Alzheimer’s & Dementia 15 (2019) 1333-1347 Featured Article Genome-wide association analysis of dementia and its clinical endophenotypes reveal novel loci associated with Alzheimer’s disease and three causality networks: The GR@ACE project Sonia Moreno-Graua,b, Itziar de Rojasa, Isabel Hernandeza,b, Ines Quintelac, Laura Montrreala, Montserrat Alegreta,b, Bego~ na Hernandez-Olasagarrea, Laura Madridd, Antonio Gonzalez-Perezd, Olalla Maro~ nas , Maitee Rosende-Rocaa, Ana Maule c ona, Liliana Vargasa, Asuncion Lafuentea, Carla Abdelnoura,b, Octavio Rodrıguez-G  omeza,b, Silvia Gila, Miguel Angel Santos-Santosa,  Ana Espinosaa,b, Gemma Ortegaa,b, Angela Sanabriaa,b, Alba Perez-Cord ona, Pilar Ca~nabatea,b, Mariola Morenoa, Silvia Precklera, Susana Ruiza,b, Nuria Aguileraa, Juan Antonio Pinedae, Juan Macıase, Emilio Alarc on-Martına,f, Oscar Sotolongo-Graua, GR@ACE consortium**, DEGESCO consortium**, Alzheimer’s Disease Neuroimaging Initiative, Marta Marquiea, Gemma Monte-Rubioa, Sergi Valeroa,b, Alba Benaquea, Jordi Clarim onb,g, Maria Jesus Bullidob,h,i, Guillermo Garcıa-Ribasj, Pau Pastork, Pascual Sanchez-Juanb,l,  Victoria Alvarez m,n nol-Ripollb,o, Jose Maria Garcıa-Albercap, Jose Luis Royof, , Gerard Pi~ Emilio Francoq, Pablo Mirb,r, Miguel Calerob,s,t, Miguel Medinab,s, Alberto Rabanob,s,u,  Jesus Avila b,v , Carmen Ant  unezw, Luis Miguel Reale,f, Adelina Orellanaa, Angel Carracedoc,x, Marıa Eugenia Saezd, Lluıs Tarragaa,b, Merce Boadaa,b, Agustın Ruiza,b,* a Research Center and Memory clinic Fundacio ACE, Institut Catala de Neurociencies Aplicades, Universitat Internacional de Catalunya, Barcelona, Spain b CIBERNED, Network Center for Biomedical Research in Neurodegenerative Diseases, National Institute of Health Carlos III, Madrid, Spain c Grupo de Medicina Xenomica, Centro Nacional de Genotipado (CEGEN-PRB3-ISCIII). Universidade de Santiago de Compostela, Santiago de Compostela, Spain d CAEBI, Centro Andaluz de Estudios Bioinformaticos, Sevilla, Spain e Unidad Clınica de Enfermedades Infecciosas y Microbiologıa. Hospital Universitario de Valme, Sevilla, Spain f Department of Surgery, Biochemistry and Molecular Biology, School of Medicine, University of Malaga, Malaga, Spain g Memory Unit, Neurology Department and Sant Pau Biomedical Research Institute, Hospital de la Santa Creu i Sant Pau, Universitat Autonoma de Barcelona, Barcelona, Spain h Centro de Biologia Molecular Severo Ochoa (C.S.I.C.-U.A.M.), Universidad Autonoma de Madrid, Madrid, Spain i Instituto de Investigacion Sanitaria “Hospital la Paz” (IdIPaz), Madrid, Spain j Hospital Universitario Ramon y Cajal, Madrid, Spain k Fundaci o per la Recerca Biomedica i Social Mutua Terrassa, and Memory Disorders Unit, Department of Neurology, Hospital Universitari Mutua de Terrassa, University of Barcelona School of Medicine, Terrassa, Spain l Neurology Service ’Marques de Valdecilla’ University Hospital (University of Cantabria and IDIVAL), Santander, Spain m Laboratorio de Genetica Hospital Universitario Central de Asturias, Oviedo, Spain n Instituto de Investigacion Biosanitaria del Principado de Asturias (ISPA), Oviedo, Spain o Unitat Trastorns Cognitius, Hospital Universitari Santa Maria de Lleida, Institut de Recerca Biomedica de Lleida (IRBLLeida), Lleida, Spain The authors have declared that no conflict of interest exists. and/or provided data but did not participate in the analysis or writing of **The full list of members of the GR@ACE consortium and DEGESCO this report. A complete listing of ADNI investigators can be found at consortium were reported in the Acknowledgment section. http://adni.loni.usc.edu/wp-content/uploads/how_to_apply/ADNI_Ack- Alzheimer’s Disease Neuroimaging Initiative: Data used in preparing nowledgement_List.pdf this article were obtained from the Alzheimer’s Disease Neuroimaging *Corresponding author. Tel: 13493.444.73.18; Fax: 13493.410.17.01. Initiative (ADNI) database (adni.loni.usc.edu). As such, the investigators E-mail address: aruiz@fundacioace.org within the ADNI contributed to the design and implementation of ADNI https://doi.org/10.1016/j.jalz.2019.06.4950 1552-5260/ Ó 2019 The Authors. Published by Elsevier Inc. on behalf of the Alzheimer’s Association. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 1334 S. Moreno-Grau et al. / Alzheimer’s & Dementia 15 (2019) 1333-1347 p Alzheimer Research Center & Memory Clinic, Andalusian Institute for Neuroscience, Malaga, Spain q Unidad de Demencias, Servicio de Neurologıa y Neurofisiologıa. Instituto de Biomedicina de Sevilla (IBiS), Hospital Universitario Virgen del Rocıo/CSIC/Universidad de Sevilla, Seville, Spain r Unidad de Trastornos del Movimiento, Servicio de Neurologıa y Neurofisiologıa. Instituto de Biomedicina de Sevilla (IBiS), Hospital Universitario Virgen del Rocıo/CSIC/Universidad de Sevilla, Seville, Spain s CIEN Foundation, Queen Sofia Foundation Alzheimer Center, Madrid, Spain t Instituto de Salud Carlos III (ISCIII), Madrid, Spain u BT-CIEN, Madrid, Spain v Department of Molecular Neuropathology, Centro de Biologıa Molecular “Severo Ochoa” (CBMSO), Consejo Superior de Investigaciones Cientıficas (CSIC)/Universidad Autonoma de Madrid (UAM), Madrid, Spain w Unidad de Demencias, Hospital Clınico Universitario Virgen de la Arrixaca, Murcia, Spain x Fundacion Publica Galega de Medicina Xenomica- CIBERER-IDIS, Santiago de Compostela, Spain Abstract Introduction: Large variability among Alzheimer’s disease (AD) cases might impact genetic dis- coveries and complicate dissection of underlying biological pathways. Methods: Genome Research at Fundacio ACE (GR@ACE) is a genome-wide study of dementia and its clinical endophenotypes, defined based on AD’s clinical certainty and vascular burden. We as- sessed the impact of known AD loci across endophenotypes to generate loci categories. We incorpo- rated gene coexpression data and conducted pathway analysis per category. Finally, to evaluate the effect of heterogeneity in genetic studies, GR@ACE series were meta-analyzed with additional genome-wide association study data sets. Results: We classified known AD loci into three categories, which might reflect the disease clinical heterogeneity. Vascular processes were only detected as a causal mechanism in probable AD. The meta-analysis strategy revealed the ANKRD31-rs4704171 and NDUFAF6-rs10098778 and confirmed SCIMP-rs7225151 and CD33-rs3865444. Discussion: The regulation of vasculature is a prominent causal component of probable AD. GR@ACE meta-analysis revealed novel AD genetic signals, strongly driven by the presence of clin- ical heterogeneity in the AD series. Ó 2019 The Authors. Published by Elsevier Inc. on behalf of the Alzheimer’s Association. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/ 4.0/). Keywords: Alzheimer’s disease; Vascular pathology; Cerebral amyloid angiopathy; GWAS; Biological pathway 1. Background representing the coexistence of both neurodegenerative and vascular components [7]. Despite that, whether there Dementia is an age-related clinical syndrome that devas- are differential biological routes operating under different tates cognitive abilities and interferes in elderly people’s levels of vascular burden in clinical AD patients remains daily activities. Although its incidence is decreasing due to mostly unknown. improvements to public health systems and control of cardio- In the search for the etiology of AD, genetic factors play a vascular risk factors [1], its prevalence is steadily increasing pivotal role. Two forms of the disease can be differentiated ac- due to rising life expectancy of human populations [2]. cording to individual genetic background. The Mendelian Dementia is linked to many underlying pathologies, with form is an uncommon disorder that mainly affects families Alzheimer’s disease (AD) being the most common condi- with early-onset AD (,65 years), whereas the polygenic tion. Clinical AD is a heterogeneous syndrome. Brain au- form is a complex disorder mainly appearing in sporadic cases topsies have shown that roughly 80% of clinical AD with late-onset AD (LOAD) (.65 years). Highly penetrant patients present with brain vascular pathology [3] in addition mutations detected in families with early-onset AD have to the common neuropathological AD hallmarks: amyloid- been pinpointed to three genes: APP [8], PSEN1 [9], and osis, neurofibrillary tangles, and cerebral amyloid angiop- PSEN2 [10], leading to the establishment of the amyloid athy (CAA) [4]. In fact, brain vascular pathology has been hypothesis as a potential causal mechanism for the disease shown to be an important risk factor for AD that accelerates [11]. cognitive decline [5] and lowers the threshold for clinical LOAD heritability falls in the range of 13%‒80% [12,13]. diagnosis of AD [6]. In that context, it has been suggested Although APOE ε4 was the first to be discovered and still that dementia is represented by a gradient of neurodegener- remains the strongest genetic risk factor for AD [14], almost ative and vascular components [7], from pure AD forms, 40 additional genetic variants have been identified [15–17] with a strong neurodegenerative component, to pure vascular using genome-wide association studies (GWAS) and large dementia (VaD) cases and in-between mixed pathologies, S. Moreno-Grau et al. / Alzheimer’s & Dementia 15 (2019) 1333-1347 1335 sequencing projects. Among the biological pathways underly- riched cases. This approach was feasible due to Fundaci o ing genetic hits, the roles of the immune system, cholesterol, ACE’s endorsement of a primary and a secondary etiologic amyloid, and tau metabolism have been highlighted [18]. diagnosis as well as routine follow-up evaluations [21] Despite these, current genetic findings account for 31% of (Supplementary Methods). Using the entire clinical chart LOAD heritability [19], and the biological picture of AD is of each subject, we differentiated five clinical subgroups still poorly understood. Several reasons can explain this. of patients representing the GR@ACE endophenotypes: Among them, the presence of clinical and neuropathological (1) the AD111 endophenotype comprises individuals with heterogeneity between AD cases in genetic studies, as a last clinical diagnosis of probable AD in both primary recently demonstrated [12], might compromise the power to and secondary diagnoses (n 5 1854); (2) the AD11 includes detect genuine genetic associations and decrease the estimates individuals diagnosed with probable AD in the primary diag- of risk attributed to genetic variation. nosis and probable or possible AD in the secondary diag- The neuropathological variability in clinical AD cases com- nosis (n 5 2611); (3) the AD1 encompasses patients prises a wide spectrum, from those with concomitant vascular diagnosed with probable or possible AD either in the pri- brain disease to those with a pure AD phenotype, as previously mary diagnosis or in the secondary diagnosis (n 5 3797); proposed by Viswanathan et al. [7]. This large heterogeneity (4) the VaD1 includes patients diagnosed with VaD or might hamper the identification of functional categories of possible AD in the primary diagnosis (n 5 1168); and (5) genes underlying differential biological routes to dementia the VaD11 comprises patients diagnosed with probable or and might impact AD genetic studies. To gain insight into possible vascular dementia in the primary diagnosis the causality networks behind AD clinical subgroups and to (n 5 373) (Table 1) (Supplementary Fig. 1). Patients with explore their impact in large GWAS, we conducted the VaD were defined according to the Neuroepidemiology Genome Research at Fundacio ACE (GR@ACE) study. This Branch of the National Institute of Neurological Disorders is a GWAS of dementia and its clinical endophenotypes and Stroke and the Association Internationale pour la Re- defined based on AD’s clinical certainty and the burden of cherche et l’Enseignement en Neurosciences criteria [22]. vascular comorbidity. The GR@ACE is a unique genomic Control individuals were recruited from three centers: resource comprising the largest number of dementia cases Fundacio ACE (Barcelona, Spain), Valme University Hospi- diagnosed in a single memory clinic to date. First, we deter- tal (Seville, Spain), and the Spanish National DNA Bank mined whether we could identify categories of known Carlos III (University of Salamanca, Spain) (www. LOAD genes linked to clinical subgroups of AD cases. Next, bancoadn.org). Written informed consent was obtained we explored whether these categories suggested different bio- from all participants. The Ethics and Scientific Committees logical routes. Finally, to assess the impact of these clinical have approved this research protocol (Acta 25/2016. Ethics subgroups of AD cases in GWAS findings, we meta- Committee. H. Clinic i Provincial, Barcelona, Spain). analyzed the GR@ACE data with independent GWAS series. 2.1.2. Replication sample With the objective to successfully replicate novel GWAS 2. Methods findings, we used an independent Spanish sample of 1943 AD cases (mean age 5 79.2; standard deviation 2.1. Subjects [SD] 5 7.6; 66.3% women) and 3016 controls (mean 2.1.1. GR@ACE cohort and phenotype definitions age 5 52.8; SD 5 15.2; 46% women) presenting similar The GR@ACE study comprises 4120 AD cases and 3289 characteristics to the GR@ACE cohort and with available control individuals (Table 1). Cases were recruited from genetic data. All AD cases were examined at a single site, Fundacio ACE, Institut Catala de Neurociencies Aplicades Fundacio ACE, Institut Catala de Neurociencies Aplicades (Catalonia, Spain). Diagnoses were established by a (Catalonia, Spain), and were assessed by applying the multidisciplinary working group, including neurologists, neu- same criteria previously explained. The sample composition ropsychologists, and social workers, according to the Diag- of dementia cases comprised 30.1% of AD111 (n 5 584), nostic and Statistical Manual of Mental Disorders–IV criteria 53.9% of AD11 (n 5 1048), 91.9% of AD1 (n 5 1783), for dementia and to the National Institute on Aging and Alz- 23.0% of VaD1 (n 5 447), and 4.6% of VaD11 (n 5 89) heimer’s Association’s (NIA-AA) 2011 guidelines for defining cases. Control individuals were selected from the Spanish AD [20] (see Supplementary Material). In the present study, we population available at three centers, previously described. considered AD cases as dementia individuals diagnosed with probable or possible AD at any moment of their clinical course. We took advantage of this wide clinical definition to 2.2. GWAS genotyping, quality control, imputation, and refine AD cases. Considering the dementia spectrum pro- statistical analysis posed by Viswanathan et al. [7], we classified Gr@ACE AD patients according to the degree of clinical certainty Participants were genotyped using the Axiom 815K for AD phenotype and the presence of vascular comorbidity, Spanish Biobank Array (Thermo Fisher). Genotyping was from “pure” clinical AD cases to mixed and vascular en- performed in the Spanish National Center for Genotyping 1336 S. Moreno-Grau et al. / Alzheimer’s & Dementia 15 (2019) 1333-1347 Table 1 GR@ACE demographic characteristics and endophenotype definitions Phenotype Primary diagnostic Secondary diagnostic N Mean age 6 SD Women % APOE ε4 % Controls – – 3289 54.3 6 14.4 48.9 21.4 VaD11 VaD Pss AD 373 80.1 6 5.5 54.9 25.0 VaD1 VaD/Pss AD VaD/Pss AD 1168 80.4 6 6.3 65.0 32.8 AD Pr/Pss AD at any time in medical history 4120 79.0 6 7.5 69.6 40.1 AD1 Pr/Pss AD Pr/Pss AD 3797 79.2 6 7.5 70.6 41.2 AD11 Pr AD Pr/Pss AD 2611 78.8 6 7.9 72.8 44.6 AD111 Pr AD Pr AD 1854 79.0 6 8.0 74.6 47.0 Abbreviations: AD, Alzheimer’s disease; SD, standard deviation; VaD, vascular dementia; Pss AD, possible AD; Pr AD, probable AD. (CeGEN, Santiago de Compostela, Spain) (Supplementary GR@ACE GWAS data have been deposited into the Eu- Methods). ropean Genome-phenome Archive (https://ega-archive.org), We removed samples with genotype call rates ,97%, which is hosted by the European Bioinformatics Institute excess heterozygosity, duplicates, samples genetically and the Center for Genomic Regulation, under the accession related to other individuals in the cohort, or sample number EGAS00001003424. mix-up (PIHAT .0.1875). If a sex discrepancy was de- tected, the sample was removed unless the discrepancy 2.3. Genetic exploration of GR@ACE clinical was safely resolved. To detect population outliers of endophenotypes and enrichment analysis non-European ancestry (.6 SD from European popula- With the objective to explore whether different biological tion mean), principal component analysis was conducted routes operate under different levels of vascular burden in using SMARTPCA from EIGENSOFT 6.1.4 (Fig. 1) clinical AD patients, first, we classified GR@ACE dementia (Supplementary Methods). cases to cover the dementia spectrum previously proposed We excluded variants with a call rate ,95% or that by Viswanathan et al. [7] (see Section 2.1). Second, we ex- grossly deviated from Hardy-Weinberg equilibrium in con- tracted the effect (odds ratio [OR]) for known LOAD genetic trols (P value  1 ! 1026); we also excluded markers variants (MAF .1%). See included variants in with a different missing rate between case and control (P Supplementary Table 1. Then, we explored whether previ- value , 5 ! 1024 for the difference) or a minor allele fre- ously identified LOAD variants were more strongly associ- quency (MAF) , 0.01. Imputation was carried out using ated with a specific subgroup of AD patients. We the Haplotype Reference Consortium panel in the Michigan quantified the strength of the association for each variant Imputation Server (https://imputationserver.sph.umich.edu). across endophenotypes, named henceforth global effect Only common markers (MAF .0.01) with a high imputation change. It was calculated as the absolute difference between quality (R2 . 0.30) were selected for downstream analyses. variant OR for extreme endophenotypes (VaD11 vs. The GWAS was performed for GR@ACE as a whole AD111). According to the global effect change and direc- and for each endophenotype, ([N AD111 5 5143]; tion of the enrichment, i.e., from VaD11 to AD111 or [N AD11 5 5900]; [N AD1 5 7086]; [N AD from AD111 to VaD11, we classified LOAD genetic vari- dementia 5 7409]; [N VaD1 5 4487]; and [N ants into three categories. Thus, category A includes variants VaD11 5 3662]). The GWAS was conducted for genotype with an increase in the association effect from VaD11 to dosages using an additive genetic model with PLINK 1.9. AD111 endophenotypes and a global effect change A model including the top four PCs as covariates was used .0.05, and category B includes variants with an increase for the discovery stage because it exhibited the lowest infla- in the association effect from AD111 to VaD11 and a tion and optimal power compared with alternative tested global effect change .0.05. Category C comprises variants models (Supplementary Fig. 2). Further description is pro- not fulfilling criteria for categories A or B (Supplementary vided in Supplementary Methods. Power analysis was per- Table 1). Finally, we assessed the biological pathways un- formed using QUANTO software v1.2.4 [23] to model the derlying each category. We incorporated data from gene co- impact on statistical power at different MAFs and effect expression for each specific gene category using the sizes in available case-control cohorts. Results were de- GeneFriends tool (http://genefriends.org/) and performed picted using the ggplot2 package in R. Analyses were per- pathway analysis of top coexpressed genes using the over- formed for a GWAS experiment that would meet criteria representation enrichment method in WebGestalt (http:// for genome-wide significance (P , 5 ! 1028) and for a www.webgestalt.org/option.php) (see Supplementary replication experiment that would meet P , .05 Methods). With the objective to further explore potential (Supplementary Fig. 3). Calculations were performed additional trends in category C, we performed a specific sub- considering a disease prevalence of 1.73%, according to reg- analysis in this category, widely described in Supplementary isters in the Spanish population. Methods. To validate previous gene classification, which S. Moreno-Grau et al. / Alzheimer’s & Dementia 15 (2019) 1333-1347 1337 Fig. 1. Results of genome-wide association analysis for the GR@ACE data set (n 5 7409). (A) Principal component analysis; (B) principal component analysis centered in European population; (C) QQplot for the discovery model, adjusted for first four PCs; (D) Manhattan plot for genome-wide results. Abbreviations: AFR, African; AMR, Admixed American; EAS, East Asian; EUR, European; SAS, South Asian. strongly determines the pathway analysis results, we con- 2.5. Replication of genome-wide significant findings ducted a stringent subanalysis (see Supplementary Methods and Supplementary Fig. 4). We then explored genome-wide significant (GWS) signals in an independent cohort of Spanish ancestry (N 5 4959). We extracted variants of interest from GWAS data, which were genotyped and processed applying similar methods to those 2.4. Meta-analysis: Data sets and association analysis explained for the GR@ACE study (Section 2.3). Finally, To explore the impact of the different clinical endopheno- meta-analysis including the discovery stage, named stage I, types in GWAS findings, we combined the whole GR@ACE and the replica data set, stage II, was performed as previously GWAS data set, which represents a dementia set of samples described. Results were interpreted according to the Amer- and its endophenotypes with (1) genotype-level data from ican Statistical Association guidelines [25,26]. nine additional GWAS series (N 5 13,826), available through dbGaP (https://www.ncbi.nlm.nih.gov/gap) that 2.6. Biological interpretation of meta-GWAS signals we processed by applying identical quality control and imputation procedures to those described for the GR@ACE Gene expression quantitative trait locus (eQTL) analysis cohort (Supplementary Table 2) and (2) aggregated sum- was conducted to link meta-GWAS top signals to genes. mary statistics available from the International Genomics Markers with moderate-to-high LD (r2  0.6) with the novel Alzheimer’s Project (IGAP) (http://web.pasteur-lille.fr/en/ lead markers were identified using LDlink [27] for European recherche/u744/igap/igap_download.php) [24], including population and were included in this analysis. We used brain IGAP stages I (N final 5 61,571) and IGAP I and II (N (n 5 11) and whole-blood (n 5 1) tissues from the GTEx v7 final 5 81,455) (Supplementary Methods). Meta-analyses repository (https://www.gtexportal.org/home) for mapping were conducted using the inverse variant method in METAL cis-eQTLs (Supplementary Table 3). As an extension of software (https://genome.sph.umich.edu/wiki/METAL). GTEx tissue eQTL mapping, we explored brain eQTLs for The linkage disequilibrium (LD) score calculations, clump- GWS genomic regions using additional databases available ing, and conditional analysis are described in Supplementary via Functional Mapping and Annotation of Genome-Wide Methods. Association Studies (FUMA) [28]. We also performed 1338 S. Moreno-Grau et al. / Alzheimer’s & Dementia 15 (2019) 1333-1347 NOTE. The threshold for genome-wide significance was 5 ! 1028. Position 5 GRCh37/hg19 coordinates Stage I corresponds to the discovery stage; stage II corresponds to the replica; stage I 1 II corresponds functional annotation for GWS markers, chromatin 1.79 ! 1024 1.15 ! 1028 2.32 ! 1028 1.38 ! 1028 9.45 ! 1029 Abbreviations: AD, Alzheimer’s disease; VaD, vascular dementia; chr:bp, chromosome:base pair; A1, minor allele; A2, mayor allele; MAF, minor allele frequency obtained from the GR@ACE study interaction, and gene-based analysis using similar criteria to those previously described by Jansen et al. [17] (see Supplementary Methods). P 2.08 (1.42–3.07) 1.18 (1.11–1.24) 0.94 (0.92–0.96) 1.10 (1.06–1.14) 1.11 (1.07–1.15) 3. Results Stage I 1 II (95% CI) 3.1. GR@ACE genome-wide association study OR After quality control and imputation, the GR@ACE to the meta-analysis of the discovery and the replica. Stage I 1 II for rs117834366 corresponds to the meta-analysis of GR@ACE VaD11 and the entire replica (N 5 4959). study encompassed 7409 unrelated individuals from the .73 .09 .40 .04 .49 P Spanish population and 7.7 million variants (lGC 5 1.03). The APOE-rs429358 marker was the only Stage II (replica) 1.09 (0.66–1.78) 1.10 (0.98–1.25) 0.96 (0.89–1.05) 1.14 (1.00–1.30) 1.07 (0.88–1.30) one to have a GWS association (OR 5 2.27 [2.06–2.50]; P 5 1.25 ! 10262) (Fig. 1). Four additional LOAD (95% CI) variants displayed statistically significant evidence of OR replication (BIN1-rs6733839, MAPT-rs2732703, MS4A2- rs983392, and PICALM-rs10792832) and nine additional 28 2.78 ! 1028 2.54 ! 1028 8.98 ! 1028 1.12 ! 1028 markers presented a consistent direction for the effect 1.91 ! 10 (Supplementary Table 1). MAPT marker association remains significant in APOE ε4 noncarriers, but the effect P size was stable in both strata (Supplementary Table 4). GWAS of clinical endophenotypes showed that 6.03 (3.22–11.2) 1.19 (1.12–1.27) 0.94 (0.91–0.96) 1.10 (1.06–1.14) 1.11 (1.07–1.15) CNTNAP2-rs117834366 was associated with the VaD11 OR (95% CI) endophenotype (Table 2). This marker is in complete linkage equilibrium with CNTNAP2-rs114360492 (r2 5 0), previously reported in GWAS of AD by proxy [17]. See results in Supplementary Results GR@ACE 1 IGAP I & II GR@ACE 1 IGAP I & II (Supplementary Figs. 5 and 6; Supplementary Table 5). AD111 1 IGAP I & II GR@ACE 1 dbGaP 11 3.2. Genetic exploration of GR@ACE clinical Association results for lead single-nucleotide polymorphisms reaching genome-wide significance Discovery stage GR@ACE VaD endophenotypes and enrichment analysis GR@ACE To explore whether clinical AD subgroups, representing Stage I GR@ACE endophenotypes, reflected variations in the underlying biological pathways driving dementia, we 0.011 0.123 0.470 0.126 0.126 MAF classified LOAD genetic variants into three categories. Category A comprised variants strongly related to the purest form of clinical AD (i.e., subjects with probable AD in A1/A2 primary and secondary diagnoses). The most prominent A/G A/G A/G C/T C/T locus of this category was APOE-rs429358 (AD111 OR 5 2.92 [2.60–3.27], P value 5 9.26 ! 10275; VaD11 7:147634891 5:74368254 8:95992020 17:5137047 17:5137047 OR [95% confidence interval] 5 1.27 [1.02–1.59], P (N 5 7409); OR, odds ratio; CI, confidence interval. Position value 5 .04). Other loci included in category A were CR1, Chr:bp BIN1, MEF2C, MS4A2, PICALM, MAPT, and CD33. In *Imputed variant; rs117834366 r2 5 0.68. contrast, category B comprised variants with the strongest TP53INP1/NDUFAF6 effect observed in subjects with AD mixed with vascular disease (SORL1, ADAM10, CASS4, ATP5H, and ACE) (Supplementary Table 1). Further description is provided Near locus ANKRD31 CNTNAP2 in Supplementary Results. Category C comprised a group SCIMP SCIMP of variants with effects in all clinical endophenotypes Genotyped variant. (Fig. 2). Subanalysis for category C is shown in Supplementary Results and Supplementary Table 6. rs117834366* Next, we explored biological pathways for each gene rs10098778y rs4704171y rs7225151y rs7225151y category. Note that the regulation of vasculature develop- Table 2 Marker ment and blood vessel morphogenesis were only detected y for genes in category A, which is more closely related to S. Moreno-Grau et al. / Alzheimer’s & Dementia 15 (2019) 1333-1347 1339 Fig. 2. (A) Enrichment trend per genetic marker and gene category across GR@ACE endophenotypes. (B) Graph centered in effect change range 0–0.5. Effect change per endophenotype 5 Variant Odds ratio–variant null effect; Enrichment trend per category was obtained applying a linear regression using ggplot2 in R. Abbreviations: AD, Alzheimer’s disease; VaD, vascular dementia. pure AD (P 5 2.03 ! 1027, P 5 1.90 ! 1026, respectively) identified 12 LOAD genomic regions reaching GWS. (Table 3). Additional categories indicated immune system CD33-rs3865444, which did not reach GWS in the IGAP pathways (category B, P 5 2.07 ! 1027; category C, meta-analysis, was significantly associated with LOAD P 5 5.77 ! 10215) (Table 3). Finally, with the aim of (OR 5 0.92 [0.89–0.95]; P 5 3.61 ! 1028) validating previous results, we conducted a subanalysis by (Supplementary Fig. 7). Among the top suggestive signals, classifying LOAD genetic variants with more stringent we detected HBEGF-rs4150233 (OR 5 0.92 [0.90–0.95]; classification criteria (Supplementary Methods). Again, P 5 5.10 ! 1028) previously identified by a transethnic APOE, CR1, MEF2C, MS4A2, and PICALM loci were found GWAS [29]. in category A; SORL1 and CASS4 were in category B; and Next, meta-analysis of the whole GR@ACE data set with additional AD loci were in category C (Supplementary IGAP I and II enabled the identification, for the first time, Fig. 4). Regulation of vasculature development was of NDUFAF6-rs10098778 as a GWS signal (OR 5 0.94 exclusively identified as the top pathway in category A [0.91–0.96]; P 5 2.54 ! 1028). When we combined (P 5 2.14 ! 1027) when we restricted the analysis to GR@ACE AD111 endophenotype with IGAP I and II, we include those loci coexpressing with at least 4 LOAD genes also detected SCIMP-rs7225151 (OR 5 1.11 [1.07–1.15]; (Supplementary Table 7). Further replication in cohorts with P 5 1.12 ! 1028) (Table 2) (Supplementary Fig. 8). It available neuropathological data would be recommended. was previously reported as a genome-wide suggestive signal by IGAP [24]. Recently, SCIMP-rs113260531, which is in complete LD with our lead marker (r2 5 1), was associated 3.3. Meta-analysis of GR@ACE study with other data sets with AD [17]. To assess the impact of sample composition in AD GWAS, and look for new AD loci, we first combined 3.4. Replication of genome-wide significant findings the GR@ACE data set with nine additional genomic databases that had genotypic level data available. Subtle Finally, we tested for replication of the new signals in an genomic inflation was detected, mainly explained independent sample of 4959 Spanish individuals. The by polygenicity (lGC 5 1.10; LD score CNTNAP2-rs117834366, detected in the GWAS of intercept 5 1.04). Five regions were associated with GR@ACE VaD11 endophenotype, had a P value of 0.79 LOAD (Fig. 3); of these, four (APOE-rs429358, with a similar effect direction to that reported previously PICALM-rs10792832, MS4A2-rs983392, and BIN1- but strongly deflated in the entire replica sample rs6733839) have been previously linked to AD (OR 5 1.09 [0.66–1.78]; P 5 .79) (Table 3). Analysis of (Supplementary Table 8), and one is a new GWAS the subspecific VaD phenotype in the replica (N 5 89) would finding (ANKDR31-rs4704171; OR 5 1.19 [1.12–1.27]; be highly inaccurate. P 5 2.78 ! 1028) (Table 2). In the exploration of meta-GWAS findings, we observed Then, we conducted a genome-wide meta-analysis that the ANKDR31-rs4704171-C marker increased the risk combining the GR@ACE study with IGAP stage I. We of AD (OR 5 1.10 [0.98–1.25]; P 5 .09; power 5 33%). 1340 S. Moreno-Grau et al. / Alzheimer’s & Dementia 15 (2019) 1333-1347 Table 3 11% to a risk increase of 5% (OR 5 0.96 [0.89–1.05]; Top ten biological pathways per gene category P 5 .40; power 5 16%). This signal had a Gene ontology Top 10 coregulated pathways P 5 2.32 ! 1028 in the final meta-analysis, including the pathway for category A P value whole GR@ACE data set, the replica, and IGAP I and II GO:1901342 Regulation of vasculature 2.03 ! 1027 (n 5 91,373) (Supplementary Fig. 9). development SCIMP-rs7225151 showed a risk effect in the whole GO:0060326 Cell chemotaxis 2.59 ! 1027 replica. Limits of the interval were consistent with a positive GO:0048771 Tissue remodeling 6.77 ! 1027 GO:0050865 Regulation of cell activation 1.14 ! 1026 association (OR 5 1.14 [1.01–1.29]; P 5 .047; GO:0007159 Leukocyte cell-cell adhesion 1.21 ! 1026 power 5 39%; n cases 5 1943). In the AD111 endopheno- GO:0048514 Blood vessel morphogenesis 1.90 ! 1026 type, the marker presented the same positive risk effect GO:0003012 Muscle system process 2.54 ! 1026 direction (OR 5 1.07 [0.88–1.30]), although had P 5 .49, GO:0002764 Immune response-regulating 3.48 ! 1026 which could be mainly explained by a reduction of the signaling pathway GO:0032103 Positive regulation of response 3.91 ! 1026 sample size (power 5 19%; n cases 5 584). Our results to external stimulus for both meta-analyses, the final meta-analysis of the whole GO:0010959 Regulation of metal ion transport 4.36 ! 1026 GR@ACE data set (n 5 91,373) and the final meta-analysis Gene ontology Top 10 coregulated pathways of GR@ACE AD111 endophenotype, (n 5 85,055) were pathway for category B P value compatible with a potential effect of this marker in AD (Table 3). See forest plots in Supplementary Fig. 9. GO:0009620 Response to fungus 2.02 ! 1027 GO:0050886 Endocrine process 3.58 ! 1027 GO:0002443 Leukocyte mediated immunity 5.47 ! 1027 3.5. Biological interpretation of meta-GWAS signals GO:0050865 Regulation of cell activation 1.52 ! 1025 GO:0031349 Positive regulation of 8.42 ! 1025 To identify candidate genes and potential causal variants defense response within novel meta-GWAS regions, we conducted cis-eQTL GO:0032103 Positive regulation of 1.00 ! 1024 response to external stimulus mapping. The rs2335107 marker located in the ANKRD31 GO:0002250 Adaptive immune response 1.30 ! 1024 locus (chr5:74,451,693) was associated with the cortical GO:0098542 Defense response to 2.00 ! 1024 expression of the long noncoding RNA (lncRNA) CTD- other organism 2235C13.3 (P 5 1.26 ! 1025). This variant is located GO:1901568 Fatty acid derivative 2.24 ! 1024 83.4 kb from the meta-GWAS lead single nucleotide metabolic process GO:0050900 Leukocyte migration 2.57 ! 1024 polymorphism (rs4704171, chr5:74,368,254), and both are in complete LD (r2 5 1). The CTD-2235C13.3 gene is Gene ontology Top 10 coregulated pathways located 1.6 kb from the HMGCR locus, and its function is pathway for category C P value unknown. The NDUFAF6 region mapped for NDUFAF6 GO:0007159 Leukocyte cell-cell adhesion 5.77 ! 10215 RNA cortical expression (P 5 5.56 ! 1026) and for 4.37 ! 10214 GO:0050865 Regulation of cell activation TP53INP1 RNA blood expression (P 5 1.17 ! 10210). GO:0002764 Immune response-regulating 1.33 ! 10212 signaling pathway Finally, rs73976325 (chr17:5,123,227), located in the GO:0002253 Activation of immune response 3.96 ! 10212 SCIMP locus, to 13.8 kb from the meta-GWAS top signal GO:0002443 Leukocyte mediated immunity 4.34 ! 10212 (rs7225151, chr17:5,137,047), mapped to brain cis-acting GO:0002274 Myeloid leukocyte activation 7.78 ! 10212 eQTL for AC012146.1 lincRNA (P 5 2.15 ! 1027). Two GO:0002250 Adaptive immune response 1.24 ! 10211 additional markers were pinpointed to blood eQTLs, GO:0002263 Cell activation involved 7.07 ! 10211 in immune response SCIMP-rs6502851 (P 5 3.89 ! 10208) and RABEP- GO:0022407 Regulation of cell-cell 5.40 ! 1029 rs59277121 (P 5 3.89 ! 1028) (Supplementary Table 3). adhesion In an additional prioritization strategy, combining informa- GO:0070661 Leukocyte proliferation 1.22 ! 1028 tion from positional mapping, eQTL, chromatin interaction, and gene-based genome-wide association analysis via FUMA, results pointed to ANKDR31 and POLK for the Although the expected effect is in line with previous data, the ANKDR31 genomic region, as well as NDUFAF6 and precision of the estimate in this replica differs, ranging from TP53INP1 for the NDUFAF6 region (Supplementary a 2% decrease (a small negative association) to a 25% Table 9). Further description is provided in Supplementary increase. Of note, the result emerging from the Information and Supplementary Table 9. meta-analysis of the replica with the discovery sample (n 5 26,194) is compatible with its potential role in dementia 4. Discussion (OR 5 1.18 [1.11–1.24]; P 5 1.15 ! 1028) (Table 3). See forest plot in Supplementary Fig. 9. We present a comprehensive GWAS of AD dementia We observed a similar effect direction in the NDUFAF6- cases. This represents the first pilot study exploring the rs10098778 marker to that reported in the discovery stage, genetics and underlying biological pathways of subgroups with interval estimates ranging from a risk decrease of of patients with AD, defined based on vascular burden. We S. Moreno-Grau et al. / Alzheimer’s & Dementia 15 (2019) 1333-1347 1341 Fig. 3. A) Results of genome-wide association analysis for GR@ACE meta-analysis with nine additional databases (n 5 21,235). (B) QQplot. (C) Associations of the region centered on rs4704171 located in the ANKRD31 locus and containing the HMGCR locus. showed differential biological routes underlying clinical en- classification of AD has evolved. In 2018, the NIA-AA pro- dophenotypes and demonstrated how these differential sub- posed a novel research framework for the biological classi- groups of patients with AD impact GWAS discoveries. The fication of AD based on the presence in vivo of biomarker GR@ACE study represents a unique genomic resource evidences for amyloid (A), tau (T), and neurodegeneration because all affected cases were diagnosed in a single mem- (N), as surrogate of the neuropathological state of an individ- ory clinic using the same screening and diagnostic tech- ual [31]. The AT(N) biomarker system allows the classifica- niques. This might limit potential sources of clinical tion of individuals into three categories: those with a normal variation between study participants, recently demonstrated biomarker profile, those with biomarkers compatible with in a large meta-GWAS [12]. AD change, and those with biomarkers compatible with Based on the increase in evidence suggesting that non-AD pathological changes [31]. Using the NIA-AA vascular brain pathology can act concomitantly with AD approach, the generation of subgroups of AD patients to produce a more rapid cognitive decline [5], we explored considering vascular pathology would be unfeasible, as the effect of known LOAD loci across different levels of nowadays the ATN classification does not integrate mea- vascular burden in dementia patients using only clinical def- sures of vascular dysfunction. Taking into account that initions. Our basic idea was to dissect, from a molecular most of dementia cases are caused by mixed pathologies, point of view, the model previously proposed by Viswana- the current system seems deeply incomplete to study the than et al. [7]. We observed three categories of loci, which probable interaction between neurodegeneration and might reflect the disease’s clinical heterogeneity, from vascular dysfunction. This idea has also been claimed by vascular and mixed forms to a “purer” AD phenotype. others [32,33]. Thus, we encourage other groups to Intriguingly, we detected vascular processes to be the main contrast the proposed loci classification, which was based causal mechanism in clinically pure AD and found the im- on the GR@ACE clinical endophenotypes, but using well- mune system pervasively detected across the three cate- powered GWAS cohorts with available neuropathological gories. Although both pathways have been previously data. associated with LOAD by network analysis [30], this is the Silent changes occur in brain microvasculature during first study to show that the association with the vascular sys- AD progression. In fact, CAA is a well-recognized AD path- tem is conducted by AD-specific clinical subgroup. ological feature characterized by the accumulation of amy- It should be noted that the present study used clinical loid proteins in the walls of small cerebral vessels. CAA criteria to define the AD cases [20], but recently the has been proposed to compromise the perivascular drainage 1342 S. Moreno-Grau et al. / Alzheimer’s & Dementia 15 (2019) 1333-1347 of Ab from the brain to the peripheral system [34]. Almost empower genetic studies to detect genes associated with spe- all AD brains harbor CAA pathology to some extent, cific disease axes. although in vivo most CAA cases remain undiagnosed, An alternative strategy is taking advantage of clinical het- even using the validated Boston criteria [35]. Mendelian mu- erogeneity. Specifically, heterogeneity might play a dual role tations of the APP gene have been found in CAA and AD in genetic studies. Although it might decrease the power to [8,36]. APOE ε4 and CR1 have been associated with an detect genes associated with more specific clinical sub- increased risk of CAA [37,38]. In particular, distinct AD groups, incorporating detailed clinical AD definitions can loci have been associated with capillary and noncapillary also promote identifying genes shared with other conditions CAA [39]. Between them, APOE ε4 was strongly related or copathologies such as small vessel disease (SVD). In fact, to capillary CAA [39]. These links make it conceivable this was the case for the ATP5H locus, which was previously that a potential genetic overlap exists between CAA and found to be associated with AD [41] and more recently found AD and suggest that CAA pathology could represents an un- in relation to SVD [42]. We think that the same could derlying process for AD. In that context, we think that apply to the ANKRD31 finding. ANKRD31 encodes a intrinsic alterations to the vasculature could contribute to protein containing ankyrin repeats, which is linked with disease pathogenesis in more pure forms of AD, explaining neurodevelopmental disorders [43]. Of note, ANKRD31 our results. Conversely, in AD individuals with evident cere- GWAS signal mapped to the brain eQTL of a lncRNA, brovascular lesions comprising mixed forms, the additional located 1.6 kb from the HMGCR locus and residing in the role of cardiovascular risk factors such as hypertension, COL4A3BP gene. The HMGCR locus is one of the most atherosclerosis, or arteriosclerosis should be considered, as important coregulators of cholesterol biosynthesis, and it is these could point to a systemic pathological state leading the therapeutic target of statins. The COL4A3BP gene is to vascular damage and dementia. This would agree with involved in lipid transport [44]. Several studies have linked the limited genetic correlation between neurodegenerative HMGCR polymorphisms and AD risk or age at onset for and other neurologic disorders [12], as well as with the re- AD [45], and the cholesterol pathway has been identified sults obtained from heterochronic parabionts in aging to be a biological route shared between AD and SVD. models [40]. Interestingly, markers in the POLK locus, associated by Understanding the role of vasculature pathology in AD gene-based genome-wide association analysis on this study seems to be a pertinent step. In that scenario, CAA would and located in the same disequilibrium block of ANKDR31 be a key AD hallmark. CAA represents the unique identified (Fig. 3), jointly conferred risk for AD and plasma levels of link between the vascular and amyloid hypotheses, but it has LDL [46]. Considering prior findings, our results are been completely neglected in the original hypothesis consistent with this genomic region having a role in mixed formulation. dementia. The reported genetic signal should be considered From a clinical point of view, placing each patient some- a highly probable finding, although independent replications where along the disease spectrum proposed by Viswanathan are still required. et al. [7] is complex. A deep understanding of heterogeneity In the present work, NDUFAF6 signals reached GWS for in AD seems necessary to design better genetic studies, the first time. This finding presented the same effect direc- which must drive the discovery of novel loci and, ultimately, tion in the independent sample (power 5 16%) and remains innovative targets for AD therapies. In this study, we as GWS after the final meta-analysis. Our lead GWAS explored how clinical heterogeneity might impact GWAS marker is in high LD with NDUFAF6-rs4735340, the top findings by integrating distinct GWAS data sets with either suggestive signal reported by Kunkle et al. [18] at this region the GR@ACE cohort as a whole or its endophenotypes. (r2 5 0.95, for Utah residents of European ancestry popula- We found several new GWS signals that seem strongly tion). Despite that, there are subtle differences in LD esti- dependent on the sample composition. For example, after mates for the Iberian population (r2 5 0.87), suggesting combining IGAP stages I and II with the entire GR@ACE that the genetic architecture of the Spanish population could data set, we identified genetic signals in the NDUFAF6 be helping to pinpoint the region of interest. This region is in genomic region but not in the SCIMP region. When this ex- the close vicinity of TP53INP1, previously associated with ercise was conducted using GR@ACE endophenotypes, the AD by a gene-based approach [47]. TP53INP1 is involved SCIMP signal was detected using the clinically “pure” AD in mitochondrial function. Considering these findings and GR@ACE endophenotype. It should be noted that the power results emerging from eQTL analysis, it would be feasible to detect SCIMP signal in the meta-analysis with that a regulatory element for NUDFAF6 resides upstream GRA@ACE dementia and GR@ACE AD111 was 75 and of the TP53INP1 locus. We also detected that the SCIMP 70%, respectively. We tried to replicate this finding in a purer signal was mainly conducted by a specific group of AD AD data set without clinical mixed dementia cases, but the cases. This signal was reported to be a suggestive signal available number of clinical AD111 cases might be by IGAP [24], and a proxy of it (SCIMP-rs113260531) compromising the statistical power to replicate (N reached GWS recently [17]. SCIMP regions has been cases 5 584; power 5 19%). Despite that, we think that us- involved in several eQTLs, from uncharacterized cortical ing specific clinical subgroups of the AD population might lncRNA to blood eQTLs in SCIMP and RABEP1 loci, both S. Moreno-Grau et al. / Alzheimer’s & Dementia 15 (2019) 1333-1347 1343 associated with immune system function [48,49]. The CD33 subsets of patients within the disease spectrum and promote locus remains a controversial LOAD locus because large novel therapy targets for AD. meta-GWAS were unable to replicate this signal [24], but here it reached GWS. We previously proposed that cryptic Acknowledgments population substructure could explain the divergent observa- tions for this locus [50]. The authors would like to thank patients and controls who Note that the lack of definitive neuropathological data for participated in this project. The Genome Research @ AD cases is a severe limitation of the present study. Clinical Fundacio ACE project (GR@ACE) is supported by definitions have important uncertainties, and diagnosis mis- Fundacion bancaria “La Caixa”, Grifols SA, Fundaci o classifications sometimes occur. Hence, some AD individ- ACE, and ISCIII (Ministry of Health, Spain). They also uals included in enriched AD endophenotypes may present want to thank the private sponsors who support the basic concomitant vascular brain disease. The generation of large and clinical projects of our institution (Piramal AG, Labora- histopathological GWAS cohorts with associated quantita- torios Echevarne, Araclon Biotech S.A., and Fundaci o tive data on each pathological hallmark is the ultimate solu- ACE). They are indebted to the Trinitat Port-Carbo legacy tion to tackling the intrinsic heterogeneity in AD. and her family for their support of Fundacio ACE research Unfortunately, there are few examples of neuropathological programs. Fundacio ACE is a participating center in the cohorts: only one GWAS has investigated the genetics of Dementia Genetics Spanish Consortium (DEGESCO). CAA, with APOE being the unique GWS signal [37]. In A.R. and M.B. receive support from the European Union/ this study, a small number of AD cases evolved to vascular EFPIA Innovative Medicines Initiative Joint undertaking dementia during follow-up. Large cross-sectional clinical ADAPTED and MOPEAD projects (grant numbers GWAS cannot control diagnostic changes occurring in clin- 115975 and 115985, respectively). M.B. and A.R. are also ical practice. Clinical diagnosis is a dynamic variable, so un- supported by national grants PI13/02434, PI16/01861, and derstanding the genetic profiles of subgroups of patients PI17/01474. Accion Estrategica en Salud is integrated into evolving to other pathologies would provide powerful infor- the Spanish National R 1 D 1 I Plan and funded by ISCIII mation. (Instituto de Salud Carlos III)-Subdireccion General de It should be considered that there is a limitation in Evaluacion and the Fondo Europeo de Desarrollo Regional reducing the sample size by splitting the cohort into different (FEDER- “Una manera de Hacer Europa”). L.M.R. is endodophenotypes, instead of combining them. Despite that, supported by Consejerıa de Salud de la Junta de Andalucıa spanning the spectrum of dementia individuals to generate (grant PI-0001/2017). Control samples and data from clinical endophenotypes provided us a versatile design, patients included in this study were provided in part by the which let us explore the effect of heterogeneity in GWAS National DNA Bank Carlos III (www.bancoadn.org, and replicate the main findings of pathway analysis using University of Salamanca, Spain) and Hospital Universitario an alternative strategy. The limited number of VaD cases Virgen de Valme (Sevilla, Spain); they were processed after in subgroup analysis limits the accuracy of gene categoriza- standard operating procedures with the appropriate approval tion and pathway analysis. Finally, the exact effector genes of the Ethical and Scientific Committee. The present work for LOAD genetic findings remain unclear. This is a severe was performed as part of the Biochemistry, Molecular limitation to pathway analysis that can only be circumvented Biology, and Biomedicine doctoral program of S. Moreno- by isolating the causative mutations. Independent replication Grau at Universitat Autonoma de Barcelona (Barcelona, will be needed to corroborate our new GWS signals. In that Spain). sense, the selection of specific patient groups might lead to Data collection and sharing for this project was partially successful replication studies. funded by the Alzheimer’s Disease Neuroimaging Initiative The assessment of heterogeneity has important implica- (ADNI) (National Institutes of Health grant U01 AG024904) tions for gene discovery, the development of treatments, and DOD ADNI (Department of Defense award number and their appropriate use in individual patients. The W81XWH-12-2-0012). The ADNI is funded by the National GR@ACE cohort provides useful genomic information, as Institute on Aging and the National Institute of Biomedical it accounts for potential sources of variability and contains Imaging and Bioengineering, as well as through generous different subgroups of cases. This enabled us to analyze contributions from the following: AbbVie; the Alzheimer’s the LOAD genetic landscape in terms of clinical endopheno- Association; the Alzheimer’s Drug Discovery Foundation; types. Our efforts to disentangle the mechanistic pathways Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers operating under clinical subgroups of patients revealed that Squibb Company; CereSpir, Inc.; Cogstate; Eisai Inc.; vasculature regulation may be an essential part of the caus- Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroIm- ative mechanism of LOAD. Finally, our exploration of AD mun; F. Hoffmann-La Roche Ltd and its affiliated company genetics highlights the relevance of sample composition in Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; genetic discoveries. Considering sample composition in Janssen Alzheimer Immunotherapy Research & Develop- the design of genetic studies might lead to the identification ment, LLC.; Johnson & Johnson Pharmaceutical Research of genetic profiles, which can help clinicians distinguish & Development LLC.; Lumosity; Lundbeck; Merck & 1344 S. Moreno-Grau et al. / Alzheimer’s & Dementia 15 (2019) 1333-1347 Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Diabetes y Enfermedades Metabolicas Asociadas (CIBER- Research; Neurotrack Technologies; Novartis Pharmaceuti- DEM); CIBERDEM is an ISCIII Project. cals Corporation; Pfizer Inc.; Piramal Imaging; Servier; The ROS/MAP study data were provided by the Rush Alz- Takeda Pharmaceutical Company; and Transition Therapeu- heimer’s Disease Center, Rush University Medical Center, tics. The Canadian Institutes of Health Research provides Chicago. Data collection was supported through funding funds to support ADNI clinical sites in Canada. Private by NIA grants P30AG10161, R01AG15819, sector contributions are facilitated by the Foundation for R01AG17917, R01AG30146, R01AG36836, U01AG32984 the National Institutes of Health (www.fnih.org). The and U01AG46152, the Illinois Department of Public Health, grantee organization is the Northern California Institute and the Translational Genomics Research Institute. for Research and Education, and the study was coordinated The TGEN study was supported by Kronos Life Science by the Alzheimer’s Therapeutic Research Institute at the Laboratories, the National Institute on Aging (Arizona Alz- University of Southern California. ADNI data are dissemi- heimer’s Disease Center grants P30 AG19610 and RO1 nated by the Laboratory for NeuroImaging at the University AG023193, the Mayo Clinic Alzheimer’s Disease Center of Southern California. grant P50 AG16574, and the Intramural Research Program), The AddNeuroMed data are from a public-private partner- the National Alzheimer’s Coordinating Center (U01 ship supported by EFPIA companies and SMEs as part of In- AG016976), and the state of Arizona. noMed (Innovative Medicines in Europe), an integrated The results published here are in part based on the data ob- project funded by the European Union of the Sixth Frame- tained from the AMP-AD Knowledge Portal accessed at work program priority FP6-2004-LIFESCIHEALTH-5. https://doi.org/10.7303/syn2580853. Clinical leads responsible for data collection are Iwona The authors thank the International Genomics of Alz- K1oszewska (Lodz), Simon Lovestone (London), Patrizia heimer’s Project (IGAP) for providing summary result Mecocci (Perugia), Hilkka Soininen (Kuopio), Magda Tso- data for these analyses. The investigators within IGAP laki (Thessaloniki), and Bruno Vellas (Toulouse). Imaging contributed to the design and implementation of IGAP leads are Andy Simmons (London), Lars-Olad Wahlund and/or provided data but did not participate in analysis (Stockholm), and Christian Spenger (Zurich). Bioinformat- or writing of this report. IGAP was made possible by ics leads are Richard Dobson (London) and Stephen New- the generous participation of the control subjects, the pa- house (London). tients, and their families. The i–Select chip was funded by Funding support for the Alzheimer’s Disease Genetics Con- the French National Foundation on Alzheimer’s disease sortium (ADGC) was provided through the NIA Division of and related disorders. European Alzheimer’s Disease Neuroscience (U01-AG032984). Initiative (EADI) was supported by the LABEX (labora- The genotypic and associated phenotypic data used tory of excellence program investment for the future) in the study “Multi-Site Collaborative Study for DISTALZ grant, Inserm, Institut Pasteur de Lille, Uni- Genotype-Phenotype Associations in Alzheimer’s Disease versite de Lille 2, and the Lille University Hospital. (GenADA)” were provided by GlaxoSmithKline, R&D GERAD was supported by the Medical Research Council Limited. The data sets used for the analyses described (grant no. 503480), Alzheimer’s Research UK (grant no. in this manuscript were obtained from dbGaP at 503176), the Wellcome Trust (grant no. 082604/2/07/Z), http://www.ncbi.nlm.nih.gov/gap through dbGaP accession and German Federal Ministry of Education and Research number phs000219. (BMBF): Competence Network Dementia (CND) grant The Mayo Clinic Alzheimer’s Disease Genetic Studies, led no. 01GI0102, 01GI0711, 01GI0420. CHARGE was by Dr. Nil€ufer Ertekin-Taner and Dr. Steven G. Younkin at partly supported by the NIH/NIA grant R01 AG033193 the Mayo Clinic in Jacksonville, FL, used samples and the NIA AG081220 and AGES contract N01–AG– from the Mayo Clinic Study of Aging, the Mayo Clinic 12100, the NHLBI grant R01 HL105756, the Icelandic Alzheimer’s Disease Research Center, and the Mayo Heart Association, and the Erasmus Medical Center and Clinic Brain Bank. Data collection was supported through Erasmus University. ADGC was supported by the NIH/ funding by NIA grants P50 AG016574, R01 AG032990, NIA grants: U01 AG032984, U24 AG021886, U01 U01 AG046139, R01 AG018023, U01 AG006576, U01 AG016976, and the Alzheimer’s Association grant AG006786, R01 AG025711, R01 AG017216, and R01 ADGC–10–196728. AG003949, NINDS grant R01 NS080820, the CurePSP The GR@ACE study group: Abdelnour C1,2, Aguilera N1, Foundation, and support from the Mayo Foundation. Alarcon E1,3, Alegret M1,2, Benaque A1, Boada M1,2, Buen- The Neocodex-Murcia study was funded by the Fundacion dia M1, Ca~nabate P1,2, Carracedo A4,5, Corbaton A6, de Ro- Alzheimur (Murcia), the Ministerio de Educacion y Ciencia jas I1, Diego S1, Espinosa A1,2, Gailhajenet A1, Garcıa (Gobierno de Espa~ na), Corporacion Tecnol ogica de Gonzalez P1, Gil S1, Guitart M1, Gonzalez Perez A7, Andalucıa, Agencia IDEA (Consejerıa de Innovacion, Junta Hernandez I1,2, Ibarria, M1, Lafuente A1, Macias J8, de Andalucıa), the Diabetes Research Laboratory, and the Maro~nas O4, Martın E1, Martınez MT6, Marquie M1, Biomedical Research Foundation. University Hospital Mauleon A1, Monte-Rubio G1, Montrreal L1, Moreno- Clınico San Carlos has been supported by CIBER de Grau S1,2, Moreno M1, Orellana A1, Ortega G1,2, Pancho S. Moreno-Grau et al. / Alzheimer’s & Dementia 15 (2019) 1333-1347 1345 A1, Peleja E1, Perez-Cordon A1, Pineda JA8, Preckler S1, Catalunya. Barcelona, Spain; 4. Dep. of Surgery, Biochem- Quintela I3, Real LM3,8, Rodrıguez-G omez O1,2, Rosende- istry and Molecular Biology, School of Medicine. Univer- Roca M , Ruiz A , Ruiz S , Saez ME7, Sanabria A1,2, 1 1,2 1,2 sity of Malaga. Malaga, Spain; 5. Fundacio per la Santos-Santos MA1, Serrano-Rios M6, Sotolongo-Grau O1, Recerca Biomedica i Social Mutua Terrassa, and Memory Tarraga L1,2, Valero S1,2, Vargas L1 (1. Research Center Disorders Unit, Department of Neurology, Hospital Uni- and Memory clinic Fundaci o ACE. Institut Catala de Neuro- versitari Mutua de Terrassa, University of Barcelona ciencies Aplicades. Universitat Internacional de Catalunya. School of Medicine, Terrassa, Barcelona, Spain; 6. Labora- Barcelona, Spain; 2. CIBERNED, Center for Networked torio de Genetica Hospital Universitario Central de Astu- Biomedical Research on Neurodegenerative Diseases, Na- rias, Oviedo; 7. Instituto de Investigacion Biosanitaria del tional Institute of Health Carlos III, Ministry of Economy Principado de Asturias (ISPA); 8. Department of and Competitiveness, Spain; 3. Dep. of Surgery, Biochem- Neurology, Hospital Universitario Son Espases, Palma, istry and Molecular Biology, School of Medicine. University Spain; 9. Unidad de Demencias. Hospital Clınico Universi- of Malaga. Malaga, Spain; 4. Grupo de Medicina Xen omica, tario Virgen de la Arrixaca; 10. Servei de Neurologia, Hos- Centro Nacional de Genotipado (CEGEN-PRB3-ISCIII). pital Universitari i Politecnic La Fe; 11. Unidad de Universidade de Santiago de Compostela, Santiago de Com- Demencias, Servicio de Neurologıa y Neurofisiologıa. In- postela, Spain; 5. Fundaci on P ublica Galega de Medicina stituto de Biomedicina de Sevilla (IBiS), Hospital Univer- Xen omica- CIBERER-IDIS, Santiago de Compostela, sitario Virgen del Rocıo/CSIC/Universidad de Sevilla, Spain; 6. Centro de Investigaci on Biomedica en Red de Dia- Seville, Spain; 12. Memory Unit, Neurology Department betes y Enfermedades Metab olicas Asociadas, CIBERDEM, and Sant Pau Biomedical Research Institute, Hospital de Spain, Hospital Clınico San Carlos, Madrid, Spain; 7. la Santa Creu i Sant Pau, Universitat Autonoma de Barce- CAEBI. Centro Andaluz de Estudios Bioinformaticos., Sev- lona, Barcelona, Spain; 13. Centro de Biologia Molecular illa, Spain; 8. Unidad Clınica de Enfermedades Infecciosas y Severo Ochoa (C.S.I.C.-U.A.M.), Universidad Autonoma Microbiologıa. Hospital Universitario de Valme, Sevilla, de Madrid, Madrid, Spain; 14. Instituto de Investigacion Spain) Sanitaria ‘Hospital la Paz’ (IdIPaz), Madrid, Spain; 15. DEGESCO consortium: Adarmes-G omez AD1,2, Alarcon- CIEN Foundation, Queen Sofia Foundation Alzheimer 3,4  5  Martın E , Alvarez I , Alvarez V , Amer-Ferrer G8, An- 6,7 Center, Madrid, Spain; 16. Instituto de Salud Carlos III (IS- tequera M9, Ant unez C9, Baquero M10, Bernal M11, Blesa CIII), Madrid, Spain; 17. Hospital Universitario Ram on y R , Boada M2,3, Buiza-Rueda D1,2, Bullido MJ2,13,14, 2,12 Cajal; Madrid, Spain; 18. BIOMICs, Paıs Vasco; Centro Burguera JA10, Calero M2,15,16, Carrillo F1,2, Carri on-Claro de Investigacion Lascaray. Universidad del Paıs Vasco M1,2, Casajeros MJ17, Clarim on J2,12, Cruz-Gamero JM4, UPV/EHU; 19. Neurology Service, Hospital Universitario de Pancorbo MM18, de Rojas I3, del Ser T14, Diez-Fairen La Paz (UAM), Madrid, Spain; 20. Alzheimer Research M5, Fortea J2,12, Franco E11, Frank-Garcıa A2,14,19, Center & Memory Clinic. Andalusian Institute for Neuro- Garcıa-Alberca JM20, Garcia Madrona S16, Garcia-Ribas science. Malaga, Spain; 21. Neurology Service, Marques G16, G omez-Garre P1,2, Hernandez I2,3, Hevilla S20, Jesus de Valdecilla University Hospital (University of Cantabria S , Labrador Espinosa MA1,2, Lage C2,21, Legaz A9, 1,2 and IDIVAL), Santander, Spain; 22. Hospital Donostia de Lle o A2,12, Lopez de Munain A22, L opez-Garcıa S2,21, Ma- San Sebastıan; 23. Fundacion para la Formacion e Inves- cias D , Manzanares S , Marın M11, Marın-Mu~ 1,2 9,23 noz J9, tigacion Sanitarias de la Region de Murcia; 24. Servicio 20 3 2,13,19 Marın T , Marquie M , Martın Montes A , Martınez de Neurologıa -Hospital de Cabue~nes-Gijon; 25. Centro B9, Martınez C7,24, Martınez V9, Martınez-Lage Alvarez  de Investigacion y Terapias Avanzadas. Fundacion CITA- P25, Medina M2,14, Mendioroz Iriarte M26, Menendez- alzheimer; 26. Navarrabiomed; 27. Servicio de Neurologıa Gonzalez M7,27, Mir P1,2, Molinuevo JL28, Monte-Rubio -Hospital Universitario Central de Asturias, Oviedo; 28. G3, Montrreal L3, Moreno-Grau S2,3, Orellana A3, Pastor Barcelona beta Brain Research Center – Fundacio Pasqual AB15, Pastor P5, Perez Tur J2,29,30, Peri~ nan-Tocino T1,2, Pi- Maragall; 29. Unitat de Genetica Molecular. Institut de ~ 2,31 nol Ripoll G , Rabano A 2,15,32 , Real de As ua D33, Ro- Biomedicina de Valencia-CSIC; 30. Unidad Mixta de Neu- drigo S , Rodrıguez-Rodrıguez E , Royo JL4, Ruiz 11 2,21 rologia Genetica. Instituto de Investigacion Sanitaria La A2,3, Sanchez del Valle Dıaz R34, Sanchez-Juan P2,21, Fe; 31. Unitat Trastorns Cognitius, Hospital Universitari Sastre I2,13, Sotolongo-Grau O3, Tarraga L2,3, Valero Santa Maria de Lleida, Institut de Recerca Biomedica de S2,3, Vicente MP9, Vivancos L9 (1. Unidad de Trastornos Lleida (IRBLLeida), Lleida, Espa~na. 32. BT-CIEN; 33. del Movimiento, Servicio de Neurologıa y Neurofisiologıa. Hospital Universitario La Princesa, Madrid, Spain; 34. Instituto de Biomedicina de Sevilla (IBiS), Hospital Uni- Hospital Clınic Barcelona). versitario Virgen del Rocıo/CSIC/Universidad de Sevilla, Seville, Spain; 2. CIBERNED, Network Center for Biomedical Research in Neurodegenerative Diseases, Na- Supplementary Data tional Institute of Health Carlos III, Spain; 3. Research Center and Memory clinic Fundaci o ACE. Institut Catala Supplementary data related to this article can be found at de Neurociencies Aplicades. Universitat Internacional de https://doi.org/10.1016/j.jalz.2019.06.4950. 1346 S. Moreno-Grau et al. / Alzheimer’s & Dementia 15 (2019) 1333-1347 cursor protein gene with familial Alzheimer’s disease. Nature 1991; RESEARCH IN CONTEXT 349:704–6. [9] Sherrington R, Rogaev EI, Liang Y, Rogaeva EA, Levesque G, Ikeda M, et al. Cloning of a gene bearing missense mutations in early-onset familial Alzheimer’s disease. Nature 1995; 1. Systematic review: Recent literature search, con- 375:754–60. ducted using PubMed database, revealed that the [10] Sherrington R, Froelich S, Sorbi S, Campion D, Chi H, presence of Alzheimer’s disease (AD) with vascular Rogaeva EA, et al. Alzheimer’s disease associated with mutations brain pathology is the most common form of demen- in presenilin 2 is rare and variably penetrant. Hum Mol Genet 1996;5:985–8. tia. Regardless, it is unclear whether subgroups of [11] Hardy JA, Higgins GA. Alzheimer’s disease: the amyloid cascade hy- AD patients present differential genetic profiles or pothesis. Science 1992;256:184–5. develop dementia through differential biological [12] Brainstorm Consortium V, Anttila V, Bulik-Sullivan B, Finucane HK, routes. Walters RK, Bras J, et al. Analysis of shared heritability in common disorders of the brain. Science 2018;360:eaap8757. 2. Interpretation: Our findings showed three gene cate- [13] Avramopoulos D. Genetics of Alzheimer’s disease: recent advances. gories operating differently across subgroups of pa- Genome Med 2009;1:34. tients with AD and highlighted the role of vascular [14] Corder E, Saunders A. Gene dose of apolipoprotein E type 4 allele and the risk of Alzheimer’s disease in late onset families. Science 1993; pathways in pure forms of AD. We identified novel 8:41–3. genome-wide significant signals, which seem [15] Kunkle BW, Grenier-Boley B, Sims R, Bis JC, Naj AC, Boland A, strongly dependent on the AD subset used for meta- et al. Meta-analysis of genetic association with diagnosed Alzheimer’s analysis. disease identifies novel risk loci and implicates Abeta, Tau, immunity and lipid processing. BioRxiv 2018:294629. 3. Future directions: Our findings suggest the impor- [16] Sims R, van der Lee SJ, Naj AC, Bellenguez C, Badarinarayan N, tance of investigating the role of cerebral amyloid an- Jakobsdottir J, et al. Rare coding variants in PLCG2, ABI3, and giopathy, a unique identified AD hallmark that TREM2 implicate microglial-mediated innate immunity in Alz- heimer’s disease. Nat Genet 2017;49:1373–84. connects both the vascular and amyloid hypotheses [17] Jansen IE, Savage JE, Watanabe K, Bryois J, Williams DM, for AD. Accounting for sample composition in the Steinberg S, et al. Genome-wide meta-analysis identifies new loci design of genetic studies might lead to the identifica- and functional pathways influencing Alzheimer’s disease risk. Nat tion of genetic profiles, which help clinicians to Genet 2019;51:404–13. distinguish subsets of patients and establish novel [18] Kunkle BW, Grenier-Boley B, Sims R, Bis JC, Damotte V, Naj AC, et al. Genetic meta-analysis of diagnosed Alzheimer’s disease iden- therapy targets. tifies new risk loci and implicates Ab, tau, immunity and lipid process- ing. Nat Genet 2019;51:414–30. [19] Ridge PG, Hoyt KB, Boehme K, Mukherjee S, Crane PK, Haines JL, et al. Assessment of the genetic variance of late-onset Alzheimer’s dis- ease. Neurobiol Aging 2016;41:200.e13–200.e20. [20] McKhann GM, Knopman DS, Chertkow H, Hyman BT, References Jack CR, Kawas CH, et al. The diagnosis of dementia due to Alzheimer’s disease: recommendations from the National Insti- [1] Baumgart M, Snyder HM, Carrillo MC, Fazio S, Kim H, Johns H. tute on Aging-Alzheimer’s Association workgroups on diag- Summary of the evidence on modifiable risk factors for cognitive nostic guidelines for Alzheimer’s disease. Alzheimers Dement decline and dementia: a population-based perspective. Alzheimers De- 2011;7:263–9. ment 2015;11:718–26. [21] Boada M, Tarraga L, Hernandez I, Valero S, Alegret M, Ruiz A, et al. [2] Brookmeyer R, Johnson E, Ziegler-Graham K, Arrighi HM. Fore- Design of a comprehensive Alzheimer’s disease clinic and research casting the global burden of Alzheimer’s disease. Alzheimers Dement center in Spain to meet critical patient and family needs. Alzheimers 2007;3:186–91. Dement 2014;10:409–15. [3] Toledo JB, Arnold SE, Raible K, Brettschneider J, Xie SX, [22] Roman GC, Tatemichi TK, Erkinjuntti T, Cummings JL, Masdeu JC, Grossman M, et al. Contribution of cerebrovascular disease in autopsy Garcia JH, et al. Vascular dementia: diagnostic criteria for research confirmed neurodegenerative disease cases in the National Alz- studies. Report of the NINDS-AIREN International Workshop. heimer’s Coordinating Centre. Brain 2013;136:2697–706. Neurology 1993;43:250–60. [4] Serrano-Pozo A, Frosch MP, Masliah E, Hyman BT. Neuropatholog- [23] Gauderman WJ. Sample size requirements for matched case-control ical Alterations in Alzheimer Disease. Cold Spring Harb Perspect studies of gene-environment interaction. Stat Med 2002;21:35–50. Med 2011;1:a006189. [24] Lambert JC, Ibrahim-Verbaas CA, Harold D, Naj AC, Sims R, [5] Regan C, Katona C, Walker Z, Hooper J, Donovan J, Livingston G. Bellenguez C, et al. Meta-analysis of 74,046 individuals identifies Relationship of vascular risk to the progression of Alzheimer disease. 11 new susceptibility loci for Alzheimer’s disease. Nat Genet 2013; Neurology 2006;67:1357–62. 45:1452–8. [6] Arvanitakis Z, Capuano AW, Leurgans SE, Bennett DA, Schneider JA. [25] Wasserstein RL, Lazar NA. The ASA’s statement on p -values: context, Relation of cerebral vessel disease to Alzheimer’s disease dementia process, and purpose. Am Stat 2016;70:129–33. and cognitive function in elderly people: a cross-sectional study. Lan- [26] Wasserstein RL, Schirm AL, Lazar NA. Moving to a World Beyond “ p cet Neurol 2016;15:934–43. , 0.05”. Am Stat 2019;73:1–19. [7] Viswanathan A, Rocca WA, Tzourio C. Vascular risk factors and de- [27] Machiela MJ, Chanock SJ. LDlink: a web-based application for mentia: how to move forward? Neurology 2009;72:368–74. exploring population-specific haplotype structure and linking corre- [8] Goate A, Chartier-Harlin MC, Mullan M, Brown J, Crawford F, lated alleles of possible functional variants: Fig. 1. Bioinformatics Fidani L, et al. Segregation of a missense mutation in the amyloid pre- 2015;31:3555–7. S. Moreno-Grau et al. / Alzheimer’s & Dementia 15 (2019) 1333-1347 1347 [28] Watanabe K, Taskesen E, van Bochoven A, Posthuma D. Functional tive function and synaptic plasticity in mice. Nat Med 2014; mapping and annotation of genetic associations with FUMA. Nat 20:659–63. Commun 2017;8:1826. [41] Boada M, Antunez C, Ramırez-Lorca R, DeStefano A, Gonzalez- [29] Jun GR, Chung J, Mez J, Barber R, Beecham GW, Bennett DA, et al. Perez A, Gayan J, et al. ATP5H/KCTD2 locus is associated with Alz- Transethnic genome-wide scan identifies novel Alzheimer’s disease heimer’s disease risk. Mol Psychiatry 2014;19:682–7. loci. Alzheimers Dement 2017;13:727–38. [42] Traylor M, Adib-Samii P, Harold D, Alzheimer’s Disease Neu- [30] Zhang B, Gaiteri C, Bodea L-G, Wang Z, McElwee J, roimaging Initiative the ADN, International Stroke Genetics Podtelezhnikov AA, et al. Integrated systems approach identifies ge- Consortium (ISGC), UK Young Lacunar Stroke DNA resource netic nodes and networks in late-onset Alzheimer’s disease. Cell TISGC, Dichgans M, et al. Shared genetic contribution to Is- 2013;153:707–20. chaemic Stroke and Alzheimer’s Disease. Ann Neurol 2016; [31] Jack CR, Bennett DA, Blennow K, Carrillo MC, Dunn B, 79:739. Haeberlein SB, et al. NIA-AA Research Framework: toward a biolog- [43] Lucariello M, Vidal E, Vidal S, Saez M, Roa L, Huertas D, et al. Whole ical definition of Alzheimer’s disease. Alzheimers Dement 2018; exome sequencing of Rett syndrome-like patients reveals the muta- 14:535–62. tional diversity of the clinical phenotype. Hum Genet 2016; [32] Sweeney MD, Montagne A, Sagare AP, Nation DA, Schneider LS, 135:1343–54. Chui HC, et al. Vascular dysfunction-The disregarded partner of Alz- [44] Hanada K, Kumagai K, Yasuda S, Miura Y, Kawano M, Fukasawa M, heimer’s disease. Alzheimers Dement 2019;15:158–67. et al. Molecular machinery for non-vesicular trafficking of ceramide. [33] Hachinski V. Dementia: paradigm shifting into high gear. Alzheimers Nature 2003;426:803–9. Dement 2019;15:985–94. [45] Leduc V, De Beaumont L, Theroux L, Dea D, Aisen P, Petersen RC, [34] Weller RO, Boche D, Nicoll JAR. Microvasculature changes and cere- et al. HMGCR is a genetic modifier for risk, age of onset and MCI con- bral amyloid angiopathy in Alzheimer’s disease and their potential version to Alzheimer’s disease in a three cohorts study. Mol Psychiatry impact on therapy. Acta Neuropathol 2009;118:87–102. 2015;20:867–73. [35] Charidimou A, Shoamanesh A, Al-Shahi Salman R, Cordonnier C, [46] Broce IJ, Tan CH, Fan CC, Jansen I, Savage JE, Witoelar A, et al. Perry LA, Sheth KN, et al. Cerebral amyloid angiopathy, cerebral mi- Dissecting the genetic relationship between cardiovascular risk crobleeds and implications for anticoagulation decisions: the need for factors and Alzheimer’s disease. Acta Neuropathol 2019; a balanced approach. Int J Stroke 2018;13:117–20. 137:209–26. [36] Sakai K, Yamada M. Cerebral amyloid angiopathy. Brain Nerve 2014; [47] Escott-Price V, Bellenguez C, Wang L-S, Choi S-H, Harold D, Jones L, 66:827–35. et al. Gene-wide analysis detects two new susceptibility genes for Alz- [37] Beecham GW, Hamilton K, Naj AC, Martin ER, Huentelman M, heimer’s disease. PLoS One 2014;9:e94661. Myers AJ, et al. Genome-wide association meta-analysis of neuropath- [48] Luo L, Bokil NJ, Wall AA, Kapetanovic R, Lansdaal NM, Marceline F, ologic features of Alzheimer’s disease and related dementias. PLoS et al. SCIMP is a transmembrane non-TIR TLR adaptor that promotes Genet 2014;10:e1004606. proinflammatory cytokine production from macrophages. Nat Com- [38] Biffi A, Shulman JM, Jagiella JM, Cortellini L, Ayres AM, Schwab K, mun 2017;8:14133. et al. Genetic variation at CR1 increases risk of cerebral amyloid angi- [49] Delunardo F, Margutti P, Pontecorvo S, Colasanti T, Conti F, Rigano R, opathy. Neurology 2012;78:334–41. et al. Screening of a microvascular endothelial cDNA library identifies [39] M€akel€a M, Kaivola K, Valori M, Paetau A, Polvikoski T, rabaptin 5 as a novel autoantigen in Alzheimer’s disease. J Neuroim- Singleton AB, et al. Alzheimer risk loci and associated neuropa- munol 2007;192:105–12. thology in a population-based study (Vantaa 851). Neurol Genet [50] Moreno-Grau S, Hernandez I, Heilmann-Heimbach S, Ruiz S, Rose- 2018;4:e211. nde-Roca M, Mauleon A, et al. Genome-wide significant risk factors [40] Villeda SA, Plambeck KE, Middeldorp J, Castellano JM, Mosher KI, on chromosome 19 and the APOE locus. Oncotarget 2018; Luo J, et al. Young blood reverses age-related impairments in cogni- 9:24590–600.