hsa-miR-582-3p in umbilical cord blood is negatively associated with maternal exposure to childhood maltreatment

Abstract

Aim: Childhood maltreatment (CM) may affect not only directly exposed individuals but also their offspring. However, the underlying biological mechanisms remain unclear. microRNAs (miRNAs) may play a regulatory role in this process. This study investigates the relationship between maternal exposure to CM and miRNA expression in maternal and perinatal tissues.

Methods: We enrolled 43 pregnant women and assessed their CM exposure. We collected maternal blood, cord blood and placental tissue samples during childbirth and performed miRNA profiling using next generation sequencing.

Results: Maternal CM was inversely associated with hsa-miR-582-3p levels in cord blood. Pathway analysis revealed that this miRNA regulates genes involved in intrauterine development.

Conclusion: Our findings highlight the potential impact of maternal CM exposure on offspring epigenetic mechanisms.

Plain Language Summary

Child maltreatment (CM) includes physical, sexual and emotional abuse, as well as physical and emotional neglect. CM not only harms those directly exposed but can also negatively impact their offspring. However, the biological reasons behind this are not well understood. To explore this further, our study investigates how CM affects the biology of pregnant women and their newborns through changes in small regulatory molecules called microRNAs (miRNAs). We recruited 43 pregnant women and assessed their exposure to CM. During childbirth, we collected blood samples from the mothers, blood from the umbilical cord and placental samples. We then analyzed the levels of miRNAs in these samples using advanced sequencing technology. We observed that more severe maternal exposure to CM was associated with lower levels of a miRNA named hsa-miR-582-3p in umbilical cord blood. This miRNA regulates genes involved in fetal development in utero and has been linked to spontaneous preterm birth. It may also influence immunologic and stress-related processes. Thus, newborns of mothers who had been exposed to CM may be more vulnerable to adverse effects on their brain development and overall health. Despite our small sample size, our study highlights the importance of addressing CM as an intergenerational concern and provides new insights into the biological mechanisms through which maternal CM can affect offspring.

Plain language summary

Article highlights.

Background

• Child maltreatment (CM), encompassing physical, sexual and emotional abuse, as well as physical and emotional neglect, may affect not only directly exposed individuals but also their offspring.

• Epigenetic alterations, such as changes in miRNA expression profiles, may play a regulatory role in this process.

• We investigate the association between maternal CM and miRNA expression profiles in maternal peripheral blood, umbilical cord and placenta.

Methods

• We enrolled 43 women in the third trimester of a low-risk gestation and their newborns in the public maternity hospital in Brazil.

• We assessed maternal history to CM using the Childhood Trauma Questionnaire and exposure to recent abuse using the Abuse Assessment Screen.

• We collected the samples during childbirth and performed miRNA expression profiling using Next Generation Sequencing.

• Associations between maternal CM and miRNA expression were tested using linear regression models, controlling for potential covariates.

• miRNA target genes were predicted using miRwalk or TargetScan and enrichment for biological pathways was analyzed using the “gene2func” function of FUMA GWAS.

Results

• hsa-miR-582-3p in umbilical cord blood is negatively associated with maternal history of CM.

• The enrichment analysis of target genes associated with hsa-miR-582-3p in GWAS for spontaneous early preterm birth and pathways related to embryonic development underlines its potential role in intrauterine development.

• 112 candidate miRNAs were identified in our data, although none were associated with maternal history of childhood maltreatment.

Discussion

• The enrichment analyses showed important pathways for fetal development and point to the possible role of hsa-miR-582-3p as a molecular regulator during this period.

• Possibly, the altered profile of this miRNA could render the newborns an altered physiology, which could be contributing to the impact on offspring's health associated with maternal history of CM.

• To our knowledge, this is the first study to identify in humans changes in offspring's miRNA expression associated with maternal exposure to CM.

Conclusion

• hsa-miR-582-3p has significant negative correlation with maternal exposure to CM in umbilical cord blood.

• Pathway enrichment analysis of hsa-miR-582-3p target genes pointed to their roles in important pathways for fetal development.

• Despite the limitations imposed by our sample size, our study provides valuable insights into the potential impact of maternal CM exposure on offspring epigenetic mechanisms.

1. Introduction

Childhood maltreatment (CM) encompasses experiences of sexual, physical, or psychological abuse, as well as physical or emotional neglect inflicted by parents and/or caregivers [1]. The global prevalence of these forms of maltreatment is estimated at 26.6, 22.6, 36.3, 16.3 and 18.4%, respectively [2]. Exposure to adverse events in childhood has been linked to increased risks of physical, emotional, mental, sexual and reproductive health problems in adulthood [3,4,5,6,7], possibly as early childhood and adolescence are periods of greater brain and immune system plasticity [3].

Growing evidence supports the hypothesis that parental exposure to CM can impact the next generation [8,9,10], manifesting in negative emotional and behavioral dysregulation [11], increased externalizing behavior [12], low birth weight [13], small-for-gestational-age newborns [13], altered cortisol levels [14] and—as suggested by preclinical studies—changes in the fetal hypothalamic-pituitary-adrenal (HPA) axis [15]. However, the mechanistic underpinnings of maternal-fetal influence remain unclear. The ‘fetal programming hypothesis’ posits that maternal exposure to CM may create an altered fetal environment during pregnancy, potentially affecting infant development [15,16,17]. Epigenetic modifications, such as micoRNAs (miRNAs), represent the potential mechanisms supporting the fetal programming hypothesis.

miRNAs are small non-coding RNA fragments (∼22 nucleotides) that regulate the expression of specific target genes at the mRNA level by inhibiting translation or inducing messenger RNA degradation [18,19]. Although miRNAs have been implicated in Alzheimer's disease, schizophrenia, major depressive disorder and autism spectrum disorders in individuals exposed to CM [20,21], their long-term impact on directly exposed individuals and their offspring remains poorly understood.

Studies integrating maternal tissues and perinatal tissues offer valuable insights into how maternal CM may impact offspring through fetal programming. Perinatal tissues, such as umbilical cord blood and placenta, provide a snapshot of the in utero environment without the direct influence of external factors. The placenta, acting as the conduit for maternal-fetal communication, plays a pivotal role in shaping the uterine environment, while cord blood analysis offers insights into the physiological changes directly affecting newborns.

This study is grounded on the understanding that exposure to maternal childhood maltreatment can have enduring effects on multiple biological systems, including inflammatory and stress-response systems, as well as epigenetic mechanisms. These effects may persist into adulthood and, during pregnancy, impact the fetal environment, impairing infant neurodevelopment and behavior. Our primary objective is to investigate the association between maternal CM and miRNA expression profiles in maternal and perinatal tissues, including candidate miRNA sequences. Additionally, we aim to explore target genes of associated miRNAs and their pathways to better understand their biological function.

2. Materials & methods

2.1. Study sample

We enrolled 43 women aged 18 to 43 years in the third trimester of a low-risk pregnancy. Women were approached during prenatal visits in the Hospital Municipal do Campo Limpo Dr. Fernando Mauro Pires da Rocha, in the city of São Paulo, Brazil. This hospital is affiliated with the Brazilian Health System (Sistema Único de Saúde, SUS) [22], a nationwide public healthcare system that offers free-of-charge healthcare services to all individuals. Pregnant mothers were ineligible for participation if they: exhibited a severe psychiatric condition, encompassing schizophrenia, persistent delusional disorder, bipolar disorder, obsessive-compulsive disorder, dementia or suicidal ideation; had a prior history of head trauma resulting in diagnosed brain injury, received treatment for epilepsy or underwent neurosurgery; manifested decompensated clinical diseases, such as diabetes, hypertension and hypo/hyperthyroidism; engaged in illicit drug use, excluding cannabis; experienced during gestation any TORCH infections (toxoplasmosis, others [such as syphilis and hepatitis b], rubella, cytomegalovirus and herpes simplex).

Infants born to participating women were included in the study upon birth. The mother-infant dyad was excluded if the newborn met criteria such as preterm birth (born before 37 weeks of gestation), a 5-minute Apgar score below 7, admission to the neonatal intensive care unit, or the presence of kernicterus or inborn errors of metabolism.

2.2. Clinical assessments

Maternal history of CM was assessed with the Questionário sobre Traumas na Infância (QUESI) [23], the validated version of the Childhood Trauma Questionnaire (CTQ) [24] for Portuguese and adapted for the Brazilian population. This questionnaire addresses exposure to emotional abuse, physical abuse, sexual abuse, physical neglect and emotional neglect during childhood and adolescence. The Brazilian Portuguese version of the Abuse Assessment Screen (AAS) [25] was used to screen mother's exposure to physical and sexual abuse in the past 12 months. The socioeconomic status was determined by the Brazilian Economic Classification Criteria 2015, ranging from A (highest income) to D-E (lowest income). Newborns' clinical information was obtained from medical records or during pediatrician evaluation up to 4 weeks after birth.

2.3. Biospecimen collection & RNA isolation

Samples of maternal peripheral blood (n = 43), umbilical cord blood (n = 43) and placenta (n = 43) were collected during the childbirth. Maternal peripheral blood was collected through a venipuncture using PAXGene^® Blood RNA tubes (PreAnalytiX, Switzerland) immediately before the delivery procedures. Cord blood was collected using a standard 20-ml syringe with luer lock tip. The plunger was removed, and the unclamped umbilical cord was inserted into the barrel. Once the barrel was filled with cord blood, the plunger was reinserted, and the blood was promptly transferred through an 18 G (1 1/2”) needle from the syringe into a PAXgene^® Blood RNA tube.

Placenta collection was conducted within 10 minutes of placental delivery and immediately stored in RNAlater^® (Qiagen, Germany) to prevent sample degradation, following collection recommendations previously described [26]. The placenta underwent inspection for abnormalities, and membranes were removed to fully expose its fetal surface. A placental tissue sample measuring 5–10 mm³ was then collected from each of the four quadrants, washed with sterile phosphate-buffered saline (PBS) 1x pH 7.4 and individually placed in 2-ml microtubes containing 1 ml of RNAlater^® (Qiagen, Germany) for immediate stabilization RNA stabilization. Subsequently, the samples were maintained at 4–8°C overnight before being stored at -20°C.

Total RNA isolation (including miRNAs) from blood samples was performed with the PaxGene^® Blood miRNA kit (Qiagen, Germany). For a better representation of the placenta, around 5 mg of tissue was taken from each quadrant and pooled together (~20 mg). The pooled tissue lysis was performed with the TissueLyser LT (Qiagen, Germany) using 5-mm Stainless Steel Beads (Qiagen, Germany). Total RNA (including miRNAs) was isolated with the AllPrep DNA/RNA/miRNA Universal kit (Qiagen, Germany), including treatment with RNase-free DNase (Qiagen, Germany). All protocols were executed according to manufacturers' instructions.

2.4. Library preparation & sequencing

Total RNA isolated from blood and placental samples were subjected to high-throughput sequencing using Illumina technology. Library preparation was performed using QIAseq miRNA Library Kit (Qiagen, Germany) from an input of 100–500 ng of total RNA. Library fragments were analyzed with Agilent High Sensitivity DNA Kit (Agilent Technologies, USA) and quantified with Qubit^® 2.0 fluorometer (Life Technologies, USA). All libraries were diluted to a concentration of 5 nM and combined in three pools (one for each tissue). Each pool underwent cluster generation and single-end sequencing (1 × 75) on an Illumina NextSeq 500 instrument using the NextSeq 500/550 High Output kit v2 75 cycles (Illumina, USA), aiming for an expected coverage of around 10 million reads per sample. The sequencing of the library pools yielded on average 7.9 million (±4.5), 7.9 million (±3.3) and 13 million (±6.0) reads for maternal blood, umbilical cord blood and placenta, respectively.

Raw sequencing data was processed using miRge 3.0 [27] pipeline. We filtered out duplicate reads using the unique molecular identifiers (UMI). Reads were aligned to miRBase reference v22 using Bowtie v1.3.0—65.9%, 65.6% and 73.6% of them were successfully aligned in the datasets of maternal blood, umbilical cord blood and placenta, respectively. The parameters used were: -on human -db miRBase -a AACTGTAGGCACCATCAAT -qumi -umi 0.12. The following data processing, statistical analyses and data visualization were conducted using the R statistical programming language (version 4.3.1) within the RStudio environment. Read counts were normalized by tissue using the variance stabilizing transformation (VST) method provided within the DESeq2 package [28]. Only miRNAs with at least three read counts in more than 80% of samples were retained for downstream analysis. We conducted Principal Component Analysis (PCA) using miRNA normalized read count data. To detect possible aberrant expression patterns and clustering indicating bias in our results, we plotted the principal component (PC)1 vs PC2, PC1 vs PC3, PC2 vs PC3 labeling samples by maternal age, sex, socioeconomic status and maternal educational attainment.

2.5. Identification of candidate novel miRNAs

The reads not aligned to miRBase were subsequently aligned to the human genome (hg38) to identify probable candidate miRNAs using the miRge3.0 tool. The detection of candidate miRNAs is carried out using a machine learning algorithm built using support vector machine (SVM) [27]. Similar sequences are grouped into clusters, and the most conserved region of each cluster is identified. The coordinates of the putative miRNA are defined based on the cluster where the frequency of each base over the total number of reads is greater than 80%. Using the RNAfold method, the folding energy of the sequence is calculated. A SVM model is applied, and a probability value is calculated—we assumed ≥95% probability. 52,285, 31,089 and 59,571 reads were aligned to the genome and reached ≥95% probability in the maternal blood, cord blood and placenta datasets, respectively. We filtered candidate miRNAs with two or more read counts in at least 50% of the samples of its tissue for downstream analysis. The read normalization parameters were the same described for known miRNAs.

2.6. Statistical analyses

We used the CTQ total score, a continuous variable, as a measure of the maternal history of CM. To test the association of maternal exposure to CM with the expression of known miRNAs, we employed linear regression models (LM). For each tissue, covariates were selected from our dataset if they had been described as related to miRNA expression/CM and/or were used as covariates by other studies, and were not strongly correlated with our main predictor (maternal CM). We used Spearman's test to examine the correlation between maternal exposure to CM score and covariates, including maternal age, newborn's birth weight, gestational age at birth, educational attainment and socioeconomic status. This analysis aimed to identify potential issues of collinearity among predictors and determine which variables could be added as covariates to the models. To ensure that the use of Spearman's correlation, a rank-based method, did not distort our conclusions about the multicollinearity, we additionally calculate the variance inflation factor (VIF) for models of miRNAs associated with maltreatment using the R package “car”. In crafting the models, we tested all combinations of covariates for each tissue. We used Quantile-Quantile (Q-Q) plot of -log₁₀(p-value), within-tissue adjusted R² and Akaike criterion (AIC) to choose the best model. The selected models are shown in Table 1. The false discovery rate (FDR) was controlled at a level of 0.05 within tissue using the Benjamini-Hochberg procedure. The chosen models to explore known miRNAs were also used to test the association of maternal CM with candidate miRNAs.

Table 1.

Chosen linear regression models.

Tissue	Outcome	Predictor	Covariates	Number of models/miRNAs
Maternal peripheral blood	miRNA normalized counts	CTQ total score	maternal age + educational level + socioeconomic status + newborn's birth weight	450
Cord blood	miRNA normalized counts	CTQ total score	maternal age + newborn's sex assigned at birth + newborn's birth weight + delivery type + gestational age at birth	561
Placenta	miRNA normalized counts	CTQ total score	newborn's sex assigned at birth + newborn's birth weight + delivery type + gestational age at birth + maternal socioeconomic status	667

CTQ: Childhood trauma questionnaire.

2.7. Prediction of target genes & pathway analysis

To predict target genes of the miRNAs associated with maternal exposure to CM, we used the miRWalk database [29]. We included only target genes with a binding probability ≥95% to the miRNA of interest, regardless of the binding region (3' untranslated region [UTR], 5' UTR or coding DNA sequence [CDS]). The pathway enrichment analysis was performed using “gene2func" tool in the Functional Mapping and Annotation of Genome-wide Association Studies (FUMA GWAS) [30] v1.5.2, including the GWAS catalog, the Gene Ontology (GO) (for Biological Process, Cellular Component and Molecular Function) and the Kyoto Encyclopedia of Genes and Genomes (KEGG). Pathways were listed as significant if at least five target genes were present and had an enrichment adjusted p-value ≤ 0.05 (FDR). The graphical representation of enriched pathways was produced with “enrichplot” and “clusterProfiler” packages, in the R environment.

3. Results

3.1. Sample description

Our study comprised 43 mother-infant dyads, all residing in the same low-resource area in São Paulo, Brazil. Mothers ranged in age from 18 to 43 years old, with a mean age of 28,6 years (±6.2). All mothers were classified within low socioeconomic status—most of them (67.4%) in the lowest status (D-E)—, with only 12% having attained higher education. On average, the gestational age at birth was 39.7 weeks (±1.4) and the occurrence of vaginal and cesarean deliveries had an equitable distribution (48.8% and 46.5%, respectively), while two deliveries required assistance with forceps. Detailed demographic information is provided in Table 2.

Table 2.

Study sample demographics.

Subject		N	Overall
Mother	Age (years)	43
	Mean ± SD		28.6 ± 6.2
	Educational level	43
	Primary education		12/43 (27.9%)
	Secondary education		26/43 (60.5%)
	Higher education		5/43 (11.6%)
	Socioeconomic status	43
	C		14/43 (32.6%)
	D-E [lowest status]		29/43 (67.4%)
	Self-declared race	43
	Black		11/43 (25.6%)
	Brown		18/43 (41.9%)
	Indigenous		0/43 (0%)
	White		13/43 (30.2%)
	Yellow		1/43 (2.3%)
	CTQ total score	43
	Mean ± SD		39.8 ± 12.7
	CTQ categories	43
	No trauma		14/43 (32.5%)
	Mild trauma		10/43 (23.2%)
	Moderate trauma		7/43 (16.2%)
	Severe trauma		12/43 (27.9%)
	Exposed to recent abuse	34
	Yes		2/34 (5.9%)
Newborn	Sex assigned at birth	43
	Female		23/43 (53.5%)
	Gestational age at birth (wks)	41
	Mean ± SD		39.7 ± 1.4
	Weight at birth (g)	43
	Mean ± SD		3345.8 ± 435.8
	Delivery type	43
	Vaginal		21/43 (48.8%)
	C-section		20/43 (46.5%)
	Assisted [Forceps]		2/43 (4.7%)
	5-minute Apgar score	42
	Mean ± SD		9.8 ± 0.5

CTQ: Childhood trauma questionnaire; SD: Standard deviation; g: Gram; wks: Week.

Mothers in our study reported childhood maltreatment scores ranging from 25 to 86. Upon categorizing the scores into levels reflecting none, mild, moderate and severe trauma, we found that 19/43 mothers were categorized as highly exposed to CM, exhibiting moderate or severe trauma in at least one trauma domain (Supplementary Figure S1).

Out of the 43 dyads assessed, data on exposure to recent abuse was available for 34 of them. Two mothers reported experiencing recent physical abuse in the 12 months preceding their enrollment, with one of them reporting instances of physical abuse during pregnancy. Thus, we inferred a low prevalence of recent trauma in our sample and initially conducted our analyses treating the missing data as negative (n = 9). However, considering the potentially high impact of recent trauma on an individual's physiology and the likelihood of masking the effects of early-life trauma, we removed the dyads of the two mothers exposed to recent abuse from all further analyses.

Furthermore, two newborns presented low birth weight (below 2.5 kg) and were also removed from our investigation. The Spearman's test revealed a weak correlation between gestational age at birth and newborn's birth weight (rho = 0.35, p = 0.02, Supplementary Figure S2), thus they were both kept as covariates in LM in perinatal tissues.

3.2. Association of miRNA expression with maternal exposure to CM

Our investigation comprised a total of 450 miRNAs in maternal tissue, 561 miRNAs in umbilical cord blood and 667 miRNAs in the placenta. Principal component analysis of VST-normalized reads for each tissue is shown in Supplementary Figures S3–S7. This analysis revealed a slight clustering for socioeconomic level in maternal blood sample and for newborn's biological sex in umbilical cord blood sample. These variables were included in the respective models as covariates, aiming to account for these variations. Additionally, one umbilical cord sample and two placental samples had a slightly different expression pattern compared with the others in the same tissue, however we decided to retain these samples in the analyses as no technical variable explained such profiles, suggesting this could be related to a biological diversity. The 30 most abundant miRNAs in our samples are shown in the Supplementary Figure S8.

Considering: i) the removal of dyads whose mother had a history of recent trauma (n = 2); ii) the removal of newborns with low birth weight (n = 2); and iii) missing values for gestational age (n = 2); out of a sample of 43 mother-infant dyads, a total of 41 maternal blood samples, 37 umbilical cord blood samples and 37 placental samples proceeded to the association tests between the maternal exposure to CM and miRNA expression.

The summary of top 5 miRNAs ranked by p-value for each tissue is listed in Table 3 and the results for all models are provided in detail in the Supplementary Spreadsheet S1–S3. Q-Q plots of -log₁₀(p) for the models are shown in Figure 1A (for umbilical cord blood) and Supplementary Figure S9 (for the other tissues).

Table 3.

Top 5 miRNAs in association tests with child maltreatment by tissue.

Tissue	miRNA	Estimate (β)	Standard error	t	p-value	Adjusted p-value	Adjusted R2
Maternal blood	hsa-miR-6868-3p	-0.0898	0.0261	-3.4412	0.0015	0.3480	0.1887
	hsa-miR-4766-3p	-0.0513	0.0149	-3.4340	0.0015	0.3480	0.1707
	hsa-miR-4521	-0.0255	0.0085	-2.9832	0.0052	0.6629	0.2695
	hsa-miR-151a-3p	0.0166	0.0061	2.7114	0.0103	0.6629	0.1141
	hsa-miR-485-3p	-0.0386	0.0147	-2.6291	0.0126	0.6629	0.1649
Cord blood	hsa-miR-582-3p	-0.0401	0.0088	-4.5826	0.00008	0.0453	0.3337
	hsa-miR-6511b-3p	0.0221	0.0061	3.6364	0.0011	0.2937	0.3107
	hsa-miR-654-3p	-0.0252	0.0072	-3.4887	0.0016	0.2937	0.2968
	hsa-miR-185-5p	0.0188	0.0057	3.2764	0.0027	0.3001	0.3311
	hsa-miR-6803-3p	0.0178	0.0055	3.2297	0.0031	0.3001	0.1778
Placenta	hsa-miR-20b-5p	-0.0261	0.0071	-3.6916	0.0009	0.3518	0.3090
	hsa-miR-142-5p	-0.0193	0.0053	-3.6393	0.0011	0.3518	0.3065
	hsa-miR-7976	-0.0113	0.0034	-3.2741	0.0027	0.6101	0.2968
	hsa-miR-409-5p	-0.0081	0.0026	-3.1114	0.0042	0.6596	0.2131
	hsa-miR-21-3p	-0.0174	0.0057	-3.0425	0.0049	0.6596	0.1439

Significant models after adjustment for multiple comparisons (adjusted p-value < 0.05) are listed in bold.

Figure 1.

Plots of the linear regression model for hsa-miR-582-3p in umbilical cord blood. Dyads whose mothers reported recent physical/sexual abuse (n = 2) were not included in this model. (A) Scatterplot showing the significant association between CTQ score (x-axis) and hsa-miR-582-3p expression (y-axis) in umbilical cord blood (β = -0.04, adjusted p = 0.045, adjusted R² = 0.33). In the upper right, the Q-Q plot of -log₁₀(p) of all regression models ran for umbilical cord blood. (B) Forest plot of the linear regression model for hsa-miR-582-3p in umbilical cord blood, showing that no covariate was associated with hsa-miR-582-3p. All models in umbilical cord blood were covaried for maternal age, newborn biological sex, delivery type, newborn weight at the birth and gestational age at birth.

Notably, we observed that the hsa-miR-582-3p was negatively associated with maternal exposure to CM (F(29) = 3.576, β = -0.04, FDR = 0.048, adjusted R² = 0.33) in the umbilical cord blood, a pattern also observed as a tendency in the other two tissues. The decreased expression of hsa-miR-583-3p in umbilical cord blood in relation to the maternal childhood maltreatment score is depicted in Figure 1A and the model quality parameters are shown in Supplementary Figure S10. The VIF scores for all independent variables of hsa-miR-582-3p model were below 5, indicating no multicollinearity concern (Supplementary Table S1). Other 34 miRNAs presented p-values <0.05 in umbilical cord blood but did not withstand correction for multiple comparisons. A total of 22 and 39 miRNAs presented p-value < 0.05, respectively, in maternal blood and placenta. However, none remained significant after controlling for FDR.

To confirm that our findings were not biased by missing data for maternal exposure to recent abuse (n = 9 dyads), we rerun the previously presented models for all tissues including only those dyads whose mothers were screened for recent abuse and reported having no exposure (n = 32 in maternal tissue, n = 29 in perinatal tissues). The hsa-miR-582-3p remained negatively associated with maternal exposure to CM in umbilical cord blood after adjusting for 561 comparisons (β = -0.03, adjusted p = 0.012, R² = 0.52) (Supplementary Figure S11). A summary of the results of this association test for each tissue is provided in Supplementary Table S2 & Supplementary Spreadsheet S4–S6.

3.3. hsa-miR-582-3p targets genes & their pathways

To explore the pathways and mechanisms regulated by hsa-miR-582-3p, we investigated the target genes using the miRWalk database. A total of 704 genes were found (Supplementary Spreadsheet S7). We found 181 GO Biological Processes, 18 GO Molecular Function, 24 GO Cellular Component, 2 KEGG pathways and 33 GWAS catalog gene sets statistically significant for the target genes of hsa-mir-582-3p. Among the most interesting, we highlight the following terms: regulation of cell differentiation (FDR = 1.50^-6 and FDR = 1.62^-4), cell development (FDR = 1.16^-3 and FDR = 2.72^-3), cell cycle (FDR = 1.16^-3, FDR = 1.45^-3 and FDR = 2.16^-3), cellular response to stress (FDR = 5.76^-4 and FDR = 4.64^-3), apoptotic process (FDR = 1.85^-3), embryo development (FDR = 2.02^-5 and FDR = 4.62^-4), in utero embryonic development (FDR = 0.001), regulation of response to DNA damage (FDR = 3.17^-3) and early spontaneous preterm birth (FDR = 0.024) in GWAS catalog (Supplementary Spreadsheet S8). The main enrichment pathways are shown in Figure 2.

Figure 2.

Gene ontology pathway enrichment by hsa-miR-582-3p regulated genes. (A) Treeplot showing the clustering of Gene Ontology Biological Process (GOBP) terms of miR-582-3p target genes. 20 terms from the top 35 (p-adjusted < 0.005) GOBP terms were selected for graphic representation. (B) Concept-network plot of selected GOBP terms and their corresponding hsa-miR-582-3p target genes.

3.4. Association of candidate miRNA expression with maternal maltreatment

Across all tissues, we identified 112 unique sequences with the potential to be annotated as novel miRNAs, with an average length of 22 nucleotides consistent with Drosha and Dicer processing. Out of these sequences, 56, 68 and 15 sequences were identified in maternal, umbilical cord and placental samples, respectively.

Among these candidate miRNAs, 27 were found with sequence overlap in maternal and umbilical cord samples, suggesting they are expressed in blood. In contrast, all candidate miRNAs in the placenta were exclusively found in this tissue. A candidate for novel miRNA (UGACUUCUUAUUCUUUCCUGUG) was detected in 98.8% of the maternal and umbilical cord blood samples. In maternal blood, 22 (39.29%) candidate miRNAs were expressed in at least three samples, while, in umbilical cord blood, 33 (48.53%) candidate miRNAs were detected in three or more samples. The genomic coordinates, genomic region, sequence, host gene and number of samples expressing these candidates are detailed in the Supplementary Spreadsheet S9–S11.

To test the association of candidate miRNAs with maternal exposure to CM, we selected the miRNAs present in at least half of the samples to proceed to the LM models. Thus, there were five, three and two candidate miRNAs remaining for analysis in maternal blood, umbilical cord and placenta, respectively. However, none of these candidate miRNAs were associated with maternal exposure to CM (Supplementary Table S3 & Supplementary Spreadsheets S12–S14).

4. Discussion

In our study, we explored the relationship between maternal exposure to CM and miRNA expression levels in tissues from both mothers directly exposed to CM and their newborns. Our analysis revealed a negative association between maternal history of CM and hsa-miR-582-3p in umbilical cord blood.

Analyses on FUMA GWAS showed enrichment of hsa-miR-582-3p target genes in a GWAS for spontaneous early preterm birth (EFO_0006917) and pathways related to embryonic development, pointing to the role of this miRNA in the in utero period. Indeed, hsa-miR-582-3p has been associated with maternal and fetal complications during pregnancy, such as preterm birth, preeclampsia and small-for-gestational-age newborns [31]. However, the exact biological mechanisms underlying the effect of this miRNA in these conditions are unclear.

The same FUMA GWAS analysis also pointed to multiple target genes of hsa-miR-582-3p involved in essential pathways for fetal development, such as cell differentiation and development, regulation of the cell cycle, apoptotic process and cellular response to DNA damage and to stress. These are important pathways for fetal development and point to the possible role of hsa-miR-582-3p as a molecular regulator during this period.

Aligned with these potential roles of hsa-miR-582-3p, previous studies have revealed that lower expression of hsa-miR-582-3p is associated with increased levels of components of TGF-β signaling pathway in metastatic bone tissue of prostate cancer, resulting in the promotion of tumor invasion and migration [32]. In our study, we did find enriched predicted target genes for the TGF-β signaling pathway, which suggests that miR-582-3p does regulate genes within this pathway. Additionally, the high expression of hsa-miR-582-3p has been shown to reduce levels of TNF-α and IL-1β [33], pro-inflammatory cytokines known to play a crucial role in promoting low cortisol levels [34] and negatively regulating the HPA axis through the modulation of glucocorticoid receptors [35]. Such physiological changes have been documented in individuals exposed to early-life adverse events [36]. Consequently, lower levels of this miRNA in newborns could lead to higher cortisol levels, which have been associated with alterations in behavior, stress response and overall development [37], suggesting a potential impact of hsa-miR-582-3p on neurodevelopment. Additionally, changes in cortisol can lead to dysregulation of immunological processes [36]. Indeed, hsa-miR-582-3p regulates TGF-β, a crucial immune system regulator [38], indicating both direct and indirect (via changes in cortisol) effects of this miRNA on inflammation.

Therefore, our findings, combined with the literature, suggest that newborns born to mothers with a history of CM exhibit a downregulated hsa-miR-582-3p profile in umbilical cord blood. This could compromise newborn health and brain development by impacting immunologic and stress-related regulation during a critical developmental period.

In our analysis of maternal blood and placenta samples, we did not observe any significant associations between the examined miRNAs and maternal history of CM. However, it is noteworthy that the expression of miR-582-3p exhibited a consistent negative relationship with maternal CM across these two tissues, aligning with the association observed in umbilical cord samples, reinforcing this observation.

In placental tissue, while hsa-miR-20b-5p did not remain significant after FDR correction, its relevance is underscored by its associations with small-for-gestational-age infants [39] and preeclampsia conditions [40,41]. This miRNA plays a pivotal role in regulating placental angiogenesis, trophoblastic differentiation, host defense mechanisms, cell-cell communication [40] and modulating trophoblastic cell invasion [41]. These are essential roles for conducting a healthy pregnancy and proper fetal development. Although our study did not find a statistically significant association between hsa-miR-20b-5p and maternal exposure to CM, its biological significance warrants further investigation in studies with larger sample sizes.

We also investigated whether the sequence fragments not aligned to miRBase reference could be candidate miRNAs and tested their association with maternal history of CM. Since these miRNAs are shared among different samples and even among different tissues, we believe they are true miRNAs and have biological roles. Although we did not find candidate sequences associated with maternal exposure to CM, we believe this data could be valuable for the scientific community interested in identifying and annotating novel miRNAs.

Regarding contextual factors, the women in our study predominantly presented low education attainment and fall within the two lowest socioeconomic strata. These two factors are often intertwined [42] and have been shown to influence health behaviors [43], potentially affecting dynamic biological phenotypes such as miRNA expression levels [44]. However, in our sample, these factors were not associated, likely due to the small sample size. Thus, we treated them as individual covariates in our models, recognizing their independent influence on miRNA profile. Furthermore, our chosen maternity hospital serves as a reference point for mothers residing in nearby areas, promoting homogeneity in exposure to community violence levels.

In our examination of perinatal tissue, two newborns were excluded due to low birth weight. While low birth weight has been suggested to correlate with our primary predictor, childhood trauma [45], it may also be indicative of unassessed underlying conditions [46,47]. Given the focus of this study on low-risk pregnancies to avoid the influence that such conditions may exert over miRNA expression, we opted to exclude the two newborns with low birth weight from our analyses.

Similar considerations guided our approach to the two mothers who reported experiencing physical abuse within the 12 months preceding their enrollment, one of whom also disclosed exposure to physical abuse during pregnancy. Acute stress induced by recent trauma elicits various biological coping and adaptation responses in the mother, impacting fetal exposures, including heightened cortisol levels [48], and mainly miRNA profile [49]. These physiological responses have the potential to obscure the effects of early-life trauma and introduce bias into our analyses. To ensure precise capture of the effects of early-life adverse experiences, without the biased influence of recent events, we opted to exclude the samples from these mothers and their newborns from all analyses. Additionally, accounting for the bias of including dyads whose mothers had not been assessed for recent trauma, we excluded them from our analyses and the association of maternal exposure to CM with hsa-miR-582-3p levels remained significant, reaffirming the robustness of our findings.

The miRNAs are fine and transient regulators of gene expression, suggesting that early-life stress may not directly induce long-term changes in the miRNA profile. However, previous studies have documented enduring impacts of CM on miRNA profiles [20,21,50], although the precise mechanisms underlying this phenomenon remain unclear. Considering the life-long impact of CM on HPA axis [51] and immunological dysregulations [52,53], we hypothesize that those physiological alterations may contribute to the observed changes in miRNA profiles among individuals exposed to CM, and even their next generation.

Regarding the impact of maternal history of CM on the next generation, a study in a mouse model suggested a transgenerational effect of paternal history of CM through the transmission of altered miRNA profiles in sperm [54]. Nevertheless, research on the impact of parental exposure to CM on offspring's miRNA expression profile in humans remains limited.

In our study, we identified long-term effects of maternal exposure to CM on miRNA levels in umbilical cord blood. To our knowledge, this is the first study to identify, in humans, alterations in miRNA expression levels in the offspring of women exposed to CM. Nonetheless, the precise biological mechanisms underlying the link between maternal CM and the expression of hsa-miR-582-3p require further investigation. Moreover, future studies could verify if the hsa-miR-582-3p profile could serve as a biomarker for identifying at-risk newborns. If they indicate a clear impact of hsa-miR-582-3p on newborn's health and development, early identification of an altered expression pattern of this miRNA in the umbilical cord blood of newborns born to mothers with history of CM could prompt closer monitoring and early interventions, aiming to mitigate potential adverse outcomes for the offspring.

Additionally, our study highlights the long-term impact of maternal CM on offspring, underscoring the importance of addressing CM as an intergenerational concern. Preventive strategies aimed at reducing CM and supporting affected individuals could have far-reaching benefits. Integrating comprehensive mental health assessment and support into prenatal and postnatal care for affected mothers could help improve both maternal and child health outcomes.

We acknowledge several limitations in this study. First, the retrospective feature of measuring maternal childhood traumatic events, while validated and reliable, may introduce memory bias. Second, replicating this study in a larger sample and accounting for the number of included tissues in the correction for multiple comparisons would enhance the robustness of the findings and enable more detailed hypothesis testing. Third, future studies should account for the influence of other potentially relevant variables, such as maternal consumption of tobacco and other substances. Last, our investigation focused on the fetal programming hypothesis, thus we did not explore the impact of the paternal contribution and other environmental features on the offspring's miRNA profile.

5. Conclusion

The association between maternal exposure to CM and miRNA expression levels in maternal and perinatal tissues, as explored in our study, revealed a significant negative correlation between hsa-miR-582-3p and maternal exposure to CM in umbilical cord blood. However, we observed no associated miRNAs in maternal blood and placenta. The enrichment analysis of target genes associated with hsa-miR-582-3p in GWAS for spontaneous early preterm birth and pathways related to embryonic development underlines its potential role in intrauterine development. Despite the limitations imposed by our sample size, our study provides valuable insights into the potential impact of maternal history of CM on the miRNA expression profile of the offspring. Further research is needed to elucidate the precise biological mechanisms underlying this association and their implications for child development and health.

Funding Statement

This manuscript was funded by the Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP, São Paulo Research Foundation) under grant #2019/21612-0. Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) provided support to Danilo Micali and Joice Santos Rosa (Finance Code 001). FAPESP also supported Pedro Alexandre Favoretto Galante and Rafael Luiz Vieira Mercuri under grants #2018/15579-8 and #2020/02413-4, respectively.

Supplemental material

Supplemental data for this article can be accessed at https://doi.org/10.1080/17501911.2024.2401318

Author contributions

D Micali and J Santos Rosa drafted the manuscript and designed the figures with input from all authors. These authors contributed equally to this work. D Micali, I Silva, CS Duarte, J Posner, AP Jackowski and SI Belangero contributed to the design of the study sample. D Micali, JS Rosa, VK Ota, AVG Bugiga, AC Coelho Milani, CS established methods and logistics for data and sample collection under supervision and guidance of I Silva, SI Belangero, CS Duarte, J Posner and AP Jackowski. D Micali, C Salmeron, AC Coelho Milani performed data collection and biospecimen collection and processing. D Micali, JS Rosa, VK Ota and PF Asprino performed required bench experiments. D Micali, JS Rosa, RL Vieira Mercurid performed formal data analysis, with contributions of VK Ota, L Spindola, GS Kajitani and PA Favoretto Galante. All authors critically reviewed and approved the final version of the manuscript, as well as, agreed upon the journal for its publication.

Financial disclosure

This manuscript was funded by the Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP, São Paulo Research Foundation) under grant #2019/21612-0. Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) provided support to Danilo Micali and Joice Santos Rosa (Finance Code 001). FAPESP also supported Pedro Alexandre Favoretto Galante and Rafael Luiz Vieira Mercuri under grants #2018/15579-8 and #2020/02413-4, respectively. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests disclosure

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Writing disclosure

The authors have used the OpenAI ChatGPT v3.5 for proofing of specific sections of the manuscript. All suggested changes have been individually and extensively reviewed by the authors to guarantee that no unwanted changes in meaning or writing style would be made.

Ethical conduct of research

Informed written consent was acquired from eligible women following their comprehension of the participant information sheet. Enrolled women also assented the participation of their newborns in the study. Enrollment of newborns occurred only after confirming the mothers' ongoing interest in participating, a confirmation obtained during hospital admission for childbirth. Throughout the study, participants received no compensation or financial incentives, except for reimbursing transport expenses for attendance at neuroimaging or other behavioral assessments conducted in our laboratory facilities. The ethical approval for the study was granted by the Comissão Nacional de Ética em Pesquisa (CONEP, National Commission of Ethics in Research, 78018417.2.0000.5505) and Comitê de Ética em Pesquisa da Universidade Federal de São Paulo (CEP/UNIFESP: 1200/2017).

Data availability statement

Summary statistics generated by the analyses presented are provided in the Supplementary Spreadsheets. Other raw data that support the findings of this study are available from the corresponding author, SIB, upon reasonable request. Raw data are not publicly available due to restrictions in the written consent signed by the participants of our study.