Individual differences in amygdala-medial prefrontal anatomy link negative affect, impaired social functioning, and polygenic depression risk

Abstract

Individual differences in affective and social processes may arise from variability in amygdala-medial prefrontal (mPFC) circuitry and related genetic heterogeneity. To explore this possibility in humans, we examined the structural correlates of trait negative affect in a sample of 1050 healthy young adults with no history of psychiatric illness. Analyses revealed that heightened negative affect was associated with increased amygdala volume and reduced thickness in a left mPFC region encompassing the subgenual and rostral anterior cingulate cortex. The most extreme individuals displayed an inverse correlation between amygdala volume and mPFC thickness, suggesting that imbalance between these structures is linked to negative affect in the general population. Subgroups of participants were further evaluated on social (n = 206) and emotional (n = 533) functions. Individuals with decreased mPFC thickness exhibited the poorest social cognition and were least able to correctly identify facial emotion. Given prior links between disrupted amygdala-mPFC circuitry and the presence of major depressive disorder (MDD), we explored whether the individual differences in anatomy observed here in healthy young adults were associated with polygenic risk for MDD (n = 438) using risk scores derived from a large genome-wide association analysis (n = 18,759). Analyses revealed associations between increasing polygenic burden for MDD and reduced cortical thickness in the left mPFC. These collective findings suggest that, within the healthy population, there is significant variability in amygdala-mPFC circuitry that is associated with poor functioning across affective and social domains. Individual differences in this circuitry may arise, in part, from common genetic variability that contributes to risk for MDD.

Figure 1.

Anatomical measures are reliable. Scatter plots represent values across two separate scanning sessions for the critical volumetric and cortical thickness estimates. A–D, Estimated intracranial volume (A), left amygdala volume (B), left hippocampal volume (C), left medial prefrontal cortex thickness (D). Individual circles represent estimated volume and thickness measurements for each participant (n = 91). Reported r values reflect Pearson correlations of values from Visit 1 and Visit 2.

Figure 2.

Amygdala and medial prefrontal cortex are associated with negative affect. A, Representative segmentations of the left amygdala and right amygdala. The scatter plot displays the correlation between trait negative affect and left amygdala volume. B, The targeted medial prefrontal subregion defined using FreeSurfer is displayed. The region includes portions of the left rostral anterior cingulate cortex and the subgenual anterior cingulate cortex (see Materials and Methods, Statistical analysis). Map effects do not limit exploration to the medial prefrontal cortex but also converged on the same general region, reinforcing that it was appropriate to target this subregion for hypothesis-driven analyses. The scatter plot is displayed as in A.

Figure 3.

Trait negative affect shows opposing effects on the amygdala and medial prefrontal cortex. A–C, Coronal slices display the correlation strengths between the volumes of each subcortical structure (based on automated segmentations; Fischl et al., 2002) and trait negative affect across the entire sample (A), as well as for the female (B) and male (C) participants. Correlations partial out variance associated with collection site, scanner software, estimated IQ, age, sex, and estimated intracranial volume (Buckner et al., 2004). D–F, Surface-based renderings of the left midline reflect the strength of the correlation between each vertex and negative affect across the entire sample (D), as well as for the female (E) and male (F) participants. Reported correlations are after partialing out the variance associated with collection site, scanner software, estimated IQ, age, and sex. Cortical thickness was estimated using the procedures of Fischl and Dale (2000) and then displayed on the inflated surface (Van Essen, 2005). Display threshold is set at p < 0.005 to allow complete visualization of the effect pattern. Color bars reflect Pearson correlations.

Figure 4.

Rank ordering of individuals by negative affect reveals an inverse correlation between amygdala volume and medial prefrontal thickness. Data rank ordered by negative affect, averaged (n − n + 249), and smoothed with a lowess filter (span = 250). Reported values are after partialing out the variance associated intracranial volume from the left amygdala as well as collection site, scanner software, estimated IQ, age, and sex from both structures. A, Rolling average of left amygdala volume (solid line) and left medial prefrontal thickness (dotted line). B, Rolling Pearson correlation between amygdala volume and medial prefrontal thickness. Dotted lines reflect 95% confidence interval. C, Resulting p values (−log10). Individuals reporting the most severe symptoms show the strongest negative correlation between the amygdala and the medial prefrontal thickness.

Figure 5.

Opposing structural differences in the amygdala and medial prefrontal cortex are present in the young adults with the most extreme negative affect. A–C, Scatter plots represent the distribution of values for the left amygdala volumes and left medial prefrontal thickness estimates for the low (A), medium (B), and high (C) negative affect groups. Reported r values reflect Pearson correlations after partialing out variance associated with collection site, scanner software, estimated IQ, age, and sex. Estimated intracranial volume was additionally partialed from the amygdala volume estimate. D–F, Surface-based analyses reflect the correlation strength between left amygdala volume and thickness estimates at each vertex for the low (D), medium (E), and high (F) negative affect groups. Color bar reflects Pearson correlations.

Figure 6.

Poor social functioning is associated with reduced medial prefrontal thickness. A, The correlation strength between thickness estimates at each vertex on the left medial surface and a composite score of social functioning. B, Correlation strengths for each independent measure within the composite score. These include the aloof-introverted subscale from the Interpersonal Adjective Scales (Wiggins et al., 1988), the aloof personality component of the Broad Autism Phenotype Questionnaire (Hurley et al., 2007), and the social/school factor from the Retrospective Self-Report of Inhibition (p < 0.01) (Reznick et al., 1992). Reported r values reflect Pearson correlations after partialing out variance associated with collection site, scanner software, estimated IQ, age, and sex. Color bar reflects Pearson correlations.

Figure 7.

Poor emotion perception is associated with reduced medial prefrontal thickness. A, Correlation strengths for each emotional condition. Reported r values reflect Pearson correlations after partialing out variance associated with collection site, scanner software, estimated IQ, age, sex, and the median response time for each condition of interest. Error bars reflect 95% confidence interval. B, The correlation strength between thickness estimates at each vertex and error rates in the emotional faces task. Color bar reflects Pearson correlations.

Figure 8.

Heightened polygenic depression risk is associated with reduced mPFC thickness in clinically healthy individuals. A, Variance in cortical thickness explained on the basis of scores derived from four psychiatric and six nonpsychiatric illnesses across varying significance thresholds (P_T < 0.1, 0.2, 0.3, 0.4 and 0.5, plotted left to right). BP, Bipolar Disorder; SCZ, Schizophrenia; CAD, coronary artery disease; CD, Crohn's disease; HT, hypertension; RA, rheumatoid arthritis; T1D, type I diabetes; T2D, type II diabetes. The selected mPFC region is displayed in the figure legend. Reported r values are after partialing out variance associated with collection site, scanner software, estimated IQ, multidimensional scaling components of genetic ancestry, number of nonmissing SNPs, age, and sex. Equivalent variance is explained across the MDD polygenic scores (Z < 0.12, p > 0.91). No other psychiatric or nonpsychiatric polygenic scores approached significance (p > 0.59). B, Surface-based rendering reflects the strength of the correlation between each vertex and polygenic risk for MDD (P_T < 0.5) after partialing out the variance associated with the collection site, scanner software, estimated IQ, potential for population substratification, number of nonmissing SNPs, age, and sex. Display threshold is set at p < 0.01 uncorrected for multiple comparisons. Color bar reflects Pearson correlations.