Computational protocol: Interaction of childhood urbanicity and variation in dopamine genes alters adult prefrontal function as measured by functional magnetic resonance imaging (fMRI)

Similar protocols

Protocol publication

[…] Whole-brain BOLD fMRI data were acquired at 3Tesla (General Electric Signa Scanners, Milwaukee, WI) using gradient-echo echo planar imaging (TR = 2000 msec, TE = 30 msec, flip angle = 90°, field of view = 24 cm, matrix = 64x64, 24 interleaved slices, 128 volumes) []. The fMRI images were registered to high-resolution anatomical images and analyzed using Statistical Parametric Mapping 5 (SPM5) [] and MATLAB 7.3.0(R2006b) on a Linux operating system. For the U.S. COMT replication, the Italian COMT, and DRD1/DRD2 samples, fMRI data acquisition and analysis were perfomed in the same way as for the COMT discovery sample. The data were corrected for head motion artifacts, with six motion parameters (three translational (X, Y, Z) and three rotational (pitch, roll, yaw)) used as covariates of no interest in first level analyses. The fMRI data were then anatomically normalized into a standard space Montreal Neurological Institute (MNI) template [] using affine and nonlinear transformations, and then smoothed with an 8mm full-width at half-maximum Gaussian filter to minimize noise and to account for residual inter-subject differences. As a result, all images are reported in MNI space. Images were resampled to a 2mm isotropic voxel size and signal modelled using a boxcar convolved with the hemodynamic response function at each voxel. In addition to motion, we assessed data quality and excluded subjects based on average signal-to-noise ratio, ghosting index, signal regularity, scaled variance, slice-by-slice variance, scaled mean voxel intensity, and maximal/mean/minimal slice variance. Data with movement in any one direction (> 2mm translation or > 1.5 degrees rotation) were excluded. We did not find any significant differences in movement when comparing framewise displacement across urbanicity, genotype, or urbanicity-genotype groupings [].For first level analyses, the average BOLD signal from the 2-back task was contrasted with the signal from the 0-back task for each subject in order to identify the brain regions activated during the WM aspect of the task. The timings used for each of the blocks (2-back and 0-back) were from the task output, and the BOLD signal during these blocks of time was then contrasted between the 2-back and 0-back conditions. At the second level, multiple regression analyses were used to determine main effects of childhood urbanicity, dopamine genotype, as well as GxE interaction effects (dopamine gene-by-urbanicity) in one multiple regression using these three factors. These analyses were whole brain, but we report only DLPFC results as this was our primary hypothesized region for these tests. Post hoc SPM ANOVAs were performed in the discovery sample to confirm that each subgroup used in the multiple regression significantly differed from each other in the same direction found in the multiple regression—i.e., urban showed significantly greater fMRI activation than town, the latter significantly greater than rural. Sex and age were used as covariates of no interest in all analyses. For the US discovery, US replication, and the Italian replication samples, we performed the same fMRI data acquisition and analysis. Combining US discovery and US replication samples for DRD1 and DRD2, we also analysed these data in the same fashion. To illustrate the COMT replication results in terms of anatomical overlap with the results from the discovery COMT-by-urbanicity interaction, regions of interest were defined from the discovery results and then applied to each replication’s activation maps using MRIcron [–] at a threshold of p<0.05 uncorrected for display purposes. In addition, we created a bilateral DLPFC ROI restricted to BA 9–10, 46) using the Wake Forest University (WFU) PickAtlas for further illustration of potential overlap.For determining significant activation we used a small volume correction with family wise error correction (SVC-FWE). We created our ROIs using meta-analytically derived activation in Neurosynth []. Neurosynth automatically gathers coordinates from the literature using text-mining and and organizes pre- and post-analyses via key words common to many papers, such as for ‘working memory’(n = 901). The Neurosynth database is publically available (http://neurosynth.org/). Neurosynth meta-analyses are easily searchable for the papers contributing to any given meta-analysis and surfable using a 3-D graphical depiction of significant results within brain (Panel H in ). Neurosynth can determine the strength of association between the search term’s coordinate measures across these studies, for example comparing studies with and without a given search term. Resultant statistics are performed on a whole brain basis. We used the reverse inference as this is recommended. Reverse inference measures can be evaluated by z-scores based on the selectivity of regional activation to working memory. Said another way, meta-analytic results link brain activity patterns with measurable cognitive states. All Neurosynth meta-anaylses are adjusted for multiple corrections at a false discovery rate of P FDR corrected <0.01.Using the key words ‘working memory,’ we identified regional peak activations using the 3-D brain representation of the meta-analysis in Neurosynth. These Neurosynth peaks were used to generate spherical ROIs for small volume correction within SPM (Panels F and H in ). Neurosynth results were similar to other meta-analyses of working memory [–]. We analyzed the significance of activation in our study against small volume corrections using Neurosynth-based spherical ROIs in SPM8 (Panels F and H in ). In summary, we analyzed significant foci of activation from our study for small volume correction using spherical ROIs based on Neurosynth (Tables –, Panels F and H in ). Cohen’s d were calculated in SPM using the VBM5 toolbox []. The VBM5 toolbox transforms SPM T maps into effect sizes, correlation coefficients, p-values, or log p-values. These effect sizes that are not explicitly defined by a primary peak within the SPM results. This statistical transformation produced images composed of clusters thst were were now surfable where voxel intensity across significant clusters was replaced by Cohen’s d. We located the maximum and range of d by surveying within each cluster. Finally, all fMRI coordinates are reported in standard MNI space.We also tested for differences between urbanicity groups in age, education, socioeconomic status (SES) [], IQ, handedness (Edinburgh Handedness Inventory) [], and N-back performance. For any demographic factor that differed between groups, analyses were repeated using it as a covariate of no interest and compared to the initial results. To further confirm that the results related to urbanicity were not confounded by socioeconomic status (SES), analyses were repeated using current and childhood SES as covariates. Additional analyses in the COMT discovery sample were performed to address confounding effects of current urbanicity. If individuals had moved to another urbanicity category after childhood, then the childhood measures could have represented current urbanicity (urban environment at the time of study). We investigated this possibility in two ways. First, we repeated the analyses substituting current urbanicity for childhood urbanicity and using only subjects who had changed urbanicity categories. Second, we calculated the difference between childhood and current urbanicity and added this change in urbanicity into the multiple regression design. Next, given evidence elsewhere of a potential interaction between COMT and sex [–], all fMRI analyses used sex as a covariate of no interest. In addition, the discovery data were re-analyzed in male and female subjects separately. As some cell sizes were small (although never the same cells across all three samples), we estimated effects ‘as if’ the COMT discovery results depended upon smaller subsamples subsumed in the multiple regression analyses at discovery and collapsed to a two-sample t test. […]

Pipeline specifications

Software tools SPM, MRIcron, Neurosynth
Applications Magnetic resonance imaging, Functional magnetic resonance imaging
Organisms Homo sapiens
Chemicals Dopamine