The common APOE2 gene variant is neuroprotective against Alzheimer’s disease (AD) and reduces risk by nearly 50%. However, the mechanisms by which APOE2 confers neuroprotection are largely unknown. Here we showed that ApoE protein abundance in human postmortem cortex follows an isoform-dependent pattern (E2>E3>E4). We also identified a unique downstream transcriptional profile determined by microarray and characterized by downregulation of long-term potentiation (LTP) related transcripts and upregulation of extracellular matrix (ECM)/integrin-related transcripts in E2 cases and corroborated this finding at the protein level by demonstrating increases in ECM collagens and laminins. In vivo studies of healthy older individuals demonstrated a unique and advantageous biomarker signature in E2 carriers. APOE2 also reduced the risk of mild cognitive impairment to AD conversion by half. Our findings suggest that ApoE2 protein abundance, coupled with its inability to bind to LDLRs, may act to increase amyloid-beta (Ab) clearance. In addition, increased ECM and reduced LTP-related expression results in diminished activity-dependent Ab secretion and/or excitotoxicity, and thus also promotes neuroprotection.
The APOE gene is triallelic at two SNPs in its exon 4 (rs429358, rs7412), resulting in amino acid substitutions of arginine or cysteine at positions 112 and 158. E2 corresponds to Cys/Cys, E3 to Cys112/Arg158, and E4 to Arg/Arg.1,2 Although E4 is the major genetic risk variant for late onset AD (OR=3.81), the E2 variant is neuroprotective and reduces risk for AD by nearly 50% (OR=0.54) when contrasted to the referent variant E3.3,4 The allele frequency of APOE2 in the general Caucasian population is about 8%. In contrast, APOE2 may be modestly associated with cerebral amyloid angiopathy and has been inconsistently associated with risk for several other neurologic diseases.5, 6, 7, 8, 9 APOE2 homozygosity is associated with type III hyperlipoproteinemia.5,10 Nevertheless, APOE2 has been associated with reduced cardiovascular disease and greater longevity.11, 12, 13, 14
ApoE is expressed primarily in astrocytes and delivers cholesterol and other lipids to neurons through the low-density lipoprotein receptor family (including LDLR, LRP1, VLDLR and ApoER2).15,16 Several features of the E2 isoform distinguish it from the E3 and E4 isoforms. These include dramatically reduced binding to LDLRs in comparison to E3 and E4 and conformational stability in comparison to E4.2,17 APOE isoforms differentially regulate products of amyloid precursor protein processing.18,19 Some amyloid-beta (Ab) peptides, for example, Ab1–42, cleaved from amyloid precursor protein , are thought to be toxic and may have key roles in AD neurodegeneration. E2 and E3 bind to Ab with greater affinity than does E4, and so isoforms may differentially regulate Ab clearance through neurons, microglia or delivery to the blood-brain barrier (BBB).20, 21, 22
Despite the potential importance of APOE2, there have not been comprehensive studies of its molecular properties, downstream impact on biological pathways, modulation of brain function parameters and neurodiagnostic impact. Here we will examine these questions using molecular techniques to assess mRNA and protein levels of E2 in comparison to E3 and E4 in human postmortem neocortex, examine the transcriptional profile of E2 carrier cases, determine if an E2 signature is present in combined biomarker and cognitive analyses in healthy older individuals, and last, examine the effects of E2 on neurodiagnosis in the context of MCI to AD conversion.
Materials and methods
Human Brain Samples
Fresh-frozen human tissue was obtained from the National Institute of Mental Health Clinical Brain Disorders Branch, GCAPP, Neuropathology Section. For all molecular studies, we studied younger cases (who were 60 years or less at death) and included 5 APOE2, 13 APOE3/E3, and 13 APOE4 carriers (two of whom were E4 homozygotes). Demographic characteristics of the postmortem cases are in Supplementary Table 1. All specimens underwent neuropathological examination, including Bielschowsky silver staining, for plaque and tangle density, vascular infarcts and other neurodegenerative conditions; they were free of diagnosable AD and CVD pathology. For each case, we collapsed assay results from tissue obtained from two cortical regions (BA 1/2/3 and BA 21). As noted, no case had diagnosable AD neuropathology. Further details about the samples can be found in Conejero-Goldberg et al.23 DNA from brain tissue of the above cases was used for genotyping.16 The APOE genotype groups did not differ significantly in demographic or tissue quality variables as shown in Supplementary Table 1.
The RT-qPCR reactions were carried out in an ABI Prism 7900HT thermal cycler to determine the ΔΔCt as described previously.23
More detailed information about methods of supernatant collection and western blotting is in Supplementary Methods. Western Blotting. Blots were probed with rabbit anti-ApoE (H-223) polyclonal antibody (1:400 v/v dilution) and mouse anti- β-actin monoclonal antibody (1:2000 v/v dilution) (Santa Cruz Biotechnology, CA, USA). Detection was done using enhanced chemiluminiscence (ECL, GE Healthcare Life Sciences, Pittsburgh, PA, USA). Signal intensity was quantified using NIH ImageJ software. We assessed protein levels of ApoE isoforms (at 34–39 kDa) and assessed proteolytic fragments (at 17–22 kDa), insofar as they were present/absent. ELISA. Levels of human APOE in brain were measured by human-specific APOE ELISA (MBL International, Woburn, MA, USA) according to the manufacturer’s instructions. Diluted homogenates were run in triplicate and detected using Pan-APOE antibody in a TECAN Genios Pro Microplate Reader. All APOE measurements were normalized to total protein in the brain homogenates; ELISA values were expressed as ng ml−1.
We utilized an analytic strategy that examined APOE genotype differences in a GLM framework (SAS Proc GLM) with reference molecules/gene products serving as covariates as necessary, followed by planned contrasts in which E2 samples were compared to E3 samples and to E4 samples. Using western blot data, we also analyzed the number of cases that demonstrated proteolytic fragments and compared the distribution by Fisher’s test across genotype groups. Here and elsewhere in the report, we studied E2/E3 cases (hereafter E2), E3/E3 (hereafter E3) and E4 carriers (hereafter E4) in human samples or humans. E2/E4 cases were excluded. E3/E3 cases were considered the referent variant (as by convention) for the purposes of this study.
Human Brain Samples
As detailed above, with the exception that we included 11 additional E3 cases. These cases did not show substantively different tissue quality or demographics from those cases listed in Supplementary Table 1. We utilized only tissue from BA 21 in this experiment.
The platform used was the WG6 BeadChip array (Illumina, San Diego, CA, USA), generating whole-genome expression profiles for six samples in parallel on a single array.
Data pre-processing and analysis
The data were pre-processed using BeadStudio software (Illumina). Technical replicates of the same sample were averaged after it was determined that correlations between the replicates were consistently very high (>0.98). We identified genes that were differentially expressed among the two classes (APOE3-carriers/APOE2-carriers) computing an ANOVA blocked by age and sex. A stringent significance threshold was used to limit the number of false-positive findings. In comparing the APOE3/E2 types, we also controlled for the age and sex of the patient.
For selected genes showing differential expression between APOE2 and APOE3 cases, cDNA synthesis was generated for each sample, and we determined the ΔΔCt and significance of the reported cases-control differences.
Immunoassay (Western Blotting)
For laminin and collagen VI, blots were probed with rabbit anti-laminin polyclonal antibody (1:500 v/v dilution; Sigma-Aldrich, St Louis, MO, USA) and rabbit anti-collagen VI polyclonal antibody (1:1000 v/v dilution; Abcam, Cambridge, MA, USA). More information is in Supplementary Methods.
Methods used in ADNI, including subject selection, biomarker assays, imaging parameters and cognitive testing, can be found on the ADNI website and in our previous work.24 Subjects. We utilized 197 subjects who met ADNI inclusion criteria for HCs.17 APOE genotype was available in the ADNI database. Twenty-four subjects were E2s (including 1 homozygote), 107 were E3/E3 carriers, and 47 were E4 carriers (including 3 homozygotes). For the cluster analysis, in which only subjects whose cognitive and biomarker data sets were complete were included, seventeen subjects were E2s, 61 were E3/E3s and 21 were E4 carriers.
We analyzed data from the following tests: WMS-R Logical Memory delay; Auditory Verbal Learning Test Trial 5 score, Delay Score (free recall), trails b.
Structural MRI. We used the Dale semi-automated method (FreeSurfer) for determining cortical thickness (middle temporal lobe), hippocampal volume, entorhinal thickness and total brain volume.24 CSF Ab1–42 and Tau. We examined CSF Abeta and total tau and phospho tau levels in ADNI healthy control samples that were collected at baseline (‘UPENNBIOMK’). Concentrations were log transformed for analysis.
We first conducted GLM-based procedures to determine if differences among the APOE genotype groups were present at baseline. We then conducted an unsupervised cluster analysis (PROC CLUSTER Method=AVE in SAS) which we considered central to identifying an E2 ‘signature,’ that is, the presence of clusters of cases (subjects) on the basis of cognitive and biomarker level and profile. APOE genotype was excluded from these analyses, so that clusters could be aggregated independent of genetic information. After identification of clusters, we then examined whether any one cluster contained a disproportionate number of E2 carriers by X2.
The ADNI aMCI sample of 310 patients was followed for more than 4 years. Criteria for MCI were the same as defined by Petersen and outlined in Leduc et al.17 Fifteen patients were carriers of the APOE2 allele, 128 patients were E3 homozygotes and 167 patients were APOE4 carriers (including 41 E4 homozygotes). Demographic and brain volumetric variables are reported in Supplementary Table 7.
We utilized Cox Proportional Hazard Regression (HR) models in order to determine the unique multiplicative effect of genotype differences (that is, a ‘unit increase’) on hazard rate (conversion from MCI to AD).
Study 1. Molecular properties of ApoE isoforms
mRNA and Protein Levels
Protein levels may offer a window to partially understand protective effects as ApoE protein level may be an important determinant of efficient and robust lipid transport, thought to be necessary for synaptic growth, remodeling and repair, and for Ab clearance.17,25 It has also been suggested that protein fragments are cleaved differentially among isoforms in neurons, with toxic downstream consequences including mitochondrial dysfunction most frequent in APOE4.26,27 E2 protein levels and fragmentation have not been examined in human postmortem cortical samples. In contrast, in human plasma, there are well-replicated differences in protein level, such that E2>E3>E4 (for example, Soares et al.28).
We first sought to determine if APOE genotype differences in mRNA expression were present in human postmortem cortex. We then examined protein level and tendency toward fragmentation as a function of APOE genotype in these samples (see Supplementary Table 1 for demographics and tissue quality of cases). In transgenic mice with human APOE gene variants knocked-in, differences between E2, E3 and E4 mRNA isoform levels were not found.29,30
Using RT-qPCR, we compared ApoE expression in the three genotype groups (E2/E3, E3/E3, E4 carriers), and we did not find significant isoform-related differences.
Using a polyclonal antibody for ApoE, we compared protein levels of E2, E3 and E4 cases, and identified three bands by western blot: band 1 at 39-kDa, band 2 at 37-kDa and band 3 at 34-kDa (Figure 1a). The band with the highest molecular weight (band 1) is thought to reflect heavily sialylated ApoE; 34 kDa ApoE is thought to reflect the native species.31 We also identified 17–22 kDa ApoE that reflects protein fragments (Figure 1b). Total protein differed significantly among the genotype groups (E2>E3>E4) as described in Figure 1c. Differences were largest in the native ApoE and lightly sialylated band (data not shown). By ELISA, the overall difference in protein level among the APOE genotype groups approached significance (Figure 1d). Western blot-based fragmentation rates did not differ among genotypes by Fisher’s exact test. Numerically, E2 had the lowest proportion of fragments (see Supplementary Table 2).
For mRNA, effect sizes for E2-E3 and E2-E4 contrasts were small as shown in Figure 1e. For polyclonal-based measures of protein abundance, effect sizes shown in Figure 1e were large in the E2-E3 and E2-E4 contrasts. For ELISA-based contrasts, effect sizes were medium for E2-E3 and E2-E4 contrasts as in Figure 1e.
We found a mismatch between mRNA expression levels and protein levels in APOE genotype contrasts. While mRNA expression levels did not differ, protein levels were higher in E2 cases, suggesting that the isoform was relatively impervious to post-translational modifications and in keeping with several in vivo and in vitro studies that have implicated differences in proteolytic degradation among the isoforms.2,17 In particular, in several studies of APOE targeted replacement (TR) transgenic mice, mRNA expression levels did not differ among APOE variants, but protein levels were lowest in E4 and highest in E2 mice.29,30 Pulse chase experiments in astrocyte-derived cell cultures have indicated enhanced degradation of newly synthesized E4 protein.29 Associated with these differences in the AD transgenic mice were differences in brain amyloid burden, such that E2 mice had the lowest level of amyloid and E4 mice had the highest level. Thus, this pattern of findings in human brain recapitulates with great fidelity several transgenic animal studies using human APOE TR.
Study 2. Transcriptional profiling of APOE2 in human cortex
In order to examine downstream consequences of APOE2 on the transcriptome, we conducted a microarray study that contrasted the transcriptional profiles of E2 and E3 carrier cases in human postmortem cortex. To the best of our knowledge there have been no prior studies in this area. We sought to identify key biologic pathways that might plausibly confer neuroprotection or enhanced neuronal function.
Group Comparison of Transcription Profiles
We identified 131 transcripts that differed significantly (at P<0.001) between APOE2 and APOE3-carriers using ANOVA for fixed effects in the BRB analysis suite. A large proportion of these individual transcripts (73%) were upregulated in E2 carriers with respect to E3 carriers. Results are listed in Supplementary Table 3.
Hierarchical Clustering of Expression Changes Identified by ANOVA
For the hierarchical clustering only, we excluded the 16 LOC genes, 43 ESTs and 6 open reading frames that have been identified informatically from the human transcriptome, leaving a total of 66 transcripts for further analysis, of which 41 were upregulated.
As shown in the heat map image (Figure 2), the hierarchical clustering of expression changes derived from these transcripts demonstrated a cluster of four cases in the APOE2 group. The APOE3 group demonstrated a less distinct pattern of transcriptional upregulation and downregulation.
In addition, based on recent work32 identifying Cyclophilin A (CypA) as an important mediator of BBB integrity and differentially modulated by ApoE isoforms, we examined its expression post hoc. We found that CypA mRNA was reduced in E2 carrier group (P=0.01), in keeping with the view that E2 is associated with more robust clearance of Ab at the perivasculature.33
Biological and Signaling Pathways Analysis
Two Kyoto pathways were identified with LS/KS and Efron-Tibshirani’s GSA test using P-values of P<0.005: the long-term potentiation (LTP) and the extracellular matrix (ECM)-receptor interaction.34,35 Genes within these pathways are listed in Supplementary Tables 4(a-b).
We examined transcripts related to LTP (HOMER2 and DLG2) and the ECM (COL6A1, LAMC1, ITGB1) by RT-qPCR. We validated the microarray results by RT-qPCR in four genes (HOMER2, DLG2, COL6A1, LAMC1, all Ps<0.05 by t-test). See Supplementary Table 5 for fold changes and comparison of microarray probe to RT-qPCR data and Supplementary Results for a brief overview of the molecules in question.
To further extend these results, we also conducted Western blot studies to assay levels of select ECM related proteins in the E2 group. Using a polyclonal antibody for laminin, we found a highly significant increase in protein level in the E2 group contrasted with the E3 group. Using a polyclonal for collagen VI, we also found a highly significant increase in protein level in the E2 group. Details of the analyses are in Supplementary Methods and Supplementary Figure 1(a-b).
In addition to multiple individual transcripts that differed between the E2 and E3 groups, we identified two pathways that may protect neurons in the E2 group. They involved LTP and ECM. E2 cases had lower expression levels of transcripts associated with LTP. This is in keeping with prior work in APOE TR mice that demonstrated that E2 carriers showed reduced LTP in comparison to other APOE genotypes.36 Because of findings of hypermetabolism in nodes of the so-called brain default network, activity-dependent Ab deposition, and excitotoxicity,37, 38, 39 reductions in LTP-related activity may be protective in the E2 context. Similarly, a recent study demonstrated that brain activity may be related to DNA double-strand break neurodegeneration and that such degeneration worsened in the presence of Abeta.40 Reductions of activity might also attenuate this type of neurodegenerative process. Furthermore, our findings of ECM integrin-related upregulations in E2 carriers serve to provide a mechanism for activity reduction. Increases in the ECM reduce plasticity, perhaps specifically by impacting motility of AMPA receptors, crucial for early phase LTP.41,42 As noted, insofar as Ab is secreted in an activity-dependent manner, reduced LTP might attenuate Ab levels. In addition, increases in the ECM (‘perineural nets’) at synapses reduce tangle formation in situ.43
Another neuroprotective mechanism may involve our finding of lowered CypA expression; reductions of this are thought to have positive consequences for Ab clearance across the BBB.33 ECM collagen and laminin increases (as found in the protein experiments described above) may also promote BBB integrity.44
Interestingly, the pathways did not overlap with the signaling pathways and biological processes associated with E4 in our earlier microarray study.23 We thus suggest that the pathways toward neurodegeneration and neuroprotection are not simply the inverse of one another.
Study 3. Biomarker profile of APOE2
We sought to determine if E2 carriers had better neurocognition and more advantageous biomarker profiles.45, 46, 47, 48, 49, 50 We used the ADNI public database because of the large samples of healthy controls (HCs) (mean age 75 years), APOE characterization, reasonable numbers of E2/E3 carriers in the HC group in conformity with population allele frequencies and the wide range of cognitive markers and biomarkers collected in a uniform and prospective manner.
Genotype Differences in Biomarkers
We first compared the APOE genotype groups on a panel of biomarkers. Significant differences were present for CSF Ab1–42, p-tau, and middle temporal cortical thickness (Table 1). Effect sizes were large to moderate (Figure 3a). No significant differences were present for cognitive markers (Table 1); effect sizes were generally small (Figure 3b).
We next conducted a cluster analysis. In a combined panel of cognitive and biomarkers, we identified a robust solution indicating the presence of three clusters (cubic clustering criterion=1.71, R2=0.65, estimated r2=0.61, pseudo-F=89.00). Nine cases were in cluster 1, 59 in cluster 2, and 31 in cluster 3 (Figure 3c). As noted, APOE genotype was not utilized as a clustering variable. We then interrogated each of the aggregates in order to determine their respective proportion of E2, E3 and E4 cases. Genotypes were unequally distributed by Fisher’s exact test (see Supplementary Table 6). Cluster 2 had the highest proportion of E2 genotypes (15/17) and lowest of E4 genotypes (5/21). We also examined Z-score profiles of the clusters (based on the grand means/variance of the healthy control group). Cluster 2 included nearly every E2 case, a majority of E3 cases, and relatively few E4 cases, and was associated with an advantageous biomarker profile (high Ab, low p-tau, increased middle temporal cortical thickness, modestly higher cognitive scores) compared with clusters 1 and 3.
APOE2 was associated with dramatically increased CSF Ab and lower p-tau. In models of neurodegeneration, low-CSF Ab are considered to be an early ‘preclinical’ phase indicator of increased risk for AD pathogenesis.51 In principle, high Ab associated with E2 status should be associated with reduced risk. APOE2 also reduced p-tau (but not t-tau), suggesting that it had a specific role in reducing phosphorylation at threonine 181. An E2 signature was identified and derived independent of genotype in an unsupervised cluster analysis in healthy older individuals. Cluster 2, which contained the majority of E2 carriers, reflected an in vivo version of exceptional aging and a protective profile that is quite dissimilar to the MCI at risk or AD biomarker profiles described in this sample and in other work.24,52
Study 4. APOE2 and risk of MCI to AD conversion
We next sought to determine if the E2 genotype had clinical and neurodiagnostic consequences. While it is known that E2 reduces the probability of incident AD, it is unknown if E2 can delay or reduce the risk of dementia, even in those individuals with a diagnosable amnestic memory impairment (that is, amnestic MCI). We examined the proportion of E2 cases in MCI to AD converters and nonconverters (in contrast to E3 and E4 genotypes) in an ADNI sample in Cox regression hazard analyses over a 4-year period.
Cox Regression Analyses
During follow-up, 3 (20%) of MCI E2 carriers, 46 (35.9%) of MCI E3, and 96 (57.5%) of MCI E4 carriers were uncensored (developed AD). The mean follow-up time in months for MCI APOE E2 carriers who developed AD was 34.18 (range 24.31–41.36; s.d.=8.84), for MCI APOE E3 homozygotes was 18.59 (range 5.75–42.25; s.d.=9.38), and for MCI APOE E4 carriers was 20.29 (range 5.98–52.63; s.d.=10.84).
Cox HR models to estimate risk of conversion from MCI to AD according to APOE genotype status were fitted. The survival curves of the probability for conversion to AD in each of the three APOE groups were plotted in Figure 4. In the model, APOE3 had more than 2 × the hazard of conversion per unit time as compared with APOE2, while APOE4 had nearly 4 × the hazard.
We conducted multiple experiments to demonstrate that molecular properties of ApoE2 have protective effects in biomarkers and neurodiagnostics relevant to AD. We discuss our main findings in turn.
First, mRNA expression levels of ApoE did not differ, but protein levels differed among ApoE isoforms, with E2 being most abundant. This in turn suggests the presence of differential post-translational modifications. There is ample evidence that while E4 is especially subject to proteolytic cleavage and degradation,10,53 E2 may be relatively resistant to degradation as found in a conformational computational study, TR mouse studies, and now, our findings in human brain.
We found differences in the transcriptional profile in postmortem human cortex between E2 and E3 carriers in multiple individual transcripts and pathways. The latter findings indicated that LTP-related transcripts were downregulated, while ECM-related transcripts, including integrins, were upregulated in E2 carriers. Insofar as Ab secretion is activity dependent, reductions in LTP may attenuate release and by inference, the possibility of aggregation.39,54 Increases in ECM-related transcripts and proteins reduce plasticity phenomena (linked to LTP) and maintain the integrity of the BBB, The ‘profile’ suggests reduction in activity-dependent Ab production/deposition or excitotoxicity and in the context of dendritic spine spacing and stability, relative synaptic efficiency.
Several biomarkers were modulated by APOE, with E2 carriers demonstrating a ‘signature’ characterized by higher levels of CSF Ab, greater middle temporal lobe cortical thickness and lower levels of p-tau in cluster analysis. Cognitive differences among genotypes were modest and nonsignificant in this cohort. This profile is consistent with a model in which E2 is specifically protective against Alzheimer’s type neurodegeneration, but is not generally pro-cognitive.
E2 also had an impact at the neurodiagnostic level. It dramatically reduced the number of individuals with MCI who converted to AD. Thus, even in a group with documented memory impairment (amnestic MCI), further functional and cognitive decline was slowed or prevented by the presence of E2.
On the basis of our studies, we suggest that more abundant ApoE2 protein, in combination with its known inability to bind to the LDLR, has implications for its ability to clear Ab in an advantageous manner. Protein level may be important in driving neuroprotection, as higher levels of the ApoE2 isoform may be associated with greater cholesterol delivery to APOE receptors and perhaps, more relevantly, increase the likelihood of interactions with and subsequent clearance of Ab. In the brain, LDLRs are found on neurons and glia and account for the greater part of APOE binding, while LRP1s are highly expressed on blood vessel outer walls.15,55 Although ApoE2’s binding to LDLRs is dramatically reduced, its binding to other low-density lipid receptors (including LRP1) is unaffected. Because ApoE2 cannot transport Ab species to cellular LDLRs, where it might be endocytosed inefficiently or proven toxic, it appears likely that it differentially transports Ab to the BBB and there it binds to LRP1 receptors, thus providing more robust Ab clearance, as shown in Deane et al.21
In summary, ApoE2 may have pleiotropic effects on multiple protective signaling pathways. Evidence suggests that several fundamental molecular properties of ApoE2 isoform, including its high protein level in comparison to other isoforms as shown here and its inability to bind to LDLRs, may allow more robust interaction with Ab and/or efficient clearance of Ab through the cerebral vasculature. APOE2 carrier’s transcriptional profile in cortex was characterized by reductions in expression of LTP-related molecules, perhaps reducing excitotoxic lesions or activity-dependent Ab production. Concomitant ECM/integrin-related upregulations, CypA downregulation, and increases in collagen VI and laminin protein also might reduce activity-dependent plasticity, reduce tau aggregation, and/or further maintain the integrity of the BBB. The molecular properties of ApoE2 have unusual but interpretable consequences at the neurobehavioral/neurodiagnostic level. APOE2 genotype was associated with an advantageous biomarker profile in healthy older controls, including high-CSF Ab and equivocally decreased p-tau, but not general increases in cognition. This signature likely reduces risk of AD; and as shown here, a reduced risk of MCI to AD conversion in E2 carriers. Furthermore, we have shown that high levels of the ApoE2 isoform are not disadvantageous in terms of brain function, and may be advantageous. We think that our set of observations has important implications for an E2-based drug discovery platform and understanding individual variations in cognitive aging.
Funding for this study was provided by the Litwin Zucker Alzheimer’s Disease Center (CCG, TEG, JJG, TBB, SC, PD) and GCAPP, IRP, NIMH (TMH, JEK, MH). Data collection and sharing for this project was funded by the Alzheimer's Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: Abbott, AstraZeneca AB, Bayer Schering Pharma AG, Bristol-Myers Squibb, Eisai Global Clinical Development, Elan Corporation, Genentech, GE Healthcare, GlaxoSmithKline, Innogenetics, Johnson and Johnson, Eli Lilly, Medpace, Merck, Novartis AG, Pfizer, F. Hoffman-La Roche, Schering-Plough, Synarc, as well as non-profit partners like the Alzheimer’s Association and Alzheimer's Drug Discovery Foundation, with participation from the US Food and Drug Administration. Private sector contributions to ADNI are facilitated by the Foundation for the National Institutes of Health (http://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, CA, USA. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of California, Los Angeles, CA, USA. This research was also supported by NIH grants P30 AG010129, K01 AG030514, and the Dana Foundation.The ADNI investigators contributed to the design of ADNI, diagnoses, and data collection of biomarkers, but not to the analyses and writing of this report.
About this article
Supplementary Information accompanies the paper on the Molecular Psychiatry website (http://www.nature.com/mp)
Cellular and Molecular Life Sciences (2018)