Intergenerational neuroimaging study: mother–infant functional connectivity similarity and the role of infant and maternal factors

Abstract

Mother and infant neural and behavioral synchrony is important for infant development during the first years of life. Recent studies also suggest that neural risk markers associated with parental psychopathology may be transmitted across generations before symptoms emerge in offspring. There is limited understanding of how early similarity in brain functioning between 2 generations emerges. In the current study, using functional magnetic resonance imaging, we examined the functional connectivity (FC) similarity between mothers and newborns during the first 3 months after the infant’s birth. We found that FC similarity between mothers and infants increased as infant age increased. Furthermore, we examined whether maternal factors such as maternal socioeconomic status and prenatal maternal depressive symptoms may influence individual differences in FC similarity. For the whole-brain level, lower maternal education levels were associated with greater FC similarity. In previous literature, lower maternal education levels were associated with suboptimal cognitive and socioemotional development. Greater FC similarity may reflect that the infants develop their FC similarity prematurely, which may suboptimally influence their developmental outcomes in later ages.

Introduction

The prenatal period and the first few months after birth are when the brain develops at an unparalleled rate (Stiles and Jernigan 2010; Gilmore et al. 2018). Along with rapid structural development, the fetus develops long-range functional connectivity (FC) that is organized with adultlike core network properties (Thomason 2020). After birth, different functional networks continue to develop at a different maturation rate (Gao et al. 2015a). Furthermore, at this early age, prenatal adversities have been associated with neural risk markers for negative developmental outcomes (Posner et al. 2016; Jha et al. 2019; Graham et al. 2021). For example, lower maternal socioeconomic status and maternal negative mood have been associated with a less mature sensorimotor network connectivity and greater resting-state FC of the amygdala and medial prefrontal cortex (PFC) (Gao et al. 2015a; Qiu et al. 2015). These neural profiles are like those found in adults who experience lower socioeconomic status and mood disorders (Sripada et al. 2014; Kaiser et al. 2015; Mulders et al. 2015). Therefore, similarity in FC between mothers and infants may help identify how the neural risks for negative health outcomes can be intergenerationally transmitted.

In recent years, several studies examined neural activation in both mothers and their children during dyadic interactions (Atzaba-Poria et al. 2017; Reindl et al. 2018). The findings of the studies have informed our understanding of how mother–child behavioral synchrony is associated with brain synchrony for infants’ socioemotional development (Bell 2020). However, because brain synchrony was assessed during active interactions or while using a stress paradigm, the design limits investigations with very young infants who cannot actively interact. Thus, it remains unknown how early a similarity in brain activation between mothers and infants may emerge. Therefore, in the current study, we assessed the similarity of resting-state FC between mothers and newborn infants.

A few existing studies suggest there is a similarity between mothers and children in their brain structure and FC. Gray matter volumes in the amygdala, hippocampus, anterior cingulate cortex, and medial PFC were positively correlated between mothers and daughters at age 5–13 (Yamagata et al. 2016). The sulcal pattern particularly in the right hemisphere was more similar between children and their mothers compared with unrelated females (Ahtam et al. 2021). Using electroencephalography, a study reported that a resting-state frontal alpha asymmetry of 12-month-old infants was positively correlated with a frontal alpha asymmetry of their mothers (Hill et al. 2020).

Importantly, studies using an intergenerational approach suggest that the neural similarity may be stronger among mother–child dyads at risk for psychopathology. For example, cortical thickness of the left fusiform was correlated between mothers with a history of depression and their never depressed daughters at ages 9–15 but not between mothers and daughters without a history of depression (Foland-Ross et al. 2016). A similar pattern was found in cortisol levels that reflect the neurobiological regulation of stress regulation across generations. The cortisol levels were correlated between mothers and their children only if mothers reported a history of childhood adversity (Fuchs et al. 2017) or a history of depression (LeMoult et al. 2015; Merwin et al. 2017). Strong similarities across generations may be evidence of early transmission of neural risk markers for psychopathology that can be detected before symptoms are observed in children.

The similarity between mothers and their children can be influenced by both genetic and environmental factors. Although most studies have used a twin study design, a few have used an intergenerational design and suggested heritability in brain structure and function (Bas-Hoogendam et al. 2018; Bas-Hoogendam et al. 2020). During the first year, the resting-state FC patterns across networks rapidly develops (Gao et al. 2015a; Gao et al. 2017; Gilmore et al. 2018). Thus, in infants, an older age is associated with a more mature pattern of the FC and likely with a greater similarity with a pattern of their mothers’ FC.

However, to date, little is known about the role of environmental factors on neural similarity across generations. Furthermore, to the best of our knowledge, no study has yet examined the similarity of FC using resting-state functional magnetic resonance imaging (fMRI). Lower maternal socioeconomic status (lower income or lower maternal education) have been associated with altered resting-state fMRI such as less mature network connectivity profiles among infants at 6 months of age, particularly in the sensorimotor and default mode network (Gao et al. 2015a). On the other hand, prenatal maternal stress such as depressive symptoms and inflammatory markers has also been associated with a resting-state network profile. Prenatal maternal stress was linked to accelerated development in the salience networks such as increased FC between amygdala and medial PFC in neonates (Qiu et al. 2015; Graham et al. 2018). In adults, low socioeconomic status and a history of psychopathology have been associated with alterations in the resting-state fMRI such as decreased default mode network connectivity and aberrant salience network connectivity (Sripada et al. 2014; Kaiser et al. 2016). Therefore, in mother–infant dyads, lower socioeconomic status or higher depressive symptoms may be linked to similarities in brain network connectivity. Strong similarity between mothers and infants may suggest the transmission of neural risk markers, and brain network patterns at this age may be a predictor of infants’ later cognitive development and behavioral regulation (Chen et al. 2021).

Thus, in the current study, we examined the similarity of resting-state FC between mothers and their infants. We focused on infants in the first few months to examine the similarity of brain connectivity between mothers and their infants soon after birth. This approach will help to identify how early the similarity of FC between mothers and infants emerges. We hypothesized that brain areas that mature earlier such as the primary sensorimotor, visual, and auditory networks would show higher similarity than other higher order functional networks at the neonatal stage (Gao et al. 2015a). We further hypothesized that the similarity in FC patterns between mothers and their infants would increase with infant age. We further investigated the role of maternal factors during the perinatal period. Specifically, we hypothesized that fewer years of maternal education and maternal depressive mood would be associated with greater similarities in FC across generations.

Materials and Methods

Participants

Pregnant women were recruited from the Department of Obstetrics and Gynecology at hospitals in the Metro Denver area. Eligibility criteria for the pregnant women were: 1) over 18 years of age; 2) singleton intrauterine pregnancy; 3) prior to 16 weeks of gestation; 4) fluency in English; and 5) a family income-to-needs ratio below 8. Exclusion criteria include: 1) current psychotropic medication use; 2) current or lifetime psychiatric/neurological illness other than depression and anxiety diagnosis; 3) maternal substance use except for occasional use of alcohol, cigarettes, or cannabis (assessed using maternal reports and urine toxicology); 4) obstetric risk conditions such as systemic maternal disease, placental or cord abnormalities, uterine anomalies, infection, chromosomal abnormalities; 5) corticosteroid medication usage during their pregnancy; or 6) nonremovable ferromagnetic metal in or on the body (for safety in the magnetic resonance imaging [MRI] scanner). We included participants with a history of depression and anxiety for a representative community sample, as they are the 2 most common diagnoses of mental disorders during the perinatal period (Dennis et al. 2017; Shorey et al. 2018; Meltzer-Brody and Rubinow 2021). Because the overarching goal of the research project was to examine the role of stress exposure in low- and middle-income women, very high-income women were excluded. After the infants were born, additional exclusion criteria for newborns include: 1) major delivery or neonatal complications, congenital, genetic, or neurologic disorders and 2) nonremovable ferromagnetic metal in or on the body.

A total of 65 typically developing infant–mother dyads participated in the neuroimaging sessions. Of those, 46 infants provided neuroimaging data. The other infants either did not fall asleep or woke up before any neuroimaging sequence started. Among the 46 infants and 65 mothers, resting-state fMRI data of 7 infant participants and 14 mother participants did not survive the fMRI data quality control (detailed below in resting-state fMRI data processing). Of the remaining 39 infant and 51 mother participants, 30 mother–infant dyads (infant and his/her own mother) were identified. The data have not been included in other publications.

Procedures

The overall research study had 4 home visits (12–16, 22, 32 weeks of pregnancy, and 2–4 weeks of postpartum) and 1 neuroimaging visit. At each home visit, 2 trained research assistants visited the participant’s home. At the initial home visit, participants were provided information about the study and the informed consent. The procedures were undertaken with the understanding and written consent of each participant. The visit included self-report questionnaires on maternal mood and an interview measure of demographic information and life events. Soon after the postnatal home visit, the mother–infant dyads participated in the neuroimaging portion of the study located at the Center for Innovation & Creativity at the University of Colorado—Boulder. The visit included fMRI scan sessions (1.5 h for mother and 1 h for infant from entry to completion). The visits typically occurred around the infant’s regular afternoon nap time or the infant’s regular evening sleep time. The scanner room was equipped with a rocking chair for mothers to facilitate getting their infant to fall asleep. A trained researcher always remained inside the scanning room in case the infant woke up during the scan. If the infant woke up during the scan, scanning was immediately stopped, and the infant was brought out of the MRI scanner. Childcare support was provided, and monetary compensation was provided for the participant’s time and participation at the end of each visit. All procedures were approved by the University of Denver Institutional Review Board.

Imaging Acquisition

All MRI data were collected on a Siemens 3 T MAGNETOM Prisma scanner (32-channel head coil). Anterior-posterior/posterior-anterior (AP/PA) resting-state fMRI data were acquired using a T₂^*-weighted echo planar imaging (EPI) sequence: time repetition (TR) = 864 ms, time echo (TE) = 42.20 ms, 72 slices, voxel size = 2 × 2 × 2 mm, 210/517 volumes. Structural images were acquired using a 3D magnetization-prepared rapid acquisition gradient-recalled echo sequence: TR = 3200 ms, TE = 564 ms, voxel size = 1 × 1 × 1 mm for infants’ T₂-weighted images; TR = 2400 ms, TE = 2.22 ms, inversion time = 1000 ms, voxel size = 0.8 × 0.8 × 0.8 mm for mothers’ T₁-weighted images.

Infant and Maternal Factors

Infant sex, gestational age at the scan, and other demographic information were reported by mothers at the postnatal visit (Table 1). Maternal demographic information was based on the maternal report at the first prenatal home visit and a postnatal home visit. Maternal education levels were based on years of schooling at the time of the first prenatal home visit. Maternal depressive mood was assessed using the Edinburgh Postnatal Depression Scale (EPDS) (Cox et al. 1987). The self-reported questionnaire included 10 items rated from 0 to 3 (e.g., 0 = most of the time, 3 = no, never). The items reflect the depressive symptoms that participants have felt in the past 7 days. The EPDS has been used extensively to assess prenatal and postnatal depressive symptoms (Cox et al. 1987; Harris et al. 1989; Murray and Carothers 1990; Sit and Wisner 2009). The questionnaire was administered at every home visit. The prenatal depressive symptoms were an average score across the 3 prenatal visits and the postnatal depressive symptoms were the score at the time of the postnatal visit. The overall levels of depressive symptoms were consistent across the pregnancy and postnatal visits (Table 1).

Table 1.

Characteristics of the 30 mother–infant dyads

	N (%)	Mean ± standard deviation	Range
Child characteristics
Gestational age at birth (weeks)	—	39.25 ± 1.22	37.14–41.57
Birth weight (pounds)	—	6.88 ± 0.99	5.13–9.19
Infant age at MRI (days)	—	34.17 ± 17.65	9–79
Infant sex (female)	14 (46.7)	—	—
Maternal characteristics
Age (years)	—	29.88 ± 6.11	19–42
Education (years)	—	15.29 ± 2.84	11–20
Income-to-needs ratio	—	3.14 ± 2.23	0.00–8.81
Race/Ethnicity
White non-Hispanic	17 (56.7)	—	—
Hispanic	8 (26.7)	—	—
Native Hawaiian or Other Pacific Islander	1 (3.3)	—	—
Multiracial	4 (13.3)	—	—
Depressive symptoms (EPDS)
Prenatal period	—	4.67 ± 3.68	0.33–14.33
Postnatal period	—	4.47 ± 4.24	0–15
History of depression or anxiety diagnosis	15 (50.0)	—	—
Anxiety and depression medication use during pregnancy (yes)	5 (16.7)	—	—
Interval between home and fMRI visits (days)	—	14.93 ± 11.72	2–53
Right handedness	23 (76.7)	—	—

Resting-State fMRI Data Preprocessing

Functional imaging data were preprocessed using FMRIB’s Software Library (Smith et al. 2004) and analysis of functional neuroimages (Cox 1996). Slice-timing correction was not implemented due to the short TR (Wu et al. 2011). Preprocessing included rigid-body motion correction, registration, bandpass filtering (0.01–0.08 Hz), scrubbing, nuisance signal regression, AP/PA concatenation, truncation, and global signal regression. For motion correction, the single-band reference image served as the target image. Frame-wise displacement (FD) was estimated from the 6 motion parameters (displacements and rotations). The full functional datasets were aligned to the template image using per-volume transformations; rigid-body within functional + nonlinear distortion correction + nonlinear functional-to-anatomical + nonlinear anatomical-to-standard. Both nonlinear registrations used advanced normalization tools (ANTs) (Avants et al. 2008). The Montreal Neurological Institute (MNI)-152 adult template (Grabner et al. 2006) served as the standard target for mothers. For infants, the neonate-specific template (Shi et al. 2011) served as an intermediate target space, and after the above-mentioned steps, the infants’ data in the neonate-specific template space were aligned to the MNI-152 adult template space using ANT to prepare for the BrainSync analysis between infants and mothers. Each subject was visually checked to ensure successful registration.

The DVARs calculation was done immediately after the alignment to anatomical space, which is a measure of how much the intensity of a brain image changes in comparison to the previous time-point (Power et al. 2012). Data scrubbing was performed as an additional motion correction step other than the standard rigid-body motion correction procedures; volumes with DVARs signal changes higher than 0.5% or FD higher than 0.5 mm were removed (“scrubbed”) from the data; one volume immediately preceding and 2 volumes following the scrubbed volume were also removed (Power et al. 2012). Participants (infant N = 7, mother N = 14) with <360 volumes remaining (5.184 min) were excluded from further analysis.

The nuisance signal regression model included 32 parameters (32P); 8 regressors corresponding to white matter and cerebral spinal fluid signals, as well as the 6 motion estimates, plus their derivative, quadratic, and squared derivative terms (Power et al. 2014). The data were warped to standard space, spatially smoothed using a Gaussian kernel of 6-mm full width at half maximum. For subjects with >1 AP/PA fMRI data, all successfully processed AP/PA sessions were concatenated into 1 session. Then, all fMRI data from mothers and infants were truncated to 360 volumes to improve consistency across subjects and facilitate the pair-wise BrainSync analysis as described below. Lastly, global signal regression was performed to regress out the mean gray matter signal from the data.

fMRI Data Analysis

BrainSync Analysis

BrainSync (Joshi et al. 2018) is a novel method capable of synchronizing rsfMRI time series across subjects and/or sessions so that similarities in FC between the 2 subjects/sessions can be quantified through calculating correlations of the synchronized time series. It uses an orthogonal transformation to synchronize rsfMRI time-series so that they become similar at the same voxels across subjects and/or sessions when their connectivity patterns are similar. Based on this property, we used BrainSync to synchronize time series of 2 subjects (infant and his/her mother or other mother) and voxel-wise Fisher’s Z transformed correlation was calculated to quantify similarities of FC pattern across the whole brain between infants and mothers after synchronization. An additional temporal nonlocal means filtering (Bhushan et al. 2016) and normalization were performed as presteps for BrainSync transformation, as recommended in Joshi et al. (2018).

Relationship of Similarity and Infant and Maternal Factors

We first examined the relationship between mom–infant whole-brain similarity and demographic variables including maternal age, race, education, depressive symptoms, and infant sex, and age at scan. Pearson correlation coefficients were calculated between whole-brain similarity and each demographic variable across mom–infant dyads (N = 30). The whole-brain similarity value was demonstrated by the mean similarity value across the whole-brain voxels. Considering the effect of other demographic variables, the partial correlation between whole-brain similarity and each variable of interest (infant age, maternal education, maternal depressive symptoms) was calculated to control for the effect of potentially confounding variables, for example, maternal age and race.

Permutation tests were performed to test the significance of correlations and partial correlations. Specifically, infant and maternal factor values across mom–infant dyads were permutated 1000 times to generate 1000 correlation values between similarity and permutated demographic variable to form the null distribution. If the “true” correlation was positive, the number of cases where correlation values in the null distribution > “true” correlation among 1000 permutations was counted; if the “true” correlation was negative, the number of cases where correlation values in the null distribution < “true” correlation among 1000 permutations was counted. The P value of the permutation test was obtained by the counted number/1000.

To examine the network-level FC patterns, seed-based FC analyses of 8 regions of interest (ROIs) including orbital frontal cortex (OFC; attachment network), posterior cingulate cortex (PCC; the default-mode network), amygdala (emotion regulation network), anterior insula (salience network), left superior temporal gyrus (auditory/language network), left precentral gyrus (sensorimotor network), right cuneus (visual network 1), and left inferior occipital gyrus (visual network 2) were defined for later network-level analysis. Among them, the OFC was selected as a combined region of 3-sphere areas of 6-mm radius, centers of which were defined using coordinates from the previous study (Nitschke et al. 2004). PCC, amygdala, and anterior insula (anterior part of regular insula region) were selected using corresponding bilateral regions in the automated anatomical atlas (AAL) template (Tzourio-Mazoyer et al. 2002). The left superior temporal gyrus, left precentral gyrus, right cuneus, and left inferior occipital gyrus were selected as sphere regions of 8-mm radius, centers of which were defined as the maximum point of the corresponding independent component analysis (ICA) network map (Smith et al. 2009). To derive network masks, ROI-based FC maps were obtained for each ROI using all mother subjects (N = 51). Specifically, mean fMRI time series were extracted from 8 ROIs. For each ROI, the mean time series from ROI were correlated with that of every other voxel in the brain, then the resulting correlation values were normalized using Fisher’s Z transformation. Network masks corresponding to 8 ROIs were derived from ROI-based FC maps through a t-test (voxel-wise P < 0.001, cluster correction alpha <0.05, t > 0). Consistent with whole-brain values, the network-level similarity was calculated as the mean similarity value across the voxels within the network mask.

The relationship between the similarity level and maternal factors—maternal education levels and prenatal depressive symptoms—was examined at the whole-brain as well as network levels. The whole-brain level analysis of the associations with maternal education levels was performed controlling for infant age at scan, maternal age, maternal race, and prenatal depressed mood. The whole-brain level analysis of the associations with prenatal depressive symptoms was performed controlling for infant age at scan, maternal education level, and postnatal depressed mood. If the whole-brain level analysis was significant, we explored the associations of the maternal factor in the network-level analysis. At the network-level analysis, significance was defined as P < 0.05 after false discovery rate (FDR) correction of 8 networks.

Potential sex-specific effects in resting-state fMRI and the impact of prenatal stress on infant’s brains by sex have been suggested (Wheelock et al. 2019). Thus, we examined the interaction between infant sex and the association among infant and maternal factors and the FC similarity. Finally, to further understand effects on FC similarity, the effects of maternal factors on the network-level FC patterns in mothers and infants were also examined.

Results

Demographic Variables and Correlations

The correlation table for the demographic variables is shown in Table 2 for the 30 mother–infant dyads. Maternal education (in years) was negatively correlated with infant age at MRI, r = −0.76, P = 0.0001, and positively correlated with maternal age at scan, r = 0.50, P = 0.02. Maternal education was also positively correlated with gestational age at birth (in weeks), r = 0.57, P = 0.01, and positively correlated with income-to-needs ratio, r = 0.70, P = 0.0007, and anxiety or depression medication use during pregnancy, r = 0.47, P = 0.04.

Table 2.

Correlation table of the demographic variables for the sample (N = 30)

	GA at birth (weeks)	Birth weight (lbs)	Infant age at MRI (days)	Infant sex (female)	Maternal age at scan (years)	Maternal education (years)	INR	Race/ethnicity	EPDS (prenatal)	EPDS (postnatal)	History of diagnosis	Medication (yes)	Interval between home and fMRI visit (days)
GA at birth (weeks)
Birth weight (lbs)	0.36
Infant age at MRI (days)	−0.36	0.10
Infant sex (female)	0.22	−0.30	−0.28
Maternal age at scan (years)	0.14	−0.10	−0.48*	−0.11
Maternal education (years)	0.57*	0.03	−0.76***	0.27	0.50*
Income-to-needs ratio	0.33	−0.18	−0.70***	0.35	0.54*	0.70***
Race/Ethnicity	−0.14	0.19	0.07	−0.29	−0.32	−0.27	−0.42
EPDS (prenatal)	0.19	0.12	−0.34	0.44	−0.38	0.16	−0.05	0.40
EPDS (postnatal)	0.09	−0.15	−0.17	−0.14	−0.04	0.00	−0.08	0.10	0.12
History of diagnosis	0.48*	0.54*	−0.07	−0.16	−0.05	0.18	0.08	0.07	0.09	−0.03
Medication (yes)	0.38	0.20	−0.33	0.25	−0.33	0.47*	−0.04	0.33	0.62**	0.11	0.22
Interval between home and  fMRI visit (days)	0.02	0.18	0.70***	−0.32	−0.31	−0.40	−0.45	0.04	−0.34	−0.07	0.33	−0.25
Right handedness	−0.09	−0.30	0.03	−0.02	−0.31	0.04	−0.02	0.05	−0.04	−0.06	0.02	0.08	0.17

^*p < 0.05, ^**p < 0.01, ^***p < 0.001

Mother–Infant Connectivity Similarity Analysis

Based on 30 mother–infant dyads, the whole-brain mother–infant similarity level was positively associated with infant age at scan (Figs 1 and 2). Figure 1a depicts the global patterns of mother–infant FC similarity. The higher FC similarity between mothers and infants was shown in the brain areas that mature earlier such as the sensorimotor, visual, and auditory networks (Gao et al. 2015a). At the network level, the mean correlation values were significantly higher in those networks compared with the networks that mature relatively later including the salience, attachment, or default mode networks, Ps < 0.05, FDR corrected (Fig. 1b,c).

Figure 1.

Global mother–infant similarity pattern. (a) Whole-brain mean correlation map (Fisher-Z transformed) between infants and their matching mothers after BrainSync synchronization. The red and the yellow area means high correlation (more similarity), whereas the blue and green area means low correlation (less similarity). (b) Network-level ranking of mean correlation (Fisher-Z transformed) between infants and their matching mothers after BrainSync synchronization. Mean values are plotted with standard error means. Significant differences between top 4 networks and other networks after FDR correction are labeled (^*P < 0.05). Primary networks (sensorimotor, visual 1, visual 2, and auditory networks) show more similarity than high-level networks (emotion regulation, default mode, attachment, and salience networks). (c) ROI-based FC maps of 8 networks (described in b) using all 51 mothers’ fMRI data with voxel-wise cutoff of P = 0.001 and cluster-size correction of α = 0.05. The red area means higher FC, whereas the yellow area means lower FC.

Figure 2.

A scatterplot of a correlation between whole-brain similarity levels and infants gestational age at scan (r = 0.40, P = 0.01).

To further understand the developmental pattern of the FC similarity, we analyzed the associations between FC similarity and infant age. As shown in Figure 2, infant age was positively associated with higher mother–infant similarity, r = 0.40, P = 0.01. When the correlations between mother–infant similarity and infant age were examined in each network, a positive correlation was confirmed in the visual network, P < 0.05, FDR corrected. We also conducted a test of the interaction between infant sex and age, but the interaction term was not significant.

Maternal Factors and Mother–Infant Connectivity Similarity

We explored the relationship between whole-brain similarity level and maternal factors—maternal education levels and depressive symptoms during the prenatal period. Maternal education was negatively associated with whole-brain similarity level after controlling for gestational age at scan, maternal age, maternal race, and maternal prenatal depressive mood, r = −0.33, P = 0.047 (Fig. 3a). This relationship suggests that infants from higher education mothers show lower levels of FC similarity with their mothers compared with those from lower education moms.

Figure 3.

(a) A scatterplot of a correlation between whole-brain similarity levels and maternal education, r = −0.33, P = 0.047. Covariates included infant gestational age at scan, maternal age, maternal race, and maternal prenatal depressive symptoms. (b) Scatterplots of correlations between maternal education levels and the FC similarity between mothers and infants in the emotion regulation, salience, auditory, and sensorimotor networks, Ps < 0.05, FDR corrected.

At the network level, negative correlations were observed in the 4 networks—emotion regulation (amygdala), salience (anterior insula), auditory, and sensorimotor networks, P < 0.05, FDR corrected (Fig. 3b; Supplementary Table 1). In these networks, similar to the whole-brain FC similarity analysis, lower maternal education level was associated with higher mother–infant FC similarity.

Maternal prenatal depressive symptoms were not significantly associated with the whole-brain similarity level controlling for gestational age at scan, maternal education, and postnatal depressive symptoms. The moderating effect of infant sex was tested at the whole-brain level as well as the network-level analysis. The interaction term between infant sex and maternal education was not detected in either analysis.

Last, the main effect of maternal factors on the FC patterns was examined separately in mothers and newborns. At the network-level analysis, neither maternal education levels nor prenatal depressive symptoms revealed significant effects on the FC patterns in mothers and newborns.

Discussion

Overall, the findings of the current study suggest that similarity in FC between infant and mother can be identified within the first few months after birth. First, across the brain networks, we found those that matured more rapidly such as the sensorimotor, auditory, and visual networks exhibited, on average, a greater similarity between mothers and infants (Gao et al. 2015b). Furthermore, our results show that similarity between mothers and infants increases with the infant’s age, suggesting that FC may become more similar between generations as an infant’s brain develops. Next, when we examined the role of maternal factors, we found that fewer years of maternal education were associated with a greater FC similarity between mothers and infants. Prenatal maternal depressive symptoms were not associated with the FC similarity. One potential interpretation is that lower maternal education level may influence the infant’s brain development, influencing developmental trajectories to cause premature similarity to the mothers’ brain profiles. This may be a mechanism by which neural risks for negative cognitive and emotional outcomes are transmitted across generations. The current study is one of the first studies to provide evidence that the maternal environment may be associated with FC similarity between mothers and newborns. A longitudinal design is needed to understand the impact of intergenerational similarity on the next generation’s brain and behavioral development at later ages.

In the current study, within the first few months of age, a whole-brain similarity pattern featuring higher similarity in primary networks (i.e., sensorimotor, visual, and auditory) compared with higher order networks was observed. This finding was in line with our previous results reporting the developmental sequence of infant brain functional networks from primary to higher order networks (Gao et al. 2015b). This pattern was expected since primary networks such as sensorimotor, auditory, and visual networks already show relatively mature FC profiles at birth, and they also develop rapidly during the first year after birth (Gao et al. 2015a; Gao et al. 2017; Gilmore et al. 2018). Thus, in infants, the relative maturity in these networks may contribute to the higher similarity with their mothers’ FC. The functions of the primary networks are critical for infant survival and provide the foundation for further cognitive and emotional development. On the other hand, the higher order networks likely need more postnatal experience-dependent growth and enrichment from mom–infant interactions than the primary networks, to become more similar with moms’ functional configurations. Our finding that infant age was positively associated with overall whole-brain similarity between mothers and infants underscores the general importance of postnatal environment/experience for all the brain’s functional networks, which likely reflects an evolutional optimization to allow experience- and mom-dependent fine-tuning during this process to better adapt to the postnatal environment.

Next, we examined whether the similarity between mothers and infants is associated with maternal factors, specifically maternal education and prenatal depressive symptoms. Maternal educational level was not directly associated with FC patterns in mothers and infants. However, we found that infants of mothers whose education levels were lower showed greater similarity in the overall FC with their own mothers. It is important to note, in the current study, the range of maternal education levels (ranged 11–20 years) included a mid to high range of education and did not include the low range of the education levels. However, previous studies suggest that maternal education levels with a similar range (i.e., mid to high education levels) were associated with children’s brain development (Noble et al. 2012; Lawson et al. 2013; Greenwood et al. 2021).

The greater FC similarity that was linked to lower maternal education may be consistent with previous studies of mother–child neurobiological similarity. Brain structure and cortisol levels were correlated between mothers and their children only when mothers reported childhood adversity (Fuchs et al. 2017). Moreover, in this context, the higher FC similarity between mothers and infants may implicate accelerated developmental trajectories in infants (Callaghan and Tottenham 2016). On the other hand, the lower similarity exhibited by infants of mothers with higher education levels may afford more opportunities for postnatal environmental input that may subsequently lead to better adaptation. Slower developmental trajectories in brain structure have been found in children with higher intelligence (Shaw et al. 2006).

When we further evaluate the relations of maternal education level in each network, stronger negative correlations were observed in auditory and sensorimotor networks and the salience and emotion regulation networks. The auditory and sensorimotor networks are involved in language and motor development that further support cognitive development in infants (de Bie et al. 2012). The salience and emotion regulation networks are involved in emotional response and regulation that further support socioemotional development in children (Menon 2011). Low maternal education levels have been associated with delayed language and motor development and greater difficulties in emotional and behavioral regulation in children (Noble et al. 2015; Farah 2018; Kim et al. 2018). Therefore, the current study provides evidence that previous findings of greater similarity in the neurobiological systems of at-risk dyads of mothers and their children may begin in early infancy. The high FC similarity may be a mechanism by which neural markers associated with environmental adversities such as low socioeconomic status may be transmitted to the next generations.

Although not examined directly in the current study, there are several potential mechanisms by which lower maternal education levels become associated with greater FC similarity between mothers and infants and less optimal brain development in infants. Mothers with lower education levels are more likely to be exposed to childhood adversity (Font and Maguire-Jack 2016; Mitchell et al. 2018) and higher levels of prenatal life stressors (Ward et al. 2017). The high level of stress in mothers may negatively influence fetal brain development trajectories via prenatal stress physiology (e.g., inflammatory system, cortisol) (Buss et al. 2012; Moog et al. 2018; Graham et al. 2019). Thus, the relatively higher exposure to childhood or prenatal stress may be a mechanism by which lower maternal education is transmitted to the next generation. During the postnatal period, low maternal education levels have been associated with higher parenting stress and lower cognitive stimulation in the home environment (Hoff-Ginsberg and Tardif 1995; Brooks-Gunn and Markman 2005). Thus, lower maternal education and relevant stress can lead to less optimal parenting styles that may further influence brain development among older infants.

The findings should be interpreted with caution. First, the current study design does not allow us to determine the extent to which the FC similarity is due to genetic versus environmental factors. The greater FC similarity between mothers and children can reflect increased heritability. To delineate the 2 types of factors, cross-fostering designs such as children born via in vitro fertilization with surrogate parents may be required (Rice et al. 2009). These kinds of studies would be particularly important to identify the genetic versus environmental mechanisms by which risks for psychopathology may be transmitted across generations (Toth 2015). Second, there was the limited range of depressive symptoms and maternal education levels in the sample. Few mothers with high depressive symptoms may be why the current study did not identify the impact of maternal prenatal or postnatal mood on brain similarity. The sample also did not include mothers with very low education levels. Therefore, the findings of the current study should be interpreted with caution that the maternal factors do not represent severe adversity during the prenatal period. Thus, a future study with a larger sample including high depressive symptoms and/or low maternal education levels would be needed to examine whether high maternal adversity may be associated with the similarity between mothers and infants. Third, the current study did not include fathers. Some studies suggest stronger structural correlations between mothers and daughters compared with other dyads (mother–son, father–son, father–daughter) (Yamagata et al. 2016). The inclusion of fathers will support a more complete understanding of how neural profiles may be transmitted to the next generation from parents (mothers and fathers) to their children.

Fourth, infants were sleeping while mothers were awake during the scans. Different resting-state connectivity while being awake versus asleep has been reported in adults (Tagliazucchi and Laufs 2014). FC during sleep in infants has been observed to be more similar to FC during sleep in adults as compared with FC during wakefulness (Mitra et al. 2017). Thus, future studies should match the sleep versus wakefulness status between parents and infants. Fifth, the current study had a cross-sectional study design and included a relatively small sample and a relatively wide range of infant age. Thus, the associations between maternal factors, infant age, and the FC similarity between mothers and infants could not convey the directionality. The information on whether the similarity between mothers and infants would be associated with infant cognitive or behavioral outcomes is unknown. The larger sample will also help to examine the unique role of the heritability and environmental factors including the analysis of the similarity between mothers and own infants compared with the similarity between unrelated mothers and infants. Therefore, long-term follow-ups with a larger sample of mother–infant dyads are critically needed to understand the prospective link among maternal genetic and environmental factors, brain FC similarity, and later infant outcomes.

In the current study, we examined the intergenerational similarity of resting-state FC. Mothers and their biological infants become more similar in their FC patterns as the infant grows older. Mother–infant synchrony has been suggested to be important for mother–infant attachment as well as the infant’s brain and cognitive development (Feldman 2007; Reindl et al. 2018; Bell 2020). It is around 3 months after birth that infants become more interactive with their mothers. The current study provides evidence that the increasing intrinsic FC similarity between mothers and infants during the first 3 months may support mother–infant brain and behavioral synchrony. Furthermore, lower maternal education level was associated with greater similarity between mothers and their infants. Greater similarities suggest that adverse environments may accelerate brain development in infants that may potentially limit brain plasticity at an early age. Thus, the current findings provide evidence of how neural profiles may be transmitted to the next generation, which was observed during the first few months after the infant’s birth. The findings highlight the importance of prevention and intervention for mothers who live in adverse environments during the perinatal period to reduce the potential transmission of neural risks across generations.

Notes

The authors thank the families that participated in the study and the individuals that supported recruitment. The authors also acknowledge Brian Bello, Samantha Buxbaum, Christian Capistrano, Ximena Calderon, Christina Congleton, Andrew Erhart, Leah Grande, Hana Gulli, Melissa Hansen, Akram Ibrahim, Isabella Jaramillo, Claire Jeske, Jacqueline Martinez, Aviva Olsavsky, Shannon Powers, and Chhorda Vuth for research assistance. Conflict of Interest: None declared.

Funding

The National Institute of Health [R01HD090068; R21HD078797; R21DA046556; R01DA042988, R01DA043678]; the National Center for Advancing Translational Science (NCATS) [CTSA TL1 TR001864]; The Brain & Behavior Research Foundation’s NARSAD Independent Investigator Grant. Victoria S. Levin Award for Early Career Success in Young Children’s Mental Health Research, Society for Research in Child Development (SRCD) CTSA (TL1TR001864 to A.J.D.) from the National Center for Advancing Translational Science (NCATS), a component of the National Institutes of Health (NIH).