Introduction

Exposure to prenatal maternal depression increases the susceptibility for depression in the offspring1, 2 and associates with an increase in neurobehavioral,3 cognitive4 and socio-emotional problems.5 Prenatal maternal depression is also associated with elevated maternal cortisol levels6 that, in turn, predict an enlarged amygdala in 7-year-old children.7 As increased amygdala volume is associated with depression,8, 9, 10, 11 these findings appear to reflect the prenatal, transgenerational transmission of individual differences in vulnerability for depression. However, these findings may be confounded by postnatal influences, such as parental care, which is strongly affected by parental mental health.12 We recently found a significant relation between prenatal maternal depression and the amygdala’s microstructure of infants in the imaging study performed shortly after birth.13 But, there is still a lack of knowledge on whether the amygdala functions of infants are also influenced by prenatal maternal depression.

The amygdala is a key brain region closely associated with stress reactivity and vulnerability for depressive disorder.14 Structural magnetic resonance imaging (MRI) studies showed alterations in the amygdala volume in children, adolescents and adults with major depressive disorder (MDD).8, 9, 10, 11 Functional MRI (fMRI) studies found the selective involvement of the amygdala in mood disorders, as well as in the processing of emotionally negative information and the accompanying stress responses.15, 16 Metabolic activity in the amygdala was found to be correlated with negative mood.17 Increased amygdala activation in response to novelty or threatening faces in children was associated with measures of negative emotionality,18, 19 which in turn predicted a greater risk for mood disorders.20, 21, 22 More interestingly, similar findings can also be shown in terms of amygdala functional connectivity. Using resting-state fMRI (rs-fMRI), researchers found that patients with major depression showed altered connectivity of the amygdala with the anterior cingulate,23 prefrontal gyrus24 and insula.25

In this study, we aimed to investigate the amygdala functional connectivity of 6-month-old infants as a function of prenatal maternal depressive symptoms using rs-fMRI. In line with the literature mentioned above, we expected that the amygdala's functional connectivity with key brain structures regulating emotion processes, such as the prefrontal cortex, cingulate and insula, as well as the sensory structures relevant to emotion generation, is a function of prenatal maternal depressive symptoms. We particularly used the rs-fMRI technique to examine amygdala functional networks as it enables a summary of complex patterns of brain functional organization. Moreover, it is free of tasks, imposes a minimal cognitive burden on participants and takes relatively little time in the scanner compared with task fMRI studies. Therefore, it becomes a valuable tool to examine infant brain functional networks26 and provides a unique opportunity to understand the brain functional organization as a function of prenatal maternal depressive symptoms. The results of this study provide, to our knowledge, the first direct link between prenatal maternal depressive symptoms and the amygdala functional networks in infants.

Materials and Methods

Subjects

Forty-two infants of mothers who participated in a birth cohort study, GUSTO (Growing Up in Singapore Towards Healthy Outcomes), were recruited for the neuroimaging study at 6 months of age. The GUSTO cohort consisted of pregnant Asian women attending the first trimester antenatal ultrasound scan in a clinic at the National University Hospital and KK Women's and Children's Hospital in Singapore. The parents were Singapore citizens or Permanent Residents of Chinese, Malay or Indian ethnic background. Birth outcome and pregnancy measures were obtained from hospital records. Socioeconomic status (household income) and prenatal exposure to alcohol (regular alcohol drinking) and tobacco (regular smoking, daily exposure to smoking at home and job) were extracted from survey questionnaires conducted as part of a scheduled appointment during pregnancy. The detailed recruitment criteria and measures of the GUSTO cohort were described in Eberhard-Garn et al.27 The GUSTO study was approved by the National Healthcare Group Domain Specific Review Board and the Sing Health Centralized Institutional Review Board.

Edinburgh Postnatal Depression Scale

The Edinburgh Postnatal Depression Scale (EPDS) questionnaire was administered to mothers at 26 weeks of pregnancy and at 3 months after delivery and used to quantify maternal prenatal and postnatal levels of depressive symptomatology. The EPDS is a widely used 10-item self-report scale designed as a screening instrument for postnatal depression and has been well validated for use in both prenatal and postnatal depression.27, 28, 29 In the EPDS, each item is scored on a four-point scale (0–3). The reliability of the EPDS score was 0.82, which was assessed using Cronbach’s analysis for our cohort. Higher scores on the EPDS indicate a greater severity of depressive symptoms.

MRI acquisition

Infants underwent MRI scans at 6 months of life using a 1.5-Tesla GE scanner at the Department of Diagnostic and Interventional Imaging of the KK Women's and Children's Hospital. The imaging protocols were (i) fast spin-echo T2-weighted MRI (axial acquisition; TR=3500 ms; TE=110 ms; FOV=256 mm × 256 mm; matrix size=256 × 256; 50 axial slices with 2.0 mm thickness), (ii) fast spin-echo T2-weighted MRI (coronal acquisition; TR=3500 ms; TE=110 ms; FOV=256 mm × 256 mm; matrix size=256 × 256; 50 axial slices with 2.0 mm thickness) and (iii) echo planar rs-fMRI (axial acquisition; TR=2500 ms; TE=40 ms; FOV=192 mm × 192 mm; matrix size=64 × 64; 40 axial slices with 3.0 mm thickness; 120 volumes). The scans were acquired when the subjects were sleeping in the scanner. No sedation was used and precautions were taken to reduce exposure to the MRI scanner noise. A neonatologist was present during each scan. A pulse oximeter was used to monitor heart rate and oxygen saturation through the entirety of the scans.

Amygdala delineation

We manually delineated the amygdala on the image created by averaging the axial and coronal T2-weighted images within individual subjects through a halfway affine transformation.30 This new image had better signal-to-noise ratio and better spatial resolution than the acquired coronal or axial T2-weighted image alone. The manual tracing of the amygdala was performed on the coronal view of this new image closely following a standard protocol reported elsewhere.31 Five subjects’ images were manually delineated twice by one rater. The intra-class correlation coefficient was 0.84 for the amygdala of 6-month-old infants.

Amygdala functional connectivity analysis

The rs-fMRI data were first processed with slice timing, motion correction, skull stripping, band-pass filtering (0.01–0.1 Hz) and grand mean scaling of the data. We only included subjects with framewise displacement (head motion characteristics, range from 0.05 to 0.47 in our sample) less than 0.5 as suggested by Power et al.32 However, three subjects showed sudden ‘jerk-like’ head movements at the first three volumes. We manually removed these three volumes to create the new fMRI data sets for these three subjects that satisfied the head motion criteria. Within each subject, the six parameters of the head motion and CSF and white matter signals were regressed out from the fMRI images. Finally, the fMRI images were aligned to the T2-weighted image.

To analyze the amygdala functional networks, the time-series signal averaged across the amygdala was first computed. Pearson’s correlation coefficients between the amygdala fMRI mean signal and the fMRI signals of the rest of the brain were subsequently calculated and transformed to ‘z-values’ using Fisher’s z-transformation. The z-map images were finally aligned to the atlas of infant at 6 months on the basis of the nonlinear transformation obtained from the large deformation diffeomorphic metric mapping algorithm.33

Statistical analysis

Voxel-based analysis was examined to first determine the amygdala functional connectivity networks and then investigate the influences of maternal depression on the amygdala functional connectivity using SPM8. The fMRI z-map images were smoothed with a Gaussian kernel with a full width half maximum of 6 mm. First, regression analysis was applied to the smoothed z-map images to identify the amygdala functional networks with the post-conceptual age on the MRI day (gestational age+days of life since birth to the MRI scan) as a covariate. Second, regression analysis was used to examine the associations between prenatal maternal depression and the amygdala functional networks with the post-conceptual age on the MRI day and postnatal maternal depression as covariates. All statistical results at each voxel were thresholded at the level of significance (P<0.01), the size of each cluster was >100 voxels with the significance level of each cluster <0.05. The statistical results were then corrected for multiple comparisons using false discovery rate.

We further examined whether any other confounding variables, such as household income, birth weight and sex, altered the association between prenatal maternal depression and amygdala functional networks. We first extracted the brain regions with significant associations with prenatal maternal depression and averaged the z-values within each of these brain regions. Regression analysis was subsequently performed with the mean z-value of one brain region as the dependent variable and prenatal maternal depression as the independent variable. The post-conceptual age on the MRI day, postnatal maternal depression and one of the confounding variables were entered as covariates. We repeated this analysis for every brain region and every confounding variable.

Results

Demographics

Among the 42 infants in this study, 15 with missing T2 or fMRI data and 3 with large motion in the fMRI data were excluded. Hence, only 24 subjects were included in this study. All the 24 subjects had gestational age ⩾35 weeks (range=35.4–40.2 weeks), birth weight larger than 2000 g (range=2330–4070 g) and a 5-min APGAR score >9 (Table 1).

Table 1 Demographics
Full size table

The pairwise correlation between measures of prenatal and postnatal maternal depression scores exceeded 0.5 (r=0.663, P=0.002). Student's t-test scores also showed that the magnitude of the maternal depression level was comparable between the assessments at pregnancy and the early postpartum period (t=0.236, P=0.816). Maternal depression did not vary as a function of infant sex (prenatal: t=0.720, P=0.479; postnatal: t=−0.533, P=0.601) or maternal ethnicity (prenatal: F=2.512, P=0.105; postnatal: F=0.083, P=0.921). Further examination on the variables listed in Table 1, via Pearson’s correlation, revealed that monthly household income was negatively correlated with prenatal (r=−0.475, P=0.026) but not postnatal maternal depression (r=−0.167, P=0.509).

Amygdala functional networks

We first identified the amygdala functional networks (Table 2). As shown in Figure 1, both left and right amygdala functionally connected common brain regions forming multiple amygdala networks. These networks included (i) an emotional regulation network that involved the bilateral insula, anterior cingulate cortex(ACC), lateral and medial orbitofrontal cortices, ventromedial, dorsal and ventrolateral prefrontal cortices; (ii) sensory and perceptual systems that comprised the bilateral precentral and postcentral gyri, superior, middle and inferior temporal cortex, lingual gyrus, lateral occipital cortex, thalamus, basal ganglia, midbrain and anterior cerebellum, as well as the right supramarginal gyrus and cuneus, left angular gyrus; (iii) an emotional memory network that consisted of the bilateral hippocampus, parahippocampus and entorhinal cortex. Moreover, the left amygdala sensory and perceptual systems also involved the left angular gyrus, whereas the right amygdala sensory and perceptual systems also included the left supramarginal gyrus and left cuneus.

Table 2 The amygdala functional networks
Full size table
Figure 1
figure 1

The amygdala functional networks. (a) left amygdala functional networks; (b) right amygdala functional networks; (c) amygdala functional networks. ACC, anterior cingulate cortex; AG, angular gyrus; Amyg, amygdala; antCereb, anterior cerebellum; BG, basal ganglia; dlPFC, dorsolateral prefrontal cortex; EC, entorhinal cortex; Hipp, hippocampus; lOFC, lateral orbitofrontal cortex; ITC, inferior temporal cortex; L, left; LG, lingual gyrus; LOC, lateral occipital cortex; mOFC, medial orbitofrontal cortex; MTC, middle temporal cortex; PHG, parahippocampus; PoCG, postcentral gyrus; PrCG, precentral gyrus; R, right; SMG, supramarginal gyrus; STC, superior temporal cortex; vlPFC, ventrolateral prefrontal cortex; vmPFC, ventromedial prefrontal cortex.

Full size image

Our analysis further revealed positive associations between prenatal maternal depression and the left amygdala functional connectivity with multiple regions in the amygdala functional networks identified above (Table 3, Figure 2). Infants born to mothers with higher levels of depressive symptoms during pregnancy showed greater functional connectivity of the left amygdala with (i) the bilateral medial prefrontal cortex (mPFC), including ACC, medial orbitofrontal cortex and ventromedial prefrontal cortex, and left insula in the emotional regulation network; (ii) the left superior, middle and temporal cortex in the sensory and perceptual systems; (iii) the left entorhinal cortex in the emotional memory network. These findings remained significant even after controlling for the post-conceptual age on the MRI day and postnatal maternal depressive symptoms. This suggests that the above findings were not confounded by postnatal maternal depression. Further examination confirmed that none of the other confounding variables, such as household income, birth weight and sex, altered the aforementioned associations between prenatal maternal depressive symptoms and the left amygdala functional connectivities. We did not find any association between prenatal maternal depressive symptoms and the right amygdala function connectivity.

Table 3 Prenatal maternal depression in relation with the amygdala functional networks
Full size table
Figure 2
figure 2

The association between prenatal maternal depression and amygdala functional networks. ACC, anterior cingulate cortex; EC, entorhinal cortex; ITC, inferior temporal cortex; L, left; mOFC, medial orbitofrontal cortex; MTC, middle temporal cortex; R, right; STC, superior temporal cortex; vmPFC, ventromedial prefrontal cortex.

Full size image

Discussion

We showed that amygdala functional networks in infancy are similar to those seen in childhood and adulthood, suggesting that the functional connections of the amygdala exist in infancy and persist into later life. Interestingly, independent of postnatal maternal depressive symptoms, prenatal maternal depressive symptoms were positively associated with the left amygdala functional connectivity with the key brain regions involved in the activation and regulation of emotional states. This finding suggests that the developmental origins of neural circuits associated with an increased vulnerability for depression may be apparent in infancy and established as a function of intrauterine maternal influences.

We observed that at 6 months of age the amygdala was functionally connected to a number of brain regions, forming the emotional regulation, sensory and perceptual, and emotional memory networks. These functional connections of the amygdala are well-known anatomical connections in nonhuman primates.34 The high degree of concordance with those functional networks observed in our study, and those seen in children35 and adults36 suggests that these functional connections of the amygdala appear in early life and persist into adulthood. The most consistent findings are the functional connections of the amygdala with the prefrontal cortex, including the insula, ACC, orbitofrontal cortex (OFC), ventromedial prefrontal cortex and ventrolateral prefrontal cortex, which provides the basis for the top-down emotional regulation.37, 38, 39 Moreover, the amygdala functionally connects with multi-modal sensory regions, suggesting the amygdala’s central role in perception and response to fearful stimuli presented in autonomic, somatomotor, visual and auditory modalities, such as the anticipatory arousal and uncertainty,40 unpleasant olfactory stimuli,41 unfamiliar faces42 and fearful vocalizations.43 Furthermore, the functional connections of the amygdala and medial temporal lobe emphasize the role of the amygdala in modulating the consolidation of medial temporal lobe-dependent emotional memories.44 Hence, even at infancy, the amygdala is positioned for a crucial role in the top-down regulation of emotion through direct modulation of the prefrontal cortex and bottom-up emotional processes, such as emotional perception and encoding affective, motivational and social salience of environmental stimuli, generation of affective state and autonomic responses, and formation of emotional memory and fear conditioning.

Our study showed that prenatal maternal depression alters the infants’ amygdala functional connectivity with specific brain regions involved in both bottom-up and top-down processes of emotion generation and regulation. In the bottom-up network of the amygdala, higher prenatal maternal depressive symptoms were associated with its greater functional connectivity with the temporal cortex and insula, regions that are important for processing fearful facial expressions of emotion. It has been shown that viewing emotional faces increases the coupling between the fusiform and amygdala.45 Interestingly, a resting-state functional connectivity analysis recently identified links the insula to the middle and inferior temporal cortex as well as the ACC.46 The insula also shares extensive reciprocal connections with the amygdala and is implicated in the ability to differentiate positive versus negative emotion processing and hence making judgments about emotions on the basis of facial expression and conveying the representation of aversive sensory information to the amygdala.47, 48 Elevated responses to the presentation of fearful or sad facial expression in the amygdala, insula and temporal cortex associate with depression, possibly mediated via rich connections with amygdala circuitry.49, 50 These emotional processing biases in depression have been suggested to be important in the underlying etiology of this disorder.36 Furthermore, aberrant functional connectivity of the amygdala with the insula and temporal cortex is also found in association with depression.25, 51, 52 Thus, with evidence of a greater risk for depression in the offspring born to depressed mothers, our findings of increased amygdala-temporal/insular connectivity in infants born to mothers with greater prenatal maternal depressive symptoms suggest that the amygdala bottom-up network could be the neural origin for the transgenerational transmission of vulnerability for depression.

In the top-down network of the amygdala, higher prenatal maternal depressive symptoms were associated with its greater functional connectivity with the mPFC, which is in line with studies of amygdala–mPFC connectivity in adolescents with internalizing symptoms53 and early-onset MDD,54 as well as in adults with MDD.55 The ventromedial prefrontal cortex and OFC connect to and receive inputs from the amygdala and temporal cortex and then, in turn, send signals from these regions as well as the cingulate. The ACC has strong connections with limbic structures, such as the amygdala and OFC. Moreover, the amygdala–PFC connections have been consistently observed in task-based37 and rs-fMRI.36 Hence, the top-down network of the amygdala is believed to provide the basis for emotion regulation processes.

Interestingly, our results showed only the left amygdala functional connectivity with the mPFC in the association with prenatal maternal depressive symptoms. This lateralized pattern of the amygdala was supported by evidence that top-down responses may modulate only the left amygdala.25 Although the literature regarding the laterality of amygdala dysfunction in MDD is mixed,25, 56 our results may emphasize the leftward laterality of the amygdala because the left hemisphere has the more rapid growth of gray and white matters than the right during infancy.57 Hence, the left amygdala and its connections may be more vulnerable to prenatal influences. Taken together, the altered amygdala–mPFC connectivity in infants born to mothers with increased prenatal maternal depressive symptoms is similar to that which is observed in patients with MDD, reflecting the neural basis for the familial transmission of phenotypes associated with mood disorders.

Our findings also support growing evidence suggesting that the origins of many brain abnormalities in neuropsychiatric and neurodevelopmental disorders can be traced back to the intrauterine period of life, when the fetus is influenced by environmental conditions during sensitive periods of cellular proliferation, differentiation and maturation to produce structural and functional changes in the brain. In animal models, changes in brain morphology have been observed in offspring of mothers exposed to prenatal stress. Significantly expanded dimensions of the lateral nucleus of the amygdala were found in prenatally stressed offspring.58 Reduced spine densities and dendritic length of pyramidal neurons in the dorsal ACC and OFC have also been reported in rats born to prenatally stressed mothers.59 Depressed mothers exhibit increased cortisol level during pregnancy. Converging evidence from a variety of species suggests that cortisol exerts multiple effects on the brain, which include altering neurotransmitter functioning,60 affecting the development of oligodendrocytes and the glia cells that manufacture myelin sheets in the central nervous system,61 and influencing glucocorticoid receptor rich regions such as the amygdala, the activation of which may further alter neurotransmitter function.62 In humans, elevated prenatal cortisol level was associated with a larger amygdala volume in children.7 Moreover, prenatal maternal depression was associated with difficult temperament in infancy63 even when controlling for postnatal mood.64 Such fetal programming enhances traits such as threat sensitivity and prepares the fetus to develop within a harsh postnatal environment, while influening children’s mental health trajectories.65, 66 It is also important to note that prospective studies underscore the importance of individual differences in stress reactivity as predictors of the risk for depression.67, 68

Our study has several limitations. First, the sample size in this study is relatively small, although comparable to infant imaging studies. In addition, our study only included an Asian population. Hence, the results from this study require replication and generalization to other populations. Moreover, considering both scientific importance and subjects burden, our study only assessed postnatal maternal depressive symptoms at 3 months and 12 months but not at 6 months after delivery when the infant brain imaging was conducted. Though additional measurements may have allowed for a better understanding of postnatal maternal effects on infant brain function at 6 months of age, our sample showed a high correlation of maternal depression scales at 3 months and 12 months after delivery (r=0.544, P<0.001).

In conclusion, our findings provide novel evidence that prenatal maternal depressive symptoms alter the amygdala functional connectivity in early postnatal life, suggesting that the neural correlates of the familial transmission of phenotypes associated with MDD are apparent in infants at 6 months of age. Our findings suggest the fetal programming of the transgenerational transmission of maternal mental well-being to the offspring.