fNatural disaster stress during pregnancy is linked to reprogramming of the placenta transcriptome in relation to anxiety and stress hormones in young offspring

Abstract

Prenatal stress can lead to long-term adverse effects that increase the risk of anxiety and other emotional disorders in offspring. The in utero underpinnings contributing to such phenotypes remain unknown. We profiled the transcriptome of placental specimens from women who lived through Hurricane Sandy during pregnancy compared to those pregnant during non-Sandy conditions. Following birth, longitudinal assessments were conducted in their offspring during childhood (3 to 4 years old) to measure steroid hormones (in hair) and behavioral and emotional problems. This revealed a significant link between prenatal Sandy Stress (PNSS) and child HPA dysfunction, evident by altered cortisol, dehydroepiandrosterone (DHEA), and cortisol:DHEA levels. In addition, PNSS was associated with significantly increased anxiety and aggression. These findings coincided with significant reorganization of the placental transcriptome via vascular, immune, and endocrine gene pathways. Interestingly, many of the most prominently altered genes were known to be uniquely expressed in the syncytiotrophoblast (STB)-subtype of placental cells and harbored glucocorticoid response elements in promoter regions. Finally, several vascular development and immune-related placental gene sets were found to mediate the relationship between PNSS and childhood phenotypes. Overall, these findings suggest that natural disaster-related stress during pregnancy reprograms the placental molecular signature, potentially driving long-lasting changes in stress regulation and emotional health. Further examination of placental mechanisms may elucidate the environment’s contribution to subsequent risk for anxiety disorders later in life.

Introduction

Major lifetime stressors such as natural disasters can present significant mental health challenges with lifelong consequences. Pregnant women are especially vulnerable to stressors^{1, 2} and their physical and emotional health can impact in utero development. Indeed, many psychiatric disorders are associated with aberrant prenatal or early child development^3–6. However, delineating the contribution of specific prenatal stressors to health outcome is challenging given that subjective stressors are experienced differently by each pregnant woman. Thus, natural quasi-experimental events, such as natural disasters, have been leveraged to provide unique insights regarding the impact of stressful or traumatic events during pregnancy on offspring development. One common finding of such studies is enhanced anxiety and anxiety-related mood disorders in prenatally exposed offspring^7–9. Despite this documented relationship, the in utero mechanisms underlying the developmental effects of prenatal stress remain to be fully elucidated. Such knowledge is crucial for developing possible early life interventions to mitigate future mental health risk.

The placenta plays a pivotal role in neurodevelopment, orchestrating complex maternal–fetal interactions supporting nutrient regulation, endocrine production, and immune tolerance^{10, 11}. As such, the placenta provides a critical window into the prenatal environment directly relevant to fetal development. Longitudinal studies related to natural disaster stress have repeatedly documented alterations in stress-related genes in the placenta relevant to offspring health outcome^11–14. However, no information has been obtained about the thousands of transcripts within the placental transcriptome to provide an unbiased assessment of the genome-wide impact of maternal stress on the intrauterine environment. To this end, we profiled the placental transcriptome from a cohort of mothers who were exposed to Hurricane Sandy, the deadliest hurricane to hit the northeast US. The results documented a protracted impact of prenatal-Sandy-stress (PNSS) on offspring anxiety during childhood in addition to altered steroid hormone levels. These findings coincided with transcriptomic reprograming in placenta providing unique insights about placental gene signatures relevant to PNSS-associated developmental outcomes.

Materials and Methods

Procedures and participants

The study included mother-child dyads from the Stress in Pregnancy project, an on-going longitudinal study investigating the effects of stress exposures during pregnancy; see Finik and Nomura for a full cohort profile¹⁵. Participating pregnant women were recruited at two New York metropolitan area prenatal obstetrics and gynecological clinics—Mount Sinai Hospital and New York Presbyterian Queens which draws patients from locations that were significantly impacted during Hurricane Sandy. The study was approved by the institutional review boards at Mount Sinai Hospital, New York Presbyterian Queens Hospital, and CUNY Queens College. Criteria for study exclusion included HIV infection, maternal psychosis, cocaine use, maternal age <15 years, or life-threatening medical complications of the fetus. Written informed consent was obtained from eligible women for all procedures. The specific gestational timing during which Superstorm Sandy occurred was calculated based on the date of birth of the child and the day the storm hit the metropolitan New York area (October 29, 2012). Of 361 participants who consented, 21 were lost to follow-up and 18 excluded due to cocaine use, resulting in a total of 322 children who completed at least 3 annual assessments (Supplementary Figure 1). The remaining 322 participants were grouped in two categories: PNSS(+) consisting of children whose mothers were pregnant during the storm (exposed, 31% of the sample) and PNSS(−) consisting of children whose mothers were pregnant before or after the storm between Oct-2010 and Dec-2013 (unexposed, 69%).

Placental samples were acquired and processed as previously described¹⁵. Briefly, biopsies from the outer layer of the blastocyst, free of maternal decidua, were collected from each placenta quadrant midway between the cord insertion and the placenta rim within an hour of delivery. Specimens were frozen in liquid nitrogen and stored at −80°C.

Demographic and clinical information were obtained at an initial face-to-face evaluation with a social worker as previously described ¹⁵. Annual assessments continued at CUNY Queens College. At each annual visit, children’s neurobehavioral development and mother’s psychological functioning were ascertained. Hair samples were also collected and stored for subsequent steroid hormone assays.

Hair hormone levels

Cortisol, cortisone, testosterone, progesterone, and dehydroepiandrosterone were measured in hair samples collected in children at ~36 months. Hair was processed and analyzed as previously described^{16, 17}. Briefly, a bundle of ~100 hairs was cut from the posterior vertex of the child’s head, close to the scalp. Samples of 3 cm section were processed using methanol extraction and assayed (Kirschbaum’s laboratory, Dresden, Germany)¹⁸.

Behavioral and emotional functioning during childhood

Clinical dimensions of child behavior and emotions were ascertained at the 48 month assessment by administering the Behavior Assessment System for Children-2 (BASC-2) to parents¹⁹. Dimensions included Hyperactivity, Aggression, Anxiety, Depression Somatization, Atypicality, Withdrawal, and Attention Problems. The standardized t-score assessment was generated based on the child’s biological age and sex. The mean age (SD) of the assessments was 44.9 (10.9) months¹⁹. Furthermore, we tested the effect of PNSS on clinically significant and at-risk conditions; based on the BASC-2 manual, scores >60 on the standardized clinical dimensions reflect patterns of problems similar to those diagnosed with a behavioral or emotional problem²⁰.

Analysis of Child Behavioral Assessments

Maternal age, education, marital status, prenatal substance use, child’s sex, child’s age, and race were a priori defined as confounders. Severity of traumatic exposure due to Sandy, measured by Traumatic Exposure Instrument²¹ and self-reports of maternal trait anxiety, derived from the State Trait Anxiety Inventory (STAI)²² and depression symptomatology, derived from the Edinburgh Postnatal Depression Scale (EPDS)²³ were additionally controlled for in the adjusted statistical model.

For the statistical analysis, descriptive statistics were first conducted to evaluate the correlation, mean, and standard deviation (SD) of outcome variables, including hair steroid hormones, and BASC-2 clinical scores. General Linear Model (GLM) was used as the primary analytical method. Univariate GLM was used for hair hormones. For neurobehavioral (BASC-2) outcomes, a multivariable GLM was used which adjusts for correlations among dependent variables. The effect of PNSS on dichotomized values of neurobehavioral outcomes (i.e., classifying children with at-risk and clinically significant problems²⁰) was tested with the ordinal logistic link function in GLM. The exponentiation of the β-coefficient was obtained to calculate the odds ratio (OR) and associated 95% confidence interval (CI). Prior to the analysis, each subscale was evaluated for normality. If the assumption of univariate normal distribution was violated, log-transformation was applied to achieve normality.

RNA extraction, sequencing and data analysis

Transcriptome analysis (RNA-sequencing; RNA-seq) was carried out on a representative subset (N=131) of the total cohort. RNA-seq was performed (Novogene Corp, Sacramento, CA) on RNA extracted from pulverized placental tissue and free of ribosomal RNA (using Ribo-Zero Kit; Illumina). Paired-end 75bp reads in fastq format were examined for read quality with FASTQC²⁴. Paired-end reads were then mapped to the hg38 genomic assembly using the STAR read aligner²⁵ with standard mapping and filtering parameters. Uniquely mapped reads (91.6 +/− 1.73% (µ+/−SD) of total reads) were annotated to genomic features from the Ensembl genome database and quantified using the –quantMode argument in STAR. Mapped reads with minimal average expression (<0.3 FPKM) were removed prior to subsequent analyses. Principal component analysis was used to confirm the absence of outlier samples.

For differential gene expression (DEG) analysis, all group comparisons were made using DESeq2 software for moderated estimation of fold-change²⁶. All analyses were performed examining main effects of PNSS with GLMs adjusting for maternal age, maternal drug use (e.g., nicotine, cannabis, alcohol), fetal weight, and fetal sex, marital status, and education. Identified differentially expressed genes (DEGs) were used for gene set enrichment analyses using the curated transcriptional, pathway, and ontology libraries managed by the Enrichr web server²⁷. All gene ontology utilized the databases from the Gene Ontology Consortium and PANTHER. Motif enrichment analysis of differentially expressed gene promoters utilized the JASPAR and TRANSFAC databases with promoters defined as −2000 and +500 from the transcriptional start site. Trophoblast and immune subtype-specific genes were defined using marker genes identified by single-cell RNA-seq in human term placenta²⁸, filtering for marker genes uniquely enriched in a single cell type.

Mediation analysis was performed with the ‘mediation’ package in R. For each child phenotype significantly associated with PNSS, all PNSS DEGs were tested as mediators. For a gene to be a significant mediator, (A) the independent variable (PNSS) needed to significantly predict the mediator (PNSS+ DEG) and (B) the mediator needed to significantly predict the dependent variable (child phenotype) while adjusting for PNSS. The same covariates described for differential expression analysis were used in the GLMs for testing A and B. The mediation effect was calculated as the product of the two regression coefficients (A*B). Spearman correlations were performed between significant positive mediators and PNSS-linked child phenotypes with coefficients in figure heatmaps. Finally, note the custom placental illustration was developed with BioRender.com.

Results

Demographics

Table 1 shows the demographics and obstetric characteristics of the population. There was a small increase in birthweight in PNSS(+) vs PNSS(−) children (3.35 +/− 0.06 (µ +/− SEM) vs 3.20 +/− 0.04 kg). Moreover, there were slightly more Hispanic (55.4 vs 44.3%) and Caucasian children (25.7 vs. 14.5%) and fewer African American (7.9 vs. 23.1%) children among PNSS(+) vs PNSS(−) offspring.

Table 1.

Characteristics of the participants in the total population and based on prenatal Superstorm Sandy stress (PNSS) status.

Demographics & reproductive characteristics	Total (N=322) Mean (SE)	PNSS(−) (N=221) Mean (SE)	PNSS(+) (N=101) Mean (SE)	p-value
Maternal age	27.90 (.33)	27.79 (.41)	27.95 (.57)	0.83
Paternal age	30.06 (.44)	29.95 (.59)	30.26 (.63)	0.74
Parity	1.92 (.09)	1.87 (.11)	2.04(.15)	0.38
Birthweight (kilograms)	3.263 (.04)	3.20 (.05)	3.35 (0.06)	0.07
Gestational age (weeks)	38.93 (.13)	38.99 (.15)	38.81 (.25)	0.53
Prenatal trait anxiety	38.93 (.63)	38.97 (.79)	39.73 (1.02)	0.57
Prenatal state anxiety	37.83 (.69)	37.81 (.88)	38.78 (1.09)	0.51
Prenatal depression	9.57 (.28)	9.75 (.35)	9.20 (.48)	0.38
Placenta Weight (grams)a	647.07 (8.95)	647.58 (11.48)	645.54 (12.79)	0.96
	N (%)	N (%)	N (%)
Child gender (male)	170 (52.8)	121 (54.8)	49 (48.5)	0.30
Marital status				0.10
Married	146 (45.3)	91 (41.2)	55 (54.4)
Common law	16 (5.0)	9 (4.1)	7 (6.9)
Single	155 (48.1)	117 (52.9)	38 (37.6)
Widowed	1 (0.3)	1 (0.5)	0
Divorced/separated	4 (1.2)	3 (1.4)	1 (1.0)
Education				0.10
Primary school	9 (2.8)	7 (3.2)	2 (2.0)
Some high school	37 (11.6)	32 (14.5)	5 (5.0)
High school/GED	62 (19.4)	43 (19.5)	19 (19.0)
Some college	77 (24.1)	52 (23.6)	25 (25.0)
Associate’s Degree	32 (10.0)	23 (10.5)	9 (9.0)
Bachelor’s Degree	55 (17.2)	37 (16.8)	18 (18.0)
Graduate Degree	50 (14.8)	26 (11.8)	22 (22.0)
Child Race				0.001
Caucasian	58 (18.0)	32 (14.5)	26 (25.7)
African American	70 (21.7)	62 (28.1)	8 (7.9)
Hispanic	154 (47.8)	98 (44.3)	56 (55.4)
Asian	29 (9.0)	20 (9.0)	9 (8.9)
Other	11 (3.47)	9 (4.1)	2 (2.0)
Prenatal substance use b
Alcohol use	22 (6.9)	11 (5.1)	11 (10.9)	0.06
Cigarette smoking	33 (10.4)	23 (10.6)	10 (9.9)	0.85
Cannabis smoking	74 (21.8)	52 (23.5)	19 (18.8)	0.34
Obstetric complications
GDM c	21 (6.6)	14 (6.4)	7 (7.0)	0.83
Preeclampsia c	23 (7.2)	19 (8.7)	4 (4.0)	0.16
Vaginal infection b, c	16 (5.0)	9 (4.1)	7 (6.9)	0.27
C-section c	82 (25.5)	59 (26.7)	23 (22.8)	0.45
Prescription medication use b	0	0	0	--

PNSS(+) = pregnant during Superstorm Sandy; PNSS(−) = pregnant before and after Superstorm Sandy. Significant p-value (< 0.05) are in bold.

^aBased on limited data availability (N=97; control=73, exposed=24).

^bBased on maternal self-report during pregnancy.

^cBased on electronic medical record.

Hair hormone levels

Hair samples were analyzed to obtain information about stable hormone concentrations accumulated over recent months ¹⁷. We examined hormone levels at approximately 36 months (mean = 34.7 months) between PNSS(+) and PNSS(−) children (Table 2, upper panel). PNSS(+) children had increased cortisol (111.11 vs. 76.49 pg/mg, adj. p = 0.001) and nominally reduced dehydroepiandrosterone (DHEA) levels (13.11 vs. 20.25 pg/mg, p = 0.02). In addition, the cortisol:DHEA ratio was significantly increased in PNSS(+) children (13.81 vs. 6.87, adj. p = 0.01).

Table 2.

Effects of prenatal Sandy stress (PNSS) on child hair steroid hormones and standardized behavior scores.

	PNSS(−)		PNSS(+)		Significance
Stress Hormones	Mean and SD		Mean and SD		p-value	Adj. p
Cortisol, pg/mg	76.49	7.11	111.11	10.85	0.006	0.001
Cortisone, pg/mg	62.18	4.27	70.67	6.03	0.24	0.19
Testosterone, pg/mg	1.53	0.87	1.29	0.78	0.86	0.84
Progesterone, pg/mg	12.81	3.34	4.91	1.63	0.085	0.13
DHEA, pg/mg	20.25	2.16	13.11	1.67	0.02	0.24
Cortisol:DHEA	6.87	0.94	13.81	2.56	0.003	0.01
Neurobehavioral Profile	Mean and SD		Mean and SD		p-value	Adj. p
Hyperactivity	49.23	9.2	49.92	10.51	0.29	0.34
Aggression	46.45	8.687	48.89	10	0.03	0.04
Anxiety	49.85	9.67	54.19	13.03	0.008	0.002
Depression	49.1	11.08	49.97	11.31	0.62	0.53
Somatization	48.64	8.62	50.81	8.89	0.19	0.24
Atypicality	51.8	10.22	49.72	9.08	0.24	0.53
Withdrawal	49.2	9.08	49.64	8.18	0.90	0.40
Attention problem	50.99	9.63	50.03	8.57	0.60	0.74

PNSS(+) = pregnant during Superstorm Sandy; PNSS(−) = pregnant before and after Superstorm Sandy. Stress hormones; N = 144 for PNSS(−), N = 85 for PNSS(+). Behavioral scores; N = 223 for PNSS(−), N = 99 for PNSS(+). Values are adjusted mean with potential confounders, including maternal age, marital status, education attainment, prenatal substance use (alcohol, cigarette, cannabis smoking), prenatal maternal anxiety trait, prenatal maternal depression, child sex and race, and severity of Superstorm Sandy exposure. Samples with undetectable values were excluded from analysis. DHEA = dehydroepiandrosterone. Clinical behavioral problems were standardized based on sex and age with the score of 50 being a mean and 10 being a standard division. Significant p values and adjusted (adj.) p values are represented in bold.

Neurobehavioral profile: clinical and adaptive behaviors

For the behavioral assessments, analysis of Cronbach’s α yielded coefficient alpha reliabilities in the high 0.80s for individual scales. PNSS(+) children showed significantly greater anxiety (54.19 vs. 49.85, p=0.002) and aggression (48.89 vs. 46.45, p=0.04) on the BASC-2 after controlling for potential confounders (Table 2, lower panel). There were no significant effects of PNSS(+) exposure on other behavioral scores including hyperactivity, depression, somatization, atypicality, withdrawal behaviors, and attention problems. When the eight clinical dimension area scores were dichotomized at BASC-2 scores ≥ 60 (i.e., clinically significant or at-risk levels), PNSS(+) children showed an over four-fold increased risk of aggression compared to PNSS(−) children (adjusted odds ratio (AOR) = 4.84, 95% confidence interval (CI) = 1.84–18.47, p = 0.003) and an over five-fold increased risk for anxiety (AOR = 5.30, CI = 2.71–10.37, p < 0.0001) (Supplementary Table 1). Notably, there was no significant interaction between PNSS and child sex associated with steroid hormone levels or neurobehavioral traits (Supplementary Table 2).

Consistent with our primary analyses of pregnancies overlapping with Hurricane Sandy, PNSS(+) remained significantly associated with increased hair cortisol and anxiety levels when Hurricane Sandy occurred early in pregnancy (i.e., first trimester) as well as later in pregnancy (i.e., second or third trimester) (Supplementary Table 3). Similarly, PNSS(+) remained associated with increased cortisol and anxiety when PNSS(−) controls were limited to pregnancies prior to Hurricane Sandy and when excluding all self-reported drug use (Supplementary Tables 4-5).

Placental transcriptome profile

To examine the impact of PNSS on the placental transcriptome, biopsies from 131 mother-child dyads were processed for RNA sequencing. Within this cohort, 38 pregnancies were PNSS(+) and demographics were consistent with the main cohort (Supplementary Table 6). Mapping statistics of sequencing data and principal component analysis of normalized gene counts confirmed the absence of technical outliers (Supplementary Table 7). Placentas derived from PNSS(+) pregnancies exhibited robust transcriptional alterations with 4016 genes found to be differentially expressed after false discovery rate (FDR) adjustment (FDR adj. p < 0.05; Figure 1A, Supplementary Table 8). Differentially expressed genes (DEGs) were enriched for gene ontology categories related to vascular development (Figure 1B, Supplementary Table 9), primarily platelet-derived growth factor signaling (PDGF), renin-angiotensin system (RAS), angiogenesis, and vascular endothelial growth (VEGF) signaling. Additional enriched pathways were associated with immune response functions, notably T-cell and B-cell activation (Figure 1B, Supplementary Table 9). Applying a more stringent significance threshold (fold change > 2) to our primary analysis, PNSS(+) was associated with 221 remaining DEGs enriched for estrogen biosynthesis, glucocorticoid metabolism, and growth hormone regulation (Figure 1B, Supplementary Table 9). Strikingly, PNSS(+) was associated with reduced expression of HSD11B2, HSD3B1, and GH2, which are critical for cortisol metabolism, progesterone synthesis, and growth hormone production, respectively^29–31. These robustly differentially expressed genes remained significant (p<0.05) when examining PNSS(+) separately for early (first trimester) and late (second/third trimester) PNSS and with expanded exclusion criteria (Supplementary Figure 2, Supplementary Tables 8, 10-11). Interestingly, mid-late PNSS(+) was associated with a greater number of DEGs vs. early PNSS(+) and showed unique upregulation of several immune-related gene pathways and cell-type markers (Supplementary Figure 3).

Figure 1: Prenatal Sandy stress (PNSS) is associated with robust differential gene expression across critical functional pathways in placenta.

(a-b) Gene expression alterations (blue points, q<0.05) in PNSS(+) placentas with many endocrine and pregnancy-specific gene regulators among the most differentially expressed genes (DEGs) (red points, q<0.05 & |log₂ fold change (FC)|>1). (c) PNSS(+) DEGs (q<0.05) were enriched for vascular and immune-related functions. (d) The most robustly altered DEGs associated with endocrine and glucocorticoid signaling functions (p<0.05).

To examine the validity of these sequencing results, we utilized findings from the same cohort analyzed on a different platform which reported reduced expression of eight stress- and neurotransmitter-related gene probes¹⁴. The current sequencing-derived analysis confirmed that six of those eight genes were significantly reduced in PNSS(+) placentas (Supplementary Table 12), substantiating the transcriptomic findings.

Differential expression of stress-related genes and cell type-specific gene signatures

To further assess the robust gene expression alterations associated with PNSS(+), we considered the unique function of the different trophoblast subtypes in placenta. The major trophoblast populations include undifferentiated cytotrophoblast and differentiated extravillous trophoblast (EVT) and syncytiotrophoblast (STB) cell types. EVTs anchor the placenta to the decidua and optimize maternal vasculature early in pregnancy while STBs produce essential pregnancy steroid hormones, regulate nutrient/gas exchange, and form a continuous epithelial barrier to maternal-borne pathogens and immunogenic cells^32–36 (Figure 2A). When we examined genes identified by single-cell RNA-seq to be uniquely enriched in these placental cell types²⁸, we found many of the trophoblast-subtype marker genes (CTB, EVT, STB) were differentially expressed in PNSS(+) placentas (Figure 2C, Supplementary Table 8). Strikingly, the majority of STB marker genes were robustly reduced in PNSS(+) placentas (Figure 2C). To examine potential upstream regulators of STB marker gene expression, we analyzed STB gene promoters for established transcription factor motifs, revealing enrichment for NR3C1, RUNX1, and ELF3 binding sites (Figure 2D, Supplementary Table 13). NR3C1 transcribes the glucocorticoid receptor (GR), and placental GR expression and methylation have been associated with maternal anxiety and depression^37–39. NR3C1 was increased in PNSS(+) placentas and negatively correlated with STB-specific gene expression (Figures 2E-F).

Figure 2: PNSS(+) is associated with reduced expression of trophoblast-subtype specific genes and targets of glucocorticoid signaling.

(a) Diagram depicting major placental cell-types contributing to the placental villi. (b) Several placental cell type specific genes were differentially expressed in PNSS(+) placentas, especially syncytiotrophoblast (STB)-specific markers. Bars = µ, blue points = differentially expressed, red points = differentially expressed with |log2FC| >1. (c) Heatmap of STB-enriched genes with unsupervised clustering of rows annotated for PNSS(−) (blue) and PNSS(+) placentas (red). (d) Several transcription factor motifs (represented by gene name) are enriched in the promoter of STB-specific genes. (e) Expression of NR3C1 was significantly increased in PNSS(+) placentas. (f) Significant negative correlation between NR3C1 and all STB-specific genes in placenta (lines represent regressions for each STB-gene and NR3C1 (normalized expression). ***=p<0.001 for each spearman correlation analyzed.

Placental genes mediating PNSS-linked developmental outcomes

To examine whether the robust transcriptome alterations may inform the link between PNSS and altered developmental phenotypes, we performed mediation analyses on PNSS(+) DEGs. This analysis revealed 10 and 157 significant genes positively mediating PNSS(+) related increases in cortisol and the cortisol:DHEA ratio, respectively (p<0.05; Figure 3A). Gene ontology analysis of placental genes mediating increased child cortisol:DHEA levels revealed significant enrichment of gene sets in antigen and T-cell signaling pathways (Figure 3B) which showed increased expression in PNSS(+) placentas and positively correlated with cortisol:DHEA levels (Supplementary Table 14).

Figure 3: Placental genes related to immune activation and vascular development mediate the relationship between PNSS(+) and offspring HPA/neurobehavioral phenotypes.

(a) Heatmap showing correlation coefficients between cortisol or cortisol:DHEA and PNSS+ DEGS with dot bar annotation representing log₂ fold change (FC) of each PNSS(+) DEG (blue=log₂FC<0, red=log₂FC>0). (b) Genes mediating increased cortisol:DHEA in PNSS(+) children were related to adaptive immune response pathways. (c) Correlation coefficients for child aggression or anxiety and PNSS(+) DEGs with dot bar and log₂FC annotation bar. (d) Genes mediating PNSS(+)-linked child aggression were nominally significant for VEGF, B-cell, and Ras signaling pathways. * = significant mediation (p<0.05).

Examining child anxiety and aggression behaviors, 28 and 5 PNSS(+) DEGs were found to significantly mediate child aggression and anxiety, respectively (Figure 3C, Supplementary Table 14). Mediators of PNSS(+) related child aggression were enriched for VEGF, B-cell, and Ras signaling pathways (Figure 3D). Among the significant mediators of PNSS(+) linked child anxiety, the most significant gene was Krüppel-Like Factor 1 (KLF1), an essential regulator of erythrocyte production and vascular development (Figure 3C).

Discussion

The present results link prenatal natural disaster (PNSS) exposure with increased anxiety and cortisol levels in early childhood. Moreover, transcriptome analysis of the placenta revealed a reduction of endocrine genes expressed by syncytiotrophoblast (STB) and regulated by glucocorticoid receptor signaling in relation to PNSS(+). Several genes related to immune and vascular functions were also identified as potential placental mediators of prenatal Hurricane Sandy stress. Overall, these findings illuminate many novel links between prenatal stress and placenta transcriptomic regulation with significant implications for early child development.

Consistent with the previous finding that PNSS(+) was associated with fearfulness temperament at 6 months¹³, children aged 3–4 in the current study showed increased anxiety and aggression. Moreover, PNSS(+) was linked with increased cortisol levels and an increase in cortisol:DHEA ratio, both characteristic of an anxiogenic phenotype^40–42. These results add to a robust literature demonstrating protracted effects of prenatal stress on subclinical psychosocial problems in children^3–5 and HPA function^43–46. Thus, the present study confirms and extends this literature, illuminating developmental risks associated with a traumatic stressor during pregnancy while controlling for important confounds such as pre-existing anxiety and depression in mothers.

Leveraging RNA-sequencing to comprehensively examine the intrauterine impact of Hurricane Sandy stress in an unbiased manner revealed several transcriptomic alterations in the placenta associated with PNSS. The most robust finding in this study was the link between PNSS(+) and critical endocrine genes during pregnancy (e.g., GH2, HSD11B2, HSD3B1) that are specifically expressed in STB. Consistent with the vital role of these STB-specific genes during pregnancy, many studies have associated aberrant STB function with pregnancy complications^47–49. Increased STB apoptosis is associated with preeclampsia and fetal growth restriction^47–49. While there was a significant difference in birthweight in the present study, it was surprisingly increased in PNSS(+) newborns with no other growth phenotypes. Moreover, weight was included in all statistical analyses. Future studies examining protein and STB-regulated steroid hormone levels in the fetal and maternal blood during pregnancy are needed to determine the functional significance of the striking reductions in STB gene expression.

While STB gene signatures have not been directly studied in relation to prenatal stress, studies have examined STB-related genes with HPA-axis function. For example, mothers with high trait anxiety or increased total stressful life exposures have reduced HSD11B2 expression in the placenta^{50, 51}. Furthermore, increased placental corticotrophin releasing hormone (CRH) is associated with physiological and psychological stress during pregnancy ^{52, 53} and glucocorticoids have been shown to induce CRH expression in STB in vitro⁵⁴. Thus, the current findings expand on this literature, revealing reduced expression of several STB-enriched genes in PNSS(+) placentas and implicating glucocorticoids (via increased NR3C1) as a potential upstream transcriptional regulator.

One potential mechanism underlying PNSS-linked effects on placental gene expression is epigenetic reprogramming. Indeed, both maternal anxiety and depression have been associated with increased HSD11B2 and NR3C1 methylation in placenta⁵⁵. Moreover, stress-induced hypermethylation of the HSD11B2 promoter coincides with reduced HSD11B2 expression⁵⁶. In rodents, the effects of prenatal stress on HPA-function are mediated by sex-specific chromatin remodeling in placenta⁵⁷. In addition to placental transcriptomic regulation, epigenetic mechanisms may be important for sustaining the effects of prenatal stress across development. For instance, maternal depression or traumatic exposure is associated with altered NR3C1 methylation in human infant blood^{39, 58, 59}. These studies suggest PNSS effects on the HPA-axis and anxiety may be maintained by long-lasting epigenetic reprogramming initiated during gestation. Future interrogation of epigenetic mechanisms in placenta and peripheral tissues from the longitudinal cohort will provide such critical information.

PNSS-linked gene expression alterations suggest major dysregulation of placental function and may offer biological insights into the impact of trauma-inducing stressors on subsequent behavioral and physiological outcomes relevant to anxiety-related disorders. Mediation analysis currently helped to identify genes potentially relevant to the relationship between PNSS and child developmental outcomes. Several genes in the immune response pathways were upregulated by PNSS(+) and positively correlated with cortisol:DHEA levels. Maternal immune activation has been shown to engage placental immune pathways and is associated with altered brain and behavioral development in offspring^{60, 61}. Moreover, prenatal stress increases placental inflammation and stress-induced hyperactivity in mice, both of which are rescued by nonsteroidal anti-inflammatory treatment⁶². Thus, the immune-related gene sets identified here warrant further investigation for their role in prenatal programming of HPA function and stress responsivity.

Although preclinical studies support the role of placenta stress-related gene pathways in stress- and anxiety-related behavior in offspring¹¹, we did not find a relationship between endocrine/STB related genes such as HSD11B2 or NR3C1 with anxiety or HPA activity in the current study. The most significant gene mediators of child anxiety were KLF1 and CA2, both important regulators of vasculogenic and angiogenic processes^{63, 64}. Similarly, aggression featured gene sets involved in vascular endothelial growth factor signaling. Future studies with mechanistic preclinical models can help to determine the direct role of these genes on stress and anxiety-related behaviors expressed during the juvenile period. It is also noteworthy that we did not find an interaction between prenatal stress and child sex in relation to stress hormone levels or neurobehavioral profiles despite many studies having reported sex-specific effects of prenatal stress on offspring stress-related phenotypes⁶⁵. Given that the children examined here were still very young, the continued longitudinal assessments of sex-specific psychosocial development will be critical through puberty and adolescence when sexually divergent phenotypes often emerge.

In summary, the current results reveal transcriptomic reorganization within the placenta associated with PNSS that relates to subsequent childhood stress and psychosocial phenotypes. The placental gene sets identified related to PNSS could help to predict which newborns exposed to prenatal maternal stress may be at risk of developing stress- and anxiety-related disorders later in life. Such identification along with other markers generated from additional datasets would allow early interventions to mitigate or delay the development of more severe disorders in adolescence and adulthood.