Maternal depressive symptoms, neonatal white matter, and toddler social-emotional development

Abstract

Maternal prenatal depression is associated with increased likelihood of neurodevelopmental and psychiatric conditions in offspring. The relationship between maternal depression and offspring outcome may be mediated by in-utero changes in brain development. Recent advances in magnetic resonance imaging (MRI) have enabled in vivo investigations of neonatal brains, minimising the effect of postnatal influences. The aim of this study was to examine associations between maternal prenatal depressive symptoms, infant white matter, and toddler behaviour. 413 mother-infant dyads enrolled in the developing Human Connectome Project. Mothers completed the Edinburgh Postnatal Depression Scale (median = 5, range = 0-28, n = 52 scores ≥ 11). Infants (n = 223 male) (median gestational age at birth = 40 weeks, range 32.14-42.29) underwent MRI (median postmenstrual age at scan = 41.29 weeks, range 36.57-44.71). Fixel-based fibre metrics (mean fibre density, fibre cross-section, and fibre density modulated by cross-section) were calculated from diffusion imaging data in the left and right uncinate fasciculi and cingulum bundle. For n = 311, internalising and externalising behaviour, and social-emotional abilities were reported at a median corrected age of 18 months (range 17-24). Statistical analysis used multiple linear regression and mediation analysis with bootstrapping. Maternal depressive symptoms were positively associated with infant fibre density in the left (B = 0.0005, p = 0.003, q = 0.027) and right (B = 0.0006, p = 0.003, q = 0.027) uncinate fasciculus, with left uncinate fasciculus fibre density, in turn, positively associated with social-emotional abilities in toddlerhood (B = 105.70, p = 0.0007, q = 0.004). In a mediation analysis, higher maternal depressive symptoms predicted toddler social-emotional difficulties (B = 0.342, t(307) = 3.003, p = 0.003), but this relationship was not mediated by fibre density in the left uncinate fasciculus (Sobel test p = 0.143, bootstrapped indirect effect = 0.035, SE = 0.02, 95% CI: [-0.01, 0.08]). There was no evidence of an association between maternal depressive and cingulum fibre properties. These findings suggest that maternal perinatal depressive symptoms are associated with neonatal uncinate fasciculi microstructure, but not fibre bundle size, and toddler behaviour.

Fig. 1. Visual representation of the tracts in template space.

The top row shows the location of the tracts, coloured by streamline orientation (blue: superior-inferior, red: left-right, green: anterior-posterior). On the bottom row, the middle image shows an example of fibre orientation distributions in a region of crossing fibres, while the left and right images represent “glass-brain” illustrations of the tracts of interest (pink: ventral cingulum, purple: dorsal cingulum, turquoise: uncinate fasciculus).

Fig. 2. Plots showing the relationships between variables of interest.

A–D Contain plots showing the relationship between EPDS and white matter microstructure and macrostructure, controlling for effect of covariates. For mean FD left (A) and right (B) uncinate fasciculus and mean FDC right uncinate fasciculus (D), the main model is plotted. For mean FDC left uncinate fasciculus (C), the model with EPDS by Sex interaction is plotted. The relationship between EPDS and mean FD in the left and right uncinate fasciculus (A, B) survived FDR correction. E Contains a plot from robust regression, showing the relationship between mean FD in the left uncinate fasciculus and total Q-CHAT score, controlling for the effect of covariates.

Fig. 3. Dot-and-whisker plot for the multiple linear regression model predicting mean FD in the left uncinate fasciculus.

The black dots represent the regression coefficient estimate with 95% confidence intervals. The caption contains Akaike’s Information Criterion (AIC) and the Bayesian Information Criterion (BIC), with smaller values indicating a better fit. For a model without EPDS, the AIC is −2193 and BIC is −2165, indicating a worst fit.