Altered cortical structure and psychiatric symptom risk in adolescents exposed to maternal stress in utero: A retrospective investigation

Abstract

Maternal exposure to stress during pregnancy is associated with increased risk for cognitive and behavioral sequelae in offspring. Animal research demonstrates exposure to stress during gestation has effects on brain structure. In humans, however, little is known about the enduring effects of in utero exposure to maternal stress on brain morphology. We examine whether maternal report of stressful events during pregnancy is associated with brain structure and behavior in adolescents.

We compare gray matter morphometry of typically-developing early adolescents (11–14 years of age, mean 12.7) at a single timepoint, based on presence/absence of retrospectively-assessed maternal report of negative major life event stress (MLES) during pregnancy: prenatal stress (PS; n = 28), comparison group (CG; n = 55). The Drug Use Screening Inventory Revised (DUSI-R) assessed adolescent risk for problematic behaviors. Exclusionary criteria included pre-term birth, low birth weight, and maternal substance use during pregnancy. Groups were equivalent for demographic (age, sex, IQ, SES, race/ethnicity), and birth measures (weight, length).

Compared to CG peers, adolescents in the PS group exhibited increased gray matter density in bilateral posterior parietal cortex (PPC): bilateral intraparietal sulcus, left superior parietal lobule and inferior parietal lobule. Additionally, the PS group displayed greater risk for psychiatric symptoms and family system dysfunction, as assessed via DUSI-R subscales.

These preliminary findings suggest that prenatal exposure to maternal MLES may exact enduring associations on offspring brain morphology and psychiatric risk, highlighting the importance of capturing these data in prospective longitudinal research studies (beginning at birth) to elucidate these associations.

1. Introduction

A large body of animal research has demonstrated effects of prenatal stress on the structure of the developing brain in offspring, including preclinical [1–3] and non-human primate studies [4,5]. In utero stress exposure impacts development of the corpus callosum [4], hippocampus [5–7], amygdala [8], rostral anterior cingulate cortex [9], and frontal cortex [10]. Corresponding closely with the functional neuroanatomy of these regions, offspring of animals stressed during gestation exhibit deficits in attention [11], as well as an increase in behaviors associated with anxiety [12,13] and depression (e.g., altered circadian rhythms [14] and learned helplessness [15]). Critically, the effects of prenatal stress endure throughout the lifespan of the animal [16].

As in animals, research in humans has demonstrated that maternal stress during pregnancy is associated with adverse cognitive and behavioral outcomes in childhood, adolescence, and adulthood. These studies operationalize stress in various ways, including major life events [17], natural disasters [18,19], military invasion [20], and maternal psychiatric symptoms, such as depression [21] and anxiety [22,23]. Despite differences in methodologies and specific findings, research reveals a largely consistent picture showing maternal stress experienced during pregnancy is associated with an increased risk for cognitive, behavioral and mental health difficulties among offspring [24].

While associations of maternal stress with cognition and behavior are thoroughly documented, studies have only recently examined associations with child neurodevelopment. These nascent studies show that in utero stress is associated with differences in white matter organization [25,26], and gray matter development in cortical [27–29] and subcortical [30] structures. For instance, in neonates, prenatal maternal anxiety is associated with reduced cortical thickness (CT) in frontal and parietal cortices, an association modulated by genetic variants implicated in increased vulnerability to psychiatric disorders [31]. Impacts of maternal stress during gestation have also been observed in older offspring. Maternal reports of stress during pregnancy are associated with differences in white matter microstructural organization of the uncinate fasciculus in offspring aged 6–9 years [25]. Further, studies of 6–9 year-olds have shown that greater maternal pregnancy-specific anxiety at 19-weeks gestation is associated with reduced gray matter density in bilateral prefrontal cortex, temporal and parietal cortices, and cerebellum [27], and greater maternal depressive symptoms are associated with thinner cortices across the brain [28].

Adolescence is a sensitive period for brain development, during which structural and functional changes and enhanced neuroplasticity confer advantages (e.g., facility in skill learning) [32], while also presenting potential vulnerabilities (e.g., risk for psychopathology) [32,33]. Studies of normative development demonstrate decreases in cortical gray matter volume and CT during the pre- and peri-adolescent periods [34,35], through adolescence [36,37] and into young adulthood [38]. Importantly, we expect gray matter volume to continue to decrease in adolescence [39,40]. Further, accelerated cortical thinning during adolescence (relative to both childhood and young adulthood) has been reported [41], although cortical thickness and gray matter density or volume are not equivalent [42].

The neuromaturational events during adolescence suggest this developmental period may be a relatively sensitive window into earlier neurodevelopmental influences. That is, previously masked differences may be revealed by developmental processes occurring during this time [43]. Understanding associations between exposure to maternal prenatal stress and brain structure during adolescence may be particularly important, as many types of mental health problems emerge in adolescence [44]. Adversity early in life may set forth a “cascade” of effects in the developing brain [45–47], including those at the molecular, cellular and network levels, which may manifest only during a later developmental period [48–50], such as adolescence.

Although the precise mechanisms by which gestational stress increases risk for psychiatric symptoms and is associated with human neurodevelopment have yet to be elucidated [51–53], both the animal and human literature demonstrate that prenatal stress is associated with differences in offspring brain structure and function, and can play a contributory etiological role in cognitive and behavioral sequelae. Identifying differences in offspring brain structures associated with prenatal stress exposure, as well as the behavioral and psychiatric outcomes of this exposure, may inform identification of youth at elevated risk for poor outcomes [54].

The current study examined cortical gray matter density at a single time point in two groups of typically developing, early adolescents that differed with regards to retrospective maternal report of negative major life event stress (MLES) during the index pregnancy. We hypothesized that adolescents whose mothers experienced MLES during their pregnancy would have a neuroanatomical profile different from a comparison group whose mothers reported no MLES during their pregnancy. Although we did not predict specific regional differences a priori, we anticipated that the neuroanatomical profile of youth exposed to maternal MLES in utero would be suggestive of delayed gray matter development relative to their counterparts. This hypothesis is in line with existing studies in younger cohorts that are suggestive of delayed offspring brain development in the context of maternal distress during pregnancy [25,27,28,31,55]. Testing the question of delayed versus accelerated cortical development per se, however, is an empirical question that can only be addressed by prospective longitudinal data rather than cross-sectional data, as presented in the current study.

2. Methods and materials

2.1. Participants

Participants were recruited as part of the Adolescent Development Study, a prospective investigation of the neural, cognitive, and behavioral precursors and consequences of substance use initiation and escalation. Demographic, cognitive, behavioral, and imaging assessments were conducted at baseline and were repeated during two follow-up visits, approximately 18- and 36-months later. The data reported here were collected during the “baseline” visit. Complete information on the methods of this study are described in detail elsewhere [56]. Briefly, 135 substance-naïve 11–13-year-olds and their caregivers were recruited from the Washington D.C. metropolitan region. Primary exclusionary criteria were prior substance use, neurodevelopmental disorder (e.g., autism spectrum disorder, active tic disorder that would interfere with imaging), prior head trauma, left-handedness, and conditions contraindicated in MRI. The Georgetown University Institutional Review Board approved the study, and caregivers and adolescents provided consent and assent prior to all data collection.

2.2. Family/caregiver measures

2.2.1. Pre-/Perinatal questionnaire

Biological mothers were invited to complete a pre-/perinatal questionnaire retrospectively collecting information about the pregnancy with, and delivery of, their enrolled adolescent. Mothers were asked about negative MLES occurring during the index pregnancy. Questions concerning stressors were adapted from the second item of a short screening scale for posttraumatic stress disorder (PTSD) [57]. Items from the scale probing PTSD symptom domains (e.g., re-experiencing of traumatic event(s), exaggerated startle response, etc.) were omitted. The following stressful major life events were queried: relationship conflict (e.g., divorce, break up, infidelity), death of someone close (e.g., partner, parent, another child), severe illness/harm of someone close (e.g., cancer, heart attack, severe car accident), severe financial issues (e.g., major property damage, foreclosure, sudden unemployment), involvement in an accident (e.g., car, fire, other disaster), being physically or emotionally attacked (e.g., threatened with a weapon, rape), or another terrible event that most people do not experience (with space provided for elaboration). Mothers could indicate that they had experienced no negative MLES during pregnancy by selecting the option “nothing like this happened to me while I was pregnant with this child.”

Mothers also indicated their age at time of delivery, incidence of any obstetric complications, and extent of any tobacco, alcohol, prescription or illicit drug use during the pregnancy. Finally, outcomes after pregnancy were reported for both the mother (post-partum depression (PPD)) and child (birth weight and length). (Parity was not assessed.)

2.2.2. Family socioeconomic status index (SES)

An SES Index was calculated by averaging the mean of two standard scores (mean household income bracket before taxes and mean cumulative parental education), and re-standardizing these to obtain a z-score distribution with a 0-centered mean and a standard deviation of 1 for the sample analyzed (n = 83) (method adapted from [58]).

2.3. Adolescent measures

2.3.1. Risk measure

Adolescent participants completed the Drug Use Screening Inventory Revised (DUSI-R), a 159-item survey that measures current use of drugs and alcohol, risk for future problematic use, as well as experiences and behaviors known to precede and co-occur with substance use [59,60]. Although non-diagnostic in nature, previous research has shown the DUSI-R to be psychometrically valid for identifying adolescents at risk for adverse outcomes, including disrupted psychosocial development, psychiatric symptomatology, and antisocial behavior [61–66]. Survey questions are yes-no items categorized into ten domains: psychiatric symptoms, family system, behavior patterns, health status, leisure and recreation, social competence, school performance, peer relationships, substance use, and work performance. For each domain a problem density index was tabulated [65]. Higher scores suggest greater problems within a particular domain. Given limited variability and null responses (by design) in two of the DUSI subscales (work performance, substance use), the present analysis examines the remaining eight subscales.

2.3.2. Intelligence and physical development

Full-scale intelligence was estimated via the Kaufman Brief Intelligence Test, 2^nd edition (KBIT-2) [67]. In order to determine physical growth characteristics in the adolescent groups, pubertal development and body mass index (BMI z-score) were assessed. Participants completed the Scale of Physical Development [68], a self-report questionnaire highly correlated with Tanner stage [69]. The questionnaire evaluates the degree to which particular physical maturational changes have occurred, including changes in skin/voice, breast development and facial hair. Possible scores range from 1 (prepubertal) to 4 (postpubertal). BMI (kg/m²) was calculated using height measured via stadiometer (SECA 216 wall-mount mechanical measuring rod; triplicate measures within 0.5 cm, averaged) and weight measured via digital scale (Health-O-Meter Professional 394KLX). BMI norms for age and sex were used to determine z-scores and percentiles [70]. These three measures were collected during the baseline visit, with the exception of one adolescent whose IQ was assessed during the 18-months follow-up visit.

2.4. MRI acquisition and analysis

2.4.1. MRI data acquisition

During the baseline visit, high-resolution structural images were acquired on a Siemens TIM Trio 3 T scanner with a 12-channel head coil using a T1-weighted magnetization prepared rapid acquisition gradient echo (MPRAGE) sequence. A total of 176 sagittal slices were collected using the following parameters: TR/TE/TI = 920/2.52/900 ms, flip angle = 9°, slice thickness = 1.0 mm, FOV = 250 × 250 mm² and a matrix of 256 × 256 for an effective spatial resolution of 0.97 × 0.97 × 1.0 mm³.

2.4.2. Quality control for MPRAGEs

Quality of structural scans for inclusion in a Voxel-Based Morphometry (VBM) analysis was determined through visual inspection conducted by three independent raters blind to participant group status (PS vs CG). Structural scans were scored on a 0–5 point scale for degree of artifacts that may affect the tissue classification algorithm (i.e., gross quality problems including phase wrap, ringing, and ghosting), where 0 indicates a scan free from visible artifact and 5 represents a scan of the poorest quality due to prominence of the relevant artifact. For each of the 3 raters, a weighted summary score was calculated. Summary rating scores were found to have good to excellent agreement between raters (intraclass correlation coefficient = .914, 95% confidence interval = .834–.950 [SPSS 24 based on mean rating (k = 3), absolute agreement, 2-way mixed-effects model]) [71]. Structural scans with the highest 3-rater average weighted scores, indicating extremely poor quality, were flagged and reviewed by two of the authors (VLD and JVM), who confirmed them to be unsuitable for VBM analysis.

2.4.3. MRI data preprocessing and analyses

Preprocessing for VBM was performed in SPM 8 (http://www.fil.ion.ucl.ac.uk/spm/software/spm8/). The New Segment option with default parameters identified different tissue types, generating native space tissue segmentation and DARTEL imported gray and white matter tissue maps. The DARTEL toolbox was used to create a study-specific whole-brain template (i.e., semi-optimized), and this template was registered to MNI space (affine transform) and modulated by the determinant of the Jacobian transform to preserve tissue volume and create images corresponding to gray matter concentration or regional gray matter volume (GMV) [72]. The GMV images were smoothed using a 12 mm FWHM Gaussian kernel to reduce noise, compensate for spatial noise introduced during normalization, and improve validity of parametric statistics used [73]. To account for inter-individual differences in brain size, measures of GMV were normalized by total intracranial volume (ICV) [74].

2.5. Statistical analysis

2.5.1. Inclusion criteria for analyses

Given the potential impact of preterm birth [75], low birth weight [76–78], and exposure to alcohol and other substances in utero [79,80] on fetal and postnatal brain development and behavior, analyses were restricted to adolescents who were born at term (defined as > 37 weeks) with normal birth weight (≥ 2500 g, or 5 pounds, 8.2 ounces), and had no maternal report of tobacco, alcohol, regular prescription medication, or illicit drug use during pregnancy with the enrolled adolescent. Further, adolescent participants who reported substance use at baseline were excluded from analysis, given the impact of early substance initiation on adolescent neurodevelopment [81,82]. Finally, participants were excluded for poor image quality.

A total of 128 mothers completed the pre-/perinatal questionnaire. Based on the inclusion criteria detailed above, 45 participants were excluded from analyses (Supplemental Table S1). Of the remaining 83 participants with complete pre-/perinatal history, 55 (66%) mothers reported having experienced none of the queried stressful events during the index pregnancy, and 28 (34%) reported having experienced at least one stressful major life event during pregnancy (Table 1). Given the large number of respondents reporting no exposure to the queried stressful events, groups were dichotomized into those whose mothers reported experiencing at least one negative MLES during pregnancy (Prenatal Stress, PS; n = 28); and those whose mothers did not report experiencing an MLES during pregnancy (Comparison Group, CG; n = 55).

Table 1.

Summary of reported major life events occurring during pregnancy in the PS group.

Event description	# Reports
Relationship conflict (e.g., divorce, break up, infidelity)	10
Death of someone close (e.g., partner, parent, other child)	5
Severe illness/harm of someone close (e.g., cancer, heart attack, severe car accident)	4
Severe financial issues (e.g., major property damage, foreclosure, sudden unemployment)	2
Involvement in an accident (e.g., car, fire, other disaster)	3
Being physically or emotionally attacked (e.g., threatened with a weapon, rape)	4
Other terrible experience that most people never go through.	7
Total negative major life events reported	35

Seven mothers (CG = 4; PS = 3) reported the occurrence of obstetric complications during pregnancy with the enrolled adolescent (Table 2). The mean birth weight for these 7 participants was 7.8 pounds (range: 6.9–8.8 pounds). Given that complications were reported in both groups, and infants were all full-term and born with normal birth weight [83], these adolescents were included in analyses. Finally, 5 mothers (CG = 2; PS = 3) reported having been diagnosed with postpartum depression (PPD) after birth of the enrolled adolescent. As no additional details concerning PPD diagnosis were collected, and the number of diagnosed mothers was small and distributed across both groups, we included these participants in analyses reported here. Inclusion of these 11 participants (PS = 6; CG = 5) for whom PPD and/or obstetric complications were reported was to ensure adequate statistical power to detect group differences; to justify inclusion of these adolescent participants, we further provide results from a VBM analysis excluding these participants (Supplement, Section 3.2).

Table 2.

Adolescent and maternal/perinatal characteristics.

	All participants N = 83	Prenatal Stress (PS) N = 28	Comparison Group (CG) N = 55	Test statistic	p
Adolescent characteristics
Age at scan, Mean (SD)	12.7 (.7)	12.5 (.6)	12.8(.7)	t(81) = −1.90	.061
Sex				χ2(1) = 3.79	.051
Males, n (%)	38 (46%)	17 (61%)	21 (38%)
Females, n (%)	45 (54%)	11(39%)	34 (62%)
Pubertal development Mean (SD)
Girls and boys	2.27 (.72)	2.32 (.73)	2.25 (.72)	t(81) = −.40	.693
Girls	2.56 (.69)	2.93 (.64)	2.44 (.68)	t(42) = −2.09	.042a
Boys	1.97 (.54)	1.93 (.49)	2.0 (.59)	t(36) = .39	.69
BMI
z-score (SD)	.40 (.96)	.36 (1.1)	.42 (.86)	t(81) = .30	.762
Percentile (SD)	61.3 (27.5)	59.2(31.8)	62.4(25.3)	t(81) = .49	.62
Race (%)				χ2(3) = 1.26	.739
Caucasian	57%	53.5%	58.18%
African	30%	32%	29.1%
American	6%	3.5%	7.27%
Hispanic/Latino	7%	11%	5.45%
Other
Intelligence (K-BIT), Mean (SD)	108.6 (15.7)	107.1 (16.9)	109.4 (15.1)	t(81) =.63	.529
Total ICV, Cubic Liters, Mean (SD)	1.50 (.12)	1.52 (.10)	1.49 (.13)	t(81) = −1.01	.315
Total GMV, Cubic Liters, Mean (SD)	0.72 (0.06)	0.73 (0.05)	0.72 (0.06)	t(81) = −1.10	.27
Maternal/perinatal characteristics
Maternal Age at delivery, Mean (SD)	30.9 (6.1)	31.5 (7.5)	30.6 (5.3)	t(81) = −.63	.532
Elapsed time, birth and maternal survey completion Mean (SD)	13.3 (1.2)	13.1 (1.1)	13.4 (1.3)	t(81) = 1.07	.287
SES index z-score (SD) N = 80 Parental education (years, mean (SD))	16.0 (2.7)	n = 27-.237 (1.06) 16.48 (2.7)	n = 53.120 (.955) 16.48 (2.6)	t(78) = 1.5	.131
Household income (median)	$100,000 - $149,999	$50,000 - $74,999	$100,000 -$149,999
Birth weight (ounces) Mean (SD) N=79	122.8 (14.2)	n = 27 124.5 (15.1)	n = 52 121.9 (13.8)	t(77) = −.75	.455
Birth length (inches) Mean (SD) N = 59	20.3 (.96)	n = 23 20.3 (1.02)	n = 36 20.3 (.92)	t(57) = −.30	.762
Obstetric anomaly (n) Description	7 Gestational diabetes (n=4) Bleeding during 1st trimester (n=1) Breach/emergency C-section (n=1) Unspecified (n=1)	3 Gestational diabetes (n=2) Breach/emergency C-section (n=1)	4 Gestational Diabetes (n=2) Bleeding during 1st trimester (n=1) Unspecified (n=1)		–
PPD diagnosis (n)	5	3	2		–
PPD therapy	Medication (n = 2) Therapy (n = 3)	Medication (Sertraline (SSRI))(n=1) Therapy (n=2)	Medication (unknown) (n=1) Therapy (n=1)

Notes: All tests two-tailed. BMI = body mass index, ICV = intracranial volume, GMV = gray matter volume, SES = socioeconomic status, PPD = postpartum depression, SSRI = selective serotonin reuptake inhibitor.

^aSignificant at p < .05.

2.5.2. Behavioral/Demographic analyses

Variable distributions were examined prior to analyses to confirm data were free from outliers and that the assumption of normality was satisfied. The distribution of each of the eight DUSI-R subscales was non-normal (Shapiro-Wilk p < .001, df = 82, for all subscales). Because standard transformations failed to normalize distributions, non-parametric statistical tests were used for DUSI-R variables.

PS and CG groups were compared on adolescent, prenatal, and demographic characteristics using Independent Samples Student’s t-test, Mann-Whitney Utest, and Chi-square test of independence, as noted. Additionally, groups were compared on the 8 DUSI subscales using the Mann-Whitney Utest with Bonferroni adjusted alpha levels of p < .00625 per test (.05/8) to correct for multiple comparisons. Data analyses were conducted using SPSS Version 24.0.

2.5.3. Imaging analyses

Statistical modeling tested for differences in GMV [72] (measured at a single timepoint) between the PS and CG groups. Given that the distribution of males and females approximated a significant difference (p = .051) (Table 2), and because cortical thinning trajectories show sex differences [84–86], sex was included as a covariate of no interest in imaging analyses. Clusters were defined using an uncorrected threshold of p < 0.001, k = 10. Corrections for multiple comparisons were made using non-stationary cluster correction (p < 0.05) [87,88], a method which avoids known problems in the use of cluster p-values in VBM data when not accounting for non-stationarity [72,89].

3. Results

3.1. Group characteristics

Groups were similar in demographic (SES, race, and sex) and adolescent characteristics (age at MRI, IQ, pubertal development score, BMI, total intracranial volume, total GMV) (Table 2). Groups were similar in birth weight and length (Table 2), which are neonatal characteristics associated with neonatal head circumference [90] and brain size [91], suggesting groups were similar in gross brain size at birth. Further, mothers were similar in age at delivery, and groups were similar in elapsed time between the child’s birth and pre-/perinatal survey data collection. By design, the PS and CG groups differed regarding maternal retrospective report of MLES during pregnancy. Table 1 summarizes the stressful major life events reported by PS group mothers. Within the PS group, 6 mothers reported having experienced more than one event, (5 reported 2 events, 1 reported 3) (see Supplemental Table S2).

3.2. Gray matter volume analysis

Whole brain non-stationary cluster corrected analysis (p_corr < .05) revealed significantly greater GMV for the PS group compared to the CG group in bilateral parietal cortex (Table 3, Fig. 1). Specifically, clusters in the right intraparietal sulcus (IPS: 40, −45, 55) (p_corr = .019), the left IPS (−32, −43, 55) (p_corr = .004), the left inferior parietal lobule (supramarginal gyrus) (IPL: −56, −23, 34) (p_corr = .035), and the left superior parietal lobule (SPL: −19, −44, 49) (p_corr = 0.048) showed greater volume for the contrast PS > CG (Supplemental Figures S1, S2). No significant results were found for the reverse contrast, CG > PS.

Table 3.

Summary of results for contrast PS > CG, with sex as covariate of no interest. Results were thresholded at p < 0.001, k = 10 voxels. Cluster-level non-stationary corrected p-values (< .05) reported.

Region	Cluster size	MNI coordinates x y z			Z	t	Corrected p-value
R Intraparietal sulcus	2093	40	−45	55	4.57	4.90	.019
L Intraparietal sulcus	1703	−32	−43	55	3.87	4.07	.029
L supramarginal gyrus	1463	−56	−23	34	4.10	4.34	.035
L superior parietal lobule	297	−19	−44	49	4.55	4.88	.048

Fig. 1.

VBM non-stationary cluster corrected (p < 0.05) results for PS > CG, controlling for sex. R = right, L = left, IPS = intraparietal sulcus, SPL = superior parietal lobule, SMG = supramarginal gyrus.

3.3. Risk measure

The two groups differed significantly on two of the eight DUSI-R subscales at the Bonferroni adjusted alpha-level of p < .00625 (Table 4, Supplemental Figure S3). First, the PS group scored significantly higher on the DUSI psychiatric symptoms subscale (U(81) = 416.5, z = −3.339 p = .001), indicating a greater risk for mental health symptomatology among the maternal MLES-exposed group (see post hoc analysis exploring this finding in Supplement Section 3.3, Table S5). Second, the PS group scored significantly higher on the family system subscale compared to the CG group (U(81) = 486.5, z = −2.738, p = 0.006), suggesting greater interfamilial dysfunction or conflict, including within the adolescent-parent relationship(s).

Table 4.

Median (range) for DUSI-R subscales.

	All participants N = 82	Prenatal Stress (PS) N = 28	Comparison Group (CG) N = 54	Test statistic	p
Psychiatric symptoms	20 (0–75)	30 (0–75)	15 (0–70)	U = 416.5	.001*a
Family system	7 (0–71)	14 (0–71)	7 (0–42)	U = 486.5	.006*a
Behavior patterns	15 (0–65)	17 (0–65)	10 (0–65)	U = 593	.108
Health status	20 (0–50)	20 (0–50)	10 (0–50)	U = 544.5	.033
Leisure & Recreation	16 (0–58)	16 (0–58)	12 (0–58)	U = 594.0	.016
School performance	15 (0–60)	17 (0–60)	10 (0–45)	U = 494.0	.010
Social competence	14 (0–64)	21 (0–64)	14 (0–57)	U = 554.5	.046
Peer relationships	7 (0–42)	14 (0–42)	7 (0–42)	U = 545.5	.033

Notes: All tests two-tailed. DUSI-R = Drug Use Screening Inventory, Revised.

^aSignificant at Bonferroni adjusted alpha level of p < .00625.

3.4. Mediation analysis

PS adolescents exhibited increased GMV in parietal cortical regions as well as increased risk for psychiatric symptoms, relative to CG adolescents. Given that exposure to maternal stress in utero is associated with greater risk for offspring psychiatric disorders, we examined whether GMV mediated the association between prenatal stress exposure and psychiatric symptoms.

Using the MarsBaR toolbox in SPM, masks of each of the four PPC regions showing significant between-group differences were created. Mean GMV for each subject was extracted from these four regions. A parallel multiple mediation model was estimated using the Process Macro (v3.3) [92] in SPSS. A multiple mediation model was selected in lieu of individual models for each PPC region based on advantages of the multiple parallel model over distinct simple models [92]. Based on 10,000 bootstrap samples, bootstrap confidence intervals for the indirect effects crossed zero, indicating the null hypothesis cannot be rejected: there is no significant indirect effect of GMV for any of the 4 regions on the association between prenatal stress exposure and psychiatric symptoms risk (Fig. 2, Supplemental Table S3).

Fig. 2.

Parallel multiple mediator model testing for an indirect effect of GMV on the association between exposure to maternal MLES (PS, CG) and DUSI-R psychiatric symptoms risk. Bootstrap confidence intervals based on 10,000 bootstrap samples for the indirect effect intersected zero, indicating no significant indirect effect(s).

4. Discussion

This study reports an association between in utero exposure to retrospectively assessed maternal major life event stress (MLES) and brain morphology in typically-developing adolescents.

4.1. Increased GMV in PS adolescents

We identified an association between maternal MLES and gray matter volume (GMV) at a single timepoint in early adolescents: these adolescents (PS) exhibited greater GMV in parietal regions (bilateral IPS, SPL, and IPL) relative to their peers (CG). There were no regions for which CG adolescents showed increased GMV compared to PS youth. Our results align with studies showing associations between exposure to maternal distress during gestation and child brain morphology, including in parietal cortex [27–29]. Given that groups here were similar in age and the potential influence of sex was controlled for, we may expect groups to be at similar stages in cortical development; that is, post-peak gray matter volume in PPC, with age-normative cortical regressive events resulting in developmentally-appropriate cortical thinning [84–86,93,94]. Although groups did not differ significantly for age (p = .061), results of a VBM analysis controlling for both age and sex (Supplement, Section 3.1, Table S4) largely overlap with the main results reported here, suggesting the slightly younger age of PS (12.5 years) relative to CG adolescents (12.8 years) does not explain our finding of greater GMV in PS adolescents.

In addition to predicting group differences, we speculated that exposure to maternal MLES would be associated with differences suggestive of delayed brain maturation. Cortical thickness (CT) decreases in pre- and peri-adolescence, [34] and in adolescence [84–86,95] at an accelerated rate [41], including within parietal cortices [36]. Distinct from CT metrics [42], cortical gray matter volume also decreases from childhood through adolescence [84,86,95], with parietal gray matter reaching peak volume at 10.2 years for females and 11.8 years for males [84–86] (though recent longitudinal studies suggest cortical volume reaches “peak” in earlier childhood, begins decreasing in late childhood, continuing to decline in adolescence [36,96]). Of relevance to our results, a VBM study shows pregnancy-specific anxiety is associated with decreased GMV in 6–9 year-olds, interpreted as potentially reflecting accelerated development [27]. However, less GMV in children this age may alternatively indicate a lag in reaching peak volume, rather than accelerated volumetric decreases. The current study’s finding of increased GMV in PS adolescents may similarly suggest a lag in cortical development, potentially with regards to regressive (e.g., thinning) neural events. Of particular relevance is a functional study suggesting a delay in PFC development and its function in endogenous cognitive control in 17-year-olds exposed to high antenatal maternal anxiety [55]. Taken together, these studies are suggestive of an association between maternal distress and slowed cortical development. While this is an intriguing hypothesis, both the current study examining early adolescence and the earlier VBM study focused on late childhood are cross-sectional. Thus, prospective longitudinal studies are necessary to empirically test whether in utero exposure to maternal distress is associated with perturbations in neurodevelopmental trajectories, and if so whether these changes reflect delayed or accelerated development.

4.2. Increased psychiatric features in PS adolescents

Our finding that PS adolescents exhibited greater problematic behavior on the DUSI-R psychiatric symptoms subscale is consistent with a robust literature showing an association between maternal distress during pregnancy and offspring psychiatric symptoms and disorders, including impulsivity [22], ADHD [97–100], autism spectrum disorder [17,101], anxiety [21,97], depression [23], and schizophrenia [102,103]. Although a metric of risk for psychiatric features and not a diagnostic measure, DUSI-R psychiatric symptoms scores are correlated with the K-SADS (Kiddie Schedule for Affective Disorders and Schizophrenia) total score, an interview-based diagnostic assessment [65].

A post hoc mediation analysis revealed that GMV in parietal regions does not significantly mediate the association between prenatal exposure to maternal MLES and psychiatric symptoms risk in adolescent offspring. There are several possible explanations for this null finding. Although groups demonstrate differences in brain structure in PPC, it may be that brain functions supported by these regions are more pertinent mediators of the association between prenatal exposure to maternal MLES and adolescent psychiatric risk. It is also possible that differences in brain structure in PPC impact the structure and/or function of downstream regions relying on PPC input, including, importantly, prefrontal cortex (PFC), which shows a protracted course of development [94,104]. Consistent with its diverse structural and functional connectivity patterns, the PPC is a component of multiple brain networks, and PPC subregions are hubs in both dorsal and ventral fronto-parietal attention networks, which play critical roles in endogenous attention [105–107]. Of relevance, a long-term follow-up study found an association between antenatal maternal state anxiety and performance and PFC activation during an endogenous cognitive control task among 20-year-old men [108].

Further, the metric of cortical structure used here may not be a sufficiently discriminative intermediate phenotype of psychiatric risk. Cortical gray matter volume is a product of cortical thickness (CT) and surface area (SA) [95]; thus, we are unable to disambiguate whether differences identified in GMV are attributable to CT, SA, or both. Surface-based morphometric techniques, such as FreeSurfer, which measure CT, SA and cortical volume, may yield more discriminative measures. For instance, a previous study found CT in the right frontal pole mediated the association between prenatal maternal depression and child externalizing problems [28].

4.3. Limitations and strengths

The findings reported here must be interpreted within the context of limitations of the current study. The most prominent shortcoming is that the data analyzed are cross-sectional, not longitudinal. Thus, although we report group differences, we cannot characterize nor can we make inferences concerning how these differences relate to developmental trajectories of PS vs. CG adolescents. Ideally, future prospective longitudinal studies will examine MLES and associations with changes in brain structure from birth onward in order to probe whether and how MLES affects developmental trajectories.

The questionnaire-based measure of prenatal maternal exposure to MLES used in this study has limitations in that it is a retrospective assessment of traumatic events, and therefore prone to risk of recall and other biases [109–111]. Although prospectively assessed exposure to MLES is preferable [112], the questionnaire used here is comparable to measures used in retrospective studies that examine associations between maternal prepartum MLES and offspring health and behavioral outcomes [17,113–118]. Further, in contrast to daily life stress, MLES involve events likely to be impactful (e.g., death of a loved one) and thus well-remembered [119]. Measures of MLES such as the one used here have also been shown to be accurate and reliable [120–122]. Relatedly, studies comparing medical records with retrospective maternal recall of birth and pregnancy characteristics [123–126], substance use during pregnancy [127], and breastfeeding duration [128], show accuracy of recalled pre/perinatal events, even twenty or more years later [124,128].

Along with increased psychiatric symptoms, the PS group showed significantly higher scores on the family system subscale of the DUSI-R, a 14-item subscale querying parental involvement/supervision, parent-imposed boundaries/rules, occurrence of frequent or intense arguments between the adolescent and his/her parents or others in the home, as well as questions concerning regular or problematic substance use by or arrest of an immediate family member. Thus, in the current analysis, we are unable to disentangle the contribution of prenatal maternal MLES from that of current dynamics within the child’s family since some of the factors in our measure of prenatal stress could foreshadow difficulties in postnatal familial dynamics [109].

Sex differences have been demonstrated in the association between in utero exposure to maternal distress and offspring health outcomes [129,130], although the human literature shows findings that vary widely [131]. The potential influence of sex was controlled for in the present study, and overall groups did not differ for pubertal status; however, relative to CG girls, PS girls were more pubertally advanced (t(42) = −2.09, p = .042). Biological sex impacts the timing of cortical development [84]; although, it is not known whether these effects are related to pubertal status specifically [132], and more research on the influence of pubertal hormones on normative human neurodevelopment is needed [95]. Future research on exposure to maternal MLES is needed to investigate the influence of pubertal status, and whether mediation or moderation effects may differ by sex [30].

Strengths of the present analysis include the rich characterization of two, relatively large groups (for a neuroimaging study) of typically developing adolescents, including descriptors related to prenatal, postnatal, maternal and adolescent variables. Existing studies asking similar research questions have been correlational, have had smaller total sample sizes, and did not possess the detailed characterization of participants presented in the current study [27,28]. Additionally, factors known to be related to GMV, including age [84], IQ [133], BMI [37], and sex [84,86], were either controlled for or were similar between groups in the current study.

4.4. Conclusions

Our results add to the growing literature suggesting in utero exposure to maternal MLES is associated with enduring ‘scars’ [134], with our two groups of early adolescents exhibiting differences in brain morphology and increased risk for psychiatric symptoms. The considerable developmental period—from the prenatal period through adolescence—constitutes a critical window of both vulnerability and opportunity. Adolescence is of particular interest in terms of elucidating impacts of earlier experiences on the brain and psychosocial development. It is during this phase of development when risk for the emergence of neuropsychiatric disorders is heightened [44] and realized [135,136]. Animal studies, for example, have found evidence that early environmental insults may not manifest in complex disorders until adolescence [137] (see review in [50]). Human studies further suggest early life stress may not be manifested until adolescence, when neural circuits involving structures with a protracted course of development are forming for the first time [45,138]. Focusing on adolescence therefore provides an optimal opportunity for identifying modifiable environmental factors that increase vulnerabilities for psychiatric problems. On the other hand, conducting the longitudinal studies required to link early experiences with adolescent neurodevelopment and behavior, from conception onwards, presents many challenges. Prospective studies, beginning at birth, assessing multiple time points, and scrutinizing cortical metrics against the backdrop of knowledge concerning normative development are needed to illuminate whether offspring differences in brain structure associated with exposure to maternal distress reflect perturbations (i.e., delays/accelerations) in developmental trajectories. Until then, despite shortcomings of this study, our results further support the importance of providing pregnant women and their families with sufficient social support to help mitigate potential impacts of external stressors [54,134].