Mapping Genetic Influences on Brain Activation during the N-Back Working Memory Task: An fMRI Study of 315 Twins

Methods: BOLD fMRI data were acquired during the 0-back and 2-back conditions of a spatial N-back working memory task (see Fig. 1 in [1]) in 366 twins on a 4T Bruker Medspec MRI scanner. Echo Planar Images of all twins were realigned, coregistered with the twin’s 3D T1-weighted scan, spatially normalised in MNI atlas space, smoothed with an 8 mm FWHM Gaussian kernel, and detrended, using SPM5. We excluded 41 twins with <30% accuracy on either N-back condition; and ten twins due to insufficient scan quality. The final sample consisted of 315 twins (aged 23.5±1.8 years); 74 monozygotic (MZ) pairs (29M/45F), 63 dizygotic (DZ) pairs (11M/27F/25MF), and 41 unpaired subjects (contributing to means model). Z-scores were extracted on a voxel-by-voxel-basis from 2>0-back contrast images generated at the single-subject level in each of 40,136 voxels comprising a brain mask created from a group random effects analysis irrespective of zygosity. We calculated voxel-by-voxel intra-class correlations (ICCs) and corresponding significance values for MZ and DZ groups. In addition, we computed Falconer’s heritability estimate (h2 = 2(rMZ – rDZ); [2]). Finally, we assessed fMRI reproducibility, using a paired-samples t-test, in a sub-sample of 20 twin pairs (5 MZF, 5 MZM, 5 DZF, 5 DZM) rescanned 120±55 (M±SD) days after their initial scan.