|Home | About | Journals | Submit | Contact Us | Français|
Reduction in hippocampal and amygdala volume measured via structural magnetic resonance imaging is an early marker of Alzheimer’s disease (AD). Whether genetic risk factors for AD exert an effect on these subcortical structures independent of clinical status has not been fully investigated. We examine whether increased genetic risk for AD influences hippocampal and amygdala volumes in case-control and population cohorts at different ages, in 1674 older (aged >53 years; 17% AD, 39% mild cognitive impairment [MCI]) and 467 young (16–30 years) adults. An AD polygenic risk score combining common risk variants excluding apolipoprotein E (APOE), and a single nucleotide polymorphism in TREM2, were both associated with reduced hippocampal volume in healthy older adults and those with MCI. APOE 4 was associated with hippocampal and amygdala volume in those with AD and MCI but was not associated in healthy older adults. No associations were found in young adults. Genetic risk for AD affects the hippocampus before the clinical symptoms of AD, reflecting a neurodegenerative effect before clinical manifestations in older adults.
The strongest identified genetic risk factor for Alzheimer’s disease (AD) is the apolipoprotein E (APOE) 4 allele (Genin et al., 2011). Large-scale GWASMA (Genome Wide Association Study Meta Analyses) have identified an additional 19 common risk loci with small effects on AD risk (Lambert et al., 2013). A low-frequency missense variant in TREM2 (p.R47H or rs75932628) substantially increases AD risk (Guerreiro et al., 2013). Whether these variants exert an effect on disease-related phenotypes (such as brain atrophy) in the early stages of AD, or before clinical onset is largely unknown.
The earliest histopathological changes in AD are typically seen within the medial temporal lobe, where neurofibrillary tangles and amyloid depositions first form. Beginning in the preclinical phase, these lesions lead to changes in regional brain volumes, in particular, the hippocampus and amygdala (Yang et al., 2012). Brain volume reduction is evident in other disorders (e.g., depression and anxiety) as well as in healthy aging, with the hippocampus being especially vulnerable (Small et al., 2011). Determining how AD risk variants affect hippocampal and amygdala volume directly, and whether this is detectable before clinical manifestations of AD, will provide clues as to how they contribute to disease risk.
An effect of APOE status has been observed on structural brain changes in the elderly. Carriers of 4 are generally found to have smaller hippocampal and amygdala volumes than homozygous 3 subjects, but this is not consistently observed before the onset of mild cognitive impairment (MCI) or AD (Hostage et al., 2013; Khan et al., 2014; Liu et al., 2010). Effects have also been identified in young people (O’Dwyer et al., 2012), though findings have also been inconsistent (Khan et al., 2014). In addition, sex differences have been reported, with greater deleterious effect of APOE 4 on hippocampal pathology in females (Fleisher et al., 2005).
Other common AD genetic risk factors identified though GWAS studies have been investigated in relation to hippocampal and amygdala volume, through the use of polygenic risk scores (PRSs). A PRS allows the identification of phenotypic associations that would not be detectable using single variants with low effect size, as well as allowing a reduction in the number of statistical tests (Wray et al., 2014). A PRS containing the first AD GWAS findings (3 genome-wide associated variants) was found to be associated with clinical diagnosis and reduced hippocampal and amygdala volume in the AD case and/or control cohort Alzheimer’s Disease Neuroimaging Initiative (ADNI; Biffi et al., 2010). A proxy for the rare TREM2 risk variant (rs9394721) was also associated with smaller hippocampal volume and increased rate of temporal lobe atrophy in the ADNI cohort (Rajagopalan et al., 2013).
A recent study assessed the effect of 20 AD risk variants combined in a PRS with various magnetic resonance imaging (MRI) markers of brain aging (intracranial volume, total brain volume, hippocampal volume, white matter hyperintensities, and brain infarcts) in non-demented older community persons (Chauhan et al., 2015). The PRS was applied to meta-analysis summary estimates from 10 population-based studies (total N = 11,500). An association was observed with smaller hippocampal volume only, which remained significant after excluding APOE. Here, we extend on this previous work by investigating the effect of AD genetic risk variants on hippocampal volume, and also investigate amygdala volume in our large sample (N > 2000). By accessing the raw genotyping data, as opposed to meta-analysis summary estimates, we are able to test for an effect in distinct diagnostic groups, including AD, MCI, and healthy elderly to examine at which clinical stage effects can be seen. We also tested for any early effects of AD risk factors on hippocampal and amygdala volume in healthy young adults before substantial age-related atrophy. Age and sex interaction effects were also investigated. We used the most recent GWAS findings identified by the International Genomics of Alzheimer’s Project (IGAP) GWASMA which included 74,046 individuals (Lambert et al., 2013) to select the 19 genome-wide significant AD risk variants to include in the PRS. We also examine the effect of several additional PRS adding increasing numbers of single nucleotide polymorphisms (SNPs) at different p value thresholds of association. Inclusion of SNPs that do not pass the threshold for genome-wide significance, but include a proportion of truly associated variants will give increased power to detect an association up to an optimal p value cutoff (Wray et al., 2014).
Five cohorts, including two case-control and 3 population based, were used (Table 1). ADNI (Mueller et al., 2005; Alzheimer’s Disease Neuroimaging Initiative, www.adni-info.org) and AddNeuroMed (Westman et al., 2011, Innovative Medicines (InnoMed) in Europe) are comprised of AD cases, MCI, and aged-matched controls (Table 1). All AD cases met criteria for either probable or definite AD with inclusion criteria as previously described (Simmons et al., 2011 and www.adni-info.org). MCI was assessed as having an abnormal memory complaint but with general cognition and functional performance sufficiently preserved such that a diagnosis of AD cannot be made. Elderly controls were screened for dementia.
The population cohorts were the Older Australian Twins Study (OATS; Sachdev et al., 2011), the Sydney Memory and Ageing Study (MAS; Brodaty et al., 2012), and the Queensland Twin Imaging (QTIM) cohort, of which the latter consists of young adults (Hibar et al., 2013). For Sydney MAS and OATS, diagnosis of MCI and AD were made with the most recent consensus criteria (Winblad et al., 2004). For all those participants whose neuropsychological or functional profiles indicated the possibility of dementia, a diagnosis was made at a consensus meeting (for a detailed description of Sydney MAS and OATS methodologies see Sachdev et al., 2009, 2012). Both Sydney MAS and OATS are longitudinal studies; here, we used the MRI data and diagnosis of MCI or AD at baseline (i.e., on admission).
All cohorts are independent of the IGAP GWASMA except for ADNI, which contributed 1.6% of the AD cases and 0.5% of the controls (Lambert et al., 2013). In this study, we formed an all-older group, aged 53–91 years, from 4 cohorts (ADNI, AddNeuroMed, OATS, and Sydney MAS), and stratified into AD, MCI, and healthy older groups. The QTIM cohort (aged 16–30 years) formed a separate young adult group (Table 1). As OATS and QTIM cohorts contain twins, we omitted related individuals at random.
Subcortical volumes for the hippocampus and amygdala, and intracranial volume (ICV) were extracted from anatomical T1-weighted magnetic resonance images (image acquisition is described in Supplementary Methods 1 and Supplementary Table 1), using validated automated segmentation programs following the Enhancing Neuro Imaging Genetics through Meta-Analysis Consortium protocols (Stein et al., 2012).
For the ADNI, AddNeuroMed, and QTIM samples, bilateral amygdala and hippocampus volume segmentation was performed using the Freesurfer image analysis suite, previously reported in depth (Fischl et al., 2002). Briefly, the T1 structural MRI scan is corrected for intensity bias, skull stripped, and transformed to Talairach space. Each voxel within the MRI volume is then assigned a neuroanatomic label (including left and right hippocampus and amygdala) based on probabilistic information estimated from a manually labeled training set. ICV was estimated based on the determinant of the transformation matrix used when transforming the MR volume to Talairach space.
FSL FIRST was used to segment subcortical structures for the Sydney MAS and OATS datasets as previously reported (Patenaude et al., 2011). Input images were registered to MNI space through 2-stage linear transformation. Deformable mesh models, based on shape and intensity information from a manually segmented training set, were then used to segment bilateral hippocampus and amygdala volumes. ICV was calculated as the inverse of the determinant of the affine transformation matrix, multiplied by the size of the MNI template.
We calculated the mean of left and right hippocampal volume and amygdala volume. All subcortical volumes and ICV outliers were winsorized to 4 standard deviations from the mean.
A PRS was constructed from genome-wide SNP array data using 19 genome-wide significant AD risk variants (from IGAP, PRS p < 5 × 10−8; Lambert et al., 2013). Scores were calculated by summing the number of risk alleles weighted by the effect size (log odds ratio [OR]; Supplementary Table 2).
Threshold PRS was calculated with stage 1 summary data from IGAP using the method previously described (Purcell et al., 2009; see Supplementary Methods 3 for details of the IGAP discovery sample). SNPs within 500 kb either side of the APOE locus were excluded to ensure all APOE-associated signal was removed. LD-based clumping was carried out on all SNPs in the summary data, providing the most significantly associated SNP in each region of LD (using PLINK clumping command with a pairwise r2 threshold of 0.2 and a physical distance threshold of 300 kb). SNPs were checked for flip strands between the summary data and each cohort. We calculated the total score for each individual as the number of score alleles weighted by the log of the OR from the discovery SNPs from the IGAP sample (using PLINK score function). The risk score calculation was repeated for p-value thresholds of p < 1 × 10−6, p < 1 × 10−4, p < 1 × 10−3, p < 0.01, p < 0.05, p < 0.1, p < 0.5, and p < 1 (all SNPs). The number of SNPs included in each risk score is shown in Supplementary Table 3.
The PRS method assumes a polygenic disease model and is suitable for common variants with the assumptions of an additive effect and independent contribution to risk (Wray et al., 2014). Both APOE and TREM2 don’t meet these assumptions and were assessed separately from the PRS. APOE 4 allele is a diplotype acting under a codominant genetic model, and with a much larger effect size than the other common AD risk variants (Genin et al., 2011). For the rare TREM2 variant (p.R47H/rs75932628), we used rs9394721, the closest available imputed proxy (r2 = 0.492; Rajagopalan et al., 2013). APOE genotyping was carried out as previously described (Jorm et al., 2007). Haplotype 2/4 carriers were excluded from the analysis due to the potential for counteracting effects of these alleles (1.7% of individuals).
To assess how well the PRS predicted AD risk, we first tested for an association of each PRS with clinical status (excluding MCI) in the AddNeuroMed cohort which is independent from the IGAP discovery sample, using logistic regression (STATA version 11), controlling for age, sex, and 4 ancestry principal components. Discriminative improvement of each PRS was assessed using receiver operating characteristic curves. The equality of each area under the curve for the receiver operating characteristic curves were testing using STAT “roccomp” taking into account the implicit correlation between curves (as the test is applied to the same sample). A covariance matrix is estimated using the method of structural components, and the resulting test statistic has an asymptotically χ2 distribution (DeLong et al., 1988). We also tested the association between both the APOE 4 genotype and the TREM2 SNP and AD risk.
Next, in each of the 5 cohorts, we tested for an effect of PRS, the number of TREM2 rs9394721 and APOE 4 alleles on mean hippocampal and amygdala volumes, using linear regression. Covariates included ICV, age, sex, and 4 ancestry principal components. Age2, age × sex, age2 × sex interactions were also included as covariates if they showed evidence of an association (p < 0.05).
We then combined all cohorts in a mega-analysis to test for associations in the all-older group (a combined dataset incorporating the AD, MCI, and healthy older groups shown in Table 1), controlling for study and clinical status as well as stratifying by clinical status (AD, MCI, and healthy older). Where there was a significant association, we repeated the analyses testing for interaction effects of both sex and age, by including an interaction term for both sex and age with the independent variable (either number of APOE 4 alleles, TREM2 rs9394721 alleles, and PRS) in the regression equation. Nonsignificant product terms were removed and the regression repeated. Interactions were identified by a significant product term and the nature of the interaction was investigated by testing the association in separate age and gender groups.
For all regression analysis, the variance explained (R2) was calculated by taking the R2 value of the full model (covariates and genotype/PRS) and subtracting the R2 of the reduced model (covariates only). We also assessed the total variance explained by all the investigated AD risk variants by including PRS p < 1 × 10−4, APOE 4, and TREM2 together within a single multiple regression and again subtracting the R2 of the reduced model (covariates only) from the total R2.
Owing to ascertainment and measurement differences between cohorts, we also carried out a meta-analysis (STATA METAN specifying a random effects model) and tested for study heterogeneity. Meta-analyses tested for an association between the number of APOE 4 alleles, the PRS p < 0.001 and TREM2 rs9394721, with hippocampal volume and amygdala in the combined all-older group (controlling for disease status), AD, MCI, and healthy older groups.
The AD PRS was associated with AD risk in the AddNeuroMed cohort. The most significantly associated threshold was p < 1 × 10−3 (OR = 1.51; p = 0.011), though this had no more discriminative accuracy than a model with only age and sex covariates. However, APOE 4 genotype was highly associated with AD risk (OR = 2.41 p = 1.64 × 10−5) with significant improvement in discriminative accuracy over the covariates. In contrast, TREM2 rs9394721 was not associated with AD risk in this small sample (Table 2).
The effects of PRS, TREM2, and APOE on hippocampal and amygdala volume in the mega-analysis are shown in Tables 3 and and44 and Fig. 1, with Table 3 showing results for the combined all-older and the young adults, and Table 4 stratifying the older group by clinical group (AD, MCI, and healthy older). Results from each of the 4 older cohorts when analyzed separately are presented in Supplementary Table 4.
APOE 4 status was strongly associated with lower hippocampal and amygdala volumes in the all-older group. However, when stratified by clinical status, APOE 4 associated with lower volumes in the AD and MCI groups, but not in the healthy older group. In the AD group, this constitutes each APOE 4 allele resulting in an average of 143 mm3 (4.8%) reduction in mean hippocampal volume and 47 mm3 in mean amygdala volume (3.8%). For MCI, this was a 132 mm3 reduction in hippocampal volume (4.0%), and 37 mm3 in amygdala volume (2.8%). No associations were found in the young adults.
We also found an association of PRS containing common AD risk variants of small effect with reduced hippocampal volume in the all-older group. The strongest effect was for the PRS containing SNPs less than the p < 10−4 threshold (p = 0.004). This association was suggestive in the stratified MCI (N = 645 PRS p < 1 × 10−4 and p = 0.057) and healthy older (N = 723, PRS, p < 1 × 10−3 and p = 0.075) subgroups but not apparent in the smaller AD group. In the young adults, there was no effect of the PRS on hippocampal volume, and no effect of PRS was identified for amygdala volume in any group.
TREM2 rs9394721 was associated with lower hippocampal volume in the all-older group. It showed suggestive association in the MCI and healthy older groups (significant when not corrected for multiple testing), but not in the smaller AD group or young adults. No association was found between TREM2 and amygdala volume.
The variance explained (R2) by APOE 4, PRS, and TREM2 rs9394721 on hippocampal and amygdala volume are shown for each regression in Tables 3 and and4,4, and for hippocampal volume in Fig. 1. When we assessed the variance explained by all the genetic risk factors combined (by including APOE 4, PRS P<1e04 and TREM2 rs9394721 within a single multiple regression) on hippocampal volume, we found that they accounted for an R2 of 1.6% in the all-older group, 3% in the AD group, 3.2% in the MCI group, and 0.3% in the healthy older group.
Age and sex interactions were identified and are shown as footnotes in Tables 3 and and4.4. The APOE effect was stronger in females (in the all-older group N = 817, β = −0.15, p = 3.4 × 10−8), with an interaction found between APOE 4 and sex in both AD and MCI groups. There was also an interaction between PRS (testing threshold, p < 1 × 10−3) and sex, where the effect was again driven by the females (in the all-older group: N = 848, β = −0.08, p = 0.001). For TREM2, we identified an age × rs9394721 interaction in the all-older group, with an association found for those ≤75 years (N = 791, β = −0.10, p = 3.4 × 10−4) but not >75 years, and no frequency difference between the 2 age groups. When stratified by clinical status the age effect was only evident in the MCI group (≤75 years: N = 292, β = −0.19, p = 0.003).
The results for the meta-analysis are shown in Supplementary Table 5 and Supplementary Figs 1–3. The association of APOE 4 genotype and hippocampal volume in MCI and AD was confirmed but the association between APOE 4 and amygdala volume was not significant in the meta-analysis, likely due to heterogeneity across studies in the measurement of this structure. The association of TREM2 rs9394721 was confirmed in the all-older group but was not significant in any of the clinical groups. Similarly, the association between PRS and hippocampal volume was not significant, reflecting the small effect and loss of power due to the meta-analysis.
Chauhan et al. (2015) recently showed that an AD PRS constructed from 20 AD risk loci associated with reduced hippocampal volume in a large population-based meta-analysis. Here, we confirm these findings using raw genotype level data using a smaller sample size, but including a larger number of variants in the PRS identified in the largest GWASMA available [IGAP discovery sample: 17,008 cases, 37,154 controls (Lambert et al., 2013)]. We also investigate a variant in TREM2. Through stratification into AD, MCI, and healthy older groups, we show at which disease stage associations can be identified.
In our cross-study mega-analyses, we confirm the association between APOE and hippocampal volume in adults with AD and MCI (4 carriers having smaller volumes compared to noncarriers), but show no association in healthy older adults or in young adults. The same pattern of association, although to a lesser degree, was found for amygdala volume. In addition, AD PRS (excluding APOE) and the rare variant TREM2 were also found to be associated with hippocampal volume in older adults (i.e., all-older group). The AD PRS association was driven by females, whereas the TREM2 association was limited to those aged 75 years and under, and both appear associated with volume loss during healthy aging and in cases of MCI. Notably, the PRS explain substantially less of the variance in hippocampal volume than APOE genotype. Nevertheless, in the AD and MCI cohorts, the combined effects of APOE 4, the AD risk score, and TREM2 accounted for more variance in hippocampal volume than APOE 4 alone (Fig. 1). Combining the effects of the AD genetic risk variants (i.e., APOE, AD PRS, and TREM2), increases the total variance in hippocampal volume that can be explained in AD (to 3%) and in those with MCI (to 3.2%). However, in the healthy elderly TREM2 is a better predictor alone (accounting for 0.3% of variance in hippocampal volume). Longitudinal studies exploring healthy aging and transition to MCI and AD will provide further clarity regarding these genetic profiles.
In contrast to the mega-analysis findings, meta-analyses only confirmed the strong association between APOE and hippocampal volume, and the TREM2 rs9394721 in the larger combined all-older group. The reduction in significance in the groups with small sample sizes likely reflects heterogeneity across studies, and highlights the additional insights to be gained from mega-analyses.
The differing effects of AD risk variants at different disease stages may give insight into the mechanisms of how they contribute to AD risk. Shared mechanisms may drive brain aging and AD, with clinical onset resulting when brain aging surpasses a threshold (Swerdlow, 2011). Support for this hypothesis comes from gene expression studies that identify substantial overlap between expression changes with age, and in AD (Avramopoulos et al., 2011). An opposing view is that volume differences associated with AD risk may be identifiable before clinical diagnosis because they are prodromal changes. Amyloid deposition and the specific pattern and acceleration of atrophy in AD compared to normal aging suggest that early AD is different from normal aging (Fjell et al., 2014). Recent findings suggest that AD-related hippocampal atrophy can be detected 4.2 years before onset of clinical manifestations of dementia (Villemagne et al., 2013). Testing the effect of these variants in middle age is the important next step in ascertaining how early the AD risk variant effect on degeneration can be identified. We found no association in young adulthood, although in a sub-sample of the same cohort previous work has found an association between the CLU genotype and white matter microstructure (Braskie et al., 2011), suggesting other MRI measures may be more sensitive to early differences. This could include examination of the hippocampal microstructure, such as CA1 and subiculum regions, as patterns in volume reductions in subfields of the hippocampus together with entorhinal cortex may differentiate between AD and healthy aging (Wisse et al., 2014). We show that AD variants lower than genome-wide significance contribute to the variance in hippocampal volume. The PRS containing SNPs lower than the p < 10−4 threshold was the most highly associated with AD disease status and with hippocampal volume in older adults. The threshold that maximizes the variance explained in the target sample depends on the size of the GWASMA discovery sample and the underlying genetic architecture (Wray et al., 2014). Notably, in preliminary analyses, we did not identify any association with PRS and hippocampal volume when using data from an earlier AD GWASMA with a smaller sample size [GERAD discovery sample: 3941 cases, 7848 controls (Harold et al., 2009); data not shown]. Large-scale studies also using the GERAD discovery sample have found no association of AD PRS without APOE with memory or cognitive ability in people without dementia (Harris et al., 2014; Verhaaren et al., 2013). In light of our results, these should now be tested using the latest IGAP GWASMA.
Our finding that PRS has a stronger effect on hippocampal volume in females adds to the literature showing the importance of taking sex into account in genetic association analysis in AD. The stratification of AD GWASMA by sex would be useful, allowing the generation of sex-specific PRS (Altmann et al., 2014; Azad et al., 2007). All cohorts in the discovery sample had excess females (58%–68%) (Lambert et al., 2013), so the PRS may be biased toward female genetic risk factors.
TREM2 is a receptor expressed on microglia that stimulates phagocytosis of cell debris and suppresses inflammatory reactivity. Over expression in the brains of AD transgenic mice ameliorates Aβ deposition, neuroinflammation, and neuronal loss (Jiang et al., 2014b). Thus, in humans, mutation carriers may have an inflammatory phenotype with impaired tissue debris clearance resulting in increased gray matter atrophy during aging. We build on the previous finding of an association of rs9394721, a R47H proxy with hippocampal volume in ADNI (Rajagopalan et al., 2013), and show that the association is independent of clinical status and evidence of an association is detectable before onset of MCI. Indeed, we found associations to be limited to the healthy older adult and MCI groups, and the effect to be stronger in those aged ≤75 years. Previously, the R47H genotype has been associated with lower cognitive function in those without AD and has been shown to confer increased risk for Parkinson’s disease, frontotemporal dementia, and amyotrophic lateral sclerosis (Jonsson et al., 2013; Yaghmoor et al., 2014). As rs9394721 is a relatively poor proxy for the disease causing variant (r2 = 0.492) (Rajagopalan et al., 2013), it is likely that a stronger association would be identified if direct genotyping results were available.
We did not find any association of PRS, or TREM2 on amygdala volume, contrary to the association of PRS and amygdala volume reported in the ADNI case-control cohort (Biffi et al., 2010). Amygdala volume is significantly reduced in AD in a similar magnitude to the hippocampus (Klein-Koerkamp et al., 2014). However, we have previously shown that amygdala atrophy is not identified over a 2 year follow-up period in elderly population samples, which contrasts with significant atrophy in the hippocampus (Jiang et al., 2014a). Less age-related atrophy in the amygdala and potential unreliably of segmentation of this small structure may reduce power to detect an effect.
APOE 4 was associated with lower hippocampal and amygdala volumes in the combined all-older group, with a large effect in those with AD and MCI (≥4% average difference in hippocampus and ≥2.8% in amygdala per APOE 4 allele). The association is stronger in females, as previously shown in those with MCI (Fleisher et al., 2005). An increasing body of work shows an interaction between APOE and sex in AD risk which may be explained by the influence of estrogen levels acting in concert with APOE (Altmann et al., 2014; Stone et al., 1997). APOE 4 was not associated with hippocampal volume in healthy older conflicting with some previous reports (Biffi et al., 2010; Lind et al., 2006; Reiman et al., 1998; Wishart et al., 2006) but in agreement with recent findings from other large-scale studies, including both measures of volume and atrophy (Ferencz et al., 2013; Manning et al., 2014). Population studies of older people are likely to have a proportion of individuals with MCI, which may be the source of identified APOE association in conflicting reports. Even so, an effect in the healthy elderly may be detectable in a larger sample, especially as there is a reduced frequency of APOE 4 alleles compared to AD and MCI groups. In cognitively healthy adults, a large-scale meta-analysis showed that APOE 4 carriers performed worse on measures of episodic memory, and global cognitive ability, with effect sizes increasing as age increases (Wisdom et al., 2011). There are also conflicting reports on the effect of APOE on hippocampal volume in young people (Khan et al., 2014; O’Dwyer et al., 2012). In agreement with our result, a large study (N = 1400) of 14 year olds also found no association (Khan et al., 2014). Recent evidence of neurodevelopmental effects of APOE have been identified as affecting gray matter volumes in infants and neonates, suggesting that associations may be transient and could be clearer at a very young age (Dean et al., 2014; Knickmeyer et al., 2014).
In summary, in addition to the APOE4 genotype, a PRS comprised common AD risk variants of small effect, and TREM2 associate with hippocampal volume independently of clinical status in the elderly. TREM2 is associated in healthy older and MCI individuals, with the AD PRS showing a trend nearing significance. This correlation with early MRI markers of AD shows evidence for a genetic modulation of neurodegeneration, and the potential for a combination of PRS and brain biomarkers to aid in the prediction of future cognitive decline and the development of AD.
A limitation of this study is in the combining of results across several cohorts with participant ascertainment and diagnostic adjudication as well as the use of multiple scanner platforms adding variability in the volumetric measures used. This is reflected as significant between study heterogeneity when the analysis is performed as a meta-analysis (Supplementary Table 5). Replication in independent samples is required to confirm these findings, and testing the effect of these variants in middle age is the important next step to ascertain how early the prodromal AD risk variant effects on degeneration can be identified.
The authors would like to acknowledge and thank the Sydney MAS and OATS participants, their supporters, and their research teams. Sydney MAS is supported by the Program Grants 350833 and 568969 from the Australian National Health and Medical Research Council (NHMRC). The NHMRC/Australian Research Council Strategic Award 401162 and NHMRC Project Grant 1045325 support OATS, which was facilitated through access to the Australian Twin Registry (ATR). The ATR is supported by the NHMRC Enabling Grant 310667, administered by the University of Melbourne. DNA was extracted by Genetic Repositories Australia, which was supported by the NHMRC Enabling Grant 401184. OATS genotyping was partly funded by a Commonwealth Scientific and Industrial Research Organisation Flagship Collaboration Fund Grant.
For QTIM, they thank the twins for their participation, Kori Johnson and the radiographers for MRI scanning and preprocessing the images, Marlene Grace and Ann Eldridge for twin recruitment, Daniel Park for database support, Anjali Henders for DNA processing and preparation, Scott Gordon for quality control and management of the genotypes. The QTIM study was supported by the National Institute of Child Health and Human Development (R01 HD050735), and the National Health and Medical Research Council (NHMRC 486682, 1009064), Australia. Genotyping was supported by NHMRC (389875).
AddNeuroMed is part of InnoMed (Innovative Medicines in Europe), an integrated project funded by the European Union of the Sixth Framework program priority ((FP6-2004-LIFESCIHEALTH-5); the Alzheimer’s Research Trust UK; the John and Lucille van Geest Foundation; and the NIHR Biomedical Research Centre for Mental Health and Biomedical Research Unit for Dementia at the South London, Maudsley NHS Foundation Trust and Kings College London, and a joint infrastructure grant from Guy’s and St Thomas’ Charity and the Maudsley Charity; Academy of Finland, Kuopio University Hospital (HS) and funding from UEF-BRAIN (HS).
Data collection and sharing for ADNI was funded by the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2–0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; Araclon Biotech; BioClinica, Inc; Biogen Idec; Bristol-Myers Squibb Company; Eisai; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research & Development, LLC; Johnson & Johnson Pharmaceutical Research & Development LLC; Medpace; Merck & Co., Inc; Meso Scale Diagnostics, LLC; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc; Piramal Imaging; Servier; Synarc; and Takeda Pharmaceutical Company. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (www.fnih.org). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer’s Disease Cooperative Study at the University of California, San Diego. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California.
The authors thank the International Genomics of Alzheimer’s Project (IGAP) for providing summary results data for these analyses. The investigators within IGAP contributed to the design and implementation of IGAP and/or provided data but did not participate in analysis or writing of this report. IGAP was made possible by the generous participation of the control subjects, the patients, and their families. The i–Select chips were funded by the French National Foundation on Alzheimer’s disease and related disorders. EADI was supported by the LABEX (laboratory of excellence program investment for the future) DISTALZ grant, Inserm, Institut Pasteur de Lille, Université de Lille 2 and the Lille University Hospital. GERAD was supported by the Medical Research Council (Grant no 503480), Alzheimer’s Research UK (Grant no 503176), the Wellcome Trust (Grant no 082604/2/07/Z), and German Federal Ministry of Education and Research (BMBF): Competence Network Dementia (CND) grant no 01GI0102, 01GI0711, 01GI0420. CHARGE was partly supported by the NIH/NIA grant R01 AG033193 and the NIA AG081220 and AGES contract N01eAGe12100, the NHLBI grant R01 HL105756, the Icelandic Heart Association, and the Erasmus Medical Centre and Erasmus University. ADGC was supported by the NIH/NIA grants: U01 AG032984, U24 AG021886, U01 AG016976, and the Alzheimer’s Association grant ADGC–10–196728.
Michelle Lupton is supported by a Perpetual Foundation Wilson Fellowship for early career researchers. Hilkka Soininen has received funding from 7FP grant agreement no 601055 VPH-DARE@IT. Karen Mather is supported by an Alzheimer’s Australia Dementia Research Foundation Postdoctoral Fellowship and the NHMRC Capacity Building Grant 568940. Paul Thompson and Derrek Hibar are supported, in part, by an NIH Big Data Center of Excellence Grant (U54 EB020403), funded by multiple NIH institutes, including NIBIB and NCI.
The authors have no conflicts of interest to disclose.