ORIGINAL RESEARCH published: 21 March 2022 doi: 10.3389/fnagi.2022.840651 Genome-Wide Association Study of Edited by: Alzheimer’s Disease Brain Imaging Can Martin Zhang, Massachusetts General Hospital and Harvard Medical School, Biomarkers and Neuropsychological United States Reviewed by: Phenotypes in the European Medical Davide Chiasserini, University of Perugia, Italy Information Framework for Xia Li, Shanghai Jiao Tong University, China Alzheimer’s Disease Multimodal *Correspondence: Lars Bertram Biomarker Discovery Dataset lars.bertram@uni-luebeck.de † These Jan Homann 1† , Tim Osburg 1† , Olena Ohlei 1 , Valerija Dobricic 1 , Laura Deecke 1 , authors have contributed equally to this work Isabelle Bos 2,3 , Rik Vandenberghe 4,5 , Silvy Gabel 4 , Philip Scheltens 3 , ‡ Present Charlotte E. Teunissen 6 , Sebastiaan Engelborghs 7,8 , Giovanni Frisoni 9,10 , Olivier Blin 11 , address: Jill C. Richardson 12 , Regis Bordet 13 , Alberto Lleó 14 , Daniel Alcolea 14 , Julius Popp 15,16 , Johannes Streffer, AC Immune SA, EPFL Innovation Christopher Clark 15 , Gwendoline Peyratout 16 , Pablo Martinez-Lage 17 , Mikel Tainta 17 , Park, Lausanne, Switzerland Richard J. B. Dobson 18,19,20,21,22 , Cristina Legido-Quigley 23,24 , Kristel Sleegers 25,26 , Christine Van Broeckhoven 26,27 , Michael Wittig 28 , Andre Franke 28 , Christina M. Lill 1,29 , Specialty section: Kaj Blennow 30,31 , Henrik Zetterberg 30,31,32,33 , Simon Lovestone 34 , Johannes Streffer 35,36‡ , This article was submitted to Mara ten Kate 4,37 , Stephanie J. B. Vos 2 , Frederik Barkhof 37,38 , Pieter Jelle Visser 2,4,39 and Alzheimer’s Disease and Related Lars Bertram 1,40* Dementias, 1 a section of the journal Lübeck Interdisciplinary Platform for Genome Analytics (LIGA), University of Lübeck, Lübeck, Germany, 2 Department Frontiers in Aging Neuroscience of Psychiatry and Neuropsychology, School for Mental Health and Neuroscience, Alzheimer Centrum Limburg, Maastricht University, Maastricht, Netherlands, 3 Department of Neurology, Alzheimer Center Amsterdam, Amsterdam University Received: 21 December 2021 Medical Centers, Amsterdam Neuroscience, Vrije Universiteit Amsterdam, Amsterdam, Netherlands, 4 Laboratory Accepted: 15 February 2022 for Cognitive Neurology, Department of Neurosciences, KU Leuven, Leuven, Belgium, 5 Neurology Service, University Published: 21 March 2022 Hospital Leuven, Leuven, Belgium, 6 Neurochemistry Laboratory, Department of Clinical Chemistry, Amsterdam Citation: Neuroscience, Amsterdam University Medical Centers, Vrije Universiteit, Amsterdam, Netherlands, 7 Department Homann J, Osburg T, Ohlei O, of Biomedical Sciences, University of Antwerp, Antwerp, Belgium, 8 Department of Neurology and Center for Neurosciences, Dobricic V, Deecke L, Bos I, Universitair Ziekenhuis Brussel and Vrije Universiteit Brussel (VUB), Brussels, Belgium, 9 Department of Psychiatry, University Vandenberghe R, Gabel S, of Geneva, Geneva, Switzerland, 10 IRCCS Istituto Centro San Giovanni di Dio Fatebenefratelli, Brescia, Italy, 11 Institut Scheltens P, Teunissen CE, Neurosciences Timone, AIX Marseille University, Marseille, France, 12 Neurosciences Therapeutic Area, GlaxoSmithKline Engelborghs S, Frisoni G, Blin O, R&D, Stevenage, United Kingdom, 13 Lille Neuroscience and Cognition, University of Lille, Inserm, CHU Lille, Lille, France, 14 Richardson JC, Bordet R, Lleó A, Memory Unit, Neurology Department, Hospital de Sant Pau, Barcelona and Centro de Investigación Biomédica en Red en Alcolea D, Popp J, Clark C, Enfermedades Neurodegenerativas (CIBERNED), Madrid, Spain, 15 Department of Geriatric Psychiatry, University Hospital Peyratout G, Martinez-Lage P, of Psychiatry Zurich, Zurich, Switzerland, 16 Old Age Psychiatry, Department of Psychiatry, University Hospital of Lausanne, Tainta M, Dobson RJB, Lausanne, Switzerland, 17 Department of Neurology, Center for Research and Advanced Therapies, CITA-Alzheimer Legido-Quigley C, Sleegers K, Foundation, Donostia-San Sebastian, Spain, 18 Department of Biostatistics and Health Informatics, Institute of Psychiatry, Van Broeckhoven C, Wittig M, Psychology and Neuroscience (IoPPN), King’s College London, London, United Kingdom, 19 NIHR Biomedical Research Franke A, Lill CM, Blennow K, Centre at South London and Maudsley NHS Foundation Trust and King’s College London, London, United Kingdom, 20 Zetterberg H, Lovestone S, Streffer J, Health Data Research UK London, University College London, London, United Kingdom, 21 Institute of Health Informatics, ten Kate M, Vos SJB, Barkhof F, University College London, London, United Kingdom, 22 NIHR Biomedical Research Centre at University College London Visser PJ and Bertram L (2022) Hospitals NHS Foundation Trust, London, United Kingdom, 23 Steno Diabetes Center, Copenhagen, Denmark, 24 King’s Genome-Wide Association Study College London, Institute of Pharmaceutical Sciences, London, United Kingdom, 25 Complex Genetics of Alzheimer’s of Alzheimer’s Disease Brain Imaging Disease Group, Center for Molecular Neurology, VIB, Antwerp, Belgium, 26 Department of Biomedical Sciences, University Biomarkers and Neuropsychological of Antwerp, Antwerp, Belgium, 27 Neurodegenerative Brain Diseases Group, Center for Molecular Neurology, VIB, Antwerp, Phenotypes in the European Medical Belgium, 28 Institute of Clinical Molecular Biology, Christian-Albrechts-University of Kiel, Kiel, Germany, 29 Ageing Information Framework Epidemiology Research Unit, School of Public Health, Imperial College London, London, United Kingdom, 30 Department for Alzheimer’s Disease Multimodal of Psychiatry and Neurochemistry, Institute of Neuroscience and Physiology, The Sahlgrenska Academy, University Biomarker Discovery Dataset. of Gothenburg, Gothenburg, Sweden, 31 Clinical Neurochemistry Laboratory, Sahlgrenska University Hospital, Mölndal, Front. Aging Neurosci. 14:840651. Sweden, 32 Department of Neurodegenerative Disease, University College London, Queen Square Institute of Neurology, doi: 10.3389/fnagi.2022.840651 Queen Square, London, United Kingdom, 33 UK Dementia Research Institute at University College London, London, Frontiers in Aging Neuroscience | www.frontiersin.org 1 March 2022 | Volume 14 | Article 840651 Homann et al. GWAS in EMIF-AD MBD Dataset United Kingdom, 34 Department of Psychiatry, University of Oxford, Oxford, United Kingdom, 35 Reference Center for Biological Markers of Dementia (BIODEM), Institute Born-Bunge, University of Antwerp, Antwerp, Belgium, 36 Janssen R&D, LLC. Beerse, Belgium, 37 Department of Radiology and Nuclear Medicine, Amsterdam University Medical Centers, Amsterdam Neuroscience, Amsterdam, Netherlands, 38 Institutes of Neurology and Healthcare Engineering, University College London, London, United Kingdom, 39 Division of Neurogeriatrics, Department of Neurobiology, Care Sciences and Society, Karolinska Institutet, Stockholm, Sweden, 40 Department of Psychology, University of Oslo, Oslo, Norway Alzheimer’s disease (AD) is the most frequent neurodegenerative disease with an increasing prevalence in industrialized, aging populations. AD susceptibility has an established genetic basis which has been the focus of a large number of genome- wide association studies (GWAS) published over the last decade. Most of these GWAS used dichotomized clinical diagnostic status, i.e., case vs. control classification, as outcome phenotypes, without the use of biomarkers. An alternative and potentially more powerful study design is afforded by using quantitative AD-related phenotypes as GWAS outcome traits, an analysis paradigm that we followed in this work. Specifically, we utilized genotype and phenotype data from n = 931 individuals collected under the auspices of the European Medical Information Framework for Alzheimer’s Disease Multimodal Biomarker Discovery (EMIF-AD MBD) study to perform a total of 19 separate GWAS analyses. As outcomes we used five magnetic resonance imaging (MRI) traits and seven cognitive performance traits. For the latter, longitudinal data from at least two timepoints were available in addition to cross-sectional assessments at baseline. Our GWAS analyses revealed several genome-wide significant associations for the neuropsychological performance measures, in particular those assayed longitudinally. Among the most noteworthy signals were associations in or near EHBP1 (EH domain binding protein 1; on chromosome 2p15) and CEP112 (centrosomal protein 112; 17q24.1) with delayed recall as well as SMOC2 (SPARC related modular calcium binding 2; 6p27) with immediate recall in a memory performance test. On the X chromosome, which is often excluded in other GWAS, we identified a genome-wide significant signal near IL1RAPL1 (interleukin 1 receptor accessory protein like 1; Xp21.3). While polygenic score (PGS) analyses showed the expected strong associations with SNPs highlighted in relevant previous GWAS on hippocampal volume and cognitive function, they did not show noteworthy associations with recent AD risk GWAS findings. In summary, our study highlights the power of using quantitative endophenotypes as outcome traits in AD-related GWAS analyses and nominates several new loci not previously implicated in cognitive decline. Keywords: genome-wide association study, GWAS, X chromosome, Alzheimer’s disease (AD), MRI, imaging, cognitive function INTRODUCTION disorder with both monogenic and polygenic forms. The former is caused by highly penetrant but rare mutations in Alzheimer’s disease is the most common neurodegenerative three genes encoding the amyloid beta precursor protein disease in humans and the most common form of dementia. In (APP) and presenilins 1 and 2 (PSEN1/PSEN2), which only 2018, estimates were published that 50 million dementia patients make up a small fraction (<<5%) of all AD cases (Cacace exist worldwide, about two-third of whom were diagnosed et al., 2016). Most patients, however, suffer from “polygenic with AD (Patterson, 2018). Pathologically, AD is characterized AD,” which is determined by the action (and interaction) of by the accumulation of extracellular amyloid β (Aβ) peptide numerous independent genomic variants, likely in concert with deposits (“plaques”) and intracellular hyperphosphorylated non-genetic factors, such as environmental exposures (e.g., tau protein aggregates (“tangles”) in the brain, leading to head trauma) and lifestyle choices (e.g., alcohol consumption synaptic dysfunction, neuroinflammation, neuronal loss, and, and cigarette smoking) (Bertram and Tanzi, 2020). Based on ultimately, onset of cognitive decline (Sperling et al., 2014; results from the currently most recent and largest genome- Mattsson et al., 2015). Genetically, AD is a heterogeneous wide association study (GWAS) performed in AD, there are Frontiers in Aging Neuroscience | www.frontiersin.org 2 March 2022 | Volume 14 | Article 840651 Homann et al. GWAS in EMIF-AD MBD Dataset now 38 independent loci showing genome-wide significant several new genetic loci potentially involved in AD-related association with disease risk (Wightman et al., 2021). The pathophysiology. most strongly and most consistently associated AD risk gene is APOE, which encodes apolipoprotein E, a cholesterol transport protein that has been implicated in numerous MATERIALS AND METHODS amyloid-specific pathways, including amyloid trafficking, as well as plaque clearance (Holtzman et al., 2012). Although Sample Description the heritability of polygenic AD is estimated to be around Analyses were based on the EMIF-AD MBD dataset which 60–80% (Gatz et al., 2006), APOE and the other currently was collected across 11 different European study centers known 37 independent risk loci explain only part of the (Bos et al., 2018). In total, this dataset included 1,221 [563 (46%) disease’s phenotypic variance (Wightman et al., 2021). female; mean age = 67.9 years, SD = 8.3] individuals from While most AD GWAS only consider clinically diagnosed three diagnostic stages: normal controls (NC), subjects with “probable AD” cases and cognitively unimpaired controls, mild cognitive impairment (MCI) and subjects with a clinical involving a risk for mis-diagnosis of patients and inclusion diagnosis of AD. A diagnosis of AD was based on National of preclinical AD cases as controls, additional information Institute of Neurological and Communicative Disorders and about the genetic architecture of AD and additional statistical Stroke–Alzheimer’s Disease and Related Disorders Association power is also afforded by using “endophenotypes” related criteria (NINCDS-ADRDA) (McKhann et al., 1984) while MCI to AD, ideally measured on a quantitative scale such as was diagnosed using criteria of Petersen (2004) in nine centers, biomarker data, imaging, or neurocognitive performance while two centers used the criteria by Winblad et al. (2004). (Gottesman and Gould, 2003; MacRae and Vasan, 2011; Individuals showing normal performance on neuropsychological Zhang et al., 2020). assessment (within 1.5 SD of the average for age, gender, and In our study, we expand earlier work from our group education) at baseline were classified as NC (Bos et al., 2018). An (Hong et al., 2020, 2021) derived from European Medical overview of the quantitative phenotypes investigated in this study Information Framework Alzheimer’s Disease Multimodal is provided in Table 1. Due to partially missing phenotype data Biomarker Discovery (EMIF-AD MBD) sample (Bos et al., (in the neurocognitive domain), the effective sample sizes vary 2018). Specifically, in two previous GWAS we set out to for the different GWAS analyses (see Table 1). The local medical identify variants underlying variation in several cerebrospinal ethical review boards in each participating recruitment center fluid (CSF) phenotypes, such as levels of CSF Aβ and tau had approved the study prior to commencement. Furthermore, protein (Hong et al., 2020), or neurofilament light (NfL) all subjects had provided written informed consent at the time chain, chitinase-3-like protein 1 (YKL-40), and neurogranin of inclusion in the cohort for use of data, samples and scans (Ng), which reflect axonal damage, astroglial activation, (Bos et al., 2018). and synaptic degeneration, respectively (Hong et al., 2021). However, the EMIF-AD MBD dataset features several other quantitative phenotypes, including cross-sectional Magnetic Resonance Imaging MRI measurements and cross-sectional and longitudinal Phenotypes Description neuropsychological tests, which are used as outcome traits in The five MRI phenotypes were collected for 862 subjects. Brain the current study. Thus, while using the same individuals MRIs were used to assess hippocampal volume (mm3 , left and and identical genome-wide SNP genotype data as in right hemisphere, and sum of both; all adjusted for intracranial Hong et al. (2020, 2021), we substantially extended our volume), whole brain cortical thickness (in mm), and white previous work by focusing on entirely novel phenotypic matter lesions (WML; using the Fazekas scale) (Ten Kate et al., domains available in EMIF-AD MBD. Specifically, we 2018). The Fazekas scale categorizes WMLs into 4 categories: performed GWAS and polygenic score (PGS) analyses Level 0 (no or almost no lesion), level 1 (multiple punctate on seven neuropsychological (using both cross-sectional lesions), level 2 (early confluent WML), and level 3 (presence and longitudinal data) and five brain imaging phenotypes of large confluent WML). Details on the scanning procedures (using cross-sectional data from MRI scans). In the 19 and data harmonization across centers can be found in Bos et al. performed GWAS scans (which also included the X (2018) and Ten Kate et al. (2018). chromosome), we identified a total of 13 genome-wide significant loci highlighting several novel genes showing Neuropsychological Phenotypes association with the analyzed traits. While we do not see a Description noteworthy overlap in the genetic architectures underlying Cross-sectional (and follow-up) data were available for the our “endophenotypes” and AD by polygenic score (PGS) following seven neuropsychological domains within the analysis, we did observe significant correlations in PGS EMIF-AD MBD dataset: global cognition (Mini Mental State constructed from earlier GWAS on hippocampal volume Examination, MMSE), attention, executive function, language, (Hibar et al., 2017) and general cognitive function (Davies memory (immediate and delayed) and visuoconstruction et al., 2018) with the respective phenotypes in EMIF- (for a detailed description of all neuropsychological tests see AD MBD. Taken together, our novel results pinpoint Supplementary Material). For each cognitive domain, a primary Frontiers in Aging Neuroscience | www.frontiersin.org 3 March 2022 | Volume 14 | Article 840651 Homann et al. GWAS in EMIF-AD MBD Dataset TABLE 1 | Description of EMID-AD MBD datasets analyzed per phenotype. Baseline Longitudinal Category Phenotype Sample size MAF filter Sample size MAF filter Neuropsychological MMSE 867 0.02 520 0.02 Attention 806 0.01 402 0.02 Executive functioning 686 0.01 234 0.02 Language 849 0.01 409 0.02 Memory Delayed 729 0.01 337 0.02 Memory Immediate 797 0.01 345 0.02 Visuoconstruction 429 0.02 149 0.04 MRI Fazekas score 606 0.01 n.a. n.a. Cortical thickness 560 0.01 n.a. n.a. Left Hippocampus volume 605 0.01 n.a. n.a. Right Hippocampus volume 605 0.01 n.a. n.a. Summed Hippocampus volume 605 0.01 n.a. n.a. “MAF filter” denotes the applied MAF filter for each GWAS. For cross-sectional MMSE we used an MAF threshold of 0.02 due to residual inflation of the GWAS test statistics. Information on tests used for generating baseline and longitudinal phenotypes can be found in Supplementary Material. MMSE, Mini Mental State Examination. “n.a.”, not available. MRI, Magnetic Resonance Imaging. MAF, Minor Allele Frequency. test was selected by Bos et al. (2018). If the preferred test were using the Infinium Global Screening Array (GSA) with Shared not available, an alternative priority test from the same cognitive Custom Content (Illumina Inc.). Imputation was then performed domain was chosen. More details on the neuropsychological using Minimac3 (Das et al., 2016). Extensive post-imputation tests used for generating these phenotypes can be found in QC resulted in 7,464,105 autosomal SNPs with a minor allele Bos et al. (2018). Raw data on these tests were normalized frequency (MAF) ≥ 0.01 in 888 individuals of European ancestry. with the help of a z-transformation, so that the data were More details can be found in Supplementary Material. comparable within a cognitive domain despite representing For the X chromosome, QC was performed separately for partially different tests across centers. For the cross-sectional male and female subjects for non-pseudoautosomal regions, GWAS analyses, the z-scores derived from baseline data were using slightly different criteria compared to the autosomes (see used. The number of subjects used for each test can be found Supplementary Material). In contrast, pseudoautosomal regions in Supplementary Material. For all seven neuropsychological (PAR1 and PAR2) were treated analogously to the autosomal domains, follow-up data from at least one additional time SNPs. After QC, imputations were performed on the Sanger point were available for each individual and used to construct Institute imputation server1 using the extended HRC reference a longitudinal phenotype using the following formula [which panel (McCarthy et al., 2016). After imputation, we used the estimates the relative change in cognitive performance per time same QC criteria as for the autosomal SNPs but performed these interval (here: years)]: separately for female and male data sets, except the HWE test (P < 1.0E-4) which was performed on all samples combined Scorelast − Scorefirst as recommended previously (Graffelman and Weir, 2016) and Scorelast +Scorefirst 2 ∗ interval implemented in PLINK2. For males, markers were coded as 0 vs. 2 (instead of 0 vs. 1), to adjust for the missing second X When calculating longitudinal phenotypes, this formula was chromosome (as recommended in Smith et al., 2021). applied separately for each neuropsychological test. Outlying scores were determined using a false discovery rate (FDR) Genome-Wide Association Studies and threshold of 0.05 and were excluded from all subsequent analyses. Post-Genome-Wide Association Studies Only the most frequently used tests per cognitive domain were included in the final phenotypes (for more information, Analyses see Supplementary Material). Both baseline and longitudinal SNP-based association analyses were performed assuming phenotypes were adjusted for age at baseline. an additive linear model (command: –glm) using allele dosages (to account for imputation uncertainty) in PLINK2 (Purcell et al., 2007). The model is equivalent to a test for a dose- DNA Extraction, Genotype Imputation response relationship between allele dose (i.e., one or two copies and Quality Control vs. reference genotype) on the outcome trait. The covariates A detailed description of the genotyping procedures, quality included in the analyses were sex, diagnostic status and the first control (QC) and subsequent data processing can be found in three principal components from a principal component analysis Hong et al. (2020) (Supplementary Material). Here, the same (PCA) to adjust for population-specific differences. Generally, genotype data were used for the GWAS analyses. Briefly, 936 DNA samples were subjected to genome-wide SNP genotyping 1 https://imputation.sanger.ac.uk/ Frontiers in Aging Neuroscience | www.frontiersin.org 4 March 2022 | Volume 14 | Article 840651 Homann et al. GWAS in EMIF-AD MBD Dataset we excluded SNPs from the GWAS analyses with MAF < 0.01. considered genes and loci within a 1 Mb region (±500,000 bp) However, due to differences in the effective sample sizes across around the SNP of interest. In case SNPs not identical to our “top phenotypes this threshold was adapted upward (up to 0.04) to SNP” were reported to show association with an AD-relevant prevent inflation of test statistics owing to low frequency SNPs phenotype (brain imaging, cognition, etc.), the LDlink platform (see Table 1 for more details). Diagnostic status was coded (Machiela and Chanock, 2015) was used to determine pairwise with two dummy variables as follows: NC = (0, 0), MCI = (0, LD to top SNPs.7 In this context we defined relevant LD using a 1), AD = (1, 1). For four longitudinal cognitive phenotypes threshold of r2 > 0.6. an additional dummy variable was introduced to code for the neuropsychological test used, in cases where two different tests Polygenic Score Analysis were used for generating these phenotypes. Details can be found In addition to the primary GWAS analyses described above, we in Supplementary Material. also calculated polygenic scores (PGS) to estimate the extent To explore associations on the X chromosome that were of genetic correlation with the GWAS results for three other potentially driven by genetic sex, we additionally conducted the phenotypes. To this end, we used the summary statistics of a analyses separately in females and males. We then combined GWAS on AD risk (Jansen et al., 2019) as comparison to both these two additional sets of results in a meta-analysis using phenotypic domains (MRI and neurocognitive performance) of Stouffer’s method as implemented in METAL (Willer et al., 2010). our study, and the GWAS on general cognitive function (Davies As we found no noteworthy differences in the results using et al., 2018) as comparison to the GWAS on neuropsychological Stouffer’s method, only the results from the linear regression phenotypes. Finally, the GWAS on hippocampal volume (Hibar analysis in the combined sample are shown. et al., 2017) served as comparison to our GWAS analyses on The FUMA platform2 (Watanabe et al., 2017) was used for MRI phenotypes. PGS calculations were performed using PRSice- post-GWAS analyses, including gene-based association analyses 2 software (Choi and O’Reilly, 2019). Statistical analyses fitted (via MAGMA, de Leeuw et al., 2015) and to annotate and general linear regression models with PGS as predictor adjusting visualize the GWAS results. To this end, we defined genome- for the same covariates as in the primary GWAS analyses: sex, wide significance at α = 5.0E-08 for the SNP-based analyses while diagnostic status, and PC1-3 (and type of cognitive test, where genome-wide suggestive evidence was set at α = 1.0E-05. For applicable). To adjust for multiple testing of this arm of our the gene-based analyses, we adjusted for the number of protein- study, we used a conservative threshold based on Bonferroni coding genes examined (19,485) using the Bonferroni method, adjustment (α = 5.0E-03 = 0.05/(5∗ 2) for the MRI phenotypes, resulting in a threshold of α = 2.566E-06. and α = 1.8E-03 = 0.05/(14∗ 2) for the neuropsychological In FUMA, both the SNP annotation and the Combined phenotypes). However, given the (at least partial) correlation Annotation Dependent Depletion (CADD) score (Rentzsch et al., between phenotypes, we note that the true threshold is likely 2021) are provided. The main GWAS results are reported only somewhere between 0.05 and these Bonferroni-adjusted values. for “independent significant” SNPs, as defined by FUMA. These represent SNPs that are not highly correlated with one another using a threshold of r2 < 0.6 (using reference data from the 1,000 RESULTS Genomes Project). Subsequently, the top SNPs, i.e., those with the smallest Genome-Wide Association Studies on P-values per respective phenotype, were examined in more detail using additional tools. First, the Variant Effect Predictor on Magnetic Resonance Imaging Ensembl (VEP,3 McLaren et al., 2016) was used to determine Phenotypes a possibly functional effect due to changes in the coding The genomic inflation factor λ ranged between 1.004 and 1.012 sequence, e.g., missense variants. Second, SNPs were examined in all five SNP-based analyses, indicating that the results of the using data from the RegulomeDB database4 (Boyle et al., 2012) MRI GWAS analyses were not affected by substantial inflation to assess possible effects on gene expression. Third, we used of the test statistics. In the actual association analyses of the data from the Genotype-Tissue Expression (GTEx, V8) project five quantitative MRI phenotypes, we identified no genome- portal5 (Lonsdale et al., 2013) to assess whether SNPs represent wide significant (P < 5.0E-08) signals but observed 385 variants expression / splicing quantitative trait loci (eQTLs/sQTLs). While with at least suggestively significant (P < 1.0E-05) evidence of GTEx provides data on gene expression in 54 tissues, we laid association (Supplementary Tables 15–19). The lowest P-value particular emphasis in genes expressed in brain. Lastly, we was observed with SNP rs16829761 for the Fazekas phenotype interrogated the GWAS catalog6 (Buniello et al., 2019) to assess (P = 5.08E-08; Supplementary Figure 31), which only fell slightly whether any of the top SNPs were previously reported to show above the genome-wide significance threshold. According to VEP association with other phenotypes by GWAS. To this end, we (McLaren et al., 2016), this variant is located in an intron of the genes IQCJ (protein: IQ motif containing J) and SCHIP1 2 http://fuma.ctglab.nl/ (protein: schwannomin-interacting protein 1). In the GTEx 3 http://grch37.ensembl.org/Tools/VEP database (Lonsdale et al., 2013), the lead-SNP identified here 4 https://regulomedb.org/regulome-search (rs16829761) is not listed as eQTL or sQTL, which may be due 5 https://www.gtexportal.org/home/ 6 7 https://www.ebi.ac.uk/gwas/home https://ldlink.nci.nih.gov/?tab=ldpair Frontiers in Aging Neuroscience | www.frontiersin.org 5 March 2022 | Volume 14 | Article 840651 Homann et al. GWAS in EMIF-AD MBD Dataset to the comparatively low MAF (0.01). The CADD score, i.e., the regulate synapse formation (Montani et al., 2019). The third in silico predicted deleteriousness, of rs16829761 is also low at highlighted signal in this domain relates to the genome-wide approximately 0.074. In addition, none of the gene-based GWAS significant variant rs74381761 (MAF ∼0.05) on chromosome analyses using MAGMA revealed any genome-wide significant 8p23.1 (P = 1.89E-08; Table 2, Figure 1B, and Supplementary signals (P < 2.566E-06) using the MRI traits analyzed. The Figure 5) which shows association with the longitudinal MMSE genomic inflation factor λ ranged between 0.984 and 1.060 in phenotype. The lead SNP is located in an intergenic region near these five gene-based analyses. TNKS (gene-based P = 4.87E-04; Table 2). This gene encodes the protein tankyrase, which belongs to a class of poly (ADP- Genome-Wide Association Studies on ribose) polymerases and is involved in various processes in the Neuropsychological Phenotypes body, such as telomere length regulation, the Wnt/β-catenin Across the 14 GWAS performed on cross-sectional and signaling pathway, or glucose transport (Damale et al., 2020). The longitudinal neuropsychological phenotypes available in EMIF- last featured signal relates to the association observed near SNP AD MBD, there were a total of 13 genome-wide significant loci, rs9652864 (MAF ∼0.22) on chromosome 17q24.1 (P = 3.20E- two of which were identified via the gene-based analyses using 08; Table 2, Figure 1B, and Supplementary Figure 21) and the MAGMA (de Leeuw et al., 2015). Three of the genome-wide delayed recall test. This variant is located in an intron of CEP112, significant signals were observed in the analyses of cross-sectional which encodes centrosomal protein 112. Overall, there were eight phenotype data and 10 with longitudinal outcomes. Overall, none correlated SNPs in this locus all showing strongly association of the sets of GWAS results in this arm of our study appeared to (Supplementary Table 10). be strongly affected by inflation of the genome-wide test statistics as evidenced by genomic inflation factors near 1 (range: 0.969– Comparison of Cross-Sectional vs. Longitudinal 1.012 in the SNP-based analyses and 0.922–1.036 in the gene- Genome-Wide Association Studies Results based analyses). Table 2 provides a detailed summary of these After completion of the separate GWAS on cross-sectional genome-wide significant loci, and Figure 1 shows multi-trait and longitudinal outcomes, we assessed whether the results Manhattan (MH) plots of the SNP-based GWAS results for cross- of these two analysis arms showed any overlap. To this sectional (Figure 1A) and longitudinal (Figure 1B) analyses end, we followed two approaches: First, we performed a (for corresponding QQ plots: see Supplementary Figures 1, 2). look-up of top results from one paradigm in the equivalent The following two paragraphs highlight the most interesting other. Specifically, we checked whether a genome-wide results in either the analyses of cross-sectional or longitudinal significant SNP from the cross-sectional analyses also had a neuropsychological traits. low P-value in the corresponding longitudinal GWAS and vice versa. The lowest corresponding P-value was 0.015 (at Analyses of Cross-Sectional Data baseline) for rs73045828, which attained P = 5.65E-09 in the The most interesting finding in this domain was elicited by longitudinal GWAS for immediate memory (Supplementary markers in EHBP1 which showed genome-wide significant Table 21). No further signal overlaps were observed across evidence of association with the delayed recall memory corresponding cross-sectional and longitudinal phenotypes. phenotype in the gene-based analysis (P = 1.17E-07; Table 2 Second, we took a more comprehensive approach by comparing and Supplementary Figure 20). The lead SNP (rs6705798) in a larger set of SNPs across both phenotypic domains. To this region only missed the genome-wide significance threshold this end, we constructed PGS from the summary statistics by a small margin (P = 8.78E-08; Table 2 and Figure 1A). of the cross-sectional GWAS (as an approximate measure of EHBP1 is located on chromosome 2p15 and encodes EH domain “aggregated SNP effects”) and used these PGS as independent binding protein 1. variables in a linear model predicting longitudinal outcomes. Effectively, this allowed us to determine how much phenotypic Analyses of Longitudinal Data variance in the longitudinal data can be explained by top The strongest signal in the longitudinal analyses was elicited by a SNPs of the matching cross-sectional GWAS. Overall, these locus on chromosome 6q27 (rs73045836; P = 7.50E-11; Table 2, analyses did not reveal a substantial correlation in genetic Figure 1B, and Supplementary Figure 25) in the analysis using results for corresponding phenotypes (Supplementary an immediate memory recall paradigm. This SNP is located in an Table 22), in agreement with the look up of individual intron of SMOC2 coding for secreted modular calcium-binding SNPs (see above). The best model fit was observed with the protein 2, which, among other functions, promotes extracellular PGS for executive function and visuoconstruction, where matrix assembly (Gao et al., 2019). It needs to be noted that with the GWAS top SNPs from the cross-sectional data used an MAF ∼2% this SNP is rather infrequent which may increase in the PGS explained 4–9% of the phenotypic variance the possibility of representing a false-positive finding. Perhaps of the corresponding longitudinal outcomes, respectively more interesting is the association signal observed near SNP (Supplementary Table 22). We note, however, that the PGS rs5943462 (MAF ∼0.05) and the visuoconstruction phenotype method was not designed for computing genetic correlations on the X chromosome (P = 1.06E-09; Table 2, Figure 1B, and of non-independent samples (as is the case here), so this Supplementary Figure 29). This SNP is an intronic variant analysis must be considered “exploratory,” and the reported located in IL1RAPL1 encoding interleukin 1 receptor accessory results represent no more than “upper bounds” of the potential protein−like 1, which belongs to a class of molecules that genetic correlations. Frontiers in Aging Neuroscience | www.frontiersin.org 6 March 2022 | Volume 14 | Article 840651 Homann et al. GWAS in EMIF-AD MBD Dataset TABLE 2 | Genome-wide significant associations observed in GWAS of cognitive phenotypes. Study arm Phenotype Lead variant Chr Position Nearest gene A1 A2 Beta MAF P (SNP) P (Gene) Cross-sectional MemoryDelayed rs6705798 2p15 63,259,881 EHBP1 C T −0.32739 0.358 8.78E-08 1.17E-07 MMSE rs2122118 2q33.3 207,252,439 AC017081.2 G A −2.82266 0.022 3.03E-09 n.a. Visuoconstruction rs113492235 4q34.2 177,252,900 SPCS3 T C −2.17956 0.022 1.51E-08 0.15674 Longitudinal MemoryImmediate rs73045836 6q27 169,062,739 SMOC2 G T −0.36094 0.020 7.50E-11 0.0035373 MMSE rs74381761 8p23.1 9,389,761 TNKS C G −0.08453 0.048 1.89E-08 0.00048716 Attention rs116900143 10q23.31 92,588,290 HTR7 C T −0.35173 0.023 1.95E-08 0.019663 MemoryImmediate rs11217863 11q23.3 120,293,138 AP002348.1 A G −0.16626 0.080 7.81E-08 8.91E-07 Attention rs111959303 12q14.3 66,844,015 GRIP1 T C 0.37459 0.022 2.52E-08 0.81478 Attention rs34736485 16q23.2 79,272,611 RP11-679B19.2 T G 0.31924 0.022 1.59E-08 n.a. MemoryDelayed rs9652864 17q24.1 63,741,645 CEP112 A T 0.29184 0.218 3.20E-08 0.016339 MemoryImmediate rs146202660 18q21.1 45,022,937 CTD-2130O13.1 T G −0.29342 0.029 4.63E-08 n.a. Executive rs16982556 20q13.32 57,801,889 ZNF831 T C −0.29752 0.062 1.26E-08 0.0025565 Visuoconstruction rs5943462 Xp21.3 28,823,154 IL1RAPL1 G C −0.14082 0.051 1.06E-09 0.006719 Bold font indicates genome-wide significant (on SNP- or gene-level) results (see section Materials and Methods for details). “Chr” and “Position” according to GRCh37/hg19. “A1” denotes the effect allele. “P (SNP)” is the P-value of the lead SNP at this locus. “P (Gene)” is the P-value belonging to “Nearest gene.” Top results from these GWAS analyses can be found in Supplementary Tables 1–19. MMSE, Mini Mental State Examination. “n.a.”, not available. MRI, Magnetic Resonance Imaging. MAF, Minor Allele Frequency. FIGURE 1 | Multi-trait Manhattan plots for the SNP-based GWAS results on neuropsychological phenotypes (A: cross-sectional; B: longitudinal). For details on the analyzed traits see section “Materials and Methods” and Supplementary Material. Frontiers in Aging Neuroscience | www.frontiersin.org 7 March 2022 | Volume 14 | Article 840651 Homann et al. GWAS in EMIF-AD MBD Dataset Role of APOE in Genome-Wide by the Fazekas score). For this trait the AD SNPs explained 1.4% Association Studies on Magnetic variance (P = 3.7E-03; Table 3 and Supplementary Table 24). Resonance Imaging and Polygenic Score Analyses With Neuropsychological Neuropsychological Performance Phenotypes Given the substantial role that variants in APOE play in the As for the MRI data, the best fit in the PGS analyses with the genetic architecture of AD, we present findings for this locus neuropsychological phenotypes was observed with a GWAS that separately, i.e., the results for SNP rs429358 (which defines the ε4 also used neurocognitive performance as outcome (Davies et al., allele) and rs7412 (which defines the ε2 allele). In relation to the 2018). Specifically, this study defined a PCA-derived factor for common genotype ε3/ε3, the risk to develop AD is increased by a “general cognitive function” which was analyzed in > 300,000 factor of ∼3.2 for genotype ε3/ε4, while two ε4 alleles (genotype individuals. In EMIF-AD MBD, associations with four of the ε4/ε4) show ORs around 10–12 when compared to normal 14 calculated PGS fell below the multiple testing threshold of controls (Neu et al., 2017). The minor allele at rs429358 (ε4) 1.8E-03 (Table 3). The strongest association was observed with is overrepresented in the EMIF-AD MBD dataset with an MAF the longitudinal attention function for which the GWAS results ∼29% (the MAF in the general Northern European population from Davies et al. (2018) explained 2.3% of the phenotypic is ∼16%), which is due to the special design of participant variance (P = 1.79E-03, Table 3 and Supplementary Table 23). recruitment (see Bos et al., 2018). For the neuropsychological The next best associations were seen with longitudinal executive phenotypes, the P-values of rs429358 are unremarkable except functioning (r2 = 0.028; P = 9.79E-03; Supplementary Table 23) for the domain “delayed memory,” where P-values of 0.0005 and and visuoconstructional abilities (r2 = 0.058; P = 3.08E-03; 0.0042 were observed for the baseline and longitudinal analyses, Supplementary Table 23). However, these latter two associations respectively (Supplementary Table 20). In the MRI analyses, do not survive multiple testing correction (Table 3). Interestingly the only association signal observed with rs429358 was with and similar to the MRI-based results, we did not find strong hippocampal volume (Supplementary Table 20). Interestingly, evidence for a genetic overlap between the neurocognitive this was driven by an association with the volume of the left outcomes tested here and AD risk based on Jansen et al. (P = 0.0002) hippocampus, while no association was observed (2019) (Supplementary Table 23). This included the various with the corresponding data of the right hemisphere (P = 0.2956). phenotypes measuring components of “memory” performance, We note that for both traits, i.e., delayed memory and left regardless of whether or not they were ascertained cross- hippocampal volume, the effect direction the corresponding β sectionally or longitudinally. coefficient is consistent with the deleterious effect of the minor (T/ε4) allele at rs429358 known from the literature (Neu et al., 2017). For the minor allele at rs7412 (ε2) we observed no DISCUSSION noteworthy association signals in any of the analyses performed in this study (Supplementary Table 20), possibly because power This study extends previous GWAS analyses from our group for this variant was much reduced owing to its lower MAF (4.6% utilizing phenotypic data from the EMIF-AD MBD study (Hong here, 7.5% in the general western European control population). et al., 2020, 2021) using different outcome traits hitherto not analyzed by GWAS. The overarching goal of this work was Polygenic Score Analyses Using to decipher the genetic architecture of AD-related MRI and neuropsychological (endo)phenotypes to better understand AD Published Genome-Wide Association pathophysiology. Both previous EMIF-AD MBD GWAS focused Studies Results on AD biomarkers measured in CSF and, among other findings, In these analyses we aimed to estimate the degree of genetic identified variants in TMEM106B as trans-pQTLs of CSF overlap between the MRI and neuropsychological outcomes neurofilament light (NfL) levels (Hong et al., 2021). Interestingly, available in EMIF-AD MBD and other relevant traits from the same locus was subsequently highlighted as a novel AD the literature, such as AD risk, using published GWAS risk locus in a GWAS on > 1.1 million individuals (Wightman summary statistics. et al., 2021), showcasing the power of the quantitative biomarker GWAS approach that was also followed in this study. In the Polygenic Score Analyses With Magnetic Resonance current work, we focused on biomarkers / phenotypes derived Imaging Phenotypes from brain imaging and neuropsychological testing in the same As expected, the strongest overlap was observed with a prior EMIF-AD MBD individuals. Overall, we performed 19 individual GWAS also using MRI outcomes. Specifically, we used GWAS GWAS and identified a total of 13 genome-wide significant loci results by the ENIGMA group (Hibar et al., 2017) who studied highlighting several novel genes that are potentially involved 26 imaging traits in n = 33,536 individuals. Here, the best in contributing to AD pathophysiology. Our study represents overlap was seen with each of the three hippocampal MRI one of few GWAS in the literature to also include the X traits (up to 2.7% variance explained, P = 6.0E-06; Table 3 and chromosome, where we identified a genome-wide significant Supplementary Table 24). In contrast, in PGS analyses using association between markers near IL1RAPL1 and longitudinal SNPs associated with AD risk (Jansen et al., 2019), we found only visuoconstructive ability. Interestingly, neither APOE nor the one moderate correlation with white matter damage (measured other recently described AD GWAS loci appear to have a major Frontiers in Aging Neuroscience | www.frontiersin.org 8 March 2022 | Volume 14 | Article 840651 Homann et al. GWAS in EMIF-AD MBD Dataset TABLE 3 | Summary of PGS results significant after multiple testing correction. Prior GWAS Phenotype Threshold Number of SNPs R2 P-value Hibar et al. (2017) Hippocampus volume sum 0.0001 127 0.027 6.06E-06 Hippocampus volume left 0.0001 127 0.026 9.98E-06 Hippocampus volume right 0.0001 127 0.024 2.48E-05 Jansen et al. (2019) Fazekas 0.17075 22,269 0.014 3.72E-03 Davies et al. (2018) Baseline MMSE 0.248 63,792 0.016 1.66E-06 Baseline executive functioning 0.0014 4,469 0.018 1.98E-05 Baseline language 0.0061 9,031 0.010 1.25E-03 Longitudinal attention 5.0E-08 163 0.023 1.79E-03 “Threshold” refers to P-value cut-off used for PGS construction in prior GWAS summary statistics and “Number of SNPs” refers to the LD-pruned SNPs passing this threshold that are included in PGS calculations. “R2 ” denotes the phenotypic variance explained by the SNPs of the prior GWAS in the EMIF-AD MBD dataset. A full listing of results from these PRS analyses can be found in Supplementary Material. MMSE, Mini Mental State Examination. impact on the traits analyzed in our study. In summary, our protein may have a specialized role in physiological processes extensive genome-wide analyses nominate several novel loci underlying memory and learning abilities (Montani et al., 2019). potentially involved in neurocognitive functioning. Some of these Even small changes in the expression and function of these may prove informative to better understand the genetic forces proteins can provoke major alterations in synaptic connectivity, underlying AD and related phenotypes. resulting in cognitive damage (Montani et al., 2019). Moreover, In the remainder of this section, we discuss the potential IL1RAPL1 was nominated as a candidate gene for X-linked role of five loci, which we consider the most interesting mental retardation (Raymond, 2006). Although the GWAS on findings of our study. The strongest GWAS signal was elicited longitudinal visuoconstruction included only 149 individuals, we by SNP rs73045836 (P = 7.50E-11; Table 2, Figure 1B, and believe this signal to be plausible and very interesting because of Supplementary Figure 25) showing genome-wide significant the well-established role of IL1RAPL1 on human brain function. association with the longitudinal data of the immediate recall The third highlighted signal relates to the association between memory phenotype. The gene annotated to the associated variant rs74381761 (MAF ∼0.05) on chromosome 8p23.1 region on chromosome 6q27, SMOC2, encodes secreted modular (P = 1.89E-08; Table 2, Figure 1B, and Supplementary Figure 5) calcium-binding protein 2. SMOC2 is an extracellular matrix and longitudinal MMSE measurements. This SNP is located protein from the secreted protein, acidic and rich in cysteine in an intergenic region near TNKS. This gene encodes the (SPARC) family (Gao et al., 2019) recently linked to age- protein tankyrase, which belongs to a class of poly (ADP-ribose) dependent bone loss in humans (Morkmued et al., 2020). In polymerases and is involved in various processes in the body, the AD context, it is noteworthy that SMOC2 was recently such as telomere regulation, Wnt/β-catenin signaling pathway found to be altered in CSF samples of early AD in a proteomics or glucose transport (Damale et al., 2020). According to GTEx profiling study (Whelan et al., 2019). Interestingly, there is similar (Lonsdale et al., 2013), TNKS is highly expressed in brain evidence on a potential link to AD for a SMOC2 isoform, i.e., (mostly in cerebellum). Moreover, SNPs annotated to TNKS SMOC1 (gene: SMOC1, located on chromosome 14q24.2). While were associated with brain white matter hyperintensity (WMH) variants in this gene did not show strong evidence of genetic measurements (Armstrong et al., 2020; Sargurupremraj et al., association with the traits analyzed here, it is noteworthy that 2020; Zhao et al., 2021) and cortical surface area measurements SMOC1 was recently nominated as a novel AD biomarker in (Grasby et al., 2020) according to the GWAS catalog (Buniello proteomic screens of AD CSF and brain samples in various et al., 2019). With a gene-based P-value of 4.87E-04 and the studies (Bai et al., 2020; Wang et al., 2020; Sathe et al., 2021). strong functional link to brain function, we consider the signal In summary, our finding of genome-wide significant association around TNKS as plausible and very interesting. between SMOC2 and memory performance in the EMIF-AD The last highlighted finding from the longitudinal MBD datasets extends the emerging literature on the role of analyses relates to the genome-wide significant association SPARC protein family members in AD and related traits. observed between SNP rs9652864 and the delayed recall The second strongest association signal was observed near memory phenotype (P = 3.20E-08; Table 2, Figure 1B, and SNP rs5943462 (MAF ∼0.05) on the X chromosome (P = 1.06E- Supplementary Figure 21). This variant (MAF = 0.218) attained 09; Table 2; Figure 1B and Supplementary Figure 29) with a P-value of 6.73E-04 in the GWAS of Davies et al. (2018) the longitudinal data of the visuoconstruction phenotype. on cross-sectional cognitive performance, lending additional The SNP is located in an intron of IL1RAPL1. This gene support to our finding. The SNP is located in an intron of encodes interleukin 1 receptor accessory protein−like 1, CEP112 which encodes centrosomal protein 112. Centrosomal which belongs to a class of molecules that regulate synapse proteins are known as the components of the centrosome formation. IL1RAPL1 is mostly expressed in brain areas that involved in centriole biogenesis, cell cycle progression, and are involved in memory development, such as hippocampus, spindle-kinetochore assembly control (Mazaheri Moghaddam dentate gyrus, and entorhinal cortex, suggesting that the et al., 2021). Despite showing only low levels of expression in Frontiers in Aging Neuroscience | www.frontiersin.org 9 March 2022 | Volume 14 | Article 840651 Homann et al. GWAS in EMIF-AD MBD Dataset the central nervous system (CNS) according to GTEx, SNPs more recent GWAS on brain imaging phenotypes in the in this gene have been associated with cortical surface area by United Kingdom biobank, none of the hippocampus-based neuroimaging in two independent GWAS (Grasby et al., 2020; measurements exceeded that study’s multiple testing threshold van der Meer et al., 2020) according to the GWAS catalog (Smith et al., 2021). We note, that our current findings are (Buniello et al., 2019). However, none of these neuroimaging different from our earlier GWAS analyses in the EMIF-AD MBD SNPs is in relevant LD (r2 > 0.6) to the lead variant identified dataset, where the ε4 allele showed very pronounced evidence here. Notwithstanding, given that variants in this gene have of association in CSF and imaging markers related to Aβ42 shown genetic links to both cognitive function and structural (Hong et al., 2020). Taken together, there is now converging brain imaging, we consider this finding as plausible and evidence from our and previous studies that despite APOE ε4’s highly interesting. role in contributing to AD risk and Aβ42-related phenotypes, In the GWAS analyses of the cross-sectional neurocognitive the same allele does not appear to show a comparably strong phenotypes, we observed three genome-wide significant signals, influence on variation of other AD-related traits, including the of which we consider the gene-based association with EHBP1 as ones studied here. the most interesting finding (P = 1.17E-07; Table 2, Figure 1A, Extending the comparison to additional genetic variants and Supplementary Figure 20). This protein interacts with associated with AD risk in the GWAS by Jansen et al. (2019) Eps15-homology domain-containing protein 1/2 (EHD1/2) that did also not show any noteworthy or consistent overlap plays a central role in GLUT4 transport and couples endocytic with the GWAS results generated in this study. In contrast, vesicles to the actin cytoskeleton (Rai et al., 2020). It is highly highly significant overlaps by PGS analysis were observed expressed in many tissues, including the brain, according to upon using GWAS results from Davies et al. (2018) for GTEx (Lonsdale et al., 2013). While there does not appear to be an the neuropsychological and Hibar et al. (2017) for the MRI obvious link between EHBP1 and brain function in the literature phenotypes, which is not surprising given that very similar (e.g., in the GWAS catalog, Buniello et al., 2019), we note that this neuropsychological and neuroimaging traits were used as gene is located within 5 kb of OTX1 (orthodenticle homeobox outcomes in these studies. Collectively, the PGS results of this 1; gene-based P = 1.19E-05), which acts as transcription factor and previous work show that there is only very limited overlap in and plays a role in brain and sensory organ development in the genetic architecture (at least when studying common SNPs) Drosophila and vertebrates, including humans (Omodei et al., between AD on the one and neuropsychological performance or 2009). Our lead SNP in this region, i.e., rs6705798, falls just short structural brain imaging on the other hand. We note that this of attaining genome-wide significance (P = 8.78E-08; Table 2 does not preclude the possibility that certain molecular pathways and Figure 1A) and is reported to represent an eQTL of both targeted by the genes highlighted in this GWAS may be shared OTX1 and EHBP1 in various human tissues according to GTEx with AD pathophysiology. (Lonsdale et al., 2013). While our study has several noteworthy strengths (e.g., In addition to searching for novel genetic determinants of the use of highly standardized procedures in generating and the neuroimaging and neurocognitive traits analyzed in this harmonizing both the genotype and phenotype data of our study, study, we also investigated the overlap with known GWAS use of both cross-sectional and longitudinal neurocognitive findings. First and foremost, this relates to two commonly performance data, inclusion of the X chromosome in the studied alleles in the APOE gene, which appear to only play GWAS), it may also have been negatively affected by some a minor role in this setting. Specifically, SNP rs429358, which limitations. First and foremost, we note that the sample defines the ε4 allele in APOE, does not even reach genome- size used for the present analyses is comparatively small for wide suggestive significance (P < 1.0E-05) in any of the 19 “GWAS standards” and was well under 1,000 in some instances GWAS investigated here. The strongest associations with this (Table 1). Accordingly, the statistical power of these analyses allele were seen with MRI-based hippocampus volume (left was low. This limitation is at least partially countered by volume P = 0.0002, summed volume P = 0.0005; Supplementary the quantitative nature of nearly all analyzed phenotypes: it Table 20) and with the delayed recall memory test (baseline is well established that quantitative trait association analyses P = 0.0005, longitudinal P = 0.0042; Supplementary Table 20). are more powerful than those using binary phenotypes, e.g., The effect directions of these associations are consistent with in a case-control setting (Bush and Moore, 2012). Second, the deleterious influence of rs429358 on AD (Neu et al., in addition to resulting in low power, small sample sizes 2017). While this at best modest degree of association between also increase the possibility of false-positive findings, especially rs429358 and the traits investigated here could be due to for infrequent variants (i.e., those with an MAF < 5%). a number of study-specific aspects, e.g., insufficient power, In this context we note that eight of our thirteen genome- aspects of sample ascertainment, it is in general agreement wide significant signals were elicited by such variants. Thus, with the literature: In the GWAS on cognitive function by independent replication—ideally in larger datasets—is needed to Davies et al. (2018), APOE ε4 also showed only marginal confirm the main findings of our GWAS before any further- association (P = 2.2E-04), and it was not reported to show reaching conclusions can be reached. Third, we note that genome-wide significant evidence of association (P = 4.1E- the phenotype data used as outcome traits in our GWAS 07) with hippocampal volume in the GWAS by Hibar et al. analyses were collected at different participating centers at times (2017). Lastly, while rs429358 did show evidence of genome- using different types of examinations (e.g., different tests to wide significant association with several imaging traits in the study the same overarching neuropsychological domain). To Frontiers in Aging Neuroscience | www.frontiersin.org 10 March 2022 | Volume 14 | Article 840651 Homann et al. GWAS in EMIF-AD MBD Dataset alleviate potential bias resulting from this inherent phenotypic to the interpretation of the data and visualization of the results. heterogeneity, all clinical data were processed, quality-controlled IB, SV, and PV coordinated the collection and harmonization and harmonized (e.g., by normalizing most variables within of phenotypes and biosamples in EMIF-AD MBD. MW and AF centers) centrally by an experienced team of researchers (see supervised the genotyping experiments. RV, SG, JS, PS, CT, SE, Bos et al., 2018 for more details). We emphasize that this GF, OB, JR, RB, AL, DA, JP, CC, GP, PM-L, MT, RD, CL-Q, potential heterogeneity does not apply to the genetic data as these KS, CV, CL, KB, HZ, MK, and FB contributed to sample and were generated in one laboratory experiment and subsequently phenotype data collection. JS, PV, and SL were the lead PIs for the processed jointly in one analytical framework, minimizing the EMIF-AD MBD study as a whole and were in charge of designing emergence of potential batch effects. Last but not least, we and managing the platform. LB designed and supervised the emphasize that owing to its particular ascertainment design (Bos genomics portion of the EMIF-AD MBD project and co-wrote et al., 2018) the EMIF-AD MBD dataset does not (attempt to) all drafts of the manuscript. All authors critically revised all constitute a representative sample from the “general population.” manuscripts drafts, read and approved the final manuscript. Accordingly, the results presented here cannot be generalized to the general population. We note that the same is true for many GWAS in this and other fields, which typically use FUNDING clinic-based ascertainment which is not representative of the population as a whole. The present study was conducted as part of the EMIF-AD MBD project, which has received support from the Innovative Medicines Initiative Joint Undertaking under EMIF grant CONCLUSION agreement No. 115372, the resources of which were composed of financial contribution from the European Union’s Seventh In conclusion, our study delivers an entirely novel set of GWAS Framework Program (FP7/2007-2013) and EFPIA companies’ results from participants of the EMIF-AD MBD dataset. We in kind contribution. Parts of this study were made possible nominate several novel and functionally interesting genetic through support from the German Research Foundation (DFG association signals with phenotypes related to neurocognitive grant FOR2488: Main support by subproject “INF-GDAC” function and structural brain imaging. Even though independent BE2287/7-1 to LB) and the Cure Alzheimer’s Fund (to LB). RV replication is still needed, our results may prove informative acknowledges support by the Stichting Alzheimer Onderzoek to better understand the genetic forces underlying AD and (#13007, #11020, #2017-032) and the Flemish Government related phenotypes. (VIND IWT 135043). KB was supported by the Swedish Research Council (#2017-00915), the Alzheimer Drug Discovery Foundation (ADDF), United States (#RDAPB-201809-2016615), DATA AVAILABILITY STATEMENT the Swedish Alzheimer Foundation (#AF-742881), Hjärnfonden, Sweden (#FO2017-0243), the Swedish state under the agreement GWAS summary statistics for the top (P-value < 1E-05) results between the Swedish Government and the County Councils, are listed in the Supplementary Tables. Full GWAS summary the ALF-agreement (#ALFGBG-715986), European Union Joint statistics are available from the authors upon request. Clinical Program for Neurodegenerative Disorders (JPND2019-466-236), data and genome-wide genotyping data are stored on an and the Alzheimer’s Association 2021 Zenith Award (ZEN-21- online data platform using the “tranSMART” data warehouse 848495). HZ was a Wallenberg Scholar supported by grants framework. Access to the genome-wide genotyping data can be from the Swedish Research Council (#2018-02532), the European requested from the corresponding author of this study who will Research Council (#681712), and Swedish State Support for forward each request to the EMIF-AD data access team. Clinical Research (#ALFGBG-720931). SV received funding from the Innovative Medicines Initiative 2 Joint Undertaking under ETHICS STATEMENT ROADMAP grant agreement No. 116020 and from ZonMw during the conduct of this study. Research at VIB-UAntwerp The studies involving human participants were reviewed and was in part supported by the University of Antwerp Research approved by the ethics committees of all participating centers of Fund and SAO-FRA 2018 0016. The Lausanne study was this study (for a full list see Supplementary Table 4 of primary funded by a grant from the Swiss National Research Foundation publication Bos et al., 2018). The patients/participants provided (SNF 320030_141179) to JP. Research of CT was supported their written informed consent to participate in this study. by the European Commission [Marie Curie International Training Network, grant agreement No 860197 (MIRIADE), and JPND], Health Holland, the Dutch Research Council (ZonMW), AUTHOR CONTRIBUTIONS Alzheimer Drug Discovery Foundation, the Selfridges Group Foundation, Alzheimer Netherlands, Alzheimer Association. JH and TO performed all the analyses and together with LB CT was recipient of ABOARD, which was a public-private interpreted the data and wrote the manuscript. OO performed X partnership receiving funding from ZonMW (#73305095007) chromosomal analyses. VD was responsible for EMIF-AD MBD and Health∼Holland, Topsector Life Sciences and Health (PPP- DNA sample preparation and DNA extraction. LD contributed allowance; #LSHM20106). More than 30 partners participate in Frontiers in Aging Neuroscience | www.frontiersin.org 11 March 2022 | Volume 14 | Article 840651 Homann et al. GWAS in EMIF-AD MBD Dataset ABOARD. ABOARD also received funding from Edwin Research (NIHR) Applied Research Collaboration South London Bouw Fonds and Gieskes-Strijbisfonds. RD was supported (NIHR ARC South London) at King’s College Hospital NHS by the NIHR Biomedical Research Centre at South London Foundation Trust. We acknowledge financial support by Land and Maudsley NHS Foundation Trust and King’s College Schleswig-Holstein within the funding programme Open Access London, London, United Kingdom; Health Data Research Publikationsförderung. UK, which was funded by the UK Medical Research Council, Engineering and Physical Sciences Research Council, Economic and Social Research Council, Department of Health and ACKNOWLEDGMENTS Social Care (England), Chief Scientist Office of the Scottish Government Health and Social Care Directorates, Health We acknowledge the assistance of Ellen De Roeck, Naomi De and Social Care Research and Development Division (Welsh Roeck, and Hanne Struyfs (UAntwerp) with data collection. Government), Public Health Agency (Northern Ireland), British We thank Tanja Wesse and Sanaz Sedghpour Sabet at the Heart Foundation and Wellcome Trust; the BigData@Heart Institute of Clinical Molecular Biology, Christian-Albrechts- Consortium, funded by the Innovative Medicines Initiative- University of Kiel, Kiel, Germany for technical assistance with 2 Joint Undertaking under grant agreement No. 116074. the GSA genotyping. We thank Fabian Kilpert for his assistance This Joint Undertaking received support from the European with the QC and genotype imputations and Marcel Schilling Union’s Horizon 2020 Research and Innovation Programme for his assistance on visualizing the results. The LIGA team and EFPIA; it is chaired by DE Grobbee and SD Anker, acknowledges computational support from the OMICS compute partnering with 20 academic and industry partners and ESC; cluster at the University of Lübeck. the National Institute for Health Research University College London Hospitals Biomedical Research Centre; the National Institute for Health Research (NIHR) Biomedical Research SUPPLEMENTARY MATERIAL Centre at South London and Maudsley NHS Foundation Trust and King’s College London; the UK Research and Innovation The Supplementary Material for this article can be found London Medical Imaging and Artificial Intelligence Centre for online at: https://www.frontiersin.org/articles/10.3389/fnagi. Value-Based Healthcare; and the National Institute for Health 2022.840651/full#supplementary-material REFERENCES Das, S., Forer, L., Schönherr, S., Sidore, C., Locke, A. E., Kwong, A., et al. (2016). Next-generation genotype imputation service and methods. Nat. Genet. 48, Armstrong, N. J., Mather, K. A., Sargurupremraj, M., Knol, M. J., Malik, R., 1284–1287. doi: 10.1038/ng.3656 Satizabal, C. L., et al. (2020). Common genetic variation indicates separate Davies, G., Lam, M., Harris, S. E., Trampush, J. W., Luciano, M., Hill, W. D., causes for periventricular and deep white matter hyperintensities. Stroke 51, et al. (2018). Study of 300,486 individuals identifies 148 independent genetic 2111–2121. doi: 10.1161/STROKEAHA.119.027544 loci influencing general cognitive function. Nat. Commun. 9:2098. doi: 10.1038/ Bai, B., Wang, X., Li, Y., Chen, P. C., Yu, K., Dey, K. K., et al. (2020). Deep s41467-018-04362-x multilayer brain proteomics identifies molecular networks in Alzheimer’s de Leeuw, C. A., Mooij, J. M., Heskes, T., and Posthuma, D. (2015). MAGMA: disease progression. Neuron 105, 975–991.e7. doi: 10.1016/j.neuron.2019.12. generalized gene-set analysis of GWAS data. PLoS Comput. Biol. 11:e1004219. 015 doi: 10.1371/journal.pcbi.1004219 Bertram, L., and Tanzi, R. E. (2020). Genomic mechanisms in Alzheimer’s disease. Gao, Q., Mok, H. P., and Zhuang, J. (2019). Secreted modular calcium-binding Brain Pathol. (Zurich, Switzerland) 30, 966–977. doi: 10.1111/bpa.12882 proteins in pathophysiological processes and embryonic development. Chin. Bos, I., Vos, S., Vandenberghe, R., Scheltens, P., Engelborghs, S., Frisoni, G., et al. Med. J. 132, 2476–2484. doi: 10.1097/CM9.0000000000000472 (2018). The EMIF-AD multimodal biomarker discovery study: design, methods Gatz, M., Reynolds, C. A., Fratiglioni, L., Johansson, B., Mortimer, J. A., Berg, S., and cohort characteristics. Alzheimers Res. Ther. 10:64. doi: 10.1186/s13195- et al. (2006). Role of genes and environments for explaining Alzheimer disease. 018-0396-5 Arch. Gen. Psychiatry 63, 168–174. doi: 10.1001/archpsyc.63.2.168 Boyle, A. P., Hong, E. L., Hariharan, M., Cheng, Y., Schaub, M. A., Gottesman, I. I., and Gould, T. D. (2003). The endophenotype concept in Kasowski, M., et al. (2012). Annotation of functional variation in personal psychiatry: etymology and strategic intentions. Am. J. Psychiatry 160, 636–645. genomes using RegulomeDB. Genome Res. 22, 1790–1797. doi: 10.1101/gr.137 doi: 10.1176/appi.ajp.160.4.636 323.112 Graffelman, J., and Weir, B. S. (2016). Testing for Hardy-Weinberg equilibrium Buniello, A., MacArthur, J., Cerezo, M., Harris, L. W., Hayhurst, J., Malangone, at biallelic genetic markers on the X chromosome. Heredity 116, 558–568. C., et al. (2019). The NHGRI-EBI GWAS catalog of published genome-wide doi: 10.1038/hdy.2016.20 association studies, targeted arrays and summary statistics 2019. Nucleic Acids Grasby, K. L., Jahanshad, N., Painter, J. N., Colodro-Conde, L., Bralten, J., Hibar, Res. 47, D1005–D1012. doi: 10.1093/nar/gky1120 D. P., et al. (2020). The genetic architecture of the human cerebral cortex. Bush, W. S., and Moore, J. H. (2012). Chapter 11: genome-wide association Science (New York, N.Y.) 367:eaay6690. doi: 10.1126/science.aay6690 studies. PLoS Comput. Biol. 8:e1002822. doi: 10.1371/journal.pcbi.100 Hibar, D. P., Adams, H., Jahanshad, N., Chauhan, G., Stein, J. L., Hofer, M. A., et al. 2822 (2017). Novel genetic loci associated with hippocampal volume. Nat. Commun. Cacace, R., Sleegers, K., and Van Broeckhoven, C. (2016). Molecular genetics 8:13624. doi: 10.1038/ncomms13624 of early-onset Alzheimer’s disease revisited. Alzheimers Dement. 12, 733–748. Holtzman, D. M., Herz, J., and Bu, G. (2012). Apolipoprotein E and apolipoprotein doi: 10.1016/j.jalz.2016.01.012 E receptors: normal biology and roles in Alzheimer disease. Cold Spring Harb. Choi, S. W., and O’Reilly, P. F. (2019). PRSice-2: polygenic Risk Score software for Perspect. Med. 2:a006312. doi: 10.1101/cshperspect.a006312 biobank-scale data. GigaScience 8:giz082. doi: 10.1093/gigascience/giz082 Hong, S., Dobricic, V., Ohlei, O., Bos, I., Vos, S., Prokopenko, D., et al. Damale, M. G., Pathan, S. K., Shinde, D. B., Patil, R. H., Arote, R. B., and Sangshetti, (2021). TMEM106B and CPOX are genetic determinants of cerebrospinal fluid J. N. (2020). Insights of tankyrases: a novel target for drug discovery. Eur. J. Med. Alzheimer’s disease biomarker levels. Alzheimers Dement. 17, 1628–1640. doi: Chem. 207:112712. doi: 10.1016/j.ejmech.2020.112712 10.1002/alz.12330 Frontiers in Aging Neuroscience | www.frontiersin.org 12 March 2022 | Volume 14 | Article 840651 Homann et al. GWAS in EMIF-AD MBD Dataset Hong, S., Prokopenko, D., Dobricic, V., Kilpert, F., Bos, I., Vos, S., et al. (2020). Rentzsch, P., Schubach, M., Shendure, J., and Kircher, M. (2021). CADD- Genome-wide association study of Alzheimer’s disease CSF biomarkers in the Splice-improving genome-wide variant effect prediction using deep EMIF-AD multimodal biomarker discovery dataset. Transl. Psychiatry 10:403. learning-derived splice scores. Genome Med. 13:31. doi: 10.1186/s13073-021- doi: 10.1038/s41398-020-01074-z 00835-9 Jansen, I. E., Savage, J. E., Watanabe, K., Bryois, J., Williams, D. M., Steinberg, S., Sargurupremraj, M., Suzuki, H., Jian, X., Sarnowski, C., Evans, T. E., Bis, J. C., et al. et al. (2019). Genome-wide meta-analysis identifies new loci and functional (2020). Cerebral small vessel disease genomics and its implications across the pathways influencing Alzheimer’s disease risk. Nat. Genet. 51, 404–413. doi: lifespan. Nat. Commun. 11:6285. doi: 10.1038/s41467-020-19111-2 10.1038/s41588-018-0311-9 Sathe, G., Albert, M., Darrow, J., Saito, A., Troncoso, J., Pandey, A., et al. (2021). Lonsdale, J., Thomas, J., Salvatore, M., Phillips, R., Lo, E., Shad, S., et al. (2013). Quantitative proteomic analysis of the frontal cortex in Alzheimer’s disease. The genotype-tissue expression (GTEx) project. Nat. Genet. 45, 580–585. doi: J. Neurochem. 156, 988–1002. doi: 10.1111/jnc.15116 10.1038/ng.2653 Smith, S. M., Douaud, G., Chen, W., Hanayik, T., Alfaro-Almagro, F., Sharp, K., Machiela, M. J., and Chanock, S. J. (2015). LDlink: a web-based application for et al. (2021). An expanded set of genome-wide association studies of brain exploring population-specific haplotype structure and linking correlated alleles imaging phenotypes in UK Biobank. Nat. Neurosci. 24, 737–745. doi: 10.1038/ of possible functional variants. Bioinformatics (Oxf. Engl.) 31, 3555–3557. doi: s41593-021-00826-4 10.1093/bioinformatics/btv402 Sperling, R., Mormino, E., and Johnson, K. (2014). The evolution of preclinical MacRae, C. A., and Vasan, R. S. (2011). Next-generation genome-wide association Alzheimer’s disease: implications for prevention trials. Neuron 84, 608–622. studies: time to focus on phenotype? Circ. Cardiovasc. Genet. 4, 334–336. doi: doi: 10.1016/j.neuron.2014.10.038 10.1161/CIRCGENETICS.111.960765 Ten Kate, M., Redolfi, A., Peira, E., Bos, I., Vos, S. J., Vandenberghe, R., et al. (2018). Mattsson, N., Carrillo, M. C., Dean, R. A., Devous, M. D. Sr., Nikolcheva, T., Pesini, MRI predictors of amyloid pathology: results from the EMIF-AD multimodal P., et al. (2015). Revolutionizing Alzheimer’s disease and clinical trials through biomarker discovery study. Alzheimers Res. Ther. 10:100. doi: 10.1186/s13195- biomarkers. Alzheimers Dement. (Amst.) 1, 412–419. doi: 10.1016/j.dadm.2015. 018-0428-1 09.001 van der Meer, D., Frei, O., Kaufmann, T., Shadrin, A. A., Devor, A., Smeland, Mazaheri Moghaddam, M., Mazaheri Moghaddam, M., Hamzeiy, H., O. B., et al. (2020). Understanding the genetic determinants of the brain with Baghbanzadeh, A., Pashazadeh, F., and Sakhinia, E. (2021). Genetic basis MOSTest. Nat. Commun. 11:3512. doi: 10.1038/s41467-020-17368-1 of acephalic spermatozoa syndrome, and intracytoplasmic sperm injection Wang, H., Dey, K. K., Chen, P. C., Li, Y., Niu, M., Cho, J. H., et al. (2020). Integrated outcomes in infertile men: a systematic scoping review. J. Assist. Reprod. Genet. analysis of ultra-deep proteomes in cortex, cerebrospinal fluid and serum 38, 573–586. doi: 10.1007/s10815-020-02008-w reveals a mitochondrial signature in Alzheimer’s disease. Mol. Neurodegen. McCarthy, S., Das, S., Kretzschmar, W., Delaneau, O., Wood, A. R., Teumer, A., 15:43. doi: 10.1186/s13024-020-00384-6 et al. (2016). A reference panel of 64,976 haplotypes for genotype imputation. Watanabe, K., Taskesen, E., van Bochoven, A., and Posthuma, D. (2017). Nat. Genet. 48, 1279–1283. doi: 10.1038/ng.3643 Functional mapping and annotation of genetic associations with FUMA. Nat. McKhann, G., Drachman, D., Folstein, M., Katzman, R., Price, D., and Stadlan, Commun. 8:1826. doi: 10.1038/s41467-017-01261-5 E. M. (1984). Clinical diagnosis of Alzheimer’s disease: report of the NINCDS- Whelan, C. D., Mattsson, N., Nagle, M. W., Vijayaraghavan, S., Hyde, C., ADRDA work group under the auspices of department of health and human Janelidze, S., et al. (2019). Multiplex proteomics identifies novel CSF and plasma services task force on Alzheimer’s disease. Neurology 34, 939–944. doi: 10.1212/ biomarkers of early Alzheimer’s disease. Acta Neuropathol. Commun. 7:169. wnl.34.7.939 doi: 10.1186/s40478-019-0795-2 McLaren, W., Gil, L., Hunt, S. E., Riat, H. S., Ritchie, G. R., Thormann, A., et al. Wightman, D. P., Jansen, I. E., Savage, J. E., Shadrin, A. A., Bahrami, S., Holland, (2016). The ensembl variant effect predictor. Genome Biol. 17:122. doi: 10.1186/ D., et al. (2021). A genome-wide association study with 1,126,563 individuals s13059-016-0974-4 identifies new risk loci for Alzheimer’s disease. Nat. Genet. 53, 1276–1282. Montani, C., Gritti, L., Beretta, S., Verpelli, C., and Sala, C. (2019). The synaptic and doi: 10.1038/s41588-021-00921-z neuronal functions of the X-Linked intellectual disability protein interleukin- Willer, C. J., Li, Y., and Abecasis, G. R. (2010). METAL: fast and efficient 1 receptor accessory protein like 1 (IL1RAPL1). Dev. Neurobiol. 79, 85–95. meta-analysis of genomewide association scans. Bioinformatics (Oxf. Engl.) 26, doi: 10.1002/dneu.22657 2190–2191. doi: 10.1093/bioinformatics/btq340 Morkmued, S., Clauss, F., Schuhbaur, B., Fraulob, V., Mathieu, E., Hemmerlé, Winblad, B., Palmer, K., Kivipelto, M., Jelic, V., Fratiglioni, L., Wahlund, L. O., J., et al. (2020). Deficiency of the SMOC2 matricellular protein impairs bone et al. (2004). Mild cognitive impairment–beyond controversies, towards a healing and produces age-dependent bone loss. Sci. Rep. 10:14817. doi: 10.1038/ consensus: report of the international working group on mild cognitive s41598-020-71749-6 impairment. J. Intern. Med. 256, 240–246. doi: 10.1111/j.1365-2796.2004. Neu, S. C., Pa, J., Kukull, W., Beekly, D., Kuzma, A., Gangadharan, P., et al. 01380.x (2017). Apolipoprotein E genotype and sex risk factors for Alzheimer disease: Zhang, Q., Cai, Z., Lhomme, M., Sahana, G., Lesnik, P., Guerin, M., et al. a meta-analysis. JAMA Neurol. 74, 1178–1189. doi: 10.1001/jamaneurol.2017. (2020). Inclusion of endophenotypes in a standard GWAS facilitate a detailed 2188 mechanistic understanding of genetic elements that control blood lipid levels. Omodei, D., Acampora, D., Russo, F., De Filippi, R., Severino, V., Di Francia, Sci. Rep. 10:18434. doi: 10.1038/s41598-020-75612-6 R., et al. (2009). Expression of the brain transcription factor OTX1 occurs Zhao, B., Zhang, J., Ibrahim, J. G., Luo, T., Santelli, R. C., Li, Y., et al. (2021). Large- in a subset of normal germinal-center B cells and in aggressive Non- scale GWAS reveals genetic architecture of brain white matter microstructure Hodgkin Lymphoma. Am. J. Pathol. 175, 2609–2617. doi: 10.2353/ajpath.2009. and genetic overlap with cognitive and mental health traits (n=17,706). Mol. 090542 Psychiatry 26, 3943–3955. doi: 10.1038/s41380-019-0569-z Patterson, C. (2018). World Alzheimer Report 2018. The State of the Art of Dementia Research: New Frontiers. An Analysis of Prevalence, Incidence, Cost and Trends. Conflict of Interest: HZ has served at scientific advisory boards and/or as a London: Alzheimer’s Disease International. consultant for Abbvie, Alector, Annexon, Artery Therapeutics, AZTherapies, Petersen, R. C. (2004). Mild cognitive impairment as a diagnostic entity. J. Intern. CogRx, Denali, Eisai, Nervgen, Pinteon Therapeutics, Red Abbey Labs, Passage Med. 256, 183–194. doi: 10.1111/j.1365-2796.2004.01388.x Bio, Roche, Samumed, Siemens Healthineers, Triplet Therapeutics, and Wave, has Purcell, S., Neale, B., Todd-Brown, K., Thomas, L., Ferreira, M. A., Bender, D., et al. given lectures in symposia sponsored by Cellectricon, Fujirebio, Alzecure, Biogen, (2007). PLINK: a tool set for whole-genome association and population-based and Roche, and was a co-founder of Brain Biomarker Solutions in Gothenburg AB linkage analyses. Am. J. Hum. Genet. 81, 559–575. doi: 10.1086/519795 (BBS), which was a part of the GU Ventures Incubator Program. FB is supported by Rai, A., Bleimling, N., Vetter, I. R., and Goody, R. S. (2020). The mechanism of the NIHR biomedical research centre at UCLH. JP received consultation honoraria activation of the actin binding protein EHBP1 by Rab8 family members. Nat. from Nestle Institute of Health Sciences, Ono Pharma, OM Pharma, and Fujirebio, Commun. 11:4187. doi: 10.1038/s41467-020-17792-3 unrelated to the submitted work. CT has a collaboration contract with ADx Raymond, F. L. (2006). X linked mental retardation: a clinical guide. J. Med. Genet. Neurosciences, Quanterix and Eli Lilly, performed contract research or received 43, 193–200. doi: 10.1136/jmg.2005.033043 grants from AC-Immune, Axon Neurosciences, Biogen, Brainstorm Therapeutics, Frontiers in Aging Neuroscience | www.frontiersin.org 13 March 2022 | Volume 14 | Article 840651 Homann et al. GWAS in EMIF-AD MBD Dataset Celgene, EIP Pharma, Eisai, PeopleBio, Roche, Toyama, Vivoryon. She serves Publisher’s Note: All claims expressed in this article are solely those of the authors on editorial boards of Medidact Neurologie/Springer, Alzheimer Research and and do not necessarily represent those of their affiliated organizations, or those of Therapy, Neurology: Neuroimmunology and Neuroinflammation, and was editor the publisher, the editors and the reviewers. Any product that may be evaluated in of a Neuromethods book Springer. CT also holds a speaker’s contract with Roche, this article, or claim that may be made by its manufacturer, is not guaranteed or Inc. KB has served as a consultant, at advisory boards, or at data monitoring endorsed by the publisher. committees for Abcam, Axon, BioArctic, Biogen, JOMDD/Shimadzu. Julius Clinical, Lilly, MagQu, Novartis, Pharmatrophix, Prothena, Roche Diagnostics, Copyright © 2022 Homann, Osburg, Ohlei, Dobricic, Deecke, Bos, Vandenberghe, and Siemens Healthineers, and was a co-founder of Brain Biomarker Solutions in Gabel, Scheltens, Teunissen, Engelborghs, Frisoni, Blin, Richardson, Bordet, Lleó, Gothenburg AB (BBS), which was a part of the GU Ventures Incubator Program, Alcolea, Popp, Clark, Peyratout, Martinez-Lage, Tainta, Dobson, Legido-Quigley, outside the work presented in this paper. SL is currently an employee at Janssen Sleegers, Van Broeckhoven, Wittig, Franke, Lill, Blennow, Zetterberg, Lovestone, Medical UK. JS was an employee of Janssen R&D, LLC., and is currently an Streffer, ten Kate, Vos, Barkhof, Visser and Bertram. This is an open-access article employee and chief medical officer of AC Immune SA. JR was an employee of distributed under the terms of the Creative Commons Attribution License (CC BY). Neurosciences Therapeutic Area, GlaxoSmithKline R&D, Stevenage, UK. The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original The remaining authors declare that the research was conducted in the absence of publication in this journal is cited, in accordance with accepted academic practice. any commercial or financial relationships that could be construed as a potential No use, distribution or reproduction is permitted which does not comply with conflict of interest. these terms. Frontiers in Aging Neuroscience | www.frontiersin.org 14 March 2022 | Volume 14 | Article 840651