Patterns of Infant Amygdala Connectivity Mediate the Impact of High Caregiver Affect on Reducing Infant Smiling: Discovery and Replication

Abstract

Background

Behavioral research indicates that caregiver mood disorders and emotional instability in the early months following childbirth are associated with lower infant positive (PE) and higher negative emotionality (NE), but the neural mechanisms remain understudied.

Methods

Using resting state functional connectivity (rs-FC) as a measure of the functional architecture of the early infant brain, we aimed to determine the extent to which connectivity between the amygdala, a key region supporting emotional learning and perception, and large-scale neural networks mediated the association between caregiver mood and anxiety and early infant NE and PE. Two samples of infants (first sample: n=58; second sample: n=31) aged 3-months underwent an MRI scan during natural sleep.

Results

During infancy, greater rs-FC between the amygdala and the salience network, and, to a lesser extent, lower amygdala-and executive control network rs-FC, mediated the effect of greater caregiver postpartum depression and trait anxiety on reducing infant smiling (p<0.05, FWE). Furthermore, results from the first sample were replicated in the second, independent sample, to a greater extent for caregiver depression than for caregiver anxiety.

Conclusions

We provide evidence of early objective neural markers that can help identify those infants who are more likely to be at risk from, versus those who might be protected against, the deleterious effects of caregiver depression and anxiety and reduced PE.

Introduction

Although emotional regulation capacity develops into adulthood(1), high levels of emotional reactivity and rapid changes in self-regulation occur in the first years of life(2). Negative emotionality(NE), frequent crying, responding intensely to novelty and limitations, and difficulty in being soothed(3,4), and positive emotionality(PE), smiling, laughter, high pleasure(5,6), can be measured reliably in infants within the first months(7–12), with NE showing modest stability(13–15) but mean-level increases across the first years(7,16–18). Importantly, high infant NE predicts behavioral and emotional problems later in childhood(4,19–22). Low infant PE predicts behavioral inhibition in early childhood(23), and, by middle childhood, low PE is associated with later depression(24–29) and externalizing behavior problems(30). Caregiver mood disorders and emotional instability impact infant NE development(31–35), and may disrupt emotional regulation development(36–40). Similarly, low caregiver positive affect(41–43), and high negative affect and emotional instability(44–46), are associated with lower infant PE, primarily smiling and laughter.

The pathophysiological processes underlying infant NE and PE development, and the influence of caregiver affect and anxiety on these processes, remain unknown. Filling this gap and elucidating early objective neural markers of these pathophysiological processes can help identify infants at elevated or reduced risk from deleterious effects of caregiver affect and anxiety prior to emerging symptoms. Several large-scale cortical networks develop during the first two years of life(47–53). These networks include the salience network, centered on anterior insula and dorsal anterior/mid cingulate cortex, underlying attention to personally relevant stimuli(54), and with strong connectivity with the amygdala(54), implicated in positive and negative emotional reinforcement learning(55–57); the medial prefrontal cortical-posterior cingulate cortical-centered default mode network supporting self-referential processing(58,59); and the frontoparietal executive control network, supporting cognitive control processes(53). While the salience and executive control networks continue to develop during the first two years(53), functional segregation in both networks is evident in neonates and in the first year(60,61). The relationships that unfold among connectivity within these networks and cognitive functioning and self-awareness(62–67) during the first two years are thus critically important for developing emotional reactivity and regulation(47,68). Yet, few studies examined relationships among neural circuitry function and emotional reactivity during this early life period.

Resting state functional connectivity(rs-FC) studies can elucidate the functional architecture of the infant brain without administering cognitive tasks, inherently difficult in infants. The few rs-FC studies that examined networks important for emotional reactivity in infancy focused on amygdala rs-FC, and reported significant relationships between amygdala-salience network rs-FC and NE, measured by standardized behavioral rating scales(69). Specifically, positive associations were shown among neonatal amygdala rs-FC to salience network regions and fear(70), and sadness(71), at 6-months; and higher levels of such amygdala rs-FC were associated with less change in infant NE(fear) from 6 to 24-months(71), even after accounting for maternal postpartum depression(71). Neonatal amygdala rs-FC with insula and dorsal anterior cingulate cortex also predicted internalizing behavior at 2 years(72).

To our knowledge, no studies examined relationships among amygdala-salience network rs-FC and infant PE. Additionally, only a few studies with small sample sizes examined relationships between infant rs-FC and caregiver affect, reporting greater rs-FC within and among the amygdala, insula and anterior cingulate cortex in infants of mothers with higher levels of depression during pregnancy(73,74). Furthermore, while one small study(75) reported that greater default mode network rs-FC mediated the relationship between higher interparental conflict and higher infant NE, no study to our knowledge examined the extent to which infant neural network rs-FC might mediate relationships between caregiver affect and anxiety and infant NE and PE. It thus remains largely unknown whether infant amygdala rs-FC explains variance in the association between caregiver affect/anxiety-infant NE and PE relationships.

We aimed to determine whether infant amygdala rs-FC mediated associations among caregiver affect and anxiety and infant NE and PE. To achieve this, we assessed neural networks in 3-month-old infants, given the predictive utility of early emotionality for later function(9,69,76,77), early neural functional specialization for processing negative emotion(78), and the feasibility of scanning infants during natural sleep at this age. Given the pressing need to replicate neuroimaging findings(79,80) to definitively inform understanding of brain-behavioral relationships, we also aimed to replicate findings in an independent sample of 3 month-old infants.

We tested two hypotheses in the first sample, and aimed to replicate significant findings in the independent sample: Hypothesis 1: Greater infant positive amygdala-salience network rs-FC would be associated with greater infant NE, and with lower magnitudes of PE dimensions; Hypothesis 2: Infant amygdala rs-FC measures that showed significant relationships with infant NE and PE would mediate associations between caregiver negative affect/anxiety and infant NE and PE. We hypothesized that greater caregiver negative affect and anxiety would be associated with greater infant amygdala-salience network rs-FC, and greater infant NE and lower infant PE, while lower infant amygdala-salience network rs-FC would be associated with greater suppression of these caregiver affect/anxiety-infant NE and PE relationships.

Given early sex differences in NE, with NE being more common in males than females(81,82); and sex differences in interhemispheric integration of large-scale networks supporting executive function(48), and maturation rates(83,84), we explored the potentially moderating influence of infant sex on relationships between infant rs-FC and NE and PE, and caregiver affect/anxiety-infant rs-FC in the first sample.

Methods and Materials

We examined two samples of 3-month-old infants, using the same research measures. In the first sample, behavioral assessments of caregiver-infant dyads were conducted in the home at 3-months. The infant completed an MRI scan at Children’s Hospital of Pittsburgh(CHP) within 2 weeks; data collection occurred between September, 2018 and March, 2020. In the second, independent sample, mother-infant dyads completed behavioral assessments and an MRI scan at CHP at 3-months; data collection occurred between September, 2015 and July, 2017. All procedures in both studies were approved by the University of Pittsburgh IRB.

Samples

The first sample comprised 78 caregivers and their 3-month-old infants recruited from the community: postnatal wards at a local university hospital, pediatric practices supported by the University of Pittsburgh Clinical and Translational Science Institute, and a university research registry(Pitt+Me). Infant exclusion criteria were: premature birth(prior to 37 weeks), low birthweight(<5.5lb; caregiver report, medical records), abnormal morphometry(i.e. small occipitofrontal circumference(<32cm), abnormal APGAR scores(<7 at 5 mins), and extended hospitalization for physical health problems, including neurological illness. MRI metal exclusion criteria applied to infants: pacemakers, aneurysm clips, non-removable ferromagnetic material. Caregivers were excluded if they had prenatal/concurrent illicit substance use/substance use disorder(obstetric records/self-report); < 2 hours/day care of their infant; <18 yrs (as unable to give informed written consent). 19 infants were excluded for incomplete scans, excessive motion and/or artifact, and one infant did not have complete behavioral data, leaving 58 caregiver-infant dyads(57 mothers, 1 father, age 19–42,mean=30.8,SD=4.42).

The independent sample comprised 58 mothers(age 19–24) and their 3-month-old infants, recruited from the population-based, longitudinal Pittsburgh Girls Study(PGS(85,86)). Infant and caregiver exclusion criteria were as in the first sample. 25 infants were excluded for incomplete scans, excessive motion and/or artifact, and 2 mothers did not complete the postpartum depression scale(below), leaving 31 caregiver-infant dyads. The independent sample was previously examined using different methods(87).

In both samples, the caregiver was the biological mother in the majority of cases, partly due to recruitment from postnatal units.

Measures

Infant temperament was measured via caregiver report on the Infant Behavior Questionnaire Revised(IBQ-R)(69) at age 3-months. Caregiver affect and anxiety were assessed using the Edinburgh Postnatal Depression Scale(EPDS(88)) and the Spielberger State-Trait Anxiety Inventory(STAI)(89), respectively. The Personality Assessment Inventory–Borderline Features Scale(PAI-BOR(90)) assessed affective instability. Education level was a proxy for socioeconomic status in both samples. Given that parenting quality shapes infant emotional regulation development(91–93), we assessed caregiving quality using the Home Observation for Measurement of the Environment(HOME) Inventory: Responsivity, Acceptance, and Organization(94), and used these measures as covariates(Supplement; Table 1).

MRI Scanning

All infants in both samples were scanned on a Siemens 3 Tesla Skyra system using a 32-channel head coil during non-sedated sleep using the feed-and-wrap technique(95,96). Two, five-minute resting state sequences were acquired in both samples(Supplement).

Data Analyses

Preliminary Analyses

Partial correlations between caregiver affect/anxiety measures and infant temperament scores were computed in the total sample of caregivers and their infants with available behavioral data, range:n=72–75. Infant age, sex, maternal education, and HOME variables were covariates. Correlations were deemed significant at FDR corrected q<0.05(97). Images were preprocessed using previously-published methods(98)(Supplement).

Amygdala Seed-Based Computation

A neonatal-specific parcellation atlas(99) determined voxels in the amygdala(in standardized space). A reference time course was the average time course for all amygdala voxels. Nuisance parameters, including motion transformation parameters and linear and quadratic drift correction, were regressed out of the time course, which was then band-pass filtered(0.009–0.08 Hz). The seed-based rs-FC for a given voxel in the brain was defined as the regression between the amygdala time course and the voxel time course(converted to a T-score), with the nuisance parameters as covariates.

Voxelwise Analyses to Determine Relationships Among Infant Emotionality, and Caregiver Affect and Anxiety, and Amygdala rs-FC

General Linear Models were performed on a voxelwise basis, with the dependent variable being the seed-based rs-FC. Hypothesis 1: Independent variables were infant NE and PE measures that were significantly associated with caregiver affect and anxiety. Hypothesis 2: Independent variables were caregiver affect and anxiety measures that were significantly associated with infant NE and PE. Covariates included infant gender, post-conceptional age(PCA) at scan, caregiver education level, and(square root of) number of volumes(frames) remaining after censoring. In the first sample, covariates included HOME variables; these data were not available for the independent sample. Results were spatially filtered with kernel s.d(σ)=4mm. Standard methods(100) involving construction of noise maps (with identical spatial autocorrelation as the dataset) identified intensity and spatial extent thresholds corresponding to family-wise error(FWE)-corrected p<0.05.

Indirect Effect (Mediation) Analyses

These analyses were performed, using in-house-written software in IDL on a wholebrain voxelwise basis (to avoid circularity) in each model, according to standard methods(101), involving linear regression with and without the mediator as an additional covariate, and computing the difference in the regression parameters. Caregiver affect and anxiety measures that were significantly associated with infant NE and PE were independent variables, amygdala rs-FC was the mediator, and infant NE and PE measures that were significantly associated with caregiver measures were dependent variables. Covariates were as above. Bootstrapping with replacement(1000 replications) ascertained voxelwise-level statistical significance(102) using bias-corrected and accelerated confidence intervals(103). Spatial filtering and thresholding were performed as above to ensure statistical significance at p<0.05, FWE-corrected. P-values were converted to Z-score maps.

Results

First Sample

While there was substantial intra-correlation between individual caregiver affect and anxiety measures and some infant behavioral measures(Supplemental Table 1), the only caregiver affect/anxiety-infant behavior relationships that received statistical significance were negative correlations between caregiver postpartum depressed mood and infant smiling, and between caregiver trait anxiety and infant smiling(q<0.05 FDR), across all relationships among all four caregiver and all four infant behavioral measures(Figure 1). Thus, analyses testing Hypotheses 1 and 2 were restricted to these relationships(Table 2).

Figure 1. Scatterplots showing associations between caregiver affect and anxiety and infant smiling.

A. caregiver postpartum depression (EPDS) and infant smiling (r=−0.34, p= 0.003); and B. between caregiver trait anxiety and infant smiling (r=−0.34, p=0.003).

Table 2:

Partial Correlations between Caregiver Affect and Infant Emotionality

	EPDS	State Anxiety	Trait Anxiety	Affective Instability	IBQ Negative (composite)	IBQ Smiling	IBQ Pleasure	IBQ Soothability
EPDS	-	0.70* (<0.001)	0.76* (<0.001)	0.36* (0.002)	−0.06 (0.597)	−0.34* (0.003)	−0.14 (0.229)	−0.04 (0.771)
State Anxiety	0.70* (<0.001)	-	0.81* (<0.001)	0.25 (0.038)	−0.011 (0.907)	−0.25 (0.037)	−0.13 (0.293)	−0.06 (0.599)
Trait Anxiety	0.76* (<0.001)	0.81* (<0.001)	-	0.49* (<0.001)	−0.04 (0.749)	−0.34* (0.003)	−0.19 (0.115)	−0.02 (0.838)
Affective Instability	0.36* (0.002)	0.25 (0.038)	0.49* (<0.001)	-	0.14 (0.236)	−0.23 (0.056)	−0.05 (0.659)	0.01 (0.953)
IBQ Negative (composite)	−0.06 (0.597)	−0.01 (0.907)	−0.04 (0.749)	0.14 (0.236)	-	0.16 (0.173)	0.18 (0.135)	−0.14 (0.236)
IBQ Smiling	−0.34* (0.003)	−0.25 (0.037)	−0.34* (0.003)	−0.23 (0.056)	0.16 (0.173)	-	0.71* (<0.001)	0.23 (0.057)
IBQ Pleasure	−0.14 (0.229)	−0.13 (0.293)	−0.19 (0.115)	−0.05 (0.659)	0.18 (0.135)	0.71* (<0.001)	-	0.30*(0.011)
IBQ Soothability	−0.04 (0.771)	−0.06 (0.599)	−0.02 (0.838)	0.01 (0.953)	−0.14 (0.236)	0.23 (0.057)	0.30* (0.011)	-

^*Partial correlation meeting q<0.05, FDR. Observed p-values for each partial correlation are in parentheses.

Hypothesis 1: Associations Between Infant rs-FC and Infant Smiling

A negative correlation was shown(Figure 2A;Supplemental Table 3A) between infant smiling and amygdala rs-FC, predominantly with the salience network(insula) and the default mode network(precuneus, ventromedial prefrontal cortex). A positive correlation was shown between infant smiling and amygdala rs-FC, predominantly with the executive control network(middle frontal gyrus, superior frontal gyrus, angular gyrus/inferior parietal lobe), and visual cortex(all ps<0.05 FWE).

Figure 2. Associations between Infant Amygdala rs-FC and Infant Smiling.

A. First sample. Negative correlations (cool colors) are shown with the salience network and the default mode network, and a positive association (warm colors) with the executive control network.

All ps<0.05 FWE across the wholebrain.

B. Independent sample. The negative correlation between amygdala rs-FC and infant smiling in the salience network was replicated, as was the positive correlation in the executive control network. All ps<0.05 FWE across the wholebrain. Images are presented using the radiological display convention.

Hypothesis 2: Infant rs-FC Mediation of Associations Between Caregiver Affect and Anxiety and Infant Smiling

Indirect effects.

Negative indirect effects(Figure 3A,E;Supplemental Table 4A) were observed with caregiver postpartum depression as the independent variable, infant smiling as the dependent variable, and amygdala seed-related rs-FC predominantly with the salience network(insula), executive control network(middle frontal gyrus, angular gyrus/inferior parietal lobe), visual cortex, and default mode network(ventromedial prefrontal cortex) as the mediator (all ps<0.05 FWE).

Figure 3. Indirect Effects: Amygdala rs-FC as the Mediator, and Infant Smiling as the Dependent Variable.

A. Caregiver postpartum depression as the independent variable in the first sample. Negative indirect effects (warm colors because the direction is the same as for the total direct effects) are shown in the salience network and executive control network. All ps<0.05 FWE across the wholebrain.

B. Caregiver trait anxiety as the independent variable in the first sample. Negative indirect effects are shown in the salience network and executive control network. All ps<0.05 FWE across the wholebrain.

C. Caregiver postpartum depression as the independent variable in the independent sample. Negative indirect effects are shown in the salience network and in the executive control network. All ps<0.05 FWE across the wholebrain.

D. Trait anxiety as the independent variable in the independent sample. Negative indirect effects are shown in the salience network and the default mode network. All ps<0.05 FWE across the wholebrain.

Images are presented using the radiological display convention.

E. Scatterplots for Arms A and B in mediated relationships with caregiver postpartum depression as the independent variable in the first sample.

Top row: Amygdala rs-FC with the salience network as the mediator, and infant smiling as the dependent variable. The (raw) regression coefficient is provided for each pathway (and standard deviation); and the indirect effect (a × b) is provided with 95% confidence intervals.

a=0.05169 (0.01978); b=−0.435 (0.214); c=−0.08424 (0.031120; c’= −0.06174 (0.03216); a × b= −0.02250 (−0.07287, −0.00234)

Bottom row: Amygdala rs-FC with the executive control network as the mediator, and infant smiling as the dependent variable. The (raw) regression coefficient is provided for each pathway (and standard deviation); and the indirect effect (a × b) is provided with 95% confidence intervals.

a=−0.04812 (0.02248); b=0.276 (0.192); c=−0.08424 (0.03112); c’= −0.07098 (0.03216); a × b =−0.01326 (−0.11899, −0.00160)

F. Scatterplots for Arms A and B in mediated relationships with caregiver trait anxiety as the independent variable in the first sample.

Top row: Amygdala rs-FC with the salience network as the mediator, and infant smiling as the dependent variable. The (raw) regression coefficient is provided for each pathway (and standard deviation); and the indirect effect (a × b) is provided with 95% confidence intervals.

a=0.01688 (0.00937); b=−0.471 (0.207); c=−0.03438 (0.01440); c’=−0.02643 (0.01427); a × b = −0.00795 (−0.03677, −0.00120)

Bottom row: Amygdala rs-FC with the executive control network as the mediator, and infant smiling as the dependent variable. The (raw) regression coefficient is provided for each pathway (and standard deviation); and the indirect effect (a × b) is provided with 95% confidence intervals.

=a=−0.01877 (0.00884); b=0.324 (0.226); c= −0.03438 (0.01440); c’=−0.02829 (0.01487); a × b= −0.00609 (−0.03576, −0.00090).

The region of interest-based results shown in the scatterplots in Figures 3E and 3F are for illustrative purposes only. Mediation analyses were performed on a voxelwise basis across the wholebrain.

Negative indirect effects(Figure 3B,F;Supplemental Table 4B) were observed with caregiver trait anxiety as the independent variable, infant smiling as the dependent variable, and amygdala seed-related rs-FC predominantly with the salience network(insula, anterior cingulate gyrus), executive control network(middle frontal gyrus, superior frontal gyrus), visual cortex, and default mode network(ventromedial prefrontal cortex) as the mediator(all ps<0.05 FWE).

There were negative relationships between caregiver trait anxiety/postpartum depression and infant smiling(preliminary analyses). Thus, negative indirect effects reflected mediation of these negative caregiver affect/anxiety-infant behavior relationships by the patterns of amygdala rs-FC associated with lower infant smiling, i.e., greater amygdala-salience and default mode network rs-FC; lower amygdala-executive control network and visual cortical rs-FC.

Caregiver Affect and Anxiety-infant rs-FC Relationships

There were significant relationships between these caregiver measures and amygdala rs-FC with the networks showing negative indirect effects and relationships with infant smiling: the salience network, executive control network and visual cortex. Specifically, positive correlations between caregiver postpartum depression and trait anxiety and amygdala rs-FC were observed in the salience network(insula, anterior cingulate gyrus), and negative correlations in the executive control network(middle frontal gyrus, superior frontal gyrus, angular gyrus/inferior parietal lobe, superior parietal gyrus) and visual cortex. There were also some positive correlations between these caregiver measures and amygdala rs-FC with the executive control network(middle frontal gyrus, inferior frontal gyrus, superior frontal gyrus), default mode network(ventromedial prefrontal cortex, precuneus), and temporal and visual cortices and striatal regions; and negative correlations between these caregiver measures and amygdala rs-FC with temporal cortices and the default mode network(precuneus; all ps<0.05 FWE; Supplemental Tables 5A,B).

Independent Sample

There was a negative correlation between caregiver trait anxiety, although not postpartum depression, and infant smiling (q<0.05 FDR; Supplemental Table 2A). Neuroimaging analyses included postpartum depression, as we aimed to determine whether all significant first-sample mediation findings were replicated in the independent sample, irrespective of the significance of the direct caregiver-infant behavior relationships(101).

Hypothesis 1: Infant rs-FC-NE and PE Relationships

The negative correlations between infant smiling and amygdala rs-FC with the salience network(insula), and default mode network(ventromedial prefrontal cortex), were replicated, as was the positive correlation with the executive control network(middle frontal gyrus, superior fontal gyrus, angular gyrus/ inferior parietal lobe, superior parietal gyrus) and visual cortex; although there were some positive correlations with the default mode network(precuneus; all ps<0.05 FWE; Figure 2B;Supplemental Table 3B).

Hypothesis 2: Infant rs-FC Mediation of Associations Between Caregiver Affect and Anxiety and Infant Smiling

Indirect effects.

Postpartum depression as the independent variable: negative indirect effect findings were replicated in the salience network(insula, anterior cingulate gyrus), executive control network(inferior frontal gyrus, middle frontal gyrus, angular gyrus/inferior parietal lobe), visual cortex, and default mode network(precuneus, ventromedial prefrontal cortex, posterior cingulate gyrus; Figure 3C;Supplemental Table 4C; all ps<0.05 FWE).

Trait anxiety as the independent variable: negative indirect findings were replicated in the salience network(insula), visual cortex and default mode network(precuneus, posterior cingulate gyrus, ventromedial prefrontal cortex). Positive indirect effects were evident in the middle temporal gyrus and supramarginal gyrus(ps<0.05 FWE; Figure 3D;Supplemental Table 4D).

Caregiver Affect and Anxiety-infant rs-FC Relationships

The positive correlations between caregiver postpartum depression and trait anxiety and amygdala rs-FC in the salience network(insula), and negative correlations in the executive control network(inferior frontal gyrus, middle frontal gyrus, superior frontal gyrus, angular gyrus/ inferior parietal lobe, superior parietal gyrus) and visual cortex were replicated. There were also positive correlations between these caregiver measures and amygdala rs-FC with temporal and visual cortices and striatal regions; and negative correlations between these caregiver measures and amygdala rs-FC with the default mode network(precuneus, ventromedial prefrontal cortex; all ps<0.05 FWE; Supplemental Tables 5C,D).

(Supplement for exploratory gender moderation findings, and relationships among infant amygdala rs-FC and other infant NE and PE measures, and infant amygdala rs-FC and other caregiver affect and anxiety measures; and mediation analyses in the first sample after removal of the one father caregiver).

Discussion

We aimed to identify patterns of infant amygdala-wholebrain rs-FC underlying early infant emotionality, as a putative mechanism explaining the association between caregiver affect/anxiety and infant NE and PE. In a community sample of 3 month-old infants, greater amygdala-salience network rs-FC, and lower amygdala-executive control network rs-FC, more strongly mediated negative relationships between greater caregiver postpartum depression/trait anxiety and lower infant smiling, paralleling previous findings of early functional segregation in these networks(60,61). Critically, these findings were largely replicated in an independent community sample of same-age infants assessed using the same scanner, an important goal in neuroimaging research(79,80).

Infant smiles are compelling signals of positive emotions that engage caregiver attention and elicit warmth and affiliative behaviors(104–106). Caregiver-infant interactions that are characterized by reciprocated positive affect and shared smiles reinforce positive emotion in infants(107) and are associated with later functional social interaction(108,109). The centrality of infant smiling to the caregiver-infant relationship especially in the context of postpartum depression also forms the basis of effective video-feedback interventions that emphasize attention to positive emotional signals of the infant, to increase sensitive parental interactions and reduce parental perceptions of infant NE(110). Thus, smiling is a core component of infant PE that lays the foundation for future, adaptive patterns of social behavior.

The negative relationships between caregiver postpartum depression/trait anxiety and infant smiling, replicated for trait anxiety in the independent sample, parallel reported relationships between greater caregiver negative affect and emotional instability and lower infant smiling(44,111,112). Greater caregiver postpartum depression and trait anxiety were associated with stronger connections between the amygdala and the salience network, and greater infant amygdala-salience network rs-FC was associated with lower infant smiling. While a very small number of studies show positive relationships between infant rs-FC within and among the amygdala, insula and anterior cingulate cortex and higher levels of prenatal depression(74,113), we show positive relationships between caregiver depression and trait anxiety in the postpartum period and infant amygdala-salience network rs-FC, and negative relationships between amygdala-salience network rs-FC and infant smiling. Critically, greater infant amygdala-salience network rs-FC was associated with greater mediation of the relationships between greater caregiver postpartum depression/trait anxiety and lower infant smiling. The amygdala supports reinforcement learning(55–57), and the salience network underlies attention to personally relevant stimuli(54). Thus, exposure to caregiver depression/anxiety might lead to increases in rs-FC between the amygdala and this network that lead the infant to make stronger associations between negative emotional displays in the caregiver and potential environmental threat, and to lower smiling. Conversely, lower rs-FC between the amygdala and this network might protect against the effect of greater caregiver postpartum depression/trait anxiety on lowering infant smiling. Together, our findings indicate how caregiver postpartum depression and trait anxiety can shape emerging infant brain-emotional reactivity relationships, and highlight a neural mechanism in the infant that confers vulnerability to the impact of higher levels of these caregiver measures on lowering smiling, a core component of infant PE.

Lower amygdala-executive control network rs-FC mediated relationships among greater caregiver postpartum depression and lower infant smiling in both samples, and among greater caregiver trait anxiety and lower infant smiling in the first sample. In both samples, greater amygdala-executive control network rs-FC was associated with greater infant smiling, while greater caregiver postpartum depression and trait anxiety were associated predominantly with lower, although in the first sample also higher, amygdala rs-FC with this network. These between-sample differences in relationships with trait anxiety might reflect the greater range of caregiver trait anxiety in the first sample. The executive control network supports executive control processes in emotional regulation and social function(53). Thus, lower amygdala rs-FC with this network is a potential neural mechanism to diminish infant emotional regulation capacity and render the infant more vulnerable to effects of greater caregiver postpartum depression/trait anxiety on lowering smiling. This finding parallels reports of early development of negative rather than positive amygdala-prefrontal cortical FC in youth exposed to early-life adversity(114), highlighting the importance of the early caregiving environment in shaping emotional regulation neural network development.

Lower amygdala-visual cortex rs-FC mediated the relationship between greater caregiver postpartum depression/trait anxiety and lower infant smiling in both samples. Greater infant amygdala rs-FC with visual cortex was associated with greater infant smiling, and greater caregiver postpartum depression and trait anxiety were associated predominantly with lower amygdala rs-FC with visual cortex. These findings parallel previous reports of aberrant visual perception of socially salient information and associated emotional regulation in non-human-primates with neonatal amygdala lesions(115,116), and reduced visual attention to socially salient stimuli in infants later diagnosed with autism(117). Thus, greater amygdala-visual cortex rs-FC might facilitate direction of visual attention toward salient external stimuli(118), and protect against the effect of higher levels of the above caregiver measures on predisposing to negative emotion-focused internalizing behavior and lower infant smiling. The most consistent findings in both samples regarding amygdala-default mode network rs-FC were negative relationships between amygdala rs-FC with this network and infant smiling, which mediated the negative relationships between caregiver postpartum depression/trait anxiety and infant smiling. The additional positive correlations between amygdala rs-FC with this network and infant smiling in the independent sample might reflect the smaller size of this sample, and thus greater variability of statistical findings. Given the role of the default mode network in self-referential processing(58,59), greater connectivity between the amygdala and this network might enhance infant capacity for learning contingent relationships between negative emotional signals in the caregiver and personally-directed threat, and thereby render the infant vulnerable to the effect of caregiver trait anxiety on lowering smiling.

Our exploratory findings(Supplement), which need replication, indicate that main relationships among caregiver postpartum depression and trait anxiety and amygdala rs-FC, and between amygdala rs-FC and infant smiling, were largely stronger in males than females, paralleling reports of sex differences in neurodevelopment(48), and that male infants and young children might be more susceptible than their female counterparts to the impact of caregiver depression(119–121). Additional findings(Supplement) revealed predominantly positive relationships between other caregiver affect/anxiety measures: caregiver state anxiety and affective instability, and infant amygdala-salience network rs-FC; and positive relationships between infant amygdala-salience network rs-FC and infant general NE, and negative relationships between infant amygdala-salience network rs-FC and other infant PE components, high pleasure and soothability, in both samples. These findings accord with reported positive relationships between amygdala-salience network rs-FC and infant NE(70–72).

The present findings need to be considered in the context of some limitations. Infant NE and PE as well as affect and anxiety measures were based on caregiver report. Thus, infant NE and PE measures might be subject to caregiver bias. Although the measures employed are well-validated measures of infant emotionality and caregiver mood, further work should confirm these results using independent observations of infant NE and PE. The first sample was of moderate size, HOME environment variables were not available in the independent sample, and the independent sample size was smaller than that of the first sample. Thus, exact replication was not expected; nevertheless, replicated findings were shown. In two infants, the caregiver reported normal birthweight, but medical records received later indicated that birthweight just missed the normal threshold in n=1 first-sample (5lb), and n=1 independent-sample (5.41b) infants. Findings remained very similar in each sample after excluding this one infant, and excluding HOME covariates in the first sample, apart from less mediation by amygdala-executive control network rs-FC in the first sample(Supplement). Caregiver affect and anxiety, and caregiver report of infant emotional behavior, were reported from only one caregiver. Additional information from other caregivers (other parents, child care workers) would be important in future research, as would examining whether caregiver smiling impacts caregiver depression/anxiety-infant smiling relationships. The cross sectional, rather than longitudinal, nature of the mediational analyses is another limitation. It is possible that some findings might have been impacted by family history of depression and anxiety, resulting in shared liability to internalizing problems, including depression/anxiety, in caregiver and infant, or by caregiver treatment. Lifetime, or family history of, depression and anxiety, and caregiver depression and anxiety treatment, data were not available for either sample, however. This could be examined in future studies. Despite these limitations, our main findings were replicated across the samples suggesting that the results are relatively robust.

We show that greater amygdala-salience network rs-FC and, to a lesser extent, lower amygdala-executive control network rs-FC in two separate infant samples mediate the effect of greater caregiver postpartum depression and trait anxiety on lowering a core component of infant PE, infant smiling. Given that low infant PE may be associated with a range of mental health problems in childhood and beyond(23–29,122), our findings provide objective neural markers to facilitate early identification of those infants more at risk from the deleterious effects of caregiver postpartum depression and trait anxiety on emotional behavior.

Table 1:

Summary of Participant Characteristics for first and independent samples.

	First sample n=58			Independent sample n=31
	Mean	SD	Range	Mean	SD	Range
Infant variables
Gender ratio (male:female)	29:29			16:15
Age (3 mons, weeks)	14.96	2.88	10–22	14.13	2.71	9–19
Gestational age (weeks) *	39.24	1.12	37–41	38.69	1.29	37–42
Birthweight (grams) **	3366.19	527.39	2268–4655	3199	378.25	2438–3827
Negative emotionality (NE):
Negative Affectivity	3.01	.62	2–4	3.12	.62	1.64–4.02
Positive emotionality (PE):
High Intensity Pleasure	4.06	1.54	1–7	5.15	1.14	2.57–6.86
Smiling	3.84	1.30	1–7	5.07	1.33	2.71–7
Soothability	5.31	.73	4–7	5.05	.73	3.86–7
Number of resting state data frames acquired (min)	663.5	95.7	416–903	256.7	24.8	212–292
Duration of resting state data acquired (min)	8.85	1.28	5.55–12.05	8.56	0.83	7.07–9.73
Caregiver variables
Depressed mood	6.48	5.43	0–22	5.55	4.86	0–21
State Anxiety	30.93	11.26	20–61	31.48	8.72	21–53
Trait Anxiety	36.18	12.07	20–69	34.84	6.93	24–51
Affective Instability	4.44	3.52	0–13	-	-	-
HOME Responsivity	9.64	1.46	6–11	-	-	-
HOME Acceptance	6.80	.961	5–8	-	-	-
HOME Organization	4.62	1.11	2–6	-	-	-
Education level (SES) *	5.55	2.04	2–8	2.29	.91	1–4

^*Gestational age data were available on n=49 in the first sample, and n=29 in the independent sample.

^**Birthweight data were available on n=54 in the first sample, and n=31 in the independent sample.

Resource Type	Specific Reagent or Resource	Source or Reference	Identifiers	Additional Information
Add additional rows as needed for each resource type	Include species and sex when applicable.	Include name of manufacturer, company, repository, individual, or research lab. Include PMID or DOI for references; use “this paper” if new.	Include catalog numbers, stock numbers, database IDs or accession numbers, and/or RRIDs. RRIDs are highly encouraged; search for RRIDs at https://scicrunch.org/resources.	Include any additional information or notes if necessary.
Antibody
Bacterial or Viral Strain
Biological Sample
Cell Line
Chemical Compound or Drug
Commercial Assay Or Kit
Deposited Data; Public Database	All raw demographic, clincal and neuroimaging data have been uploaded into the National Mental Health Data Archive (NDA)
Genetic Reagent
Organism/Strain
Peptide, Recombinant Protein
Recombinant DNA
Sequence-Based Reagent
Software; Algorithm	In house IDL software
Transfected Construct
Other