Donor-to-donor reproducibility of gene expression patterns
Five of the AN risk genes (FOXP1, CADM1, CDH10, NCKIPSD, MGMT) ranked above the 50th percentile of all 20,737 protein-coding genes, indicating reproducible patterning irrespective of sex and ethnicity (Fig. 1). Donor-to-donor associations were also assessed via Spearman’s rank correlation coefficient, with each individual association for FOXP1, CADM1, CDH10, NCKIPSD, MGMT demonstrating statistical significance (p < 0.001), further suggesting strong differential stability (Supplementary Fig. 1).
Brain-based Region-of-interest Gene Expression
Of the five genes showing reproducible expression patterns, three demonstrated spatially specific expression patterns which reached statistical significance for specific brain regions (Fig. 2). Compared to average expression across the whole brain, statistically significantly greater expression of CADM1 mRNA levels were noted diffusely throughout the cerebellum (Cerebellum Crus 2: p = 0.021, d = 2.181; Cerebellum 3: p = 0.001, d = 5.678; Cerebellum 4/5: p = 0.001, d = 5.899; Cerebellum 6: p = 0.038, d = 1.750; Cerebellum 7: p = 0.016, d = 2.387; Cerebellum 8: p = 0.014, d = 2.633; Cerebellum 9: p = 0.013, d = 2.817; Cerebellum 10: p = 0.017, d = 2.368), in the olfactory bulb (p = 0.035, d = 1.823, and in limbic regions including the hippocampus (p = 0.007, d = 3.400) and amygdala (p = 0.022, d = 2.088) (all p values adjusted after FDR correction for 54 tests). Statistically significantly lower expression of CDH10 mRNA levels were noted throughout the basal ganglia (caudate: p = 0.002, d = 5.302; putamen: p = 0.023, d = 2.515; pallidum: p = 0.022, d = 2.509), the thalamus (p = 0.002, d = 4.761), hippocampus (p = 0.035, d = 2.154), and amygdala (p = 0.049, d = 1.913).
Expression of FOXP1 mRNA was statistically significantly greater in basal ganglia regions (caudate: p = 0.001, d = 4.785; putamen: p = 0.001, d = 4.437), frontal lobe regions (precentral gyrus: p = 0.043, d = 1.156; superior frontal gyrus: p = 0.026, d = 1.413; superior orbitofrontal gyrus: p = 0.010, d = 1.945; middle frontal gyrus: p = 0.016, d = 1.652; middle orbitofrontal gyrus: p = 0.011, d = 1.867; inferior frontal operculum: p = 0.015, d = 1.703; frontal inferior gyrus triangular region: p = 0.001, d = 1.997; inferior orbitofrontal gyrus: p = 0.010, d = 1.933; Rolandic operculum: p = 0.002, d = 2.912; anterior medial orbitofontal gyrus: p = 0.016, d = 1.639; rectus: p = 0.026, d = 1.404), in the insula (p = 0.002, d = 3.106), a range of cingulum regions (anterior cingulate: p = 0.043, d = 1.160; medial cingulate: p = 0.030, d = 1.322; posterior cingulate; p = 0.020, d = 1.524), occipital regions (calcarine sulcus: p = 0.001, d = 3.918; cuneus: p = 0.002, d = 2.960; superior occipital gyrus: p = 0.002, d = 3.048; medial occipital gyrus: p = 0.002, d = 3.058; inferior occipital gyrus: p = 0.002, d = 3.528), a range of parietal regions (angular gyrus: p = 0.002, d = 3.386; postcentral gyrus: p = 0.020, d = 1.532; superior parietal gyrus: p = 0.011, d = 1.861; inferior parietal gyrus: p = 0.001, d = 2.019), and a range of temporal regions (Heschl’s gyrus: p = 0.002, d = 3.647; superior temporal gyrus: p = 0.002, d = 3.335; superior temporal pole: p = 0.016, d = 1.631; medial temporal gyrus: p = 0.085, d = 0.908; medial temporal pole: p = 0.005, d = 2.417; inferior temporal gyrus: p = 0.015, d = 1.682). Expression of FOXP1 mRNA was statistically significantly lower in the thalamus (p = 0.001, d = 3.345), a range of cerebellar regions (cerebellum crus 1; p = 0.020, d = 1.355; cerebellum crus 2; p = 0.000, d = 7.399; cerebellum 3: p = 0.000, d = 6.438; cerebellum 4/5: p = 0.004, d = 2.521; cerebellum 6: p = 0.011, d = 1.899; cerebellum 7b: p = 0.000, d = 7.084; cerebellum 8: p = 0.000, d = 6.955; cerebellum 9: p = 0.000, d = 5.814; cerebellum 10: p = 0.000, d = 8.776), and temporal regions including the hippocampus (p = 0.030, d = 1.314), parahippocampal gyrus (p = 0.030, d = 1.312, and amygdala (p = 0.031, d = 1.293).
The relative expression of NCKIPSD and MGMT was not spatially confined to specific brain regions, and appeared to be expressed diffusely. Voxel-by-voxel volumetric gene expression maps for the additional six genes are presented in Supplementary Fig. 3. Importantly, out of sample validation indicates similar expression patterns for all genes of interest in the Genotype-Tissue Expression (GTEx) project database (Supplementary Fig. 4).
Whole-body Gene Expression
Alongside whole-brain gene expression, we additionally assessed gene expression across 30 different body tissue types in the body, by extracting normalized gene expression values (reads per kilo base per million; RPKM) from the GTEx database, via the FUMA platform (Supplementary Fig. 5). Normalized expression [zero mean of log2(RPKM + 1)] was used to assess differentially expressed gene sets21, and Bonferroni adjusted p-values were calculated using two-sided t-tests per gene per tissue against all other tissues. These analyses indicate that the aggregated set of AN risk genes are cumulatively most expressed in the brain, relative to other body tissue types.
Cognitive State Correlates
Expression maps for two of the five differentially stable genes (CADM1, NCKIPSD) were highly correlated with functional imaging maps reflecting ‘conditioning’, ‘fear’ and ‘reward’, ranking among the top 0.5% strongest associations for each of these cognitive state activation maps, respectively (Fig. 3). Moreover, when assessing the relationship between gene expression maps and functional activation maps most associated with specific mental states, these same two genes (CADM1, NCKIPSD) were among the 0.5% strongest associations with depression, anxiety, stress and addiction (Fig. 4). In addition, two of these five genes (NCKIPSD, MGMT) were correlated with functional imaging maps reflecting visual processing.