Genetic predictors of 12-month change in MRI hippocampal volume in the Alzheimer's Disease Neuroimaging Initiative cohort: Analysis of leading candidates from the AlzGene database

Background: High density genome-wide microarray data and one year longitudinal MRI data have become available for most participants of the Alzheimer's Disease Neuroimaging Initiative (ADNI), a multi-center longitudinal study assessing imaging in the diagnosis and longitudinal monitoring of Alzheimer's disease (AD) and amnestic mild cognitive impairment (MCI). The AlzGene database provides meta-analytic data on leading candidate genes for AD. Here, we analyzed whether common variants in current major candidate genes could predict longitudinal changes on MRI. Methods: Baseline and 12 month T1-weighted MP-RAGE scans acquired on 1.5 T magnets from 627 ADNI participants (141 AD, 60 MCI-Converters, 241 MCI-Stable, 185 healthy controls, HC) were analyzed using Freesurfer software for automated parcellation and SPM5 for voxel-based morphometry (VBM). Genetic data consisted of APOE alleles and the Illumina 610 Quad array that includes over 620,000 features. For this analysis, common (minor allele frequency, MAF > 0.2) single nucleotide polymorphisms (SNPs) from the top 30 candidate genes from the AlzGene database were examined. Regression models were performed using SAS/Stat 9.3 to test the ability of SNPs to predict hippocampal volume and gray matter density changes. Results: Models including diagnosis group (AD, MCI-C, MCI-S, HC), SNP (0,1,2), parental history of dementia (0,1,2), APOE epsilon 4 status (0,1) and all interactions were computed. 732 SNPs were available for analysis with MAF > 0.2. A threshold of p < 0.00001 was employed to reduce the likelihood of false discoveries. Using this model and criterion, 5 genes showed significant SNPs associated with hippocampal volume changes (NEDD9, SORL1, DAPK1, IL1B, SORCS1). In addition, SNPs from several other candidates genes showed less robust indications of possible association (at p<.0001: MYH13, TNK1; at p<.001: ACE, PRNP, MAPT, PCK1, GAPDHS and APP). Conclusions: Variation in major candidate genes was significantly related to 12 month longitudinal change in hippocampal volume in the ADNI cohort. These findings suggest that combining genetics and imaging with other clinical data may yield refined prediction models for those at highest risk of progression. Replication and further investigation are warranted. As putatively disease modifying agents are developed, the ability to determine those at highest risk for progression will be critical.