, and possible mechanisms of ABCC9 variation affecting CA1 neuron loss are unclear (17).
No population-representative analyses on GRN rs5848, TMEM106B rs1990622 and ABCC9 rs704178 variants in LATE-NC + HS have been published. Here, we investigate these proposed risk variations in three population-representative cohorts of older people, and evaluate their association with LATE-NC + HS, which could indicate the underlying pathophysiological mechanisms for CA1 neuron loss in the aging brain. We also assess if LATE-NC without hippocampal sclerosis is associated with the proposed genetic variations.

Study design
This study is set within EClipSE (Epidemiological Clinicopathological Studies in Europe), which is a collaboration of the three European population-representative longitudinal cohort studies with brain donation programs: the Cognitive Function and Ageing Studies (CFAS) (multicenter, UK), the Cambridge City over-75s Cohort (CC75C, Cambridge, UK), and the Vantaa 85+ study (Vantaa, Finland). Cohort profiles, recruiting procedures and the approach for brain donation have been described previously in detail (2,5,18). The CC75C had 2160 participants aged ≥75 years at baseline in 1985, CFAS recruited 18226 people aged ≥65 years in 1989-1993, and the Vantaa 85+ study comprised 601 people aged ≥85 years in 1991. In CC75C and CFAS, a stratified random subsample of the cohort was approached for brain donation, resulting in 11% donation in CC75C (n = 241 at the time of the study) and 3% donation in CFAS (n = 562) with known representation of the source population. In the Vantaa 85+ study, a consented autopsy was performed on 51% (n = 304) of the cohort. Of these 1107 EClipSE donations, 131 were excluded due to lack of routine neuropathological evaluation (i.e. donations were too recent) or missing hematoxylin-eosin-stained hippocampal slides with used-up corresponding tissue blocks, and further 232 donations were excluded due to missing or unreliable deoxyribonucleic acid (DNA) samples ( Figure 1). Thus, 744 EClipSE samples (CC75C n = 216, CFAS n = 244, Vantaa 85+ n=284) are included in the current study ( Figure 1). All data collection, including brain donation, followed local ethical guidelines and approval.

Neuropathological protocol
In CFAS and CC75C, brains were removed as soon as possible after death and bisected. One hemisphere was then dissected coronally and frozen at − 80°C. The contralateral hemisphere was formalin-fixed for at least six weeks, cut in coronal plane, and approximately 1-centimeter (cm)-thick samples were paraffin-embedded. In Vantaa 85+, both hemispheres were formalin-fixed, and dissected coronally. The right hemisphere was sampled for paraffin-embedding, while samples from the left hemisphere were conserved in polyethylene glycol (PEG). PEG-conserved hippocampal tissue was reprocessed into paraffin in 2013.

Hippocampal sclerosis
Six-to nine-micrometer (µm)-thick hippocampal sections were stained with hematoxylin-eosin. To standardize the diagnosis of hippocampal sclerosis across all three EClipSE cohorts, available sections were systematically assessed for severe neuron loss by an inter-rater-controlled method (6). Hippocampal sclerosis was diagnosed as present or absent based on severe neuron loss (<5 neurons/ field of view [fov] at 200x magnification) in over half of CA1 fov, gliosis, neuronal loss not being explained by an infarct, and no obvious neuron loss in other hippocampal areas. Hippocampal sclerosis was present in 10/216 CC75C and in CFAS 14/224 samples ( Figure 1). In the Vantaa 85+ study, sections were scored hippocampal sclerosis positive when at least one hemisphere met the hippocampal sclerosis criteria (n = 37) ( Figure 1). It is of note that due to the standardization method used in this study, the number of hippocampal sclerosis cases in the Vantaa 85+ cohort identified is lower than that in the previously published study which used different diagnostic criteria (10).
Anti-pTDP-43 immunoreactive solid neuronal cytoplasmic inclusions (NCI) were evaluated blind to clinical or other pathological information as present or absent in the hippocampal dentate cell layer. In all three cohorts, TDP-43 assessment was inter-rater evaluated.
LATE-NC presents with TDP-43 pathology in the amygdala at stage 1, and in the hippocampus at stage 2. Figure 2 illustrates typical hippocampal LATE-NC TDP-43 pathology without hippocampal sclerosis (Figure 2A-C) and with hippocampal sclerosis ( Figure 2D-F). Dentate NCI pathology is highlighted in Figure 2C and F, respectively. In this study, dentate cell layer NCI TDP-43 pathology was considered as LATE-NC, because TDP-43-stained amygdala sections were not available across the EClipSE cohort. Moreover, the dentate cell layer has previously been staged as the first hippocampal region affected by TDP-43 pathology after the subiculum (7). In cases with hippocampal sclerosis, CA1 and subiculum neuron loss may also be so complete that no NCI are visible in these regions. Figure S1 shows a LATE-NC case with dentate TDP-43 pathology but minimal CA1 pathology without hippocampal sclerosis ( Figure S1 A-C) and with hippocampal sclerosis ( Figure S1D-F).

SNP analysis
Single nucleotide polymorphism (SNP) analysis for CC75C/ CFAS was done at the Karolinska Institute, Sweden. A TaqMan SNP 7500 genotyping assay on real time PCR (Applied Biosystems, CA, USA) was used to determine the target SNPs. The target polymorphisms were GRN rs5848, TMEM106B rs1990622 and ABCC9 rs704178. Predesigned TaqMan SNP Genotype Assays were available from ThermoFischer Scientific. For 19 CC75C subjects, DNA was extracted from paraffin-embedded blocks, as no frozen tissue was available.
In the Vantaa 85+ cohort, genotypes of GRN rs5848 and TMEM106B rs1990622 were determined from wholegenome sequencing (WGS) data in 150 subjects (11). Sanger sequencing was performed in the remainder samples. Variant containing sequences were first amplified using primers listed in Table 1. Fragments were run on an ABI3730xl DNA Analyzer at the Institute of Molecular Medicine Finland and the sequencing data were analyzed using Sequencher 4.0 analysis software (Applied Biosystems). Genotypes for ABCC9 rs704180 were determined from the WGS data in 150 subjects and the remainder samples were Sanger sequenced for this SNP (data not shown). Genotype for ABCC9 rs704178 was imputed with Beagle 4.1 (version 27Jan18.7e1) using the population-specific SISu v3 imputation reference panel (https ://doi.org/10.17504/ proto cols.io.nmndc5e). The quality of imputed rs704178 genotypes was high and passed postimputation quality control (INFO score = 1, similar MAF in our imputed dataset and in the Finnish GnomAD population (0.44 vs. 0.41) and 98.6 % concordance between imputed and whole-genome sequencing-derived genotypes of internal control samples). The number of subjects with reliable results for each target SNP is reported in Table 2.
For the EClipSE cohort, G/C SNP rs704178 of gene ABCC9 were determined, because the genotype data were not available for recommended reference A/G SNP rs704180 of all three studies. However, these two SNPs are in near perfect linkage disequilibrium.
Hardy-Weinberg equilibrium (HWE) expected frequencies were calculated and tested using Chi2 goodness of fit (χ 2 (degree of freedom [df])). Chi2 (χ 2 (df)) (or Fisher's exact test [FET] when appropriate) was used to test the association of genotype or allele frequency and LATE-NC + HS or TDP-43 pathology. Each SNP was analyzed using an additive mode of inheritance (number of risk alleles, ie, 0, 1, or 2). ABCC9 rs704178 was furthermore analyzed using a recessive mode of inheritance (2 risk alleles = 1; 0 otherwise), as this has previously been shown to be appropriate (8,14,16). Effect size for Cramér's phi is reported. Logistic regression (Odds ratio [OR], 95% CI) was used to test genotype associations with LATE-NC + HS or TDP-43 pathology when taking the effect of age at death and sex into consideration. α was set at 0.05. Data were analyzed using STATA14 software (Stata Corporation 2015, Texas, USA).

Hippocampal sclerosis and TDP-43
All hippocampal sclerosis cases in CC75C and CFAS with available TDP-43 staining (n = 22) were positive for hippocampal dentate NCI, while in the Vantaa 85+ cohort 26 of the 34 hippocampal sclerosis cases with TDP-43-immunostained sections presented with dentate NCI. However, in the Vantaa 85+ cohort, hippocampal sclerosis was evaluated bilaterally, whereas TDP-43 was evaluated only on the right hemisphere. All Vantaa 85+ cases with hippocampal sclerosis pathology on the right hemisphere were positive for dentate TDP-43 NCI on the ipsilateral side. The eight Vantaa 85+ hippocampal sclerosis cases which were negative for dentate TDP-43 NCI presented with unilateral hippocampal sclerosis on the left hemisphere. Therefore, these cases cannot be considered as dentate TDP-43 NCI negative, and all hippocampal sclerosis cases in the EClipSE cohort are referred to as LATE-NC + HS in this study.

ABCC9 and LATE-NC + HS
LATE-NC + HS did not differ from the remainder population in ABCC9 rs704178 genotype or allele frequency, irrespective of the analyzed mode of inheritance (additive or recessive) (

LATE-NC + HS risk genes and TDP-43 pathology
Presence of dentate TDP-43 NCI was significantly associated with the GRN rs5848 genotypes T/C and T/T (χ 2 (2) = 18.59, P < 0.001, phi = 0.164), also when controlling for sex and age at death in logistic regression (OR: As all LATE-NC + HS cases were of TDP-43 positive or unknown status ipsilateral to the hippocampal sclerosis pathology, analyses on the association of dentate TDP-43 NCI and LATE-NC + HS risk genes were not conducted stratified by LATE-NC + HS. The association between dentate TDP-43 NCI and the proposed LATE-NC + HS risk genes was however evaluated while excluding LATE-NC + HS cases. Dentate TDP-43 NCI remained significantly associated with GRN rs5848 genotypes T/C and T/T (χ 2 (2) = 7.99, P = 0.018, phi = 0.112) and the GRN rs5848 T-allele (χ 2 (1) = 7.6, P = 0.006, phi = 0.077) (Table S1), but the association between TMEM106B rs1990622 genotype or allele and dentate TDP-43 NCI became insignificant when LATE-NC + HS cases were not included in the analyses (Table S1).

DISCUSSION
Previous research, based on US brain bank collections drawing on volunteer and clinical dementia cohorts, has indicated that polymorphisms in GRN, TMEM106B and ABCC9 genes are associated with hippocampal sclerosis in old age/ LATE-NC + HS, but to date no population-representative data have been published. Here, we studied three European population-representative cohorts and confirmed an association of LATE-NC + HS with polymorphisms in GRN and TMEM106B. The published association of LATE-NC + HS with ABCC9 polymorphism could not be reproduced. These detected associations of GRN and TMEM106B with LATE-NC + HS may be related to progranulin expression, as literature suggests both GRN and TMEM106B polymorphisms to be associated with variation in progranulin levels.
Limitations of this study include the overall cohort's sample size, different TDP-43 antibodies and section thickness between the UK and Finnish cohorts, unilateral instead of bilateral assessment of hippocampal sclerosis in the UK studies and of TDP-43 in all three cohorts, as well as lack of proof of causality. It is likely that we did not capture all LATE-NC cases in these cohorts as TDP-43 presence was only evaluated unilaterally in the hippocampal dentate. Further, not all hippocampal sclerosis cases could be verified as LATE-NC as TDP-43 staining was not available for each. Even though the EClipSE collaboration is the largest population-representative epidemiological cohort with brain donation, numbers of LATE-NC + HS cases may still be low for a genetic study. The lack of confirmation of the ABCC9 locus must therefore be interpreted with caution. However, as our population-representative study results give very strong independent support to the previously reported associations between LATE-NC + HS and the polymorphic sites in GRN and TMEM106B, it seems likely that functions of these two genes indeed do affect survival of the CA1 neurons, at least in older individuals. On the other hand, to confirm the hypothesis that variations in GRN and TMEM106B contribute to LATE-NC + HS neuron loss by altered progranulin levels, further studies would need to assess the expression of progranulin in association with LATE-NC + HS.
Important strengths of this study are the high mean age and population-representative origin of the cohorts, introducing minimal selection bias. Studies that previously reported polymorphisms in GRN, TMEM106B and ABCC9 as risk factors for LATE-NC + HS, obtained cases from brain banks and Alzheimer's Disease Centers (1,8,14,16), which tend to oversample people with dementia and frequently lack sufficient numbers of the oldest old to test hypotheses in this age group. Results from these studies may therefore not be transferable to the general population. Moreover, one of the pitfalls in GWAS studies is multiple testing: many candidate polymorphisms are assessed, so associations may be significant by chance, even though their clinical meaning is not clear. Independent replication of positive associations is crucial. Results from GWAS which identified proposed LATE-NC + HS risk genes (1,14) have been replicated up to now only using similar sources (16). This study's results are the first to verify or challenge previous results in an independent cohort, including samples from three population-representative cohorts from two countries.
Progranulin (encoded by gene GRN at chromosome 17) is universally expressed in the central nervous system (CNS) during early neuronal development, and is also expressed at modest levels in neurons and microglia in adults (21). Progranulin is proteolytically broken down into granulin peptides by extracellular proteases, mainly elastase, produced likely by astrocytes and microglia (3). Progranulin expression is markedly increased by neuroinflammation in association with neurodegenerative disease, where it appears to play a modulatory role in tissue damage within the CNS to suppress excessive immunity-based microglial activation and protecting neurons from reactive oxygen species and proinflammatory cytokines (21). Most currently known GRN mutations causing FTLD-TDP have been reported to create null alleles via haploinsufficiency mechanism, resulting in reduced protein levels (17). Progranulin may act as a potential neurotrophic factor whose loss of function may cause neurodegeneration in FTLD-TDP or other related diseases, and lead to the development of TDP-43positive pathology in LATE-NC + HS (3). Progranulin expression may have effects on cleavage and distribution of TDP-43 (8). Notably, the polymorphic rs5848 site in the 3'-untranslated region of GRN, which is associated with variations in progranulin levels and increased risk of LATE-NC + HS, is also part of a binding site for microRNA miR-659 (19). miR-659 may confer an increased risk for FTLD-TDP and LATE-NC + HS through an inhibition of progranulin translation, generating an effect resembling the biochemical and pathological findings observed in GRN-null mutations (19,21).
Gene TMEM106B (at chromosome 7p21) encodes a transmembrane protein, with a preferential expression in the frontal lobe (4). Genetic variation in TMEM106B may specifically modify the development of FTLD in the presence of a GRN mutation (12). The TMEM106B rs1990622 variant, where the C allele regulates GRN expression (4), is reported to be protective for LATE-NC + HS within a cohort of Alzheimer's disease cases (20). Results from this study confirmed GRN rs5848 and TMEM106B rs1990622 to be associated with LATE-NC + HS. Reduced progranulin levels could increase the vulnerability of specific CA1 neuron populations, through higher susceptibility for stress and inflammatory changes. As all LATE-NC + HS cases presented with ipsilateral dentate TDP-43 NCI, it was not possible to analyze the association between LATE-NC + HS and the risk genes independent of TDP-43. However, the association between TMEM106B and dentate TDP-43 NCI was nullified when LATE-NC + HS cases were excluded from the analyses, which indicates that the association between LATE-NC + HS and TMEM106B is dependent on LATE-NC + HS.
Previous studies evaluating the effect of GRN rs5848, TMEM106B rs1990622 and progranulin levels/lysosomal pathways on neurodegeneration were largely based on younger people than those who are at greatest risk of LATE-NC + HS (9). It is conceivable that progranulin alterations might be tolerated at younger age, but that with increased age in LATE-NC + HS, deterioration of cellular defense and clearing mechanisms lower this threshold, leading to CA1 neuron damage.
ABCC9 gene encodes an evolutionarily conserved large polypeptide sulfonylurea receptor 2 (SUR2) with multiple membrane-spanning domains and multiple levels of biologic complexity (17). Intronic SNPs that cluster in the 3′ portion of ABCC9 have been associated with a risk for human brain illnesses, including sleep disturbances and LATE-NC + HS (16). SUR2 transcript variants might have novel alternative splicing in the mRNAs' coding region and 3′ untranslated region (3′UTR) variants (17). The claimed LATE-NC + HS risk-SNP in ABCC9 is also an expression quantitative trait locus, which affects the levels and splice variants of brain mRNA transcripts derived from ABCC9 (17). However, in our study, LATE-NC + HS failed to associate with ABCC9 rs704178 genotype (evaluated both in additive and recessive mode of inheritance) or allele.
The previously reported ABCC9 rs704178 was also not associated with TDP-43 pathology in these cohorts. It is possible that differences in the cohort selection and case definition explain the discrepancy. It is also possible that sex could influence the findings, since men at this age are a minority of the population and power is therefore limited to test for sex differences. Findings by Nelson et al were not adjusted for sex (14), but given that the analyses by Katsumata et al were adjusted (8), this explanation is likely to be insufficient. A mechanistic link between LATE-NC + HS and ABCC9 variation is unclear, and this study did not replicate the previous results. Further studies on ABCC9 and LATE-NC + HS from different source populations are needed to confirm the lack or verify the association.
In conclusion, this study confirmed in population-representative cohorts that GRN rs5848 C/T and T/T as well as TMEM106B rs1990622 A/A genotype are associated with loss of CA1 neurons in the aging brain, supporting a possible link through lower progranulin levels. ABCC9 rs704178, however, was not confirmed as a genetic risk factor of LATE-NC + HS.