Abstract
Objective To test the association of distinct neuropathologic features of Alzheimer disease (AD) with risk loci identified in genome-wide association studies.
Methods Vantaa 85+ is a population-based study that includes 601 participants aged ≥85 years, of which 256 were neuropathologically examined. We analyzed 29 AD risk loci in addition to APOE ε4, which was studied separately and used as a covariate. Genotyping was performed using a single nucleotide polymorphism (SNP) array (341 variants) and imputation (6,038 variants). Participants with Consortium to Establish a Registry for Alzheimer Disease (CERAD) (neuritic Aβ plaques) scores 0 (n = 65) vs score M + F (n = 171) and Braak (neurofibrillary tangle pathology) stages 0–II (n = 74) vs stages IV–VI (n = 119), and with capillary Aβ (CapAβ, n = 77) vs without (n = 179) were compared. Cerebral amyloid angiopathy (CAA) percentage was analyzed as a continuous variable.
Results Altogether, 24 of the 29 loci were associated (at p < 0.05) with one or more AD-related neuropathologic features in either SNP array or imputation data. Fifteen loci associated with CERAD score, smallest p = 0.0002122, odds ratio (OR) 2.67 (1.58–4.49) at MEF2C locus. Fifteen loci associated with Braak stage, smallest p = 0.004372, OR 0.31 (0.14–0.69) at GAB2 locus. Twenty loci associated with CAA, smallest p = 7.17E-07, β 14.4 (8.88–20) at CR1 locus. Fifteen loci associated with CapAβ smallest p = 0.002594, OR 0.54 (0.37–0.81) at HLA-DRB1 locus. Certain loci associated with specific neuropathologic features. CASS4, CLU, and ZCWPW1 associated only with CAA, while TREM2 and HLA-DRB5 associated only with CapAβ.
Conclusions AD risk loci differ in their association with neuropathologic features, and we show for the first time distinct risk loci for CAA and CapAβ.
Glossary
- Aβ=
- amyloid-β;
- AD=
- Alzheimer disease;
- APOE=
- apolipoprotein E;
- CAA=
- cerebral amyloid angiopathy;
- CapAβ=
- capillary Aβ;
- CERAD=
- Consortium to Establish a Registry for Alzheimer Disease;
- Chr9 region=
- chromosome 9 region;
- CI=
- confidence interval;
- GWAS=
- genome-wide association study;
- IHC=
- immunohistochemistry;
- LOAD=
- late-onset Alzheimer disease;
- OR=
- odds ratio;
- SNP=
- single nucleotide polymorphism
Late-onset Alzheimer disease (LOAD) is neuropathologically characterized by cerebral accumulation of amyloid-β (Aβ) peptide containing neuritic plaques and hyperphosphorylated tau-protein, and by cerebral amyloid angiopathy (CAA) and capillary Aβ (CapAβ) deposition. LOAD is known to have a fairly strong hereditary risk,1 the apolipoprotein E (APOE) ε4 being the strongest genetic risk factor.2,3 Recently, genome-wide association studies (GWASs) have identified approximately 30 Alzheimer disease (AD)-associated risk loci,4,–,10 which are known to encode proteins involved in immune system and inflammation (CLU, CR1, ABCA7, MS4A, CD33, EPHA1, MEF2C, HLA-DRB1/DRB5, TRIP4, and TREM2), cholesterol metabolism (APP, CLU, ABCA7, and SORL1), synaptic and membrane function (PICALM, BIN1, CD33, CD2AP, EPHA1, INPP5D, PTK2B, SORL1, and SLC2A4), tau pathology (BIN1), and Aβ metabolism (APP, CLU, CR1, ABCA7, INPP5D, and SORL1).6,8,11,–,15 Most of the previous GWASs have been based on large clinically diagnosed hospital-based samples,5,–,10,14 but recently, a few GWASs have been published based on neuropathologically verified data sets.16 In those studies, ABCA7 and CD2AP as well as a variant near APP (rs2829887) and ABCG1, GALNT7 and an intergenic region on chr 9 (9:129,280,000–129,380,000) loci have been found to be associated with neuritic plaque pathology.4,16 The Consortium to Establish a Registry for Alzheimer disease (CERAD) score and Braak stage have also been associated with ABCA7, BIN1, CASS4, MEF2C, and PICALM, and Braak stage with CLU, SORL1, ZCWPW1, and CERAD score with MS4A6A, and CD33.4
Here, we analyzed possible associations of AD risk loci with each neuropathologic feature (neuritic plaque, neurofibrillary tangle and CAA, and CapAβ) in a population-based sample of very elderly Finns (Vantaa 85+ Study).
Methods
Study population
The Vantaa 85+ Study includes 601 individuals, aged at least 85 years, who were living in the city of Vantaa on April 1, 1991. Autopsy and neuropathologic examination were performed on 300 (mean age 92.4 ± SD 3.7 years, range 85–105). The clinical characteristics of the whole genotyped subpopulation (N = 512) and the whole genotyped neuropathologically examined subpopulation (N = 300) are shown in table e-1 (http://links.lww.com/NXG/A16).
Neuropathologic examination
Evaluation of Braak stages of Alzheimer-type neurofibrillary pathology17 and CERAD scores of neuritic plaques18 have been described previously.19 The percentage of CAA-affected noncapillary blood vessels was estimated using histologic Congo red as described previously and confirmed by immunohistochemistry (IHC).20 The presence of CapAβ was analyzed as described before20,21 using IHC.21
Evaluation of the APOE genotype, SNP array, and candidate gene approach
APOE genotyping was performed as described previously.22 A GWAS was conducted as previously described23 by Infinium Human370 BeadChips (Illumina, San Diego, CA) for 327,521 variants in blood samples from 512 participants. Data quality control was done using the standard PLINK v1.924,25 protocol.26 In summary, related individuals (identity by descent >0.185), individuals with discordant sex information, divergent ancestry, elevated missing data (>3%) rate, or outlying heterozygosity rate (±2 SD) were excluded. Variants with missing per person rate >10%, minor allele frequency <1% and missing data rate >5%, or significantly different genotype call rates between cases and controls (p < 0.00001) were excluded. Variants not in the Hardy-Weinberg equilibrium (p < 0.00001) were also discarded.
A PubMed search was performed to identify all the loci that have been reported in previous GWAS analyses in samples from participants with clinically or neuropathologically diagnosed AD. Besides APOE, we found reports on 44 variants at 29 loci. Variants at genes near these candidate loci were extracted from the quality-controlled genome-wide single nucleotide polymorphism (SNP) array. To cover nearby variants of possible interest, variants within 1 kb of each candidate gene were also included in the study (table e-2, http://links.lww.com/NXG/A16).
Imputation
Imputation was performed using IMPUTE2.27 1000 Genomes phase3 data (October 2014 release) supplied by IMPUTE2 were used as the reference panel. Imputation was performed on the same candidate genes as in the SNP array–based analyses and on the 44 previously reported index variants. The whole available Vantaa 85+ data set (n = 512) was imputed. We have whole-genome sequences of a subset of the Vantaa 85+ study (n = 286), and we compared the imputed genotypes to the whole-genome sequencing-derived genotypes. The median discordance between genotypes was 0.7%, which indicates successful imputation. We performed the same quality control steps and association analyses as we did to the SNP array data, but the genotyping rate threshold was not defined for the 44 index variants.
Statistical analyses
In the analyses, participants with moderate or frequent CERAD scores were compared with participants with no neuritic plaques (CERAD 0). Similarly, participants with Braak stages 0–II were compared with the high-stage group (Braak stages IV–VI). All participants without CapAβ were regarded as controls in analyses related to that pathology. The associations between APOE ε4 allele and neuropathologic features were performed using logistic or linear regression analysis with age and sex as covariates on SPSS (version 23) (table 1). Other statistical analyses were performed using PLINK. Case-control association tests were calculated using logistic regression. Quantitative trait associations were calculated using linear regression. Each regression analysis was performed twice with either age and sex or age, sex, and APOE ε4 status as covariates. In this candidate gene analysis of GWAS known AD loci, p < 0.05 was considered statistically significant.
Results of association analyses between the APOE ε4 allele and neuropathologic features (CERAD, Braak, CAA, and CapAβ)
Standard protocol approvals, registrations, and patient consents
The Vantaa 85+ study was approved by the Ethics Committee of the Health Centre of the City of Vantaa in 1991 and by the Coordinating Ethics Committee of the Helsinki University Central Hospital in 2014. The Finnish Health and Social Ministry has approved the use of the health and social work records and death certificates. Blood samples were collected only after the participants or their relatives provided written informed consent. The National Authority for Medicolegal Affairs (VALVIRA) has approved the collection of the tissue samples at autopsy as well as their use for research. Written informed consent for autopsy was obtained from the nearest relatives.
Results
SNP array and imputation
After quality control, 341 variants at 26 candidate loci remained in the SNP array data (table e-2, http://links.lww.com/NXG/A16). There were no variants in EXOC3L2, HLA-DRB1, and TREM2 loci. In the imputed data set, 6,038 variants remained in 28 loci after quality control. Imputation was not successful in INPP5D, but it was covered with 26 variants in the SNP array data. Thus, all 29 candidate loci were covered in either the original SNP array or imputed data sets. Imputation of the index variants in ABCG1, APP, and chromosome 9 region (Chr9 region) was not successful because of too small minor allele frequency or low genotyping quality. Associations between the candidate loci and neuropathologic features are summarized in table 2.
Associations of candidate loci with neuropathologic features
Of the 512 samples, 487 passed the quality control criteria. Samples from 3 individuals were excluded because of difference in reported and estimated sex, 4 because of relatedness of participants, and 18 because of excessive missing data rate or heterozygosity.
Neuropathologic findings
The characteristics of the whole Vantaa 85+ sample (n = 512) and the neuropathologically and genetically examined subpopulations (n = 256) are shown in table e-1 (http://links.lww.com/NXG/A16). Neuropathologic analysis and data details have been previously reported.19,–,21 No statistically significant differences were found in age at death or sex between the whole study population and the neuropathologically examined subpopulation, but there were slightly more females in the neuropathologically examined subpopulation.
APOE
As expected and already previously published using other types of analyses,28,29 the APOE ε4 allele was strongly associated with all AD-related neuropathologic features (table 1). Further analyses were performed with and without APOE ε4 adjustment.
Association of the 29 risk loci with distinct neuropathologic features
Overall distribution of associations between the 26 candidate loci covered by the SNP array and neuropathologic features are shown in table e-3 (http://links.lww.com/NXG/A16). EXOC3L2, HLA-DRB1, and TREM2 could not be analyzed since there were no variants at these loci in the SNP array. Variant details and p values are shown in table e-4. APOE was treated as a covariate and not included in the list of tested loci. Nine of the 26 SNP array loci were not associated with any histopathologic variables in SNP array data.
After imputation of all 29 loci, associations were found with 24 loci. The 5 loci that did not show association with any neuropathologic feature were CD33, CELF1, EPHA1, EXOC3L2, and INPP5D. We found an association at p < 0.05 with a neuropathologic feature for 7 of the previously reported 44 index variants, while 9 other variants showed a trend at 0.05 < p < 0.10 (table e-5). However, the genotyping rate of index variants was <95% for 14 variants.
CERAD score of neuritic plaques
In the SNP array data, we identified 8 loci that were associated with the CERAD score when adjusted for age at death and sex but not for APOE ε4 (table 3). When APOE ε4 was included as a covariate, all these associations remained significant, and an additional association was detected with FERMT2 (rs1112777, p = 0.01733, odds ratio [OR] 0.5587, 95% confidence interval [CI] 0.35–0.90). The strongest association was found between the CERAD score and the MEF2C—locus (rs700588) (when adjusted with age, sex, and APOE ε4, p = 0.0002122, OR 2.67, 95% CI 1.59–4.49 and without APOE ε4 p = 0.0003895, OR 2.40, 95% CI 1.48–3.88). MEF2C did not associate with any other histopathologic variables (Braak, CAA, and CapAβ).
Associations in SNP array data between the CERAD score (CERAD score 0 vs M + F) and previously known AD risk loci (341 variants)
In the imputed data, we identified 14 loci that were associated with the CERAD score (table 2, tables e-6 and e-7, http://links.lww.com/NXG/A16). The strongest association found was the same as in the SNP array data: rs700588 at MEF2C.
Braak stage
In the SNP array data, 6 loci were associated with a high Braak stage (IV–VI vs 0–II) (APP, GALNT7, PTK2B, SLC24A, SORL1, and TRIP4), and when adjusted for APOE ε4, associations were also found with ABCG1 (table 4). Overall, the associations with the Braak stage were weaker than those with the CERAD score. The strongest association was found with ABCG1 (rs532345, p = 0.02671, OR 0.5571, 95% CI 0.33–0.93 with APOE ε4 adjustment).
Associations in SNP array data between the Braak stage and previously known AD risk loci (341 variants) comparing participants with Braak stage IV–VI (n = 119) vs Braak stage 0–II (n = 74)
In the imputed data, we identified 15 loci that showed association with the Braak stage (ABCA7, ABCG1, APP, CD2AP, Chr9 region, CR1, GAB2, GALNT7, MEF2C, MS4A, NME8, PTK2B, SLC24A4, SORL1, and TRIP4) (table 2, tables e-8 and e-9, http://links.lww.com/NXG/A16). All except for MEF2C showed association regardless of APOE ε4 adjustment. Rs2512518 at GAB2 locus had the strongest association (p = 0.004372, OR 0.31, 95% CI 0.14–0.69) when adjusted for age and sex.
CAA
In the SNP array data, 7 loci were associated with CAA (ABCA7, CR1, FERMT2, NME8, SLC24A4, SORL1, and ZCWPW1), and when adjusted for APOE ε4, an additional locus (GALNT7) was found (table 5). The strongest association was with CR1 (rs65087, p = 0.004934, β 2.52, 95% CI 0.78–4.26 without APOE ε4 adjustment); CR1 and ABCA7 were not associated with any other histopathologic variable than CAA.
Associations in SNP array data between CAA and previously known AD risk loci (341 variants) adjusted for age at death and sex and with and without carrier status of the APOE ε4 allele
In the imputed data, we identified 20 loci that were associated with CAA (table 2, tables e-10 and e-11, http://links.lww.com/NXG/A16). Fifteen loci were associated regardless of APOE ε4 adjustment (ABCA7, ABCG1, APP, BIN1, CASS4, Chr9 region, CR1, FERMT2, HLA-DRB1, NME8, PICALM, SLC24A4, SORL1, TRIP4, and ZCWPW1), 4 loci when adjusted for age, sex, and APOE ε4 (CD2AP, CLU, GALNT7, and MS4A locus) and 1 locus (GAB2) when adjusted for age and sex. The strongest association was found for rs185310342 at CR1 locus (p = 7.17E-07, β 14.4, 95% CI 8.88–20) when adjusted for age and sex.
CapAβ
In SNP array data, 4 loci were associated with CapAβ (APP, BIN1, MS4A, and PTK2B), and when adjusted for APOE ε4, age, and sex, 3 additional loci were associated with CapAβ (GALNT7, NME8, and FERMT2, table 6). The strongest association was found with APP (rs1783016, p = 0.005933, OR 2.01, 95% CI 1.22–3.30 with APOE ε4 adjustment).
Associations in SNP array data between CapAβ and previously known AD risk loci (341 variants)
In imputed data, we found association for 15 loci (table 2, tables e-12 and e-13, http://links.lww.com/NXG/A16). Of these, 9 were associated regardless of APOE ε4 adjustment (APP, BIN1, FERMT2, GALNT7, HLA-DRB1, MS4A, NME8, PTK2B, and Chr9 region), 3 loci (MEF2C, PICALM, and SORL1) when adjusted for age, sex, and APOE ε4, and 3 loci (CR1, SLC24A4, and TREM2) when adjusted for age and sex. The strongest association was found for rs66962766 at HLA-DRB1 locus (p = 0.002594, OR 0.54, 95% CI 0.37–0.81).
Discussion
In this study, we focused on previously reported genetic AD risk loci identified in GWAS analyses on clinically or neuropathologically verified patients with AD and controls. We confirmed the association of 24 of the 29 previously known AD risk loci with one or more AD-related neuropathologic features (CERAD, Braak, CAA, and CapAβ) (table 2).
In this study, we found strong associations between APOE ε4 and all AD-related neuropathologic features (CERAD, Braak, CAA, and CapAβ, table 1). This is in line with previous studies.4,21 To take into account the strong effect of APOE on the other loci,30 we performed analyses in 2 ways, testing each neuropathologic feature with and without the adjustment for the APOE ε4 carrier status. Previously, certain loci have been reported to be more likely influenced by APOE ε4 than others; e.g., PICALM and EXOC3L2 have been found to show stronger associations with neuropathologically confirmed AD without APOE ε4 adjustment, whereas adjustment with APOE was reported to have no effect on the association between CR1, CLU, or BIN1 and neuropathologic AD.30 We found that the APOE adjustment did not remarkably alter the associations between neuropathologic features in most loci (table 2).
In addition to APOE, APP, Chr9 region, NME8, PICALM, and SLC24A4 were associated with all neuropathologic variables (CERAD, Braak, CAA, and CapAβ) (table 2).
On the other hand, certain loci were associated only with specific neuropathologic features. CASS4, CLU, and ZCWPW1 were associated only with CAA. TREM2 and HLA-DRB5 were associated with only CapAβ, whereas HLA-DRB1 was associated with both CAA and CapAβ but not with other pathologies (table 2). These are interesting findings suggesting that the risk loci (and mechanisms) for CAA and CapAβ may be partially distinct. It is of note that there has been only 1 previous GWAS that investigated the genetic background of CAA, in which the only significant association was found between CAA and the APOE locus.4 However, in a previous candidate gene analysis, CR1 was associated with CAA pathology burden.31 The association between CAA and CR1 was confirmed in our study. No previous GWAS has been performed using CapAβ as the phenotype. Here, we reported significant associations between CapAβ and 15 loci (APP, BIN1, Chr9 region, CR1, FERMT2, GALNT7, HLA-DRB1, MEF2C, MS4A, NME8, PICALM, PTK2B, SLC24A4, SORL1, and TREM2). Our results provide information on the partly shared and partly distinct genetic backgrounds of AD-related neuropathologic features.
Author contributions
Mira Mäkelä and Karri Kaivola: acquisition of data, analysis and interpretation of data, and drafting the manuscript. Miko Valori: acquisition of data and analysis and interpretation of data. Anders Paetau and Tuomo Polvikoski: acquisition of data and critical revision of the manuscript for intellectual content. Andrew B. Singleton: acquisition of data and analysis and interpretation of data. Bryan J. Traynor: acquisition of data and critical revision of the manuscript for intellectual content. David J. Stone and Terhi Peuralinna: acquisition of data and analysis and interpretation of data. Pentti J. Tienari: analysis and interpretation of data, study concept, and design and critical revision of the manuscript for intellectual content. Tanskanen Maarit: acquisition of data and critical revision of the manuscript for intellectual content. Liisa Myllykangas: analysis and interpretation of data, supervision, study concept and design, and critical revision of the manuscript for intellectual content.
Study funding
The study was supported by the Helsinki University Central Hospital competitive research fund, the Academy of Finland (294817), the Finska Läkaresälskapet, and in part by the Intramural Research Program of the NIH, National Institute on Ageing (Z01-AG000949-02). The whole-genome sequencing of Vantaa85+ was funded by Merck & Co., Inc., Kenilworth, NJ USA, and Intramural Research Program of the NIH. David Stone is employed by Merck & Co., Inc., West Point, PA, USA.
Disclosure
M. Mäkelä has served on the editorial board of the Journal of Alzheimer's Disease and has received research support from Helsinki University. K. Kaivola has received research support from the Finnish Medical Foundation and the Maire Taponen Foundation. M. Valori is an employee of Blueprint Genetics. A. Paetau and T. Polvikoski report no disclosures. A.B. Singleton has received travel funding from 23andMe; serves on the editorial boards of Annals of Neurology, Lancet Neurology, Neurogenetics, Neurodegenerative Diseases, Brain, and the Journal of Parkinson's Disease; holds a patent (pending) for Panel of markers to diagnose stroke; and has received research support from the NIH and the Department of Defense. B.J. Traynor has received travel funding from the Italian Football Federation (FIGC), American Academy of Neurology, Cold Spring Harbor Laboratories, the Italian ALS Association (ARISLA), Movement Disorders Society, Wellcome Trust, the Finnish Neurological Society, the NorthEast ALS Consortium (NEALS), UC Irivine, the ALS Hope Foundation, the ALS Association, the American Academy of Neurology, the Motor Neuron Association (United Kingdom), Microsoft Research, the Italian Neurological Society, the University of Toronto, the International Congress of Human Genetics, and the Society for Neuroscience; has served on the editorial boards of Neurology®, Journal of Neurology, Neurosurgery, and Psychiatry (JNNP), Neurobiology of Aging, and JAMA Neurology; holds a patent (pending in Europe and the United States) on Clinical testing and therapeutic intervention for the hexanucleotide repeat expansion of the C9ORF72 gene; is employed by the NIH; and has received research support from Merck Inc., Intramural Research Program of the NIH, the Myasthenia Gravis Foundation, the ALS Association, the Packard Center for ALS Research, FIGC (The Italian Football Federation), the Center for Disease Control and Prevention, Microsoft Research, the Italian ALS Association (ARISLA), and the Muscular Dystrophy Association. D.J. Stone is employed by Merck & Co. T. Peuralinna reports no disclosures. P.J. Tienari has received travel funding from TEVA, Merck (Europe), Biogen, Sanofi-Genzyme, and Novartis; has received speaker honoraria from Biogen, Merck (Europe), TEVA, Novartis, Sanofi-Genzyme, Orion, Roche, and Pfizer; coholds a patent (pending) on Genetic testing and of Hexanucleotide repeat expansion of C9orf72 implicated in amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD); has received research support from Biogen Finland, Roche, Novartis Finland, Sanofi-Genzyme, Merck & Co Inc. (United States), and Helsinki University Hospital. M. Tanskanen reports no disclosures. L. Myllykangas serves on the editorial board of the Journal of Alzheimer's Disease. Funding information and disclosures are provided at the end of the article. Full disclosure form information provided by the authors is available with the full text of this article at Neurology.org/NG.
Acknowledgment
The authors thank Dr. Leena Niinistö and Prof. emeritus Raimo Sulkava for establishing the Vantaa 85+ Study and Merja Haukka for technical assistance.
Footnotes
↵* These authors contributed equally to this work.
Funding information and disclosures are provided at the end of the article. Full disclosure form information provided by the authors is available with the full text of this article at Neurology.org/NG.
The Article Processing Charge was funded by the University of Helsinki (funded by Academy of Finland).
- Received May 13, 2017.
- Accepted in final form November 26, 2017.
- Copyright © 2018 The Author(s). Published by Wolters Kluwer Health, Inc. on behalf of the American Academy of Neurology.
This is an open access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License 4.0 (CC BY-NC-ND), which permits downloading and sharing the work provided it is properly cited. The work cannot be changed in any way or used commercially without permission from the journal.
References
Letters: Rapid online correspondence
NOTE: All contributors' disclosures must be entered and current in our database before comments can be posted. Enter and update disclosures at http://submit.ng.neurology.org. Exception: replies to comments concerning an article you originally authored do not require updated disclosures.
- Stay timely. Submit only on articles published within the last 8 weeks.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- 200 words maximum.
- 5 references maximum. Reference 1 must be the article on which you are commenting.
- 5 authors maximum. Exception: replies can include all original authors of the article.
- Submitted comments are subject to editing and editor review prior to posting.