Perinatal Western-style diet exposure associated with altered sensory functional connectivity in infant Japanese macaques

Abstract

Sensory processing disorder (SPD) is a neurodevelopmental condition characterized by impaired sensory discrimination and responsivity. Although the causes and neural correlates of SPD remain poorly understood, prenatal influences should be considered, as the prenatal environment is strongly implicated in the progression of neurodevelopmental disorders. One factor hypothesized to promote SPD is perinatal Western-style diet (WSD) exposure. This study explored the effects of perinatal WSD exposure on the proposed neural correlates of SPD in Japanese macaques. Functional connectivity between sensory and emotional processing areas was assessed at 4 months of age using resting-state functional magnetic resonance imaging (rs-fMRI). A machine learning model successfully predicted perinatal diet group based on functional connectivity strengths, indicating that differences in sensory connectivity exist between diet groups. Intra-somatomotor, visual-auditory, somatomotor-auditory, somatomotor-visual, and intra-visual network connections demonstrated the greatest differences between groups, with primary motor cortex connectivity being the most impacted. Connections to the amygdala were not major contributors to accurate model performance, but amygdala connectivity, especially to the somatomotor network, may still be a weak driver of model performance. These findings suggest that a proposed predictor of SPD, perinatal WSD exposure, impacts the functional connectivity of sensory processing areas relevant in SPD during early infancy.

Keywords: amygdala; motor cortex; neurodevelopment; nonhuman primates; sensory processing; somatosensory.

FIGURE 1

Macaque offspring weight at 4 months of age. Offspring weight (N = 36, female n = 18) is displayed in grams. Subjects are clustered by perinatal diet group and sex, with males and females represented by white and gray markers, respectively. CTR, control diet; WSD, Western‐style diet.

FIGURE 2

Performance metrics for the functional random forest (FRF) model in macaques at 4 months of age. The functional connectivity of 378 connections within and between all sensory network regions and the amygdala was used to predict control (CTR) or Western‐style diet (WSD) perinatal diet exposure. Distributions of overall accuracy, specificity (accurate identification of true negatives when classifying perinatal CTR subjects), and sensitivity (accurate identification of true positives when classifying perinatal WSD subjects) were constructed from the predictions of the observed and permuted models. Statistical significance across all three metrics was required for a model to be considered valid for predicting perinatal diet group. A valid model was achieved at 4 months of age (N = 39, CTR n = 22, female n = 19), with the observed model demonstrating a significant improvement in overall accuracy (p < 0.001, Z = −4.62), specificity (p = 0.050, Z = −1.96), and sensitivity (p < 0.001, Z = −3.60). The central line represents the median. Wide bars refer to the 25th/75th percentiles; thinner bars refer to the 2.5th/97.5th percentiles. Significance was determined by a Wilcoxon rank sum test.

FIGURE 3

Top 30 features driving performance of the functional random forest (FRF) model of macaque connectivity at 4 months of age. The model was trained on functional connectivity features from 39 offspring at 4 months of age (CTR n = 22, female n = 19) and validated for predicting perinatal diet group. (a) The maximum variable importance value for each of the 378 functional connectivity features is displayed. The 378 features are ordered along the x‐axis from highest to lowest maximum variable importance value. The midpoint of the range of variable importance values is marked with a horizontal line at 0.080. A vertical line at position 30 along the x‐axis visually separates the 30 features that achieved a maximum variable importance value greater than the midpoint of the range of values from the remaining features. (b) Chart displaying the percentage of the 30 most important features that belong to each of the represented network groupings. (c) Comparison of network grouping representation. Dark blue bars, on the left side of each pair, display the percentage of all 378 features that belong to the respective network grouping. Lighter blue bars, on the right side of each pair, display the percentage of the 30 most important features that belong to the respective network grouping. Abbreviations from b and c: Amyg‐Any, amygdala‐any network; Aud‐Aud, auditory‐auditory; SM‐Aud, somatomotor‐auditory; SM‐SM, somatomotor‐somatomotor; SM‐Vis, somatomotor‐visual; Vis‐Aud, visual‐auditory; Vis‐Vis, visual‐visual.

FIGURE 4

Distributions of connectivity strengths between perinatal diet groups for the 30 features with the greatest maximum variable importance from the macaque functional random forest (FRF) model at 4 months of age. Boxplot pairs, each representing one of the 30 features, are clustered by network group and are ordered along the x‐axis from highest to lowest maximum variable importance value. White boxplots, on the left side of each pair, represent the distribution for the perinatal control diet (CTR) offspring. Gray boxplots, on the right side of each pair, represent the distribution for the perinatal Western‐style diet (WSD) offspring. The central line represents the median. Wide bars refer to the 25th/75th percentiles; thinner bars refer to the 2.5th/97.5th percentiles. Significance testing was not performed on these distributions as they were already identified as being important for a statistically significant model prediction. The model included data from 39 offspring at 4 months of age (CTR n = 22, female n = 19).

FIGURE 5

Distributions of connectivity strengths between perinatal diet groups for amygdala connections within the upper 50% of features with the greatest maximum variable importance from the macaque functional random forest (FRF) model at 4 months of age. Boxplot pairs, each representing a functional connection with the amygdala, are clustered by network group and are ordered along the x‐axis from highest to lowest maximum variable importance value. White boxplots, on the left side of each pair, represent the distribution for the perinatal control diet (CTR) offspring. Gray boxplots, on the right side of each pair, represent the distribution for the perinatal Western‐style diet (WSD) offspring. The central line represents the median. Wide bars refer to the 25th/75th percentiles; thinner bars refer to the 2.5th/97.5th percentiles. Significance testing was not performed on these distributions as they were already identified as being important for a statistically significant model prediction. The model included data from 39 offspring at 4 months of age (CTR n = 22, female n = 19).