Acta Neuropathologica (2022) 144:821–842 https://doi.org/10.1007/s00401-022-02454-z ORIGINAL PAPER Genome‑wide meta‑analysis for Alzheimer’s disease cerebrospinal fluid biomarkers Iris E. Jansen1,2,3 · Sven J. van der Lee1,2,4 · Duber Gomez‑Fonseca5,6,7 · Itziar de Rojas8,9 · Maria Carolina Dalmasso10,11 · Benjamin Grenier‑Boley12 · Anna Zettergren13 · Aniket Mishra14 · Muhammad Ali5,6,7 · Victor Andrade10,15 · Céline Bellenguez12 · Luca Kleineidam10,15,16 · Fahri Küçükali17,18 · Yun Ju Sung5,6,7 · Niccolo Tesí1,2,4 · Ellen M. Vromen1,2 · Douglas P. Wightman3 · Daniel Alcolea9,19 · Montserrat Alegret8,9 · Ignacio Alvarez20,21 · Philippe Amouyel12 · Lavinia Athanasiu22 · Shahram Bahrami22 · Henri Bailly23 · Olivia Belbin9,19 · Sverre Bergh24,25 · Lars Bertram26 · Geert Jan Biessels27 · Kaj Blennow28,29 · Rafael Blesa9,19 · Mercè Boada8,9 · Anne Boland30 · Katharina Buerger31,32 · Ángel Carracedo33,34 · Laura Cervera‑Carles9,19 · Geneviève Chene14,35 · Jurgen A. H. R. Claassen36,37 · Stephanie Debette14,35,38 · Jean‑Francois Deleuze30 · Peter Paul de Deyn39 · Janine Diehl‑Schmid40,41 · Srdjan Djurovic42,43 · Oriol Dols‑Icardo9,19 · Carole Dufouil14,44 · Emmanuelle Duron45 · Emrah Düzel46,47 · EADB consortium · Tormod Fladby48,49 · Juan Fortea9,19 · Lutz Frölich50 · Pablo García‑González8,9 · Maria Garcia‑Martinez51 · Ina Giegling52 · Oliver Goldhardt40 · Johan Gobom29 · Timo Grimmer40 · Annakaisa Haapasalo53 · Harald Hampel54,55 · Olivier Hanon23,56 · Lucrezia Hausner50 · Stefanie Heilmann‑Heimbach57 · Seppo Helisalmi58 · Michael T. Heneka15,16 · Isabel Hernández8,9 · Sanna‑Kaisa Herukka59 · Henne Holstege1,2,4 · Jonas Jarholm60 · Silke Kern13,61 · Anne‑Brita Knapskog62 · Anne M. Koivisto59,63,64 · Johannes Kornhuber65 · Teemu Kuulasmaa66 · Carmen Lage51,67 · Christoph Laske68,69 · Ville Leinonen70,71 · Piotr Lewczuk65,72 · Alberto Lleó9,19 · Adolfo López de Munain9,73,74,75 · Sara Lopez‑Garcia51 · Wolfgang Maier15 · Marta Marquié8,9 · Merel O. Mol78 · Laura Montrreal8 · Fermin Moreno9,73,74 · Sonia Moreno‑Grau8,9 · Gael Nicolas77 · Markus M. Nöthen57 · Adelina Orellana8,9 · Lene Pålhaugen48,49 · Janne M. Papma78 · Florence Pasquier12 · Robert Perneczky31,79,80,81 · Oliver Peters46,82 · Yolande A. L. Pijnenburg1,2 · Julius Popp83,84 · Danielle Posthuma3 · Ana Pozueta51 · Josef Priller82,85,86 · Raquel Puerta8 · Inés Quintela33 · Inez Ramakers87 · Eloy Rodriguez‑Rodriguez51 · Dan Rujescu52 · Ingvild Saltvedt88,89 · Pascual Sanchez‑Juan90 · Philip Scheltens1,2 · Norbert Scherbaum91 · Matthias Schmid92 · Anja Schneider15,16 · Geir Selbæk25,48,62 · Per Selnes49 · Alexey Shadrin22 · Ingmar Skoog13,61 · Hilkka Soininen59 · Lluís Tárraga8,9 · Stefan Teipel93,94 · The GR@ACE study group · Betty Tijms1,2 · Magda Tsolaki95 · Christine Van Broeckhoven18,96 · Jasper Van Dongen17,18 · John C. van Swieten78 · Rik Vandenberghe97,98 · Jean‑Sébastien Vidal23 · Pieter J. Visser1,2,99,100 · Jonathan Vogelgsang101,102 · Margda Waern13,103 · Michael Wagner15,16 · Jens Wiltfang101,104,105 · Mandy M. J. Wittens18,106 · Henrik Zetterberg28,29,107,108,109 · Miren Zulaica9,73,74 · Cornelia M. van Duijn76,110 · Maria Bjerke18,106,111 · Sebastiaan Engelborghs18,106,111,112 · Frank Jessen16,113,114 · Charlotte E. Teunissen2,115 · Pau Pastor116 · Mikko Hiltunen117 · Martin Ingelsson118,119,120 · Ole A. Andreassen22,121 · Jordi Clarimón9,19 · Kristel Sleegers17,18 · Agustín Ruiz8,9 · Alfredo Ramirez10,15,16,114,122 · Carlos Cruchaga5,6,7 · Jean‑Charles Lambert12 · Wiesje van der Flier1,2 Received: 16 February 2022 / Revised: 18 May 2022 / Accepted: 7 June 2022 / Published online: 6 September 2022 © The Author(s) 2022 Iris E. Jansen, Sven J. van der Lee, Duber Gomez-Fonseca, Itziar de Rojas, Maria Carolina Dalmasso and Benjamin Grenier-Boley are co-first authors. Agustín Ruiz, Alfredo Ramirez, Carlos Cruchaga, Jean-Charles Lambert and Wiesje van der Flier are co-last authors. Members of “EADB consortium” and “The GR@ACE study group” are provided in Supplementary File 2. Extended author information available on the last page of the article 13 Vol.:(0123456789) 822 Acta Neuropathologica (2022) 144:821–842 Abstract Amyloid-beta 42 (Aβ42) and phosphorylated tau (pTau) levels in cerebrospinal fluid (CSF) reflect core features of the pathogenesis of Alzheimer’s disease (AD) more directly than clinical diagnosis. Initiated by the European Alzheimer & Dementia Biobank (EADB), the largest collaborative effort on genetics underlying CSF biomarkers was established, includ- ing 31 cohorts with a total of 13,116 individuals (discovery n = 8074; replication n = 5042 individuals). Besides the APOE locus, novel associations with two other well-established AD risk loci were observed; CR1 was shown a locus for Aβ42 and BIN1 for pTau. GMNC and C16orf95 were further identified as loci for pTau, of which the latter is novel. Clustering methods exploring the influence of all known AD risk loci on the CSF protein levels, revealed 4 biological categories sug- gesting multiple Aβ42 and pTau related biological pathways involved in the etiology of AD. In functional follow-up analyses, GMNC and C16orf95 both associated with lateral ventricular volume, implying an overlap in genetic etiology for tau levels and brain ventricular volume. Keywords GWAS · Alzheimer’s disease · Cerebrospinal fluid · Amyloid-beta · Tau Introduction fluid (CSF), the two major proteins implicated in the AD pathological process. Aβ42 pathology in the brain is nega- Resolving the genetic background of Alzheimer’s disease tively correlated with CSF Aβ42 levels, where a decrease (AD) has proven to contribute greatly to our understand- in CSF Aβ42 is indicative of AD [49, 58]. CSF (p)Tau is ing of underlying disease processes, for instance with positively correlated with (p)Tau pathology in the brain, and the discovery of APP [25], PSEN1 [53], and PSEN2 [52] therefore higher CSF (p)Tau levels are observed in patients in family-based studies, leading to the amyloid cascade with AD. CSF pTau is presumed to reflect AD-type tau- theory [38]. In addition, genome-wide association studies tangles more specifically than total tau [49, 58]. Previous (GWAS) in AD have convincingly highlighted the impor- studies on CSF amyloid-beta and (p)Tau have identified tance of microglia [33, 56], a finding previously supported genetic risk loci, the most recent one including 3,146 indi- by research from other scientific fields [18, 72, 75], and now viduals [17]. Some of the 8 discovered loci had not been also widely accepted as a genetic cause rather than a result previously associated with AD, emphasizing the potential of of AD pathogenesis. Further exploration of genetic risk endophenotypes to reveal novel genetic risk factors. Our cur- factors contributing to AD development and pathogenesis rent study aimed to further define the genetic background of might reveal more biological insights, an important step in AD by studying the genetic effects on CSF Aβ42 and pTau the quest for AD treatment that will slow down or even halt levels in a total of 13,116 individuals. disease progression. GWAS of clinically diagnosed AD patients have been successful, and current efforts largely focus on increasing Materials and methods sample size to improve the statistical power to detect genetic variants [6, 70]. An alternative approach is to study effects Participants of genetic variants on pathophysiological features of AD. The strength of such studies is based on the assumption that We combined data from 16 European cohorts, encompass- more objective measurable biological properties are more ing a total of 8074 individuals (Table 1; Online Resource strongly associated with the underlying AD pathology than 1—Table 1; Online Resource 2—Fig. 1) with both genotype the clinical diagnostic classifications (e.g., misclassifications data and CSF measurements. The majority of these cohorts or symptoms not manifested yet), thereby allowing to detect (82%) are part of the EADB consortium [6], and included larger effects by reducing heterogeneity [26]. The use of the full spectrum of clinical severity potentially leading biomarkers further enables to identify genetic effects specific to AD, from subjective cognitive decline, mild cognitive for certain AD-related biological mechanisms. This is an impairment, to dementia. Written informed consent was advantage over the conventional GWAS approach for clinical obtained from study participants or, for those with substan- AD diagnosis, where it generally remains unclear through tial cognitive impairment, from a caregiver, legal guardian, what causal gene or cellular process a locus is associated or other proxy. Study protocols for all cohorts were reviewed to AD. and approved by the appropriate institutional review boards. It is possible to measure levels of amyloid-beta-42 (Aβ42) For replication, 15 cohorts totaling 5042 individuals and (phosphorylated) tau (pTau and Tau) in cerebrospinal (Online Resource 1—Table 2) were available to attempt 13 Table 1  Demographic information on cohorts of stage 1 discovery analysis Country Cohort Age Gender Diagnoses APOE4 carrier Aβ42 pTau Mean years (sd) % male AD (%) MCI (%) Control (%) Other dem (%) % APOE4 + Mean levels (sd) n Mean levels (sd) n Acta Neuropathologica (2022) 144:821–842 Belgium DEM 78.3 (7.3) 40 72 27 1 0 48 562.7 (250.8) 587 71.6 (35.4) 585 Finland ADGEN 70.2 (8.0) 34 89 1 10 0 73 477.1 (183.2) 226 80.4 (34.8) 155 France1 BALTAZAR 77.0 (6.7) 45 43 57 0 0 43 825.7 (358.9) 420 73.0 (34.0) 419 France2 MEMENTO 69.2 (8.9) 47 0 100 0 0 30 1078.6 (406.7) 389 62.6 (29.8) 386 France3 CNRMAJ-Lille 66.0 (8.7) 47 100 0 0 0 56 664.8 (205.1) 127 101.8 (53.4) 125 Germany1 Delcode 71.7 (5.9) 52 13 23 64 0 39 965.3 (338.2) 465 63.2 (35.1) 462 Germany2 KND 67.3 (8.7) 57 18 82 0 0 49 724.0 (347.9) 309 69.4 (36.8) 305 Germany3 TUM 70.2 (9.2) 48 98 1 0 1 58 536.8 (240.0) 151 88.8 (48.7) 152 Germany4 PAGES 73.4 (7.7) 40 70 30 0 0 59 528.2 (245.5) 136 90.8 (60.0) 137 Germany5 UHB 70.3 (7.2) 42 69 30 1 0 68 545.7 (327.8) 111 82.0 (44.2) 111 Netherlands ADC & Pearl ND 64.1 (8.9) 59 42 10 24 23 50 842.8 (288.9) 2936 65.9 (35.9) 2931 Spain1 ACE 72.7 (8.2) 43 27 59 8 6 36 779.2 (314.3) 609 68.2 (35.5) 609 Spain2 SIGNAL & SPIN 70.6 (8.0) 43 34 45 19 2 39 707.2 (368.3) 394 71.1 (41.9) 370 Spain3 Valdecilla 67.0 (9.0) 39 10 37 45 8 29 887.5 (360.5) 98 56.8 (27.9) 99 Sweden1 Birth cohort & Clin. AD 75.0 (9.4) 45 51 0 49 0 52 533.5 (271.9) 856 68.9 (35.5) 694 Sweden2 Uppsala university 71.0 (6.3) 46 58 37 0 6 64 489.0 (238.1) 260 77.6 (37.2) 259 For most cohorts, one of the two CSF levels is missing for a small number of samples. The demographics for age, gender diagnoses and APOE4 carriers status are then displayed for the largest group of samples with at least one CSF measurement. AB42 levels are corrected according to known drift over time for the Dutch ADC and Pearl ND cohorts. All, except the Swedisch Birth cohort and clinical AD samples, are part of EADB Aβ42 amyloid-beta 42, pTau phosphorylated tau, AD Alzheimer’s disease, MCI mild cognitive impairment, other dem other dementia, n sample size, DEM Antwerp prospective dementia cohort 13 823 824 Acta Neuropathologica (2022) 144:821–842 replication of the association signals to Aβ42 and pTau, for were available, the details on QC steps and imputation with the variants with P value < 1e–5 in the discovery analysis. the TOPMed reference panel were previously described [6]. Data from all cohorts, except one (NorCog from Univer- For the Spanish ACE and Valdecilla cohorts, QC pro- sity of Oslo, Norway), were obtained through collaboration cedures are described in another study [16], followed by with the previous largest GWAS on CSF Aβ42 and pTau, imputation with the TOPMed reference panel. For the mostly including cohorts originating from the United States Gothenburg H70 Birth Cohort studies and clinical AD [17]. Basic demographics are described in Online Resource samples from Sweden, the QC and imputation procedures 1—Table 2, more detailed cohort information is described were described elsewhere [48]. Post‐imputation QC only elsewhere [17]. included variants with a high imputation quality (RSQ [imputation quality] > 0.8). The UCSC LiftOver program CSF measurements (https://​genome-​store.​ucsc.​edu/) and Plink v2.0 (www.​cog-​ genom​ics.​org/​plink/2.​0/) [10] were used to lift the GRCh37 Due to the multi-center approach, CSF protein levels were genomic positions to GRCh38, the genomic build for all measured with various CSF protein assays (Online Resource other datasets. All genotypes were hard called using the 1—Table 2). Aβ42 was measured with ELISA, Lumipulse default Plink v2.0 (http://​www.​cog-​genom​ics.​org/​plink/2.​ or V-PLEX, and pTau with ELISA or Lumipulse. For details 0/) settings. on specific lab procedures, see the original studies [2–4, 17, 21, 23, 27, 34, 36, 42, 47, 50, 57, 62, 78]. Protein levels Heritability and genetic correlation were log10 transformed and normalized within cohorts and CSF assay type (if multiple assays were used within a single For the estimation of the SNP-heritability, two distinct tools cohort) to approximate a normal distribution to correct for were used. With LD score regression (LDSC), it was pos- the application of various CSF assays across different stud- sible to perform the calculations with the full number of ies. Then, the normalized protein levels were used as con- samples as the input for this analysis is the summary sta- tinuous phenotypes in the association analyses. The distribu- tistics. Besides heritability estimates, genetic correlations tion of raw and normalized CSF protein levels are displayed were also calculated for Aβ42, pTau, tau (to test the simi- in Online Resource 2—Figs. 2–13, for the cohorts for which larity in genetic background to pTau), and two previously we have individual-level data available (indicated in Online published AD summary statistics [33, 43]. Precalculated Resource 1—Table 1; column D). For the stratified analyses, LD scores from the 1,000 Genomes European reference two subgroups of individuals, amyloid normal and amyloid population were obtained online. All estimates were based abnormal, were defined based on their Aβ42 status. Indi- on HapMap3 SNPs only to ensure high-quality LD score viduals with an untransformed Aβ42 level below a threshold calculations (https://​alkes​group.​broad​insti​tute.​org/​LDSCO​ were assigned to the abnormal amyloid level group. The RE/). As a rule of thumb, LD Score regression tends to yield thresholds were defined by the individual research groups very noisy results when applied to datasets with fewer than as it depends on technical circumstances, and are displayed 5000 individuals (https://​github.​com/​bulik/​ldsc/​wiki/​FAQ) in Online Resource 1—Table 2. [9]. The summary statistics for the stratified analyses were, therefore, not considered. Genotyping, quality control and imputation For comparison to SNP-heritability estimates of previous studies for Aβ42 and pTau, GCTA v1.9 [74] was applied The genetic data for the EADB cohorts have been processed to the individual-level genotype data of the largest dataset in a homogeneous approach (Online Resource 1—Table 1), (Netherlands). Other datasets were not considered as the in which the Illumina Infinium Global Screening Array sample size was too low for small standard errors, thereby (GSA, GSAsharedCUSTOM_24 + v1.0) was predominantly impossible to draw any meaningful conclusions from the used for data generation. Additional arrays included the estimates. The restricted maximum likelihood (REML) Axiom 815 K Spanish biobank array (Thermo Fisher) for analysis was performed for the log10-transformed normal- ACE (Barcelona, Spain) and Valdecilla (Santander, Spain) ized CSF Aβ42 and pTau adjusted for gender, age, and the cohorts, and the Illumina Neurochip array (Gothenburg, first 10 principal components. Variance explained could Sweden). Standard quality control (QC) procedures were not be calculated for significant loci only as p-values from performed to exclude individuals and variants with low qual- GWAS results of a large independent sample are unavailable, ity, in general followed by imputation with the Trans-Omics and calculation in the Dutch sample would be hampered by for Precision Medicine (TOPMed) reference panel [13, 59]. winners-curse, causing inflation. For the EADB cohorts for which GSA genotype level data 13 Acta Neuropathologica (2022) 144:821–842 825 Single‑marker association to stage 1. The genome-wide threshold for significance of P < 5 × ­10−8 was used to defined GWAS hits in stage 3. Genome-wide association analysis for each cohort was per- formed in PLINK v2.0 [10], using linear regression for the Colocalization continuous phenotypes Aβ42, tau and pTau. Association tests were adjusted for gender, age, assay type (if applica- All variants within 1.5 megabases (Mb) of the lead vari- ble), and ten ancestry principal components. Only variants ant of each genomic risk loci were used in the colocaliza- with a minor allele frequency threshold above 0.01 were tion analysis. The stage 1 GWAS summary statistics were tested. For smaller cohorts (n < 250 individuals) this thresh- used for the CSF loci aiming consist sample sizes across old was set to 0.05 to avoid false positive findings. the variants. Colocalization comparisons were performed Association analyses were repeated for subgroups, strati- to eQTL data for brain and immune-related tissues and cell- fied according to APOE4 status (based on the high-quality types, which were obtained from the eQTL catalog [39] (R2 > 0.8) imputed variants rs429358 and rs7412) or dichoto- release 5. The microglia data were obtained from Young mous Aβ42 status, resulting in the following groups: (1) et al. [77]. GWAS summary statistics for loci comparison to APOE4 (hetero- and homozygous) carriers; (2) APOE4 non- other GWAS studies were obtained from Kunkle et al. [43] carriers; (3) individuals with abnormal Aβ42 levels; and (4) for AD, and from Vojinovic et al. [65] for brain ventricu- individuals with normal Aβ42 levels. After stratification, lar volume. The GWAS data and eQTL data were trimmed cohorts with a minimal sample size of 100 individuals were so that all variants overlap. Colocalization was performed included. Covariates were those described for the main with the Coloc R package [24], using the coloc.abf func- analyses above. tion for the approach assuming a single causal variant, and the runsusie and coloc.susie functions to test for colocaliza- Independent replication tion relaxing this assumption to multiple variants. The latter approach was performed for the comparisons to AD and A total of 5042 samples from 15 cohorts were included for brain volume loci, and LD matrices were calculated with the replication analysis. The genetic data for NorCog were our own individual-level data for the significant CSF loci generated with 2 different genotyping assays. Extensive QC using LDstore R package [7]. Default priors were used for procedures which are detailed elsewhere [6], allowed for prior probability of association with the GWAS data and joined genetic analyses of these sub-datasets. Variant asso- eQTL data. The prior probability of colocalization was set ciation testing was performed according to the association as 5 × ­10−6 as recommended [67]. Nominal P, sample size analysis section above. For all other replication cohorts, QC, and MAF were used when beta and variance of beta were not imputation and association testing procedures are described available for the GWAS data or eQTL data. Colocalizations elsewhere [17]. In short, individual and variant QC stand- with a posterior probability > 0.8 were considered success- ards were met, and imputation was performed using the ful colocalizations. Comparisons were visualized with the 1000 Genomes Project Phase 3 reference panel. Each dataset R package LocusCompareR (https://​github.​com/​boxia​ngliu/​ was QCed and imputed independently. The additive linear locus​compa​rer). regression model in PLINK v1.9 [10] was used for single- variant analyses. Gene‑based analysis Meta‑analyses Gene-based and gene-set association tests were performed using MAGMA v1.08 [14], which was implemented by METAL [71] was used for meta-analyses in stages 1–3 of the FUMA [69]. The per variant association summary statis- per-cohort association results, applying the default approach tics for the main results served as the input, where variants that utilizes P value and direction of effect, while weighted were selected if mapped within 18,870 protein-coding genes according to sample size. For stage 1, we used the genome- (with unique ensembl ID). The mean SNP-wise model was wide threshold for significance of P < 5 × ­10−8, and a sug- implemented. The Bonferroni-corrected significance thresh- gestive threshold of P < 1 × ­10−5 to select variants to study old was set to P < 2.65 × ­10–6, based on the number of tested in Stage 2. Stage 2 variants were considered a replication genes. with P < 0.05 and same direction of effect in comparison 13 826 Acta Neuropathologica (2022) 144:821–842 Gene mapping for analysis (52 proteins were excluded). Out of the 2136 proteins quantified, 1282 (55.4%) proteins respected these The genome-wide significant loci of the main results were criteria and were included in the study. further explored for promising causal AD genes using For the Knight-ADRC data, levels of 1305 proteins FUMA [69], after lifting over the results with genomic were quantified using the SOMAscan assay, a multiplexed, build GRCh38 to GRCh37 with the UCSC LiftOver Pro- aptamer-based platform CSF (n = 717) [73]. Quality control gram (https://​genome-​store.​ucsc.​edu/). Two gene mapping was performed at the sample and aptamer levels using con- strategies were used: trol aptamers (positive and negative controls) and calibra- tor samples. As described in detail [73], additional quality – Positional mapping maps SNPs to genes based on control was performed that included limit of detection cut- physical distance (within a 10-kb window) from known off, scale factor, coefficient of variation, and outlier vari- protein-coding genes in the human reference assembly ation. Only proteins with a call rate higher than 85% call (GRCh37/hg19). rate were included. A total of 713 proteins passed quality – eQTL mapping maps SNPs to genes with which they control. pQTL analyses was performed and reported in pre- show a significant eQTL association (that is, allelic vari- vious studies [73]. ation at the SNP is associated with the expression level g:Profiler and Enrichment map [44], a Cytoscape App, of that gene). eQTL mapping uses information from 85 were used to perform pathway enrichment analyses on pro- brain- and immune-related tissue types in 11 data repos- teins with a certain level of association (EMIF-AD MBD: itories (BIOSQTL, BloodeQTL, BRAINEAC, CMC, P < 0.05; Knight-ADRC: P < 0.004, corresponding to pro- DICE, eQTLcatalogue, eQTLGen, GTEx, PsychEN- teins with a similar effect size as in EMIF-AD MBD). The CODE, scRNA_eQTLs, xQTLServer), and is based on results are shown as functionally grouped networks. We used cis-eQTLs which can map SNPs to genes up to 1 Mb GO biological processes and Reactome as ontology sources. apart. We used a false discovery rate of 0.05 to define For this explorative analysis, only pathways with P < 0.05 significant eQTL associations. (corrected for multiple testing) are shown. Phenome‑wide association studies (PheWAS) Effects of AD‑associated variants on Aβ42 and pTau We conducted phenome-wide association studies (PheWAS) We assessed the most recent GWAS [6] for AD and extracted on the top SNPs, rs4844610, rs429358, rs744373, the top loci of 83 variants (excluding APOE ɛ4 and APOE rs9877502, rs4843559. A PheWAS starts out with a single to ɛ2) that showed genome-wide significant association with a few variants of interest that are systematically being tested AD [6]. We extracted Z-scores and P values and plotted for association to many phenotypes. We used the ‘phewas’ them in a heatmap. Rows and columns were clustered using function of the R-package ‘ieugwasr’ [19, 31]. Using this Euclidean distances and average hierarchical clustering. We function, we searched traits that associate with the list of performed a gene-set enrichment analysis to find molecu- SNPs with P < 1 × ­10–7 in all GWAS harmonized summary lar pathways enriched within each cluster. The SNP-gene statistics in the MRC IEU OpenGWAS data infrastructure assignment corresponds to the one described in the recent [31]. In short, this enables us to screen for other traits to main EADB GWAS [6], including several annotation strate- which these SNPs are associated. The database (May 2021) gies. When multiple genes were reported to associate with includes the GWAS summary statistics of 19,649 traits. the same SNP (rs12590654 near SLC24A4/RIN3, rs7225151 near SCIMP/RABEP1, rs6846529 near CLNK/HS3ST1, Association with CSF proteomics rs7384878 near ZCWPW1/NYAP1 and rs10437655 near CELF1/SPI1), we considered both genes for the gene- We associated the lead variants near GMNC (rs9877502) set enrichment analysis. In addition, for SNP rs6605556, and in C16orf95 (rs4843559) with CSF proteomics data of located in the complex HLA region, we considered HLA- two different sources (EMIF-AD MBD and Knight-ADRC). DRB1 gene (eQTL in blood with rs6605556), and for SNPs For the EMIF-AD MBD data, a total of 2,136 proteins were rs7157106 and rs10131280, both located in the gene-dense quantified centrally using 11-plex tandem mass tag spec- IGH region, we considered IGHG2 and IGHV2-70 (eQTLs trometry in 366 individuals from the EMIF-AD MBD study in blood with rs7157106 and rs10131280, respectively). The [61] (subset of Amsterdam Dementia Cohort within EADB). gene-set enrichment analysis was performed specifying Bio- We selected proteins with a maximum of 50% missing val- logical Processes from Gene Ontology [1, 5] as gene-set and ues. For related proteins that had identical values due to correcting P values with Bonferroni. Biological pathways fragment aspecificity, we randomly selected one protein were considered significant at corrected P < 0.01. To help 13 Acta Neuropathologica (2022) 144:821–842 827 with the interpretation of each cluster’s function, we plot Z-statistic summarizes the magnitude and the direction of effect relative to the effect allele. For APOE stage 2 and stage 3, the statistics are reported for the most significant stage 3 variants 1.00E–379 1.84E–09 7.88E–10 9.65E–37 8.03E–14 9.59E–95 the most recurring words of the significant terms underlying each cluster using wordclouds. The following R packages P were used for these analysis: gprofiler2 [41] and wordcloud2 − 6.01 (https://​github.​com/​lchif​f on/​wordc​loud2). − 41.68 6.15 12.66 7.47 20.65 rsID ID number in SNP database, Chr chromosome, Pos genomic base pair position, A1 effect allele, A2 other allele, Freq A1 allele frequency effect allele, n sample size, P p value Z Results Stage 3 13,089 12,562 11,746 12,540 11,783 12,539 n The overview of the study design is illustrated in Online 1.57E–143 Resource 2—Fig. 1. The GWA results from 16 studies were 4.08E–04 2.48E–03 3.00E–19 1.03E–04 1.66E–43 combined in stage 1. Variants that reached a suggestive level of significance (P < 1 × ­10–5) were subsequently evaluated in P an independent sample from 15 studies in stage 2. Finally, − 3.54 − 25.51 3.03 8.97 3.88 13.83 the results of stage 1 and stage 2 analyses were combined in stage 3. Detailed information on study participants, CSF Z acquisition and genotyping is provided in Table 1 and Online Stage 2 rs4420638 (G allele) for AB42 and rs769449 (A allele) for pTau, as rs429358 was not genotyped or imputed in the replication datasets 5015 4488 3948 4742 3985 4741 Resource 1—Tables 1 and 2. The results for tau and pTau are n strongly correlated (rg = 0.94; P = 1.86 × ­10–118), and there- fore only pTau findings are reported. 2.00E–288 1.13E–06 6.99E–08 1.28E–19 1.49E–10 6.87E–75 Genetic architecture P − 4.87 − 36.29 5.39 9.06 6.41 18.31 The fraction of variance in Aβ42 and pTau protein levels that could be explained by the additive effect of the genetic Z variants tested, was estimated on 0.13 (SE = 0.06) and 0.21 8074 8074 7798 7798 7798 7798 (SE = 0.07) by LDSC, respectively. These SNP-heritabili- n ties are substantially higher than the 0.07 previously esti- mated with LDSC for the diagnosis AD [80], or similarly Freq A1 Stage 1 reported for AD by this study using the same LDSC method 0.20 0.29 0.30 0.39 0.38 0.28 on more recent public GWAS summary statistics of AD [33, 43] (Online Resource 1—Table 3). GCTA estimated A2 A G A C T the SNP-heritability to be higher for both Aβ42 and pTau, T namely 0.27 (SE = 0.13) and 0.34 (SE = 0.12), respectively. A1 A G A G C C Both methods are reporting a higher SNP-heritability for 207,629,207 127,137,039 190,951,729 pTau than for Aβ42. Genetic correlation estimates with AD 44,908,684 87,191,825 44,908,684 Pos (hg38) GWAS summary statistics are described in Online Resource 1—Results and Online Resource 1—Table 4. Table 2  Meta-analysis association results for the 3 stages Chr GWAS variants associated with CSF Aβ42 and pTau 1 19 2 3 16 19 rs4844610 rs9877502 rs4843559 The stage 1 meta-analyses (QQ plots and lambda shown rs429358 rs744373 rs429358 in Online Resource 2—Fig. 14) identified 4 independent rsID significant variant associations, 1 for Aβ42 and 3 for pTau (Table 2). The strongest associations were observed for both 16q24.2 (C16orf95) Aβ42 and pTau in the APOE locus (Table 2). The variant that determines the APOE ɛ4 allele (rs429358-C) decreased 3q28 (GMNC) 2q14.3 (BIN1) 1q32.2 (CR1) Aβ42 (Z = − 36.29; P = 2.0 × ­10–288) and increased pTau (Z = 18.31; P = 6.87 × ­10–75) in CSF. In contrast, the variant APOE APOE Locus determining the APOE ɛ2 haplotype (rs7412-T) increased Aβ42 (Z = 11.97; P = 5.09 × ­10–33) and decreased pTau (Z = − 6.59; P = 4.49 × ­10–11). The APOE ɛ2 association Protein AB42 pTau 13 828 Acta Neuropathologica (2022) 144:821–842 Fig. 1  Manhattan plots of the stage 3 GWAS results. a Results visualized for CSFAβ42; b Results visualized for CSF pTau. The y-axes are lim- ited to visualize the non-APOE loci. The lowest P values for APOE are 4.07 × ­10–355 and 3.74 × ­10–94 for Aβ42 and pTau, respectively was replicated in stage 2 (Aβ42: Z = 7.27; P = 3.73 × ­10–13, significant associations (Fig. 1, Table 2) with Aβ42 in chro- and pTau: Z = − 6.43; P = 1.26 × ­10–10), and APOE ɛ4 with mosomal region 1q32.2 (Z = − 6.01; P = 1.84 × ­10–9, CR1; rs4420638-G for Aβ42 (Z = − 25.51; P = 1.57 × ­10–143), and Fig. 1a), and with pTau for the region 2q14.3 (Z = 6.15; with rs769449-A for pTau (Z = 13.83; P = 1.66 × ­10–43), both P = 7.88 × ­10–10, BIN1; Fig. 1b). The per-cohort and zoomed variants in high linkage disequilibrium with rs429358, as the in genomic location details of all significantly associated loci original APOE ɛ4 variant was not genotyped or imputed in of stage 3 are visualized in Online Resource 2—Figs. 15–20. the replication datasets. Colocalization analyses (results detailed in Online Resource In stage 1, no other significant loci were observed for 1—Table 5; Online Resource 2—Fig. 21 and 22) showed Aβ42. For pTau, we further identified significant associations colocalization of the CR1 (posterior probability = 0.97) and mapping to two chromosomal regions at 3q28 and 16q24.2 BIN1 (posterior probability = 0.82) loci to these loci in a (Table 2). The 3q28-locus (Z = 9.06; P = 1.28 × ­10–19) also recent AD GWAS [43]. For the BIN1 locus of the AD GWAS known as the GMNC locus, was reported previously for its two independent causal signals were observed, of which our association with pTau [17]. The cohorts from this previous BIN1 locus colocalized (posterior probability = 0.84) with study are the replication cohorts of the current study, thereby the first signal that was tagged by rs6733839, which is the logically GMNC was replicated (Z = 8.97; P = 3.00 × ­10–19). most significant variant of the BIN1 locus of the AD GWAS. The 16q24.2 locus (Z = 6.41; P = 1.49 × ­10–10), which is Explorative meta-analyses were repeated stratified for novel for pTau, was replicated in the stage 2 meta-analysis APOE (APOE ɛ4 carriers (n = 3240) vs. APOE ɛ4 non- (Z = 3.88; P = 1.03 × ­10–04). carriers (n = 3201)) and amyloid status (Amyloid normal Subsequently, the results from all individual studies were levels (n = 3182) vs. amyloid abnormal levels (n = 3775)) combined in the stage 3 meta-analysis (N = 13,116). In stage for stage 1 (QQ plots and lambda shown in Online Resource 3, two well-known AD loci showed additional genome-wide 2—Figs. 23 and 24), of which the results are visualized in 13 Acta Neuropathologica (2022) 144:821–842 829 Online Resource 2—Figs. 25 and 26, and detailed in Online the FUMA annotation are detailed in the Online Resource Resource 2—Results and Online Resource 1—Table 6. 2—Results and Online Resource 1—Table 7. Ten of our Besides the APOE and GMNC loci, two novel loci are CSF Aβ42 and pTau loci colocalized with one of the brain observed that have previously not been linked to any AD or immune eQTLs from the 63 tested datasets, which are phenotype. reported in Online Resource 1—Table 8. The APOE locus for both Aβ42 and pTau colocalized with an eQTL for Functional interpretation NKPD1 in a specific immune helper T cell. The CR1 locus for AB42 colocalized with an eQTL for the CR1 gene in 6 To interpret the functional effects of the identified vari- brain tissues (hippocampus, caudate, putamen, dorsolateral ants beyond AD, we performed gene prioritization (based prefrontal cortex, frontal cortex, cortex). Additionally for on positional mapping, gene-based association results, the dorsolateral prefrontal cortex, the CR1 locus also colo- and brain and immune eQTL annotations) using FUMA calized with an eQTL for the AL137789.1 gene. The BIN1 [69], colocalization analyses and PheWAS. The results of Fig. 2  LocusZoom plots showing variant association results for a GMNC and b C16orf95 loci. In black, the pTau association signals of this study; and in orange, the lateral ventricular volume (LVV) association signals observed in other studies 13 830 Acta Neuropathologica (2022) 144:821–842 locus colocalized with an eQTL for BIN1 in lymphoblastoid Relation to AD‑associated genetic variants cell lines. For the PheWAS, using data from publicly available Because of the evident overlap in etiology with clinical AD genome-wide association studies (N = 19,649) of the five dementia, we examined the association of all known AD loci top variants yielded 529 associations at P < 1 × ­10–7 (Online (excluding the APOE locus) with CSF Aβ42 and pTau. This Resource 1—Table 9). The majority is the known wide analysis has an explorative character as the small contribu- range of 490 traits associations with the APOE ɛ4 allele. tion of each individual AD loci to CSF amyloid and pTau For the other variants 39 associations were reported for is in general reflected in moderate effect sizes and associa- 27 unique traits. These traits can be categorized in three tion signals. However, patterns of reasonable signal could groups: traits related to brain ventricular volumes in par- facilitate the generation of biological hypotheses to test in ticular the lateral-ventricle (GMNC and C16orf95), Alz- future experiments. The results are shown in the heatmap heimer’s disease diagnosis (BIN1 and CR1), and measures of Fig. 3 and Online Resource 1—Tables 15 and 16. The of blood cell/lymphocyte counts (CR1). The regional pTau variants could be clustered in 4 groups of AD-associated associations of GMNC and C16orf95 overlapped with ven- genes based on their associations with Aβ42 and pTau. tricular volume (Fig. 2). Colocalization results (Online The first cluster of 14 variants showed strong association Resource 1—Table 5; Online Resource 2—Figs. 27 and 28) with both decreased levels of Aβ42 and increased levels of imply the same causal variant for C16orf95 (posterior prob- pTau in CSF. A pathway enrichment analysis of the genes ability = 0.93). The GMNC locus cannot be convincingly associated with the variants showed 29 GO terms enriched explained by the same causal variant (posterior probabil- and ‘amyloid’ is the common denominator in the names of ity = 0.63). The coloc method combined with Susie reports these terms, of which the signal is mostly driven by BIN1, only 1 causal signal for both GWAS. Of note, the colocali- PICALM ABCA7 and CLU. The second cluster contained zation probability improves to 0.74 when first reducing the 21 variants and included genes that have also been related GMNC-locus variants to a credible set of variants. Although to other dementia types (e.g., GRN, TMEM107B, SNX1, these results are subthreshold to the predefined posterior MAPT, CTSB and CTSH). This cluster was associated with probability limit of 0.8, they do imply that the same causal decreased pTau levels, an no general effect on Aβ42 levels. variant is likely underlying the GMNC association signals Pathway analysis of the genes suggests an enrichment for for pTau and AD. 8 GO terms of which the names have ‘immune’ as a com- Because of the overlap in effect of GMNC and C16orf95, mon denominator that is mostly driven by 12 genes of which we hypothesized that these two loci affect the same bio- TREM2 and GRN are the most frequent contributors. The logical pathways. We explored this hypothesis using CSF third cluster consisted of 22 variants, which were related to proteomics datasets of the EMIF consortium with 1284 decreased levels of Aβ42 but not increased levels of pTau. quantified proteins and of Knight-ADRC with 696 quan- Nine GO terms were enriched and ‘migration’ and ‘tyrosine’ tified proteins (of which 42% overlap with the EMIF-AD are the words that occur most often in these terms, though MBD proteins). For GMNC there were 279 (22%) proteins each word only based on 2 GO terms each. The last cluster associated in the EMIF-AD MBD data (Online Resource of 20 variants group because they increased pTau, but did 1—Table 10) and 255 (36%) proteins in the Knight-ADRC not decrease Aβ42 levels. No GO terms are significantly data (Online Resource 1—Table 11). C16orf95 could only enriched in this gene cluster. be tested in the EMIF-AD MBD data in which 73 (6%) proteins were associated (Online Resource 1—Table 10). Only 2 proteins (CDH9 and DPP6) overlapped between the Discussion 2 loci. We studied the overlap in affected pathways between the associated protein lists. For GMNC, consistent func- We identified 2 loci (CR1 and APOE) for Aβ42, and 4 loci tional group networks between the 2 tested datasets were (BIN1, GMNC, C16orf95 and APOE) for pTau in a total of axon guidance and ephrin signaling (Online Resource 2— 13,116 individuals (discovery n = 8074; replication n = 5402 Fig. 29, Online Resource 1—Tables 12 and 13), while for individuals). In concordance with previous GWAS studies C16orf95 (only based on the EMIF-AD MBD data) gly- [12, 17], both proteins showed the strongest association for cosaminoglycan metabolism and ECM organization were APOE, where APOE ɛ4 decreased amyloid-beta levels and overrepresented functional groups (Online Resource 2— increased pTau levels, while APOE ɛ2 had the opposite Fig. 30, Online Resource 1—Table 14). There was little effect. We confirmed GMNC as a risk factor for CSF pTau overlap between the loci in the pathways that emerged from levels. We identified CR1 as a novel locus for CSF Aβ42 the protein lists. levels, and we observed 2 novel loci (BIN1 and C16orf95) 13 Acta Neuropathologica (2022) 144:821–842 Fig. 3  The effects of all AD-associated loci. The names of the loci are named according to their linked gene names in Bellinguez et al. (2021). Hierarchical clustering was performed on the rows and columns using Eucledian distances and the method ‘average’ for clustering Pathway enrichment analyses was performed on the four first clusters. The enrichment analyses are in Online Resource 1—Table 9. The upper bar shows the odds ratio for AD, where alleles for variants with protective effects have been flipped to show AD-risk increasing effects for all vari- ants. The increases in Aβ42 and pTau (positive Zscores) are shown in red and decreases in AB42 and PTau (negative Z scores) are shown in blue. * = P value < 0.05, X = 0.05 < P value < 0.001, # = 0.001 < P value < 5 × ­10–8, $ = P value < 5 × ­10–8 13 831 832 Acta Neuropathologica (2022) 144:821–842 for CSF pTau levels. So other than APOE, no risk loci current state are not applicable for clinical trials, we believe overlap is observed for Aβ42 and pTau, implying at least by improving the definition of the amyloid-tau clusters (e.g., partly separate genetic backgrounds for both pathological by increasing sample sizes), it will be possible to assign hallmarks. Amyloid-beta appears to be dominated by the individuals to one of the groups based on the distribution effect of APOE, while pTau is influenced by multiple genetic of their AD risk variants. For example, individuals with components. Such a divergence in genetic influences is not relatively more AD risk variants in the first cluster should in concordance with a genetic etiology where accumula- preferably be included in a clinical trial addressing amyloid tion of Tau tangles is a direct downstream effect of amyloid formation, while individuals with higher sub PRS scores for plaque formation, as proposed by the amyloid cascade the- the second immune cluster, would have a higher probability ory [29, 30]. Rather, it seems that these pathologies have an benefitting from a clinical trial targeting the immune system. independent component in its origin, as already extensively Alternatively, when sample sizes are increasing for future reviewed [64], thereby implying a dual-cascade hypothesis CSF amyloid and tau GWAS, an amyloid or tau PRS could in which sporadic AD pathogenesis is caused by defects in be calculated to identify individuals at high genetic risk for correlated but independent cellular processes. This is highly amyloid and tau, thereby defining the individuals that should relevant biological knowledge for the development of poten- be included in trials aiming to reduce amyloid plagues or tau tial AD treatments. The limited clinical efficacy of agents neurofibrillary tangles, respectively. that aim to reduce beta-amyloid plaques might potentially As mentioned above, APOE is the strongest locus for both be due to this [76]. CSF Aβ42 and pTau. The functional implications of the rela- In line with this implication is the observed difference tion between APOE and the 2 pathological hallmarks has in genetic subgroups based on CSF protein patterns as pro- been excessively studied and reviewed in the scientific field posed by the explorative cluster analysis, suggesting multiple of Alzheimer’s disease [20, 32, 60, 64]. In short, studies in Aβ42 and pTau related biological pathways to be involved human and mice models have shown higher Aβ42 levels and in the etiology AD. The first genetic subgroup is defined plaque load for APOE4 carriers. Whether this is a result of by the word ‘amyloid’, where BIN1, PICALM, ABCA7 and gain of toxic effect (e.g., enhanced ability of the APOE4 iso- CLU contribute most to this signal. The involvement of these form to bind to Aβ42 [40]), or loss of defensive mechanism genes in amyloid-beta pathology is supported by previous (e.g., less effective microglial response[20]), or a combina- studies [15, 46, 63, 79]. ‘Immune’ is labeling the second tion of both, remains to be elucidated. For Tau, studies have subgroup, thereby implying that the genes included in this shown hyperphosphorylation and faster accumulation for cluster are influencing AD by altering immune responses. APOE4 carriers, both in mice and human models [32, 64]. The largest genetic drivers of this signal are TREM2 and Again, the functional route through which APOE is affect- GRN which have been previously described for their func- ing the Tau aggregates is unknown, and multiple cellular tional role in AD via the immune system [45, 54]. A sub- processes are being proposed as mediators, including neu- stantial involvement of the immune system in AD is a gen- ronal endocytosis, lipid metabolism and glial function. An erally accepted concept, supported by clinical, functional important functional observation though is the influence of and genetic research [8]. As mentioned in the Results sec- APOE4 on tau accumulation in absence of Aβ42 pathology tion, this second cluster includes genes that have also been [55, 68], which is in agreement with the genetic implica- related to other dementia types (e.g., GRN, TMEM107B, tions of this current study. An unexpected observation for SNX1, MAPT, CTSB and CTSH). This is in line with more the APOE locus in this study is the successful colocalization recent insights where dysfunction of the immune response with an eQTL for the NKPD1 in a specific type of T cells. is seen as a common cause for multiple neurodegenerative However, due to the convincing functional involvement of diseases [28]. The third and fourth clusters are less straight- APOE in AD pathogenesis via the presence of cysteine or forward to interpret as many terms are similarly enriched, or arginine at APOE amino acid residuals 112 and 158, that we no significant enrichment is observed. The definition of the anticipate this successful NKPD1 eQTL colocalization result first 3 subclasses in its current state is the best approximate to be a false positive. within reach, though presumably not a perfect reflection The CR1-locus findings are in agreement with the well- of the genuine genetic etiology of AD. Subsequent genetic known observed association for AD risk, where rs6656401 studies with larger sample sizes are anticipated to facilitate (R2 = 0.88 with lead SNP of current study) carriers are more an improved understanding of the biological implications of susceptible for AD. The similarity in these association sig- the underlying genetic subclasses. nals is strengthened by the convincing colocalization of the The variety in subclasses of genetic contributors for AD CR1 locus between our CSF pTau observation and the CR1 etiology could mean that different patient groups might ben- locus of Kunkle et al. [43]. We furthermore observe a suc- efit from distinct AD treatment depending on the biological cessful colocalization with an eQTL for the CR1 gene in pathway that is affected. Although, our genetic results in its multiple brain regions. In concordance with this observation 13 Acta Neuropathologica (2022) 144:821–842 833 is functional work on the effect of CR1 further which sug- C16orf95 for ventricular volume are counterintuitive. For gests that CR1 is involved in AD pathogenesis by regulating both loci the allele that associated with an increase in pTau Aβ42 clearance in the brain itself, but also peripherally in pathology in our dementia cohorts associated with a smaller blood cells [81]. More recent research in red blood cells ventricular volume, implying less neurodegeneration. We of AD patients showed deficient CR1 immunoreactivity, explored if these loci work through the same biological including CR1-mediated capture of circulating amyloid- pathways using CSF proteomics data. The consistently beta [35]. They observed decreased CR1 protein levels in highlighted functional groups for the GMNC locus (axon red blood cells for CR1 SNPs that associate with higher guidance and ephrin signaling) were different than for the AD risk. The second novel locus in this study is BIN1 for C16orf95-locus (extracellular matrix components), thereby CSF pTau, for which we observe a successful colocaliza- implying at least partly distinct functional routes via which tion for an eTQL for the BIN1 gene in lymphoblastoid cell they influence pTau protein levels in CSF. The power of lines, implying BIN1 to be the causal gene for this locus. the proteomics dataset is rather limited. We anticipate that The BIN1-locus furthermore localizes with the same locus these analyses would benefit and potentially find overlapping in the AD GWAS, implying that a causal variant of the biological pathways, when AD proteomics datasets at hand BIN1-locus for AD contributes to AD pathogenesis via tau will increase in sample size, or apply the same proteomics pathology. BIN1 has already been linked to Tau pathology assay (rather than using different ones with little overlap in in several functional studies, first shown in fruit flies where proteins). a decrease in the BIN1 ortholog gene expression suppressed We furthermore identified two novel loci in the strati- Tau-mediated neurotoxicity [11]. More recently, research in fied analyses for Aβ42 levels: 7q11.22 for APOE ɛ4 non- mice showed physical protein interaction between BIN1 and carriers and 12q13.3 in individuals with abnormal amyloid Tau [51], and BIN1 involvement in Tau-dependent hyper- levels. We were unable to test for replication of these loci excitability in AD [66]. In human subjects, BIN1-carriers in independent replication datasets, as such analyses were were associated with lower memory performance, medi- not performed. Neither loci have been previously linked to ated by higher tau-PET levels [22]. Our observed BIN1-Tau AD, or any other trait. The lead SNP for the locus on chro- association contributes valuable knowledge on this topic, by mosome 7 is a common intronic variant for the lncRNA observing the same trend in a substantial larger study (89 vs LOC105375341, which according to GTEx is only expressed 13,118 individuals) using different techniques to measure in testis and prostate, and thereby not a promising causal Tau pathology (PET vs. CSF). We provide in vivo confirma- gene for AD or AD-related phenotypes. The locus on chro- tion that CR1 is associated with AD via Aβ42, while BIN1 mosome 12 consists of just one rare intronic variant that relates to Tau pathology. was only detected in the Dutch cohort with the largest sam- The third novel locus in this study is the region on ple size (n = 498). Future studies including more cohorts of genomic location 16q24.2 for pTau, of which the strongest large sample sizes are required to study this rare variant in associated variants are located within intronic regions of more detail. C16orf95. This locus has not been linked to (p)Tau pathol- Besides APOE and GMNC, no other loci from the lat- ogy or AD in previous research, though it associated to lat- est GWAS of CSF amyloid and tau [17] were replicated eral ventricular volume in the CHARGE study, including by this study. For Aβ42, it was not possible to test the 23.5 k healthy individuals [65]. Similarly, the 3q28-locus GLIS1-locus as this variant was not observed in our data, for pTau from our findings colocalizes inconclusively (pos- which is in concordance with the gnomAD browser [37] terior probability of 0.73) to lateral ventricular volume by reporting extremely low coverage and a MAF of 0.007 for the CHARGE study, implying that the same genetic risk rs185031519, the strongest GLIS1 SNP in Deming et al. For factors contribute to both phenotypes, strengthening the the other loci that we were unable to replicate (SERPINB1 notion that neurodegeneration and (p)Tau pathology are for Aβ42; and GLIS3, PCDH8, CTDP1 for pTau), there highly correlated. 3q28 has been linked to (p)Tau by pre- were no significant associations (P > 0.05) despite substan- vious CSF studies in dementia cohorts and was identified tial sample sizes (n > 7000). These differences in findings as the GMNC locus [12, 17]. In comparison to the latest might be due to differences in study design, for example GWAS of Deming et al. [17], increasing the sample size with inclusion of cohorts with other diagnoses and/or differences a small 10 k individuals in this study strengthens the associa- in analysis strategies. tion of this locus from 3.07 × ­10–11 to 1.19 × ­10–36, thereby The GCTA-based SNP-heritability estimates of 27% and turning it into a well-established locus for pTau pathology 34% for Aβ42 and pTau, respectively, are in a similar range (similar for Tau: Deming P = 3.07 × ­10−11 (n = 3146); current to the estimates calculated by Deming et al. [17] (36% for P = 9.65 × ­10–37 (n = 12,540). The C16orf95-locus convinc- Aβ42 and 25% for pTau). Notable is that the previous study ingly colocalized with the same locus in ventricular volume. estimated Aβ42 to be most heritable, while we observed The formerly reported directions of effect of GMNC and the highest estimate for pTau, though the standard errors 13 834 Acta Neuropathologica (2022) 144:821–842 were about 10% for both studies. Our LDSC-based SNP- work was carried out in the context of Parelsnoer clinical biobanks heritabilities show a similar trend with higher estimates for at Health-RI (https://​www.​health-​ri.​nl/​initi​atives/​parel​snoer). Part of the genotyping included in this work was funded by the JPND EADB pTau. Furthermore, these LDSC-based 13% for Aβ42 and grant (German Federal Ministry of Education and Research (BMBF) 21% for pTau protein levels are considerably higher than the grant: 01ED1619A). Alfredo Ramirez is also supported by the German 7% previously observed for the diagnosis AD. The higher Research Foundation (DFG) grants Nr: RA 1971/6-1, RA1971/7-1, heritability for the tested CSF protein levels strengthens the and RA 1971/8-1. We would like to thank patients and controls who participated in this project. The Genome Research @ Fundació ACE assumption that more objective measurable biological prop- project (GR@ACE) is supported by Grifols SA, Fundación bancaria erties are more strongly associated with AD pathogenesis ‘La Caixa’, Fundació ACE, and CIBERNED. A.R. and M.B. receive than the diagnostic classifications. support from the European Union/EFPIA Innovative Medicines Ini- Due to this assumption, the number of risk loci identified tiative Joint undertaking ADAPTED and MOPEAD projects (grant numbers 115975 and 115985, respectively). M.B. and A.R. are also in this study might be lower than anticipated. However, our supported by national grants PI13/02434, PI16/01861, PI17/01474, study design is suboptimal for obtaining the most powerful PI19/01240 and PI19/01301. Acción Estratégica en Salud is integrated model for genetic association identification. Meta-analyzing into the Spanish National R + D + I Plan and funded by ISCIII (Insti- 31 heterogeneous cohorts (differences in genotyping array, tuto de Salud Carlos III)–Subdirección General de Evaluación and the Fondo Europeo de Desarrollo Regional (FEDER–‘Una manera de hacer imputation references, CSF protein assays, and patient/ Europa’). The position held by I.dR. is funded by grant. FI20/00215. control ratios) is more challenging rather than analyzing PFIS Contratos Predoctorales de Formación en Investigación en Salud. a homogeneous dataset of similar sample size. Alterna- We would like to thank UCL Genomics, London, UK, for performing tively, increasing the sample size by adding more cohorts the genotyping analyses of the samples within the Gothenburg H70 Birth Cohort Studies and Clinical AD Sweden. The recruitment and to increase power is presumably a more feasible approach clinical characterization of research participants at Washington Uni- for the future. versity were supported by NIH P30AG066444, and P01AG003991. In conclusion, the current findings clearly show that This work was supported by access to equipment made possible by studying the genetic effects of AD-related endophenotypes the Hope Center for Neurological Disorders, the Neurogenomics and Informatics Center (NGI: https://​neuro​genom​ics.​wustl.​edu/)​and the has the potential to reveal novel associations, and highlight Departments of Neurology and Psychiatry at Washington University important biological insights. The clear distinction in genetic School of Medicine. Research at the Belgian EADB site is funded in findings for amyloid-beta and tau emphasizes the (partly) part by the Alzheimer Research Foundation (SAO-FRA), The Research genetic independence of these two biological mechanisms Foundation Flanders (FWO), and the University of Antwerp Research Fund. FK is supported by a BOF DOCPRO fellowship of the University in AD pathogenesis. Moreover, the identification of CR1 of Antwerp Research Fund. The work of Valdecilla was supported by and BIN1, which are the second and third strongest asso- grants from the Instituto de Salud Carlos III (Fondo de Investigación ciated AD locus after APOE, furthermore implies that by Sanitario, PI08/0139, PI12/02288, PI16/01652, and PI20/01011), the increasing sample size of genetic analysis in CSF biomark- JPND (DEMTEST PI11/03028), the CIBERNED, and the Siemens Healthineers. We thank the Valdecilla Biobank (PT17/0015/0019), ers it will become more apparent through which biologi- integrated into the Spanish Biobank Network, for their support and col- cal mechanisms certain AD loci have their effect on AD laboration in sample collection and management. Our heartfelt thanks pathogenesis. Even larger collaborative efforts with more to the participants of the Valdecilla Cohort for their generosity. The homogeneous sample definitions are, therefore, encouraged work of ADGEN was supported by EU Joint Programme—Neurode- generative Disease Research (301220) and the Academy of Finland to be undertaken to enhance our genetic understanding of (338182). The DELCODE study (Study-ID:BN012 ) was supported AD, ultimately leading to improved biological knowledge and conducted by the German Center for Neurodegenerative Diseases for the development of drug treatment. (DZNE). The data samples were provided by the DELCODE study group. Details and participating sites can be found at www.​dzne.​de/​ Supplementary Information The online version contains supplemen- en/​resea​rch/​studi​es/​clini​cal-​studi​es/​delco​de. The German Dementia tary material available at https://​doi.​org/​10.​1007/​s00401-​022-​02454-z. Competence Network (KND) is funded by the German Federal Min- istry of Education and Research (BMBF) grants Number: 01G10102, 01GI0420, 01GI0422, 01GI0423, 01GI0429, 01GI0431, 01GI0433, Acknowledgements This work was supported by a grant (European 04GI0434, 01GI0711. WF, SvdL, HHolstege, CT and PhS are recipi- Alzheimer DNA BioBank, EADB) from the EU Joint Programme, ents of ABOARD, which is a public-private partnership receiving Neurodegenerative Disease Research (JPND). Amsterdam dementia funding from ZonMW (#73305095007) and Health~Holland, Topsec- Cohort (ADC): Research of the Alzheimer center Amsterdam is part tor Life Sciences & Health (PPP-allowance; #LSHM20106). More of the neurodegeneration research program of Amsterdam Neurosci- than 30 partners participate in ABOARD (www.​aboard-​proje​ct.​nl). ence. The Alzheimer Center Amsterdam is supported by Stichting ABOARD also receives funding from de Hersenstichting, Edwin Bouw Alzheimer Nederland and Stichting VUmc fonds. The clinical data- Fonds and Gieskes-Strijbisfonds. IEJ was partially supported by NWO base structure was developed with funding from Stichting Dioraphte. Gravitation program BRAINSCAPES: A Roadmap from Neurogenet- Genotyping of the Dutch case-control samples was performed in the ics to Neurobiology (NWO: 024.004.012). AZ was supported by the context of EADB (European Alzheimer DNA biobank) funded by the Swedish Alzheimer Foundation (AF-939988, AF-930582, AF-646061, JPco-fuND FP-829-029 (ZonMW projectnumber 733051061).Part of AF-741361), and the Dementia Foundation (2020-04-13, 2021-04- the work described in this study was carried out in the context of the 17). ISk was supported by the Swedish state under the agreement Parelsnoer Institute (PSI). PSI was part of and funded by the Dutch between the Swedish government and the county councils, the ALF- Federation of University Medical Centers and has received initial fund- agreement (ALF 716681), the Swedish Research Council (no 11267, ing from the Dutch Government (from 2007-2011). Since 2020, this 13 Acta Neuropathologica (2022) 144:821–842 835 825-2012-5041, 2013-8717, 2015-02830, 2017-00639, 2019-01096), receives research support from: Biogen, EISAI, Alector and Parabon. Swedish Research Council for Health, Working Life and Welfare (no The funders of the study had no role in the collection, analysis, or 2001-2646, 2001-2835, 2001-2849, 2003-0234, 2004-0150, 2005- interpretation of data; in the writing of the report; or in the decision 0762, 2006-0020, 2008-1229, 2008-1210, 2012-1138, 2004-0145, to submit the paper for publication. CC is a member of the advisory 2006-0596, 2008-1111, 2010-0870, 2013-1202, 2013-2300, 2013- board of Vivid genetics, Halia Therapeutics and ADx Healthcare. HZ 2496), Swedish Brain Power, Hjärnfonden, Sweden (FO2016-0214, has served at scientific advisory boards for Denali, Roche Diagnos- FO2018-0214, FO2019- 0163), the Alzheimer's Association Zenith tics, Wave, Samumed, Siemens Healthineers, Pinteon Therapeutics Award (ZEN-01-3151), the Alzheimer's Association Stephanie B. and CogRx, has given lectures in symposia sponsored by Fujirebio, Overstreet Scholars (IIRG-00-2159), the Alzheimer's Association Alzecure and Biogen, and is a co-founder of Brain Biomarker Solu- (IIRG-03-6168, IIRG-09-131338) and the Bank of Sweden Tercen- tions in Gothenburg AB (BBS), which is a part of the GU Ventures tenary Foundation. SK was supported by the Swedish state under the Incubator Program (outside submitted work). KBlennow has served agreement between the Swedish government and the county councils, as a consultant, at advisory boards, or at data monitoring committees the ALF-agreement (ALFGBG-81392, ALF GBG-771071), the Swed- for Abcam, Axon, Biogen, JOMDD/Shimadzu. Julius Clinical, Lilly, ish Alzheimer Foundation (AF-842471, AF-737641, AF-939825), and MagQu, Novartis, Roche Diagnostics, and Siemens Healthineers, the Swedish Research Council (2019-02075). MW was supported by and is a co-founder of Brain Biomarker Solutions in Gothenburg AB the Swedish Research Council 2016-01590. DP was supported by (BBS), which is a part of the GU Ventures Incubator Program. OAA is BRAINSCAPES: A Roadmap from Neurogenetics to Neurobiology a consultant to HealthLytix, and received speaker’s honorarium from (grant no. 024.004.012), and a European Research Council advanced Lundbeck and Sunovion. SE has served at scientific advisory boards grant (Grant No, ERC-2018-AdG GWAS2FUNC 834057. HZ is a for Biogen, Danone, Eisai, icometrix, Pfizer, Novartis, Nutricia, Roche Wallenberg Scholar supported by grants from the Swedish Research and has received unrestricted research grants from ADx Neurosciences Council (2018-02532), the European Research Council (681712), and Janssen Pharmaceutica. CC receives research support from: Bio- Swedish State Support for Clinical Research (ALFGBG-720931), gen, EISAI, Alector, GSK and Parabon. The funders of the study had the Alzheimer Drug Discovery Foundation (ADDF), USA (201809- no role in the collection, analysis, or interpretation of data; in the writ- 2016862), the European Union’s Horizon 2020 research and innovation ing of the report; or in the decision to submit the paper for publica- programme under the Marie Skłodowska-Curie grant agreement No tion. CC is a member of the advisory board of Vivid genetics, Halia 860197 (MIRIADE), and the UK Dementia Research Institute at UCL. Therapeutics and ADx Healthcare. HHampel is an employee of Eisai KB was supported by the Swedish Research Council (#2017-00915), Inc. and serves as Senior Associate Editor for the Journal Alzheimer’s the Alzheimer Drug Discovery Foundation (ADDF), USA (#RDAPB- & Dementia; during the past three years he had received lecture fees 201809-2016615), the Swedish Alzheimer Foundation (#AF-742881), from Servier, Biogen and Roche, research grants from Pfizer, Avid, Hjärnfonden, Sweden (#FO2017-0243), the Swedish state under the and MSD Avenir (paid to the institution), travel funding from Eisai, agreement between the Swedish government and the County Councils, Functional Neuromodulation, Axovant, Eli Lilly and company, Takeda the ALF-agreement (#ALFGBG-715986), the European Union Joint and Zinfandel, GE-Healthcare and Oryzon Genomics, consultancy Program for Neurodegenerative Disorders (JPND2019-466-236), the fees from Qynapse, Jung Diagnostics, Cytox Ltd., Axovant, Anavex, National Institute of Health (NIH), USA, (grant #1R01AG068398- Takeda and Zinfandel, GE Healthcare, Oryzon Genomics, and Func- 01), and the Alzheimer’s Association 2021 Zenith Award (ZEN-21- tional Neuromodulation, and participated in scientific advisory boards 848495). CC receives support from the National Institutes of Health of Functional Neuromodulation, Axovant, Eisai, Eli Lilly and com- (R01AG044546, R01AG064877, RF1AG053303, R01AG058501, pany, Cytox Ltd., GE Healthcare, Takeda and Zinfandel, Oryzon U01AG058922, RF1AG058501, R01AG064614), and the Chuck Zuck- Genomics and Roche Diagnostics. DA participated in advisory boards erberg Initiative (CZI). from Fujirebio-Europe and Roche Diagnostics and received speaker honoraria from Fujirebio-Europe, Roche Diagnostics, Nutricia, Krka Data availability statement Genome-wide summary statistics have Farmacéutica S.L., Zambon S.A.U. and Esteve Pharmaceuticals S.A. been deposited to the European Bioinformatics Institute GWAS Cata- OG reports consulting fees from Eli Lilly, grants to his institution from log (https://w ​ ww.e​ bi.a​ c.u​ k/g​ was/) under accession IDs GCST90129599 Actelion, and prescreening activities for Julius Clinical/Toyama. ABK and GCST90129600. has been a PI in the drug trials Roche BN29553, Boehringer-Ingelheim 1346.0023 and is PI in Novo Nordisk NN6535-4730. AL has received personal fees for advisory board services and/or speaker honoraria Declarations from Fujirebio-Europe, Roche Diagnostics, Nutricia,Krka Farmacéu- tica SL, Biogen and Zambon. PSJ has received personal fees for ad- Conflict of interest Research programs of Wiesje van der Flier have visory board from Roche Diagnostics and Zambon. GS participated been funded by ZonMW, NWO, EU-FP7, EU-JPND, Alzheimer Ned- in one advisory board meeting from Biogen. MI is a paid consultant erland, CardioVascular Onderzoek Nederland, Health ~ Holland, Top- to BioArctic AB. KS is editor at Acta Neuropathologica and associate sector Life Sciences & Health, stichting Dioraphte, Gieskes-Strijbis editor at Alzheimer’s Research & Therapy. HZ has served at scientific fonds, stichting Equilibrio, Pasman stichting, stichting Alzheimer & advisory boards for Denali, Roche Diagnostics, Wave, Samumed, Sie- Neuropsychiatrie Foundation, Biogen MA Inc, Boehringer Ingelheim, mens Healthineers, Pinteon Therapeutics and CogRx, has given lec- Life-MI, AVID, Roche BV, Fujifilm, Combinostics. WF holds the Pas- tures in symposia sponsored by Fujirebio, Alzecure and Biogen, and is man chair. WF is recipient of ABOARD, which is a public–private a co-founder of Brain Biomarker Solutions in Gothenburg AB (BBS), partnership receiving funding from ZonMW (#73305095007) and which is a part of the GU Ventures Incubator Program (outside submit- Health ~ Holland, Topsector Life Sciences & Health (PPP-allowance; ted work). All other authors declare no financial interests or potential #LSHM20106). WF has performed contract research for Biogen MA conflicts of interest. Inc, and Boehringer Ingelheim. WF has been an invited speaker at Boehringer Ingelheim, Biogen MA Inc, Danone, Eisai, WebMD Neu- Open Access This article is licensed under a Creative Commons Attri- rology (Medscape), Springer Healthcare. WF is consultant to Oxford bution 4.0 International License, which permits use, sharing, adapta- Health Policy Forum CIC, Roche, and Biogen MA Inc. WF partici- tion, distribution and reproduction in any medium or format, as long pated in advisory boards of Biogen MA Inc and Roche. All funding as you give appropriate credit to the original author(s) and the source, is paid to her institution. WF was associate editor of Alzheimer, Re- provide a link to the Creative Commons licence, and indicate if changes search & Therapy in 2020/2021. WF is associate editor at Brain. CC 13 836 Acta Neuropathologica (2022) 144:821–842 were made. The images or other third party material in this article are 14. de Leeuw CA, Mooij JM, Heskes T, Posthuma D (2015) included in the article's Creative Commons licence, unless indicated MAGMA: generalized gene-set analysis of GWAS data. PLoS otherwise in a credit line to the material. If material is not included in Comput Biol 11:e1004219. https://​doi.​org/​10.​1371/​journ​al.​pcbi.​ the article's Creative Commons licence and your intended use is not 10042​19 permitted by statutory regulation or exceeds the permitted use, you will 15. De Roeck A, Van Broeckhoven C, Sleegers K (2019) The role of need to obtain permission directly from the copyright holder. To view a ABCA7 in Alzheimer’s disease: evidence from genomics, tran- copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/. scriptomics and methylomics. Acta Neuropathol 138:201–220. https://​doi.​org/​10.​1007/​s00401-​019-​01994-1 16. de Rojas I, Moreno-Grau S, Tesi N, Grenier-Boley B, Andrade V, Jansen IE et al (2021) Common variants in Alzheimer’s dis- References ease and risk stratification by polygenic risk scores. Nat Commun 12:3417. https://​doi.​org/​10.​1038/​s41467-​021-​22491-8 1. (2021) The Gene Ontology resource: enriching a GOld mine. 17. Deming Y, Li Z, Kapoor M, Harari O, Del-Aguila JL, Black K Nucleic Acids Res 49:D325–D334. https://​doi.​org/​10.​1093/​nar/​ et al (2017) Genome-wide association study identifies four novel gkaa1​113 loci associated with Alzheimer’s endophenotypes and disease 2. Alcolea D, Martínez-Lage P, Sánchez-Juan P, Olazarán J, Antúnez modifiers. Acta Neuropathol 133:839–856. https://​doi.​org/​10.​ C, Izagirre A et al (2015) Amyloid precursor protein metabo- 1007/​s00401-​017-​1685-y lism and inflammation markers in preclinical Alzheimer disease. 18. El Khoury J, Toft M, Hickman SE, Means TK, Terada K, Geula Neurology 85:626–633. https://​doi.​org/​10.​1212/​wnl.​00000​00000​ C et al (2007) Ccr2 deficiency impairs microglial accumulation 001859 and accelerates progression of Alzheimer-like disease. Nat Med 3. Alcolea D, Pegueroles J, Muñoz L, Camacho V, López-Mora D, 13:432–438. https://​doi.​org/​10.​1038/​nm1555 Fernández-León A et al (2019) Agreement of amyloid PET and 19. Elsworth B, Lyon M, Alexander T, Liu Y, Matthews P, Hallett CSF biomarkers for Alzheimer’s disease on Lumipulse. Ann Clin J et al (2020) The MRC IEU OpenGWAS data infrastructure. Transl Neurol 6:1815–1824. https://​doi.​org/​10.​1002/​acn3.​50873 bioRxiv. https://​doi.​org/​10.​1101/​2020.​08.​10.​244293 4. Almdahl IS, Lauridsen C, Selnes P, Kalheim LF, Coello C, 20. Fitz NF, Nam KN, Wolfe CM, Letronne F, Playso BE, Iordanova Gajdzik B et al (2017) Cerebrospinal fluid levels of amyloid beta BE et al (2021) Phospholipids of APOE lipoproteins activate 1–43 mirror 1–42 in relation to imaging biomarkers of Alzhei- microglia in an isoform-specific manner in preclinical models of mer’s disease. Front Aging Neurosci 9:9. https://​doi.​org/​10.​3389/​ Alzheimer’s disease. Nat Commun 12:3416. https://​doi.​org/​10.​ fnagi.​2017.​00009 1038/​s41467-​021-​23762-0 5. Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry 21. Fladby T, Pålhaugen L, Selnes P, Waterloo K, Bråthen G, Hes- JM et al (2000) Gene ontology: tool for the unification of biology. sen E et al (2017) Detecting at-risk Alzheimer’s disease cases. J The Gene Ontology Consortium. Nat Genet 25:25–29. https://d​ oi.​ Alzheimers Dis 60:97–105. https://​doi.​org/​10.​3233/​jad-​170231 org/​10.​1038/​75556 22. Franzmeier N, Rubinski A, Neitzel J, Ewers M, Weiner MW, 6. Bellenguez C, Küçükali F, Jansen IE, Kleineidam L, Moreno-Grau Aisen P et al (2019) The BIN1 rs744373 SNP is associated with S, Amin N et al (2022) New insights into the genetic etiology of increased tau-PET levels and impaired memory. Nat Commun Alzheimer’s disease and related dementias. Nat Genet 54:412– 10:1766. https://​doi.​org/​10.​1038/​s41467-​019-​09564-5 436. https://​doi.​org/​10.​1038/​s41588-​022-​01024-z 23. Geijselaers SLC, Aalten P, Ramakers I, De Deyn PP, Heijboer AC, 7. Benner C, Havulinna AS, Järvelin MR, Salomaa V, Ripatti S, Koek HL et al (2018) Association of cerebrospinal fluid (CSF) Pirinen M (2017) Prospects of fine-mapping trait-associated insulin with cognitive performance and CSF biomarkers of Alz- genomic regions by using summary statistics from genome-wide heimer’s disease. J Alzheimers Dis 61:309–320. https://​doi.​org/​ association studies. Am J Hum Genet 101:539–551. https://​doi.​ 10.​3233/​jad-​170522 org/​10.​1016/j.​ajhg.​2017.​08.​012 24. Giambartolomei C, Vukcevic D, Schadt EE, Franke L, Hingo- 8. Bettcher BM, Tansey MG, Dorothée G, Heneka MT (2021) rani AD, Wallace C et al (2014) Bayesian test for colocalisation Peripheral and central immune system crosstalk in Alzheimer between pairs of genetic association studies using summary sta- disease—a research prospectus. Nat Rev Neurol 17:689–701. tistics. PLoS Genet 10:e1004383. https://d​ oi.o​ rg/1​ 0.1​ 371/j​ ourna​ l.​ https://​doi.​org/​10.​1038/​s41582-​021-​00549-x pgen.​10043​83 9. Bulik-Sullivan BK, Loh P-R, Finucane HK, Ripke S, Yang J, Pat- 25. Goate A, Chartier-Harlin MC, Mullan M, Brown J, Crawford F, terson N et al (2015) LD Score regression distinguishes confound- Fidani L et al (1991) Segregation of a missense mutation in the ing from polygenicity in genome-wide association studies. Nat amyloid precursor protein gene with familial Alzheimer’s disease. Genet 47:291–295. https://​doi.​org/​10.​1038/​ng.​3211 Nature 349:704–706. https://​doi.​org/​10.​1038/​34970​4a0 10. Chang CC, Chow CC, Tellier LC, Vattikuti S, Purcell SM, Lee 26. Gottesman II, Gould TD (2003) The endophenotype concept in JJ (2015) Second-generation PLINK: rising to the challenge of psychiatry: etymology and strategic intentions. Am J Psychiatry larger and richer datasets. GigaScience. https://​doi.​org/​10.​1186/​ 160:636–645. https://​doi.​org/​10.​1176/​appi.​ajp.​160.4.​636 s13742-​015-​0047-8 27. Grimmer T, Riemenschneider M, Förstl H, Henriksen G, Klunk 11. Chapuis J, Hansmannel F, Gistelinck M, Mounier A, Van Cauwen- WE, Mathis CA et al (2009) Beta amyloid in Alzheimer’s disease: berghe C, Kolen KV et al (2013) Increased expression of BIN1 increased deposition in brain is reflected in reduced concentration mediates Alzheimer genetic risk by modulating tau pathology. in cerebrospinal fluid. Biol Psychiatry 65:927–934. https://d​ oi.o​ rg/​ Mol Psychiatry 18:1225–1234. https://d​ oi.o​ rg/1​ 0.1​ 038/m ​ p.2​ 013.1 10.​1016/j.​biops​ych.​2009.​01.​027 12. Cruchaga C, Kauwe JS, Harari O, Jin SC, Cai Y, Karch CM et al 28. Hammond TR, Marsh SE, Stevens B (2019) Immune Signaling (2013) GWAS of cerebrospinal fluid tau levels identifies risk vari- in Neurodegeneration. Immunity 50:955–974. https://​doi.​org/​10.​ ants for Alzheimer’s disease. Neuron 78:256–268. https://​doi.​org/​ 1016/j.​immuni.​2019.​03.​016 10.​1016/j.​neuron.​2013.​02.​026 29. Hardy JA, Higgins GA (1992) Alzheimer’s disease: the amyloid 13. Das S, Forer L, Schönherr S, Sidore C, Locke AE, Kwong A et al cascade hypothesis. Science (New York, NY) 256:184–185. (2016) Next-generation genotype imputation service and methods. https://​doi.​org/​10.​1126/​scien​ce.​15660​67 Nat Genet 48:1284–1287. https://​doi.​org/​10.​1038/​ng.​3656 13 Acta Neuropathologica (2022) 144:821–842 837 30. He Z, Guo JL, McBride JD, Narasimhan S, Kim H, Changolkar 45. Minami SS, Min S-W, Krabbe G, Wang C, Zhou Y, Asgarov R L et al (2018) Amyloid-β plaques enhance Alzheimer’s brain tau- et al (2014) Progranulin protects against amyloid β deposition and seeded pathologies by facilitating neuritic plaque tau aggregation. toxicity in Alzheimer’s disease mouse models. Nat Med 20:1157– Nat Med 24:29–38. https://​doi.​org/​10.​1038/​nm.​4443 1164. https://​doi.​org/​10.​1038/​nm.​3672 31. Hemani G, Zheng J, Elsworth B, Wade KH, Haberland V, Baird 46. Miyagawa T, Ebinuma I, Morohashi Y, Hori Y, Young Chang M, D et al (2018) The MR-Base platform supports systematic causal Hattori H et al (2016) BIN1 regulates BACE1 intracellular traf- inference across the human phenome. Elife. https://​doi.​org/​10.​ ficking and amyloid-β production. Hum Mol Genet 25:2948–2958. 7554/​eLife.​34408 https://​doi.​org/​10.​1093/​hmg/​ddw146 32. Husain MA, Laurent B, Plourde M (2021) APOE and Alzheimer’s 47. Mulder C, Verwey NA, van der Flier WM, Bouwman FH, Kok A, disease: from lipid transport to physiopathology and therapeutics. van Elk EJ (2010) Amyloid-beta(1–42), total tau, and phospho- Front Neurosci. https://​doi.​org/​10.​3389/​fnins.​2021.​630502 rylated tau as cerebrospinal fluid biomarkers for the diagnosis 33. Jansen IE, Savage JE, Watanabe K, Bryois J, Williams DM, Stein- of Alzheimer disease. Clin Chem 56:248–253. https://​doi.​org/​10.​ berg S et al (2019) Genome-wide meta-analysis identifies new loci 1373/​clinc​hem.​2009.​130518 and functional pathways influencing Alzheimer’s disease risk. Nat 48. Najar J, van der Lee SJ, Joas E, Wetterberg H, Hardy J, Guer- Genet 51:404–413. https://​doi.​org/​10.​1038/​s41588-​018-​0311-9 reiro R et al (2021) Polygenic risk scores for Alzheimer’s disease 34. Jessen F, Spottke A, Boecker H, Brosseron F, Buerger K, Catak are related to dementia risk in APOE ɛ4 negatives. Alzheimer’s C et al (2018) Design and first baseline data of the DZNE mul- Dementia: Diagn Assess Dis Monitor 13:e12142. https://​doi.​org/​ ticenter observational study on predementia Alzheimer’s disease 10.​1002/​dad2.​12142 (DELCODE). Alzheimer’s Res Ther 10:15. https://​doi.​org/​10.​ 49. Olsson B, Lautner R, Andreasson U, Öhrfelt A, Portelius E, Bjerke 1186/​s13195-​017-​0314-2 M et al (2016) CSF and blood biomarkers for the diagnosis of 35. Johansson JU, Brubaker WD, Javitz H, Bergen AW, Nishita D, Alzheimer’s disease: a systematic review and meta-analysis. Lan- Trigunaite A et al (2018) Peripheral complement interactions cet Neurol 15:673–684. https://​doi.​org/​10.​1016/​s1474-​4422(16)​ with amyloid β peptide in Alzheimer’s disease: polymorphisms, 00070-3 structure, and function of complement receptor 1. Alzheimer’s 50. Ramirez A, van der Flier WM, Herold C, Ramonet D, Heilmann S, Dementia 14:1438–1449. https://​doi.​org/​10.​1016/j.​jalz.​2018.​04.​ Lewczuk P et al (2014) SUCLG2 identified as both a determinator 003 of CSF Aβ1-42 levels and an attenuator of cognitive decline in 36. Kaipainen A, Jääskeläinen O, Liu Y, Haapalinna F, Nykänen N, Alzheimer’s disease. Hum Mol Genet 23:6644–6658. https://​doi.​ Vanninen R et al (2020) Cerebrospinal fluid and MRI biomarkers org/​10.​1093/​hmg/​ddu372 in neurodegenerative diseases: a retrospective memory clinic- 51. Sartori M, Mendes T, Desai S, Lasorsa A, Herledan A, Malman- based study. J Alzheimers Dis 75:751–765. https://​doi.​org/​10.​ che N et al (2019) BIN1 recovers tauopathy-induced long-term 3233/​jad-​200175 memory deficits in mice and interacts with Tau through Thr(348) 37. Karczewski KJ, Francioli LC, Tiao G, Cummings BB, Alföldi J, phosphorylation. Acta Neuropathol 138:631–652. https://​doi.​org/​ Wang Q et al (2020) The mutational constraint spectrum quan- 10.​1007/​s00401-​019-​02017-9 tified from variation in 141,456 humans. Nature 581:434–443. 52. Sherrington R, Froelich S, Sorbi S, Campion D, Chi H, Rogaeva https://​doi.​org/​10.​1038/​s41586-​020-​2308-7 EA et al (1996) Alzheimer’s disease associated with mutations 38. Karran E, Mercken M, De Strooper B (2011) The amyloid cascade in presenilin 2 is rare and variably penetrant. Hum Mol Genet hypothesis for Alzheimer’s disease: an appraisal for the develop- 5:985–988. https://​doi.​org/​10.​1093/​hmg/5.​7.​985 ment of therapeutics. Nat Rev Drug Discov 10:698–712. https://​ 53. Sherrington R, Rogaev EI, Liang Y, Rogaeva EA, Levesque G, doi.​org/​10.​1038/​nrd35​05 Ikeda M et al (1995) Cloning of a gene bearing missense muta- 39. Kerimov N, Hayhurst JD, Peikova K, Manning JR, Walter P, Kol- tions in early-onset familial Alzheimer’s disease. Nature 375:754– berg L et al (2021) A compendium of uniformly processed human 760. https://​doi.​org/​10.​1038/​37575​4a0 gene expression and splicing quantitative trait loci. Nat Genet 54. Shi Y, Holtzman DM (2018) Interplay between innate immu- 53:1290–1299. https://​doi.​org/​10.​1038/​s41588-​021-​00924-w nity and Alzheimer disease: APOE and TREM2 in the spot- 40. Kloske CM, Wilcock DM (2020) The important interface between light. Nat Rev Immunol 18:759–772. https://​doi.​org/​10.​1038/​ apolipoprotein E and neuroinflammation in Alzheimer’s disease. s41577-​018-​0051-1 Front Immunol. https://​doi.​org/​10.​3389/​fimmu.​2020.​00754 55. Shi Y, Yamada K, Liddelow SA, Smith ST, Zhao L, Luo W et al 41. Kolberg L, Raudvere U, Kuzmin I, Vilo J, Peterson H (2020) gpro- (2017) ApoE4 markedly exacerbates tau-mediated neurodegenera- filer2–an R package for gene list functional enrichment analysis tion in a mouse model of tauopathy. Nature 549:523–527. https://​ and namespace conversion toolset g:Profiler. F1000Res. https://​ doi.​org/​10.​1038/​natur​e24016 doi.​org/​10.​12688/​f1000​resea​rch.​24956.2 56. Sims R, van der Lee SJ, Naj AC, Bellenguez C, Badarinarayan N, 42. Kornhuber J, Schmidtke K, Frolich L, Perneczky R, Wolf S, Jakobsdottir J et al (2017) Rare coding variants in PLCG2, ABI3, Hampel H et al (2009) Early and differential diagnosis of demen- and TREM2 implicate microglial-mediated innate immunity in tia and mild cognitive impairment: design and cohort baseline Alzheimer’s disease. Nat Genet 49:1373–1384. https://d​ oi.o​ rg/1​ 0.​ characteristics of the German Dementia Competence Network. 1038/​ng.​3916 Dement Geriatr Cogn Disord 27:404–417. https://d​ oi.o​ rg/1​ 0.1​ 159/​ 57. Somers C, Struyfs H, Goossens J, Niemantsverdriet E, Luyckx 00021​0388 J, De Roeck N et al (2016) A decade of cerebrospinal fluid bio- 43. Kunkle BW, Grenier-Boley B, Sims R, Bis JC, Damotte V, Naj markers for Alzheimer’s disease in Belgium. J Alzheimers Dis AC et al (2019) Genetic meta-analysis of diagnosed Alzheimer’s 54:383–395. https://​doi.​org/​10.​3233/​jad-​151097 disease identifies new risk loci and implicates Aβ, tau, immunity 58. Sunderland T, Linker G, Mirza N, Putnam KT, Friedman DL, and lipid processing. Nat Genet 51:414–430. https://​doi.​org/​10.​ Kimmel LH et al (2003) Decreased beta-amyloid1-42 and 1038/​s41588-​019-​0358-2 increased tau levels in cerebrospinal fluid of patients with Alz- 44. Merico D, Isserlin R, Stueker O, Emili A, Bader GD (2010) heimer disease. JAMA 289:2094–2103. https://​doi.​org/​10.​1001/​ Enrichment map: a network-based method for gene-set enrichment jama.​289.​16.​2094 visualization and interpretation. PLoS ONE 5:e13984. https://d​ oi.​ 59. Taliun D, Harris D, Kessler M, Carlson J, Szpiech Z, Torres R org/​10.​1371/​journ​al.​pone.​00139​84 et al (2019) Sequencing of 53,831 diverse genomes from the 13 838 Acta Neuropathologica (2022) 144:821–842 NHLBI TOPMed Program. bioRxiv. https://​doi.​org/​10.​1101/​ (Oxford, England) 26:2190–2191. https://​doi.​org/​10.​1093/​bioin​ 563866 forma​tics/​btq340 60. Therriault J, Benedet AL, Pascoal TA, Mathotaarachchi S, Cham- 72. Wyss-Coray T, Yan F, Lin AH, Lambris JD, Alexander JJ, Quigg oun M, Savard M et al (2020) Association of apolipoprotein E ε4 RJ et al (2002) Prominent neurodegeneration and increased plaque with medial temporal tau independent of amyloid-β. JAMA Neu- formation in complement-inhibited Alzheimer’s mice. Proc Natl rol 77:470–479. https://​doi.​org/​10.​1001/​jaman​eurol.​2019.​4421 Acad Sci USA 99:10837–10842. https://​doi.​org/​10.​1073/​pnas.​ 61. Tijms BM, Gobom J, Reus L, Jansen I, Hong S, Dobricic V et al 16235​0199 (2020) Pathophysiological subtypes of Alzheimer’s disease based 73. Yang C, Farias FHG, Ibanez L, Suhy A, Sadler B, Fernandez on cerebrospinal fluid proteomics. Brain 143:3776–3792. https://​ MV et al (2021) Genomic atlas of the proteome from brain, CSF doi.​org/​10.​1093/​brain/​awaa3​25 and plasma prioritizes proteins implicated in neurological dis- 62. Tijms BM, Willemse EAJ, Zwan MD, Mulder SD, Visser PJ, orders. Nat Neurosci 24:1302–1312. https://​doi.​org/​10.​1038/​ van Berckel BNM et al (2018) Unbiased approach to counter- s41593-​021-​00886-6 act upward drift in cerebrospinal fluid amyloid-β 1–42 analysis 74. Yang J, Lee SH, Goddard ME, Visscher PM (2011) GCTA: a results. Clin Chem 64:576–585. https://d​ oi.o​ rg/1​ 0.1​ 373/c​ linch​ em.​ tool for genome-wide complex trait analysis. Am J Hum Genet 2017.​281055 88:76–82. https://​doi.​org/​10.​1016/j.​ajhg.​2010.​11.​011 63. Uddin MS, Kabir MT, Begum MM, Islam MS, Behl T, Ashraf 75. Yeh FL, Wang Y, Tom I, Gonzalez LC, Sheng M (2016) TREM2 GM (2021) Exploring the role of CLU in the pathogenesis of binds to apolipoproteins, including APOE and CLU/APOJ, and Alzheimer’s disease. Neurotox Res 39:2108–2119. https://d​ oi.o​ rg/​ thereby facilitates uptake of amyloid-beta by microglia. Neuron 10.​1007/​s12640-​020-​00271-4 91:328–340. https://​doi.​org/​10.​1016/j.​neuron.​2016.​06.​015 64. van der Kant R, Goldstein LSB, Ossenkoppele R (2020) 76. Yiannopoulou KG, Anastasiou AI, Zachariou V, Pelidou SH Amyloid-β-independent regulators of tau pathology in Alzhei- (2019) Reasons for failed trials of disease-modifying treatments mer disease. Nat Rev Neurosci 21:21–35. https://d​ oi.o​ rg/1​ 0.1​ 038/​ for Alzheimer disease and their contribution in recent research. s41583-​019-​0240-3 Biomedicines. https://​doi.​org/​10.​3390/​biome​dicin​es704​0097 65. Vojinovic D, Adams HH, Jian X, Yang Q, Smith AV, Bis JC et al 77. Young AMH, Kumasaka N, Calvert F, Hammond TR, Knights A, (2018) Genome-wide association study of 23,500 individuals Panousis N et al (2021) A map of transcriptional heterogeneity identifies 7 loci associated with brain ventricular volume. Nat and regulatory variation in human microglia. Nat Genet 53:861– Commun 9:3945. https://​doi.​org/​10.​1038/​s41467-​018-​06234-w 868. https://​doi.​org/​10.​1038/​s41588-​021-​00875-2 66. Voskobiynyk Y, Roth JR, Cochran JN, Rush T, Carullo NV, Mes- 78. Zettergren A, Höglund K, Kern S, Thorvaldsson V, Johan Skoog ina JS et al (2020) Alzheimer’s disease risk gene BIN1 induces M, Hansson O et al (2019) Association of IL1RAP-related genetic Tau-dependent network hyperexcitability. Elife. https://​doi.​org/​ variation with cerebrospinal fluid concentration of Alzheimer- 10.​7554/​eLife.​57354 associated tau protein. Sci Rep 9:2460. https://​doi.​org/​10.​1038/​ 67. Wallace C (2020) Eliciting priors and relaxing the single causal s41598-​018-​36650-3 variant assumption in colocalisation analyses. PLoS Genet 79. Zhao Z, Sagare AP, Ma Q, Halliday MR, Kong P, Kisler K et al 16:e1008720. https://​doi.​org/​10.​1371/​journ​al.​pgen.​10087​20 (2015) Central role for PICALM in amyloid-β blood-brain barrier 68. Wang C, Najm R, Xu Q, Jeong DE, Walker D, Balestra ME transcytosis and clearance. Nat Neurosci 18:978–987. https://​doi.​ et al (2018) Gain of toxic apolipoprotein E4 effects in human org/​10.​1038/​nn.​4025 iPSC-derived neurons is ameliorated by a small-molecule struc- 80. Zheng J, Erzurumluoglu AM, Elsworth BL, Kemp JP, Howe L, ture corrector. Nat Med 24:647–657. https://​doi.​org/​10.​1038/​ Haycock PC et al (2017) LD Hub: a centralized database and s41591-​018-​0004-z web interface to perform LD score regression that maximizes the 69. Watanabe K, Taskesen E, van Bochoven A, Posthuma D (2017) potential of summary level GWAS data for SNP heritability and Functional mapping and annotation of genetic associations genetic correlation analysis. Bioinformatics (Oxford, England) with FUMA. Nat Commun 8:1826. https://​doi.​org/​10.​1038/​ 33:272–279. https://​doi.​org/​10.​1093/​bioin​forma​tics/​btw613 s41467-​017-​01261-5 81. Zhu X-C, Yu J-T, Jiang T, Wang P, Cao L, Tan L (2015) CR1 in 70. Wightman DP, Jansen IE, Savage JE, Shadrin AA, Bahrami S, Alzheimer’s disease. Mol Neurobiol 51:753–765. https://​doi.​org/​ Rongve A et al (2020) Largest GWAS (N=1,126,563) of Alzhei- 10.​1007/​s12035-​014-​8723-8 mer’s disease implicates microglia and immune cells. medRxiv. https://​doi.​org/​10.​1101/​2020.​11.​20.​20235​275 Publisher's Note Springer Nature remains neutral with regard to 71. Willer CJ, Li Y, Abecasis GR (2010) METAL: fast and efficient jurisdictional claims in published maps and institutional affiliations. meta-analysis of genomewide association scans. Bioinformatics Authors and Affiliations Iris E. Jansen1,2,3 · Sven J. van der Lee1,2,4 · Duber Gomez‑Fonseca5,6,7 · Itziar de Rojas8,9 · Maria Carolina Dalmasso10,11 · Benjamin Grenier‑Boley12 · Anna Zettergren13 · Aniket Mishra14 · Muhammad Ali5,6,7 · Victor Andrade10,15 · Céline Bellenguez12 · Luca Kleineidam10,15,16 · Fahri Küçükali17,18 · Yun Ju Sung5,6,7 · Niccolo Tesí1,2,4 · Ellen M. Vromen1,2 · Douglas P. Wightman3 · Daniel Alcolea9,19 · Montserrat Alegret8,9 · Ignacio Alvarez20,21 · Philippe Amouyel12 · Lavinia Athanasiu22 · Shahram Bahrami22 · Henri Bailly23 · Olivia Belbin9,19 · Sverre Bergh24,25 · Lars Bertram26 · Geert Jan Biessels27 · Kaj Blennow28,29 · Rafael Blesa9,19 · Mercè Boada8,9 · Anne Boland30 · Katharina Buerger31,32 · Ángel Carracedo33,34 · Laura Cervera‑Carles9,19 · Geneviève Chene14,35 · Jurgen A. H. R. Claassen36,37 · Stephanie Debette14,35,38 · Jean‑Francois Deleuze30 · Peter Paul de Deyn39 · Janine Diehl‑Schmid40,41 · Srdjan Djurovic42,43 · 13 Acta Neuropathologica (2022) 144:821–842 839 Oriol Dols‑Icardo9,19 · Carole Dufouil14,44 · Emmanuelle Duron45 · Emrah Düzel46,47 · EADB consortium · Tormod Fladby48,49 · Juan Fortea9,19 · Lutz Frölich50 · Pablo García‑González8,9 · Maria Garcia‑Martinez51 · Ina Giegling52 · Oliver Goldhardt40 · Johan Gobom29 · Timo Grimmer40 · Annakaisa Haapasalo53 · Harald Hampel54,55 · Olivier Hanon23,56 · Lucrezia Hausner50 · Stefanie Heilmann‑Heimbach57 · Seppo Helisalmi58 · Michael T. Heneka15,16 · Isabel Hernández8,9 · Sanna‑Kaisa Herukka59 · Henne Holstege1,2,4 · Jonas Jarholm60 · Silke Kern13,61 · Anne‑Brita Knapskog62 · Anne M. Koivisto59,63,64 · Johannes Kornhuber65 · Teemu Kuulasmaa66 · Carmen Lage51,67 · Christoph Laske68,69 · Ville Leinonen70,71 · Piotr Lewczuk65,72 · Alberto Lleó9,19 · Adolfo López de Munain9,73,74,75 · Sara Lopez‑Garcia51 · Wolfgang Maier15 · Marta Marquié8,9 · Merel O. Mol78 · Laura Montrreal8 · Fermin Moreno9,73,74 · Sonia Moreno‑Grau8,9 · Gael Nicolas77 · Markus M. Nöthen57 · Adelina Orellana8,9 · Lene Pålhaugen48,49 · Janne M. Papma78 · Florence Pasquier12 · Robert Perneczky31,79,80,81 · Oliver Peters46,82 · Yolande A. L. Pijnenburg1,2 · Julius Popp83,84 · Danielle Posthuma3 · Ana Pozueta51 · Josef Priller82,85,86 · Raquel Puerta8 · Inés Quintela33 · Inez Ramakers87 · Eloy Rodriguez‑Rodriguez51 · Dan Rujescu52 · Ingvild Saltvedt88,89 · Pascual Sanchez‑Juan90 · Philip Scheltens1,2 · Norbert Scherbaum91 · Matthias Schmid92 · Anja Schneider15,16 · Geir Selbæk25,48,62 · Per Selnes49 · Alexey Shadrin22 · Ingmar Skoog13,61 · Hilkka Soininen59 · Lluís Tárraga8,9 · Stefan Teipel93,94 · The GR@ACE study group · Betty Tijms1,2 · Magda Tsolaki95 · Christine Van Broeckhoven18,96 · Jasper Van Dongen17,18 · John C. van Swieten78 · Rik Vandenberghe97,98 · Jean‑Sébastien Vidal23 · Pieter J. Visser1,2,99,100 · Jonathan Vogelgsang101,102 · Margda Waern13,103 · Michael Wagner15,16 · Jens Wiltfang101,104,105 · Mandy M. J. Wittens18,106 · Henrik Zetterberg28,29,107,108,109 · Miren Zulaica9,73,74 · Cornelia M. van Duijn76,110 · Maria Bjerke18,106,111 · Sebastiaan Engelborghs18,106,111,112 · Frank Jessen16,113,114 · Charlotte E. Teunissen2,115 · Pau Pastor116 · Mikko Hiltunen117 · Martin Ingelsson118,119,120 · Ole A. Andreassen22,121 · Jordi Clarimón9,19 · Kristel Sleegers17,18 · Agustín Ruiz8,9 · Alfredo Ramirez10,15,16,114,122 · Carlos Cruchaga5,6,7 · Jean‑Charles Lambert12 · Wiesje van der Flier1,2 11 * Iris E. Jansen Neurosciences and Complex Systems Unit (ENyS), i.e.jansen@vu.nl CONICET, Hospital El Cruce, National University A. Jauretche (UNAJ), Florencio Varela, Argentina * Wiesje van der Flier 12 wm.vdflier@amsterdamumc.nl Univ. Lille, Inserm, CHU Lille, Institut Pasteur de Lille, U1167 - RID-AGE / Labex DISTALZ - Facteurs de 1 Alzheimer Center Amsterdam, Neurology, Vrije Universiteit risque et déterminants moléculaires des maladies liées au Amsterdam, Amsterdam UMC Location VUmc, Amsterdam, vieillissement, F‑59000 Lille, France The Netherlands 13 Neuropsychiatric Epidemiology Unit, Department 2 Amsterdam Neuroscience, Neurodegeneration, Amsterdam, of Psychiatry and Neurochemistry, Institute of Neuroscience The Netherlands and Physiology, The Sahlgrenska Academy, Centre 3 for Ageing and Health (AGECAP) at the University Department of Complex Trait Genetics, Center of Gothenburg, Gothenburg, Sweden for Neurogenomics and Cognitive Research, Amsterdam 14 Neuroscience, VU Amsterdam, Amsterdam, The Netherlands University of Bordeaux, Inserm, Bordeaux Population 4 Health Research Center, Team VINTAGE, UMR 1219, Section Genomics of Neurodegenerative Diseases and Aging, 33000 Bordeaux, France Human Genetics, Vrije Universiteit Amsterdam, Amsterdam 15 UMC location VUmc, Amsterdam, The Netherlands Department of Neurodegenerative Diseases and Geriatric 5 Psychiatry, University Hospital Bonn, Medical Faculty, Department of Psychiatry, Washington University School Bonn, Germany of Medicine, St Louis, MO, USA 16 6 German Center for Neurodegenerative Diseases (DZNE), NeuroGenomics and Informatics, Washington University Bonn, Germany School of Medicine, St Louis, MO, USA 17 7 Complex Genetics of Alzheimer’s Disease Group, VIB Hope Center for Neurological Disorders, Washington Center for Molecular Neurology, VIB, Antwerp, Belgium University School of Medicine, St Louis, MO, USA 18 8 Department of Biomedical Sciences, University of Antwerp, Research Center and Memory Clinic, Ace Alzheimer Antwerp, Belgium Center Barcelona, Universitat Internacional de Catalunya, 19 Barcelona, Spain Sant Pau Memory Unit, Department of Neurology, Institut 9 d’Investigacions Biomèdiques Sant Pau - Hospital de Sant CIBERNED, Network Center for Biomedical Research Pau, Universitat Autònoma de Barcelona, Barcelona, Spain in Neurodegenerative Diseases, National Institute of Health 20 Carlos III, Madrid, Spain Memory Disorders Unit, Department of Neurology, Hospital 10 Universitari Mutua de Terrassa, Terrassa, Spain Division of Neurogenetics and Molecular Psychiatry, 21 Department of Psychiatry and Psychotherapy, Faculty Fundació per a la Recerca Biomèdica i Social Mútua de of Medicine and University Hospital Cologne, University Terrassa, Terrassa, Spain of Cologne, Cologne, Germany 13 840 Acta Neuropathologica (2022) 144:821–842 22 46 NORMENT Centre, Institute of Clinical Medicine, German Center for Neurodegenerative Diseases (DZNE), University of Oslo and Division of Mental Health, Oslo, Magdeburg, Germany Norway 47 Institute of Cognitive Neurology and Dementia Research 23 Université Paris Cité, EA4468, Maladie d’Alzheimer, (IKND), Otto-von-Guericke University, Magdeburg, F‑75013 Paris, France Germany 24 48 The Research-Centre for Age-Related Functional Decline Institute of Clinical Medicine, University of Oslo, Oslo, and Disease, Innlandet Hospital Trust, Brumunddal, Norway Norway 25 49 Norwegian National Centre for Ageing and Health, Vestfold Department of Neurology, Akershus University Hospital, Hospital Trust, Tønsberg, Norway Lorenskog, Norway 26 50 Lübeck Interdisciplinary Platform for Genome Analytics, Department of Geriatric Psychiatry, Central Institute University of Lübeck, Lübeck, Germany of Mental Health, Medical Faculty Mannheim, Heidelberg, 27 Germany Department of Neurology, UMC Utrecht Brain Center, 51 Utrecht, The Netherlands Cognitive Impairment Unit, Neurology Service, “Marqués 28 de Valdecilla” University Hospital, Institute for Research Department of Psychiatry and Neurochemistry, Institute “Marques de Valdecilla” (IDIVAL), University of Cantabria, of Neuroscience and Physiology, The Sahlgrenska Academy Santander, Spain, and Centro de Investigación Biomédica en at the University of Gothenburg, Mölndal, Sweden Red sobre Enfermedades Neurodegenerativas (CIBERNED), 29 Clinical Neurochemistry Laboratory, Sahlgrenska University Madrid, Spain Hospital, Mölndal, Sweden 52 Division of General Psychiatry, Department of Psychiatry 30 Université Paris-Saclay, CEA, Centre National de Recherche and Psychotherapy, Medical University of Vienna, Vienna, en Génomique Humaine, 91057 Evry, France Austria 31 53 German Center for Neurodegenerative Diseases (DZNE, A.I. Virtanen Institute for Molecular Sciences, University Munich), Munich, Germany of Eastern Finland, Kuopio, Finland 32 54 Institute for Stroke and Dementia Research (ISD), University Alzheimer Precision Medicine (APM), Sorbonne University, Hospital, LMU Munich, Munich, Germany AP-HP, Pitié-Salpêtrière Hospital, Paris, France 33 55 Grupo de Medicina Xenómica, Centro Nacional de Neurology Business Group, Eisai Inc, 100 Tice Blvd, Genotipado (CEGEN‑PRB3‑ISCIII), Universidade de Woodcliff Lake, NJ 07677, USA Santiago de Compostela, Santiago de Compostela, Spain 56 Service gériatrie, Centre Mémoire de Ressources et 34 Fundación Pública Galega de Medicina Recherches Ile de France-Broca, AP-HP, Hôpital Broca, Xenómica-CIBERER-IDIS, Santiago de Compostela, Spain F‑75013 Paris, France 35 57 Department of Neurology, CHU de Bordeaux, Institute of Human Genetics, University of Bonn, School 33000 Bordeaux, France of Medicine and University Hospital Bonn, 53127 Bonn, 36 Germany Radboudumc Alzheimer Center, Department of Geriatrics, 58 Radboud University Medical Center, Nijmegen, Institute of Clinical Medicine, Internal Medicine, University The Netherlands of Eastern Finland, Kuopio, Finland 37 59 Donders Center for Medical Neuroscience, Nijmegen, Department of Neurology, Institute of Clinical Medicine, The Netherlands University of Eastern Finland, Kuopio, Finland 38 60 Department of Neurology, Boston University School Department of Neurology, Akershus University Hospital, of Medicine, Boston, MA 2115, USA Lorenskog, Norway 39 61 Department of Neurology and Alzheimer Center Groningen, Region Västra Götaland, Sahlgrenska University Hospital, University Medical Center Groningen, Groningen, Psychiatry, Cognition and Old Age Psychiatry Clinic, The Netherlands Gothenburg, Sweden 40 62 Center for Cognitive Disorders, Department of Psychiatry Department of Geriatric Medicine, Oslo University Hospital, and Psychotherapy, Klinikum rechts der Isar, Technical Oslo, Norway University of Munich, School of Medicine, Munich, 63 Department of Neurology, Kuopio University Hospital, Germany Kuopio, Finland 41 kbo-Inn-Salzach-Hospital, Wasserburg am Inn, Germany 64 Department of Neurology, Helsinki University Hospital, 42 Department of Medical Genetics, Oslo University Hospital, Helsinki, Finland Oslo, Norway 65 Department of Psychiatry and Psychotherapy, 43 Department of Clinical Science, NORMENT Centre, Universitätsklinikum Erlangen, and Friedrich-Alexander University of Bergen, Bergen, Norway Universität Erlangen-Nürnberg, Erlangen, Germany 44 66 Pôle de Santé Publique Centre Hospitalier Universitaire Bioinformatics Center, Institute of Biomedicine, University (CHU) de Bordeaux, Bordeaux, France of Eastern Finland, Kuopio, Finland 45 67 Univerisité Paris-Saclay. Inserm 1178 MOODS, Paris, Atlantic Fellow at the Global Brain Health Institute (GBHI) France ‑, University of California, San Francisco, USA 13 Acta Neuropathologica (2022) 144:821–842 841 68 92 German Center for Neurodegenerative Diseases (DZNE), Institute of Medical Biometry, Informatics Tübingen, Germany and Epidemiology, University Hospital of Bonn, Bonn, 69 Germany Section for Dementia Research, Hertie Institute 93 for Clinical Brain Research and Department of Psychiatry German Center for Neurodegenerative Diseases (DZNE), and Psychotherapy, University of Tübingen, Tübingen, Rostock, Germany Germany 94 Department of Psychosomatic Medicine, Rostock University 70 Institute of Clinical Medicine, Neurosurgery, University Medical Center, Gehlsheimer Str. 20, 18147 Rostock, of Eastern Finland, Kuopio, Finland Germany 71 95 Department of Neurosurgery, Kuopio University Hospital, 1st Department of Neurology, School of Medicine, Faculty Kuopio, Finland of Health Sciences, Aristotle University of Thessaloniki, 72 Thessaloniki, Makedonia, Greece Department of Neurodegeneration Diagnostics, Medical 96 University of Białystok, Białystok, Poland Neurodegenerative Brain Diseases Group, VIB Center 73 for Molecular Neurology, VIB, Antwerp, Belgium Hospital Universitario Donostia-OSAKIDETZA, Donostia, 97 Spain Neurology, University Hospitals Leuven, Leuven, Belgium 74 98 Instituto Biodonostia, San Sebastián, Spain Laboratory for Cognitive Neurology, Department 75 of Neurosciences, Leuven Brain Institute, Leuven, Belgium University of The Basque Country, San Sebastian, Spain 99 76 Alzheimer Center Limburg, School for Mental Health Department of Epidemiology, ErasmusMC, Rotterdam, and Neuroscience Maastricht University, Maastricht, The Netherlands The Netherlands 77 Department of Genetics and CNR‑MAJ, Normandie Univ, 100 Department of Neurobiology, Care Sciences and Society, UNIROUEN, Inserm U1245 and CHU Rouen, Rouen, France Division of Neurogeriatrics Karolinska Institutet, Stockholm, 78 Department of Neurology and Alzheimer Center Erasmus Sweden MC, Erasmus MC University Medical Center, Rotterdam, 101 Department of Psychiatry and Psychotherapy, University The Netherlands Medical Center Goettingen, Göttingen, Germany 79 Department of Psychiatry and Psychotherapy, University 102 Department of Psychiatry, Harvard Medical School, McLean Hospital, LMU Munich, Munich, Germany Hospital, Belmont, MA, USA 80 Munich Cluster for Systems Neurology (SyNergy) Munich, 103 Region Västra Götaland, Sahlgrenska University Hospital, Munich, Germany Psychiatry, Psychosis Clinic, Gothenburg, Sweden 81 Ageing Epidemiology Research Unit, School of Public 104 German Center for Neurodegenerative Diseases (DZNE), Health, Imperial College London, London, UK Göttingen, Germany 82 German Center for Neurodegenerative Diseases (DZNE), 105 Medical Science Department, iBiMED, Aveiro, Portugal Berlin, Germany 106 83 Center for Neurosciences (C4N), Vrije Universiteit Brussel, Department of Geriatric Psychiatry, University Hospital Brussels, Belgium of Psychiatry Zürich and University of Zürich, Zurich, 107 Switzerland Department of Neurodegenerative Disease, UCL Institute 84 of Neurology, London, UK Old Age Psychiatry, Department of Psychiatry, University 108 Hospital of Lausanne, Lausanne, Switzerland UK Dementia Research Institute at UCL, London, UK 85 109 Department of Psychiatry and Psychotherapy, Charité, Hong Kong Center for Neurodegenerative Diseases, Charitéplatz 1, 10117 Berlin, Germany Hong Kong, China 86 110 Department of Psychiatry and Psychotherapy, Klinikum Nuffield Department of Population Health, Oxford rechts der isar, Technical University Munich, 81675 Munich, University, Oxford, UK Germany 111 Laboratory of Neurochemistry, Universitair Ziekenhuis 87 Department of Psychiatry and Neuropsychologie, Alzheimer Brussel, Brussels, Belgium Center Limburg, Maastricht University, Maastricht, 112 Department of Neurology, Universitair Ziekenhuis Brussel, The Netherlands Brussels, Belgium 88 Department of Neuromedicine and Movement Science, 113 Department of Psychiatry and Psychotherapy, Faculty Norwegian University of Science and Technology (NTNU), of Medicine and University Hospital Cologne, University Trondheim, Norway of Cologne, Cologne, Germany 89 Department of Geriatrics, St Olav Hospital, University 114 Cluster of Excellence Cellular Stress Responses Hospital of Trondheim, Trondheim, Norway in Aging‑Associated Diseases (CECAD), University 90 Alzheimer’s Centre Reina Sofia-CIEN Foundation-ISCIII, of Cologne, Cologne, Germany 28031 Madrid, Spain 115 Neurochemistry Lab, Department of Clinical Chemistry, 91 Department of Psychiatry and Psychotherapy, Medical Amsterdam Neuroscience, Vrije Universiteit Amsterdam, Faculty, LVR-Hospital Essen, University of Duisburg-Essen, Amsterdam UMC, Amsterdam, The Netherlands Essen, Germany 13 842 Acta Neuropathologica (2022) 144:821–842 116 120 Unit of Neurodegenerative diseases, Department Department of Medicine and Tanz Centre for Research of Neurology, University Hospital Germans Trias i Pujol in Neurodegenerative Diseases, University of Toronto, and The Germans Trias i Pujol Research Institute (IGTP) Toronto, Canada Badalona, Barcelona, Spain 121 Addiction, Oslo University Hospital, 0407 Oslo, Norway 117 Institute of Biomedicine, University of Eastern Finland, 122 Department of Psychiatry, Glenn Biggs Institute Kuopio, Finland for Alzheimer’s and Neurodegenerative Diseases, 118 Department of Public Health and Caring Sciences, Molecular San Antonio, TX, USA Geriatrics, Rudbeck Laboratory, Uppsala University, Uppsala, Sweden 119 Krembil Brain Institute, University Health Network, Toronto, Ontario, Canada 13