Purpose The study investigates the effects of genetic factors on the pathology of Alzheimer's disease (AD) and Lewy body (LB) diseases, including Parkinson's disease and dementia with Lewy bodies.
Methods A multicentre autopsy series (762 brain samples) with AD, LB or vascular pathology was examined. The effects of the tau gene (MAPT) H1 haplotype, the H1 specific SNP rs242557, APOE and the α-synuclein gene (SNCA) 3′UTR SNP rs356165 on the burden of AD and LB pathology were assessed. Neurofibrillary tangles (NFTs) were counted in four brain regions, senile plaques in five and LBs in four. Braak NFT stage, brain weight and presence of vascular pathology were also documented.
Results MAPT H1 associated with lower counts of NFTs in the middle frontal (p<0.001) and inferior parietal (p=0.005) cortices, and also with lower counts of senile plaques in the motor cortex (p=0.001). Associations of MAPT H1 with increased LB counts in the middle frontal cortex (p=0.011) and inferior parietal cortex (p=0.033) were observed but were not significant after multiple testing adjustment. The APOE ε4 allele was strongly associated with overall Alzheimer type pathology (all p≤0.001). SNCA rs356165 and the MAPT H1 specific SNP rs242557 did not associate with AD or LB pathology.
Conclusion This study shows for the first time that MAPT H1 is associated with reduced Alzheimer type pathology which could have important implications for the understanding of disease mechanisms and their genetic determinants.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Alzheimer's disease (AD) and Parkinson's disease (PD) are common age related neurodegenerative conditions that lead to significant cognitive and motor impairment. Pathologically, they are characterised by the accumulation of hyperphosphorylated tau or α-synuclein.1 Disorders with tau deposition include AD and the tauopathies progressive supranuclear palsy (PSP), corticobasal degeneration (CBD), Pick's disease, argyrophilic grain disease and tau positive frontotemporal dementia with parkinsonism linked to chromosome 17 (FTDP-17T), whereas α-synuclein pathology is a core feature of PD, dementia with Lewy bodies (DLB), multiple system atrophy and pure autonomic failure.2
In humans, six tau isoforms are generated by alternative splicing of exons 2, 3 and 10. Based on the number of conserved ∼30 amino acid repeats in the microtubule binding domain, the six tau isoforms can be divided into those with three repeats (3R-tau) and four repeats (4R-tau). In AD, tau deposition in neurofibrillary tangles (NFTs) consists of approximately equal amounts of 3R-tau and 4R-tau whereas the ‘pure’ tauopathies CBD and PSP display mostly 4R-tau deposits.3 In addition to the accumulation of tau in NFTs, neuropil threads and dystrophic neurites, AD brains display numerous senile plaques (SPs) of which aggregated β-amyloid (Aβ) is a key component. In contrast, the pathological hallmark of the synucleinopathies PD and DLB is the presence of α-synuclein positive Lewy bodies (LBs) and Lewy neurites. In PD, LBs are found prominently in vulnerable brainstem nuclei whereas they extend to neocortical areas in DLB and PD dementia.4
While early views favoured AD, tauopathies and synucleinopathies as mutually exclusive clinicopathological entities, a growing body of evidence indicates that an overlap exists at the clinical, genetic, pathological and biochemical levels.2 Clinically, the incidence of extrapyramidal symptoms such as bradykinesia, rigidity and gait impairment in AD patients is higher than expected in age matched populations.5 Conversely, dementia is a core feature of DLB and it is present in up to 80% of patients with PD 17 years after disease onset.6 Pathologically, LBs can be identified in 30–60% of AD patients, with predominance in the amygdala7; one study found that rates tend to be higher among those AD patients with a high load of neuritic plaques.8 In addition, most cases of LB disease (LBD) are accompanied by varying degrees of concurrent Alzheimer's type pathology, and the finding of an association between cortical LBs and neuritic stage has led to the hypothesis that Aβ triggers both AD and LB pathology.9 Neuronal colocalisation of α-synuclein and tau has been documented in patients with LBD, AD, and AD with LBs,10 and phosphorylated forms of α-synuclein and tau were identified in synaptic enriched fractions of frontal cortex from LBD and AD patients.11 Additional evidence from various models, including MPTP induced neurodegeneration in rodents, α-synuclein transgenic mice and in vitro studies, point to a molecular interplay between Aβ, tau and α-synuclein.12 13
In families with dominantly inherited PD due to mutations in the α-synuclein gene (SNCA) or the leucine-rich repeat kinase 2 gene, pleomorphic pathology has been reported, including LBs and tau positive inclusions.14 15 Moreover, the most consistent genetic risk factor implicated in AD (the APOE ε4 allele) has been associated with the co-occurrence of Alzheimer's type pathology in non-AD tauopathies and in synucleinopathies, as well as with the presence of LBs in AD.16 17 Mutations and copy number abnormalities in SNCA cause PD and DLB, while genetic variation at the SNCA locus (including the SNCA 3′UTR SNP rs356165) has been associated with risk of PD and, more recently, with the burden of Alzheimer's type pathology (Braak NFT stage).18 19 Furthermore, mutations in the tau gene (MAPT) cause FTDP-17T, and the H1 haplotype at the MAPT locus has been associated with several tauopathies, including PSP and CBD20; yet MAPT H1 has also been implicated in the risk of PD and PD dementia.21 22
In the present study, we examined the effects of the MAPT H1 haplotype, the H1 specific SNP rs242557, APOE and the SNCA 3′UTR SNP rs356165 on the burden of AD and LB pathology in an autopsy series of 762 cases.
Design and methods
The 762 brains in this sample of convenience were from the Mayo Clinic Jacksonville brain bank for neurodegenerative disorders and were obtained from 1993 to 2004. Some had been part of the State of Florida Alzheimer's Disease Initiative (n=414), the Einstein Ageing Study (P01AG03949-25; n=57), the Mayo Clinic Jacksonville Memory Disorders Clinic (n=103), the University of Miami/National Parkinson's Foundation Brain Endowment Bank (n=25), the Udall Center for Parkinson's Disease Research (P50NS40256; n=25), hospital autopsies or referrals from private pathologists or institutions (n=114) or the Society for Progressive Supranuclear Palsy Brain Bank programme (n=24). Clinical diagnoses were dementia disorders in 631 cases (82% considered to be Alzheimer's type dementia) while 70 had clinical diagnoses of movement disorders (71% considered to have PD) and 61 were cognitively normal or had mild cognitive impairment, history of stroke or a psychiatric disorder (depression being most common). Except for the Einstein Ageing Study, subjects were not part of a prospective, standardised or longitudinal clinical study. A family history of neurological disease was noted in 99 cases (13%) although this was also not systematically assessed. The study was approved by the Mayo Clinic institutional review board and informed consent for brain donation was obtained from legal next of kin.
Median age at death was 81 years (25th–75th percentiles 74–86 years), there were 369 men (48%) and 393 women (52%), and all 762 individuals (100%) were of Caucasian descent. Total brain weight was calculated based on weight of the fixed hemi-brain.
All brains underwent complete neuropathological macroscopic and microscopic evaluations, the latter with haematoxylin–eosin and immunohistochemical stains and thioflavin-S fluorescent microscopy. Cases were selected based on the presence of AD, LB or vascular pathology and included cases with combined pathology and cases with no significant pathology or mild pathological abnormalities of the Alzheimer's type (pathological ageing=many SPs and sparse or no NFTs; senile changes=sparse SPs and sparse or no NFTs) or LB pathology not conforming to common classifications of LBD (eg, amygdala predominant LBs7). As the method used for quantification of Alzheimer's type pathology, thioflavin-S, like thiazin red fluorescent microscopy, does not reliably detect 4R-tau inclusions, cases with PSP and CBD were not included. Also excluded were cases with multiple system atrophy, frontotemporal lobar degeneration and any other major non-Alzheimer degenerative disease process.
Histology and immunohistochemistry
All brains were evaluated for Alzheimer's type pathology using thioflavin-S fluorescent microscopy.7 SPs were counted in five brain regions (maximal density of SPs per field at ×100 magnification in the medial frontal gyrus (MF), superior temporal gyrus (ST), inferior parietal gyrus (IP), motor cortex (MTR) and visual cortex). There was a ceiling effect since SP counts were truncated at more than twice the number needed to meet Khachaturian criteria for AD23; therefore, SP counts range from 0 to 50 (all cases with ≥50 SPs per field were scored as 50). The method used (thioflavin-S) did not allow us to discriminate between diffuse plaques and neuritic plaques. NFTs were counted in four brain regions (maximal density of NFTs per field at ×400 magnification in MF, ST, IP and hippocampal CA1); the Braak NFT stage24 was also determined based on the distribution of NFTs.
All samples with Alzheimer's type pathology underwent screening for LB pathology using α-synuclein immunohistochemistry (see below). Evaluation of LB pathology in Alzheimer's type pathology negative individuals was performed in two stages. Firstly, examination of brainstem nuclei (substantia nigra, locus coeruleus and dorsal motor nucleus of the vagus nerve) with haematoxylin–eosin staining assessed the presence of neuronal loss or LBs. Those cases with evidence of neuronal loss or LB pathology underwent α-synuclein immunostaining using a polyclonal rabbit antibody (NACP; Mayo Clinic, Jacksonville, Florida, USA). In all other cases with no neuronal loss in vulnerable brainstem nuclei, a section of amygdala was screened for LBs.7 If detected in the amygdala, then these cases were also studied for cortical and brainstem LBs. Presence and distribution of LBs were scored from 0 to 4 (0, no LBs; 1, amygdala predominant; 2, brainstem predominant; 3, brainstem and limbic=transitional LBD; 4, brainstem and diffuse cortical). Cortical LBs were counted in four brain regions (maximal density of cortical LBs per field at ×200 magnification in MF, ST, IP and cingulate gyrus).
Cerebrovascular pathology was assessed on all cases using a semiquantitative scale similar to that previously reported.25 Briefly, cases with no cerebrovascular lesions were scored 0, those with minimal cerebrovascular pathology (including 1–2 small lacunes, mild cerebral amyloid angiopathy or mild leukoencephalopathy) were scored 1, those with moderate lesions (including >2 lacunes, severe cerebral amyloid angiopathy or diffuse leukoencephalopathy) were scored as 2 and those with marked cerebrovascular pathology (including old cortical infarcts, multiple microinfarcts or hippocampal sclerosis) were scored 3. For the purposes of analysis, cases with scores of 2 and 3 were grouped as having vascular pathology and those with scores of 0 and 1 were considered to lack vascular pathology.
In all 762 subjects, DNA was obtained from frozen brain tissue using standard protocols. Each sample was genotyped for MAPT H1/H2 (SNP rs1052553 A/G, A=H1, G=H2), APOE and the SNCA 3′UTR SNP rs356165 (G/A). Additionally, a subset of 690 samples were genotyped for the H1 specific (ie, only polymorphic on H1 chromosomes) SNP rs242557 (A/G), given a previously reported association with risk of AD.26 All genotypes except APOE were determined using a Sequenom MassArray iPLEX platform (Sequenom, San Diego, California, USA) and analysed with Typer 4.0 software (also from Sequenom). The APOE genotype was determined using TaqMan chemistry. All primer sequences are available on request. Genotype failure occurred in <2% of samples.
We examined the association of MAPT H1/H2, MAPT rs242557, APOE and SNCA rs356165 genotypes with a variety of different endpoints (LB count, LBD score, NFT count, Braak NFT stage, SP count, vascular disease and brain weight), and the statistical tests used to evaluate these associations were determined by the nature of the given endpoint (count, censored, binary categorical or numerical). For LB and NFT counts, associations with genotypes were evaluated using negative binomial regression models. With a ceiling of 50 counts, SP counts were censored at a value of 49 counts and as such parametric survival regression models were used to examine associations between SP count and genotypes. For LBD score (<2 vs ≥2), Braak NFT stage (<5 vs ≥5) and vascular disease, logistic regression models were used to evaluate associations with genotypes; for easier interpretation, LBD score was dichotomised as <2 vs ≥2, as LBD stage ≥2 indicates disease outside the brainstem, and Braak NFT stage was dichotomised as <5 vs ≥5 due to the fact that Braak NFT stage ≥5 is considered concrete evidence of AD. Associations between brain weight and genotypes were examined using linear regression models.
Single variable models (ie, models with no adjustment) were considered in all regression analyses in an exploratory analysis, as well as multivariable models adjusting for the known potentially confounding variables age, sex and APOE (only in models involving NFT count, SP count, Braak NFT stage and brain weight) in the primary analysis. Interactions between genotypes were also considered. Due to the high concentration of LB counts equal to 0 and SP counts at the ceiling of 50, these variables were converted into binary categorical variables (>0 and <50, respectively) for ease of presentation, although the actual counts were used in performing the statistical tests.
We adjusted for the number of statistical tests that were performed as follows. We defined a family of statistical tests as all tests evaluating the associations of outcomes with a given SNP, resulting in a different family of tests in evaluating associations of endpoints with MAPT H1/H2 genotype, with MAPT rs242557 genotype, with APOE genotype and with SNCA rs356165 genotype. For each family of statistical tests, multiple testing was adjusted for using single step minP procedure with 10 000 permutations of genotype labels to determine the level of significance that controls the family wise error rate at 5%; p values less than or equal to this level were considered statistically significant.27 All statistical analyses were performed using S-Plus (V.8.0.1; Insightful Corporation, Seattle, Washington, USA).
Overall pathological findings are summarised in table 1. There was evidence of an association between the MAPT H1 haplotype (rs1052553 A-allele) and Alzheimer's type pathology (table 2), with and without adjustment for age, sex and APOE genotype.
The H1 allele was associated with lower NFT counts in three (MF, ST and IP) of the four regions examined (all p≤0.027), and also with a lower Braak NFT stage (p=0.037). Similarly, SP counts were lower among MAPT H1 carriers compared with non-carriers in three (MF, ST and MTR) of the five regions examined (all p≤0.048). After correction for multiple testing (p≤0.0054 considered statistically significant), H1 remained associated with lower NFT counts in two regions (MF, p<0.001; IP, p=0.005) and with lower SP counts in one region (MTR, p=0.001). Stratification based on the APOE genotype did not influence the effect of MAPT H1 on pathological measures (p≥0.25 for all H1-ε4 interactions). MAPT H1 was also associated with increased LB count in two regions (MF, p=0.011; IP, p=0.033) and also with increased LB score (p=0.019), however none of these findings remained significant after multiple testing adjustment. Given the unexpected associations of MAPT H1 with lower NFT and SP counts, we further examined these associations in a sensitivity analysis involving only those cases with AD pathology (see supplementary table 1, available online only). In this smaller sample with lower power to detect associations, results were generally similar, with MAPT H1 associated with lower NFT counts (MTR and IP regions) and with lower SP counts (MTR region). The H1 specific SNP rs242557 did not show evidence of an association with pathological measures (all p≥0.063, see supplementary table 2, available online only).
There was a strong association between the APOE ε4 allele and Alzheimer's type pathology (table 3).
The ε4 allele was associated with higher NFT and SP counts in all regions examined, as well as with a higher Braak NFT stage and lower brain weight (all p≤0.001), and all of these findings remained significant after correction for multiple testing (p≤0.0056 considered statistically significant). There was no evidence of an association between APOE ε4 and LB counts or LB score (all p≥0.19).
The SNCA 3′UTR SNP rs356165 did not show association with any of the PD/LBD pathological endpoints examined, including LB counts and LB score (all p≥0.16; data not shown). As expected, SNCA rs356165 did not associate with Alzheimer's type pathology (all p≥0.41). Given previous evidence of an interaction between SNCA and MAPT H1 in determining risk of PD,21 interaction between rs356165 and MAPT H1 was examined for LB pathology by stratifying the effect of MAPT H1 based on the rs356165 genotype. We observed no evidence of an interaction between rs356165 and MAPT H1 regarding LB counts (all p≥0.58) or LB score (p=0.51) (see supplementary table 3, available online only).
Our study employed a large autopsy series to examine the role of MAPT (H1/H2 and the H1 specific SNP rs242557), APOE (ε4 allele) and SNCA (rs356165) in determining the density of LBs and Alzheimer's type pathology. We identified a novel association between MAPT H1 and reduced severity of Alzheimer's type pathology. Specifically, after correction for multiple testing, H1 was associated with lower NFT counts (MF, IP) and SP counts (MTR), and these effects were independent of APOE genotype, age and sex. Furthermore, similar trends that did not reach significance after multiple testing correction were observed for NFT counts (ST), SP counts (MF and ST) and Braak NFT stage.
Our findings suggest a protective effect of the MAPT H1 haplotype on severity of Alzheimer's type pathology, which has several major implications. Although initially surprising given strong and consistent evidence linking the MAPT H1 haplotype with an increased risk for the 4R-tauopathies PSP and CBD, recent genome wide association studies have failed to identify any association of MAPT with risk of AD.28 29 Moreover, meta-analysis of reported genetic association studies in AD did not find a significant association with MAPT.30 Data from studies in three autopsy confirmed Caucasian AD patient–control series (combined sample: 655 patients and 380 controls) also could not support H1 increasing risk for developing AD.26 31
Previous in vitro studies showed that H1 is more efficient at promoting MAPT expression compared with H2,32 yet our data showed this does not translate into pathological AD lesions and may be associated with reduced burden of NFTs and SPs. Possible explanations include interactions (eg, with glycogen synthase kinase-3β (GSK3B)) that alter MAPT transcription or propensity of tau to form aggregates. In support of this hypothesis, the previously reported 2-SNP haplotype in GSK3B only associated with an increased risk of PD and AD among carriers of at least one MAPT H2 allele and not in H1/H1 homozygotes, suggesting a protective effect of H1.33 34 Interestingly, our data indicate that MAPT H1 is associated with reduction of not only NFTs, but also SPs. This emphasises the importance of tau in AD and is consistent with the known correlation between NFT pathology and presence and severity of dementia in AD,35 indicating our findings may have clinical implications.
We identified associations of MAPT H1 with severity of LB pathology that were significant at p≤0.05; however, none of these withstood correction for multiple testing. Of note, power to detect such associations was limited due to the significant number of cases with zero LB counts, and the possibility of type II error is important to consider. Despite the lack of significance after multiple testing correction, our findings appear to be consistent with previous findings from association studies that consistently reported H1 being a risk factor for clinically defined PD,22 which would be expected to be associated with brainstem and in most cases cortical LBs.4 36 37 It should be noted that one study failed to identify an association between MAPT H1 and pathologically confirmed PD.38 This may stem from differences in clinical and pathological ascertainment of PD or indicate that these genetic factors determining risk are not always associated with end point quantitative measures. Additionally, analysis of LB score included cases with amygdala predominant LBs, which may have obscured an association. Given the association of MAPT H1 with clinical PD and the limitations of the present study, future studies with larger numbers of patients with LBD are needed to better characterise the effect of MAPT H1 on LB pathology.
Consistent with previous clinical and pathological studies, our results showed a strong association between the APOE ε4 allele and Alzheimer's type pathology, with an increase in NFT counts, Braak NFT stage and SP counts, and a decrease in brain weight among ε4 carriers, all of which remained significant after multiple testing correction.39 The stronger effect on SP than NFT is consistent with the association of APOE ε4 with amyloid, but not tau measures in imaging and CSF studies of prospectively studied humans over a wide age range.40 In contrast with the protective effect of H1, the brain regions affected by APOE ε4 included those areas known to be vulnerable to Alzheimer's type pathology (CA1 and IP).
In a recent study, Peuralinna et al identified an association between SNCA intron 4 SNP rs2572324 and Braak NFT stage in an autopsy series (n=272).19 Furthermore, there was a trend towards an association for four additional SNCA SNPs, including the 3′UTR SNP rs356165. In contrast, our results did not find any association between rs356165 and AD or LB pathology. These discrepant results are not related to sample size issues as our overall series and the number of cases with LB pathology were larger. Possible explanations include differences in study design (population based in theirs versus consecutive autopsy series in ours) and pathological evaluation, or may reflect population specificity (relatively homogeneous Finnish population in theirs versus admixed US population in ours).
Given that genetic variation at both the MAPT and SNCA loci was suggested to alter risk for PD and PD dementia by means of an interaction,21 we investigated whether rs356165 influences the effect of H1 on LB pathology. We observed no evidence of an interaction between rs356165 and MAPT H1 for LB pathology in our series.
The results of our study indicate that the MAPT H1 haplotype is associated with reduced severity of Alzheimer's type pathology and that this effect is independent of age, sex and APOE. This provides important novel mechanistic insights into the complex pathogenesis of AD.
The authors thank Linda Rousseau, Virginia Phillips and Monica Casey-Castanedes for brain histology and John Gonzales for brain banking.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
CW and OAR contributed equally to this work.
Funding The Mayo Clinic is a Morris K Udall Parkinson's Disease Research Center of Excellence supported by the National Institute of Neurological Disorders and Stroke, NS072187. The Mayo Clinic brain bank for neurodegenerative disorders is supported by P01AG017216; Einstein Ageing Study, P01AG03949; Mayo Alzheimer's Disease Research Center, P50AG16754; State of Florida Alzheimer's Disease Initiative; and the Society for Progressive Supranuclear Palsy. Additional funding included the Pacific Alzheimer's Research Foundation (PARF) (grant C06-01).
Competing interests ZKW is partially funded by P01AG017216, R01NS057567, R01AG015866 and CIHR 121849. CW is supported by the Swiss National Science Foundation (PASMP3-123268/1). OAR is supported by P01AG017216 and the family of Carl and Susan Bolch. SP is currently employed by Biogen Idec Inc (Cambridge, Massachusetts, USA) and holds stock in this company.
Ethics approval Ethics approval was provided by the institutional review board committee of the Mayo Clinic as well as from all of the centres that collaborated in the study (detailed in the article).
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement All brains used in the study are stored at the Mayo Clinic and are directly accessible/available to the senior author (DWD). All pathology, genetic and statistic data are stored at the Mayo Clinic and directly accessible/available to CW, OAR, MGH and DWD. All available data have been used for publication.