Psychiatric symptoms frequently co-occur, meaning that having one condition increases the risk of developing another (1). Moreover, psychiatry has struggled to define clear boundaries between disorders, and between normal and pathological variation, suggesting that the diagnostic categories may be inadequate (2).

This has led researchers to increasingly focus on transdiagnostic approaches, such as dimensional models, to better understand the complex presentations of mental health. These models often combine into higher-order factors, like internalizing and externalizing symptoms For instance, the internalizing dimension reflects the overlap between depression and anxiety symptoms, while the externalizing dimension captures disruptive disorders and substance abuse (1). Additionally, a general factor of psychopathology, the “p factor,” has been suggested, which captures the shared variance between these dimensions (1, 3).

Previous research found evidence for both genetic and environmental influences on psychiatric symptoms. Epidemiological studies have shown that almost all forms of psychopathology are partly heritable and polygenic, transcending diagnostic boundaries (2, 4), and supporting a transdiagnostic structure of psychopathology. Also, genome-wide association studies (GWAS) have demonstrated that genetic influences are often pleiotropic, affecting multiple psychiatric outcomes (5, 6). However, recent GWASs on internalizing (7), externalizing symptoms (8), and general psychopathology (9), have shown a low single nucleotide polymorphism (SNP) heritability (1.7% - 23.5%), indicating the influence of other genetic and environmental factors on these dimensions. For instance, data obtained from the Brazilian High-Risk Cohort Study for Mental conditions (BHRCS) demonstrated that an increase in stressful life events (SLE) (10), child maltreatment (CM) (11), or threat and deprivation (12) was associated with elevated psychopathology.

However, the molecular mechanisms underlying the complex interplay of genetic and environmental factors in the development of psychopathology remain a subject of ongoing debate. Alterations in gene expression, as well as epigenetic modifications that regulate it, have emerged as promising candidates to elucidate this intricate relationship, given that: i) genetic variants associated with psychopathology are enriched for regulatory elements, suggesting a role in gene expression modulation (13, 14); and ii) gene expression and epigenetic changes are related to both environmental factors and psychopathology (15–18). Of particular interest, a growing body of research has associated exposure to early-life adversities (ELA) with significant changes in gene expression, particularly in genes involved in inflammation and immune system function (19–21). This compelling evidence points towards altered gene expression in leukocytes as a potential mechanism linking immune dysregulation to the increased risk of poor health outcomes commonly observed in individuals who have experienced ELA (20, 22). Supporting this notion, the BHRCS identified a significant indirect effect of the aggregate expression of four genes related to the hypothalamic-pituitary-adrenal (HPA) axis and inflammation, mediating the association between CM and major depressive disorder (MDD) (16). Beyond the impact of ELA, other environmental stressors, such as food restriction and low socioeconomic status, have also been demonstrated to influence gene expression patterns (23, 24). Additionally, a transcriptome-wide association study revealed significant genotype-by-environment interactions, wherein the effects of psychosocial stressors (i.e., deprivation and cumulative lifetime stress) on the risk of MDD were modulated by imputed gene expression levels in both whole blood and brain tissues (25). However, no study to date has comprehensively investigated the longitudinal dynamics of blood gene expression and its interaction with psychopathology and life adversities, nor the interplay between these factors in a cohort of children and adolescents.

Thus, this study aims to investigate within the BHRCS the longitudinal association between blood candidate gene expression and adversity (i.e. SLE, CM, threat and deprivation) or psychopathology (total Child Behavior Checklist (CBCL) score, p factor, internalizing and externalizing problems). Additionally, we aim to investigate the potential mediation or moderation role of gene expression in the relationship between life adversities and psychiatric symptoms ( Supplementary Figure S8 ).

This study included 298 participants from the BHRCS, selected based on high-quality RNA data at both time point 1 and time point 2. The sample comprised more male participants (n=181, 60.7%) and participants from São Paulo state (n=176, 59.1%).

Comparisons between time point 1 and time point 2 are shown in Table 1 . The prevalence of individuals with high CM significantly decreased from time point 1 (66/22.1%) to time point 2 (43/14.4%; p=0.014). Participant age ranged from 5 to 14 years at time point 1 and 9 to 17 years at time point 2. Internalizing domain scores increased significantly from time point 1 (mean=-0.03, SD=0.30) to time point 2 (mean=0.03, SD=0.32; p<0.001).

Additionally, this subsample of 298 participants demonstrated a significant association between life events and psychopathology measures ( Supplementary Figure S2 ), consistent with findings from the original study population (12). Also, the threat and deprivation variables calculated for time point 1 are positively correlated with SLE calculated at time point 2 ( Supplementary Figure S3 ). Finally, individuals in the high CM group exhibit more SLE, threat and deprivation than those in the low CM group ( Supplementary Figure S4 ).

Of the 88 target genes, six endogenous genes and four with low expression (fewer than 10 reads in at least 10 samples) were excluded. This resulted in a final set of 78 genes for differential expression analysis. Principal component analysis revealed an effect of the sequencing run, leading us to include this variable as a covariate in further analysis ( Supplementary Figure S5A ). While no clustering by time point was identified ( Supplementary Figure S5B ), a total of 22 genes exhibited significant associations with time point ( Supplementary Table S5 ), suggesting that the expression of these genes might change over time.

We investigated the association between expression levels of the 78 target genes and psychopathology, as measured by the total CBCL score, the p factor, internalizing and externalizing specific factors. Analyzing each time point separately, we observed a significant negative association between USP38 gene expression and externalizing symptoms at time point 1 ( Table 2 ; Supplementary Figure S6 ). At the time point 2, we found significant positive associations between internalizing symptoms and FAR1 expression; and negative associations between internalizing symptoms and the expression of NR3C1, HSPBP1, SMAD4, CRLF3 and SIN3A genes ( Table 2 ; Supplementary Figure S6 ). However, mixed models revealed no longitudinal changes in the relationship between gene expression and psychopathology.

Regarding the association between gene expression and measures of life adversities, we identified a negative correlation at time point 1 between deprivation and the expression of both DENND11 (or KIAA1147) and PRRC1 ( Table 2 ; Supplementary Figure S6 ). No associations with other measures of life adversities were identified.

Notably, all nine genes were correlated with each other ( Supplementary Figure S7A ), suggesting potential co-expression or involvement in shared pathways. GeneMania analysis confirmed this, showing that all genes except HSPBP1 are related by co-expression networks ( Supplementary Figure S7B ). In particular, NR3C1, SIN3A, and SMAD4, all associated with internalizing symptoms, are correlated with each other (NR3C1-SIN3A: r=0.594, p=2.2E-16; NR3C1-SMAD4: r=0.673, p=2.2E-16; SIN3A-SMAD4: r=0.797, p=2.2E-16; Supplementary Figure S7A ). Furthermore, these three genes are enriched in two Reactome pathways: FOXO-mediated transcription of oxidative stress, metabolic, and neuronal genes (R-HSA-9615017; 3/29 genes, p=1.277E-07, FDR=3.946E-04) and FOXO-mediated transcription (R-HSA-9614085; 3/64 genes; p=1.447E-06, FDR=0.001).

Analysis of gene expression databases ( Supplementary Figures S7C, D ) revealed ubiquitous expression of HSPBP1 (average log2TPM from 3.527 in blood to 5.628 in the uterus; average log2TPM of 5.056 across multiple tissues), with high levels also observed in brain tissues (average log2TPM=5.345). Moreover, DENND11 (average log2TPM = 4.058), PRRC1 (average log2TPM = 3.740), SIN3A (average log2TPM = 3.870), NR3C1 (average log2TPM = 4.002), SMAD4 (average log2TPM = 4.332), and FAR1 (average log2TPM = 4.536) were also highly expressed in multiple tissues. Of note, the average expression across multiple tissues was particularly similar for NR3C1, SMAD4, and SIN3A genes ( Supplementary Figure S7C ). Interestingly, USP38 expression appears to be more tissue-specific, being highly expressed in muscle (average log2TPM=4.297), whereas CRLF3 is lowly expressed in this tissue (average log2TPM=1.398) and highly expressed in the spleen (average log2TPM=4.406).

Analysis of gene expression in the BrainSpan database ( Supplementary Figure S7D ) showed that HSPBP1 is also highly expressed in the brain across all developmental stages (average log2RPKM =5.146 in multiple developmental stages), similar to DENND11, although the latter showed more moderate expression (average log2RPKM =3.331 in multiple developmental stages). Notably, while the expression of SIN3A, SMAD4, and PRRC1 appeared higher in prenatal stages (peaking in early prenatal stages: SIN3A log2RPKM=3.878, SMAD4 log2RPKM=4.232, PRRC1 log2RPKM=3.336), NR3C1 expression increased in postnatal stages (peaking in young adulthood: log2RPKM=3.222), similar to FAR1 (peaking in young adulthood, log2RPKM=3.191).

We investigated whether the nine genes were mediators or moderators between the association of life adversities and psychiatric symptoms ( Supplementary Figure S8 ). We tested four mediation or moderation models (considering CM, SLE, threat, and deprivation) for each of the NR3C1, SIN3A, SMAD4, FAR1, CRLF3, HSPBP1 and USP38 genes, which were associated with psychiatric symptoms. We also tested four mediation or moderation models (considering total CBCL score, p-factor, internalizing, and externalizing symptoms) for each of the DENND11 and PRRC1 genes, which were associated with deprivation. This resulted in a total of 36 mediation and 36 moderation models tested. We identified HSPBP1 as a mediator of the association between CM and internalizing symptoms at time point 2 (estimate =0.013; p=0.048). Moreover, we found that DENND11 expression moderated the relationship between deprivation and the p factor (estimate = -0.080, p = 0.0407), CRLF3 expression moderated the relationship between CM and internalizing symptoms (estimate = -0.044, p = 0.043), and NR3C1 expression moderated the relationship between deprivation and internalizing symptoms (estimate = -0.049, p = 0.029) at time point 2. However, these associations were not significant after adjusting for multiple comparisons (adjusted p = 1.000).

We conducted targeted RNA sequencing to identify genes associated with changes in psychopathology and life adversities. Of the 78 genes examined, six genes (NR3C1, SMAD4, SIN3A, FAR1, CRLF3 and HSPBP1) were associated with internalizing symptoms and one (USP38) with externalizing symptoms. Additionally, we identified two genes (DENND11 and PRRC1) associated with deprivation, a latent factor constructed from variables such as socioeconomic status, family income, mother’s educational level, neglect, and father presence. Seven genes (SMAD4, SIN3A, CRLF3, USP38, FAR1, DENND11 and PRRC1) were selected based on our previous microarray analysis in the Sao Paulo subsample of the BHRCS (N=103 at baseline and N=103 at follow-up). The results from the current study align with our previous findings, with SMAD4, SIN3A, CRLF3, USP38, DENND11 (previously known as KIAA1147) and PRRC1 expression decreasing in individuals who transitioned from low to high psychopathology (as measured by the CBCL) ( Supplementary Table S2 from (15)). Consistent with this, the first three were negatively associated with internalizing symptoms, USP38 with externalizing symptoms and DENND11 and PRRC1 were negatively correlated with deprivation in the present study ( Table 2 ). FAR1, on the other hand, showed increased expression in individuals who transitioned from high to low psychopathology ( Supplementary Table S3 from (15)), contrasting with the positive correlation between FAR1 expression and internalizing symptoms observed at time point 2 in the current study. Finally, both NR3C1 and HSPBP1 genes were selected because they were differentially expressed in the brains of individuals with mental disorders (27), with NR3C1 also being differentially expressed in the blood of BHRCS individuals with major depressive disorder (16). After adjusting for multiple comparisons, no significant mediation or moderation effects of gene expression on the association between life adversities and psychiatric symptoms were found for these genes.

Adolescence is a period that is particularly vulnerable to environmental influences, and traumatic experiences that occur before or during this stage can significantly impact the likelihood of developing externalizing or internalizing behaviors. Such experiences leave a lasting impact on various biological systems, affecting neuronal, endocrine, immune, transcriptomic, and epigenetic processes (35). Emerging research in genomics, epigenomics, and transcriptomics aims to understand why some individuals are more susceptible to developing certain mental health issues after experiencing trauma (36, 37). For instance, a transcriptome-wide study found that the expression of genes related to the immune system may mediate the link between emotional abuse and MDD (38). Additionally, studies in animal models have shown that patterns of gene expression in cortical and amygdala pyramidal neurons can predict whether fear extinction will be adaptive or maladaptive. which can help to understand trauma-related disorders (39). These findings highlight the complex interactions between biological and environmental factors in shaping mental health outcomes during adolescence.

Notably, traumatic events can disrupt the activity of the HPA axis, which is reflected in cortisol levels. This disruption can lead to changes in functional connectivity in the brain and is associated with internalizing problems later in life (40). Cortisol activates glucocorticoid receptors, which are expressed throughout the brain and can function as transcription factors, regulating genes related to metabolism and immune function (41). In our study, we found the NR3C1 gene (nuclear receptor subfamily 3, group C, member 1), which encodes the glucocorticoid receptor, to be associated with internalizing. Many studies have associated it with CM and depressive symptoms, particularly those investigating NR3C1 methylation (42, 43). One of the first studies to investigate this in humans observed decreased expression of NR3C1, along with increased methylation of its promoter, in the hippocampus of suicide victims with a history of childhood abuse (44). Changes in NR3C1 gene expression in blood have also been detected, with some studies identifying increased expression in patients with depression (45) and others finding a negative correlation with depressive scores (46). Our previous study in the BHRCS at time point 1 evaluated gene expression in 20 children and adolescents with major depressive disorder (MDD), 49 with no MDD but with high levels of depressive symptoms, and 61 healthy controls. This study identified decreased expression of NR3C1 in those with MDD compared to both the depressive symptom group and healthy controls (16), corroborating our current findings at time point 2.

The SIN3A gene (SIN3 transcription regulator family member A) encodes a transcriptional regulatory protein. It appears to play an important role in cell cycle events and the proliferation of embryonic stem cells, and it is highly expressed throughout brain development (47), consistent with our analysis in BrainSpan. Indeed, functional studies have revealed that SIN3A is a key regulator of cortical expansion and maturation, and its haploinsufficiency has been associated with intellectual disability and autism spectrum disorder (47). Apart from increased SIN3A protein levels in the hippocampus of suicide victims (48), no other association of this gene with mental disorders has been reported, particularly in blood.

The SMAD4 (SMAD family member 4) gene encodes a member of the SMAD family of signal transduction proteins. It is involved in signal transduction of the transforming growth factor beta (TGFβ) superfamily and pathogenic variants in this gene cause Myhre syndrome, which can include neurobehavioral phenotypes such as developmental delay, autism spectrum disorder, attention-deficit/hyperactivity disorder, and anxiety (49). Moreover, these mutations also disrupt neuronal morphogenesis in both mouse and human neurons (50). The TGFβ pathway has been associated with depression and stress, as it is an anti-inflammatory signal that exerts neuroprotective effects and influences memory formation and synaptic plasticity (51). Preliminary analysis showed that children with lower SMAD4 expression (predicted by genetic variants) exhibited a positive correlation between prenatal maternal depressive symptoms and amygdala volumes, while those with higher SMAD4 expression presented a negative correlation between prenatal maternal depression and amygdala volumes (52).

The FAR1 gene (fatty acyl-CoA reductase 1) encodes a rate-limiting enzyme in plasmalogen biosynthesis (53). Mutations in this gene have been identified in individuals affected by severe intellectual disability, early-onset epilepsy, microcephaly, congenital cataracts, growth retardation, and spasticity (54, 55). Although little is known about this gene, plasmalogens are abundant in the myelin sheath, and thus, plasmalogen homeostasis is likely important for myelination in different regions of the central nervous system (53). Indeed, impaired plasmalogen synthesis has already been implicated in schizophrenia (56–58). Interestingly, the role of FAR1 in psychopathology appears complex. In a previous study, we found that FAR1 expression was increased in individuals who transitioned from high to low psychopathology, as measured by the total CBCL score (15). However, in the current study, FAR1 expression was positively correlated with internalizing symptoms, suggesting a potential association with increased internalizing problems. This apparent discrepancy may be due to several factors, including differences in sample size and the distinct biological underpinnings of internalizing symptoms compared to the broader measure of psychopathology used previously. It highlights the need for further research to fully elucidate the role of FAR1 in different facets of psychopathology and across different developmental stages.

The CRLF3 gene (cytokine receptor-like factor 3) encodes a cytokine receptor-like factor. Studies in human and mouse cells have revealed that it is a critical regulator of neurogenesis, neuron survival, and dendritic development, and its deletion appears to cause developmental delay, intellectual disability, and autistic traits (59, 60). Although few studies have investigated this gene, its expression appears to be associated with Alzheimer’s disease (61) and epigenetic age acceleration (62).

The HSPBP1 gene (heat-shock 70-kd protein-binding protein 1) encodes a cochaperone of heat shock protein 70 (Hsp70) that inhibits the activity of Hsp70 ATPase. It is highly expressed in neurons but lowly expressed in astrocytes. Knocking down HspBP1 in neurons rescued neuropathology in a Huntington’s disease mouse model (63). Although no study investigated this particular gene in relation to mental disorders, alterations in Hsp70 have been reported in bipolar disorder (64) and MDD (65), and medications such as lithium and clozapine appear to alter Hsp70 levels (65).

The USP38 (Ubiquitin Specific Peptidase 38) gene encodes a deubiquitinating enzyme that acts on different cellular processes, such as the DNA repair, regulation of the cell cycle, and the immune response (66, 67). Recently, the variant rs7681616-C present in the USP38 gene was associated with an increase in the risk of developing schizophrenia (68), suggesting the importance of this region in psychopathology. Another study identified that USP38 blood expression was significantly downregulated in schizophrenia patients compared to biological siblings and unaffected controls (69), aligning with our findings of decreased USP38 expression in individuals with increased externalizing symptoms. Furthermore, a CpG site (cg01769344) located in a CpG island near the USP38 gene was identified within a co-methylated module associated with general psychopathology in a sample of 440 children from the Generation R cohort (18). Based on this evidence, we suggest that the USP38 gene may play a significant role in psychopathology.

While our knowledge of the DENND11 gene (formerly KIAA1147) remains limited, its protein product is predicted to participate in regulation of catalytic activity. By similarity, DENND11 is hypothesized to play a role in neuritogenesis (70), aligning with its elevated expression in multiple brain developmental stages and tissues ( Supplementary Figures S7 C, D ). Furthermore, abnormal DNA methylation levels and the presence of risk-associated single nucleotide polymorphisms (SNPs) have led to its consideration as a candidate gene for frontotemporal dementia with amyotrophic lateral sclerosis (71). Even less is known about the PRRC1 gene. It is predicted to be involved in the activation of protein kinase A activity and was previously associated with fluid intelligence in a genome-wide association study (GWAS) (72). Interestingly, the same study found that methylation of this gene in the temporal cortex was also associated with fluid intelligence.

We observed correlations among these nine genes ( Supplementary Figure S7A ); however, HSPBP1 did not appear to be correlated within the co-expression networks ( Supplementary Figure S7B ), which may be explained by its lower expression in blood ( Supplementary Figure S6B , S7C ). Of note, SIN3A, SMAD4, and NR3C1, despite being highly correlated ( Supplementary Figure S7A ), seem to exhibit similar expression patterns in multiple tissues ( Supplementary Figure S7C ) and are involved in two Reactome pathways: FOXO-mediated transcription of oxidative stress, metabolic, and neuronal genes, and FOXO-mediated transcription. None of the genes were associated with time point ( Supplementary Table S5 ) or with psychopathology in longitudinal models. This suggests that these results might be time-specific, particularly for internalizing symptoms, considering their increase at time point 2, when the mean age was 13 years ( Table 1 ).

The results of this study should be interpreted in light of certain limitations. First, the relatively small sample size may be an issue, warranting further replication studies. While few studies have investigated gene expression in longitudinal children and adolescent cohorts (73), most existing studies in this type of cohort (such as Generation R and Avon Longitudinal Study of Parents and Children) have focused on DNA variations, such as in GWAS (7–9), or DNA methylation (18). Second, this targeted RNA sequencing approach may have missed other genes relevant to psychopathology or be biased by cell type composition. Comprehensive transcriptome analysis could provide a more complete picture in future studies. Third, gene expression is tissue-specific, and our findings do not accurately reflect gene expression changes in the brain. Therefore, these results are more likely to be potential biomarkers of trait or state, and may not be clues about the pathophysiology. Furthermore, longitudinal gene expression analysis is not feasible in brain tissues. Finally, it is uncertain whether other factors that may influence gene expression, such as puberty or infection may have, affected our results, despite adjusting our analyses for sex, age, and time point.

In conclusion, our study has shed light on the complex interplay between genes, life adversities, and psychopathology in young individuals. We identified six genes associated with internalizing symptoms: NR3C1, HSPBP1, SIN3A, SMAD4, and CRLF3 exhibited a negative correlation with these symptoms, while FAR1 showed a positive correlation. Notably, NR3C1 has been consistently associated with internalizing symptoms and traumatic events, corroborating our findings. Additionally, decreased USP38 expression was associated with more externalizing problems, suggesting its possible role in regulating behavioral control or emotional reactivity. Furthermore, we observed lower expression of DENND11 and PRRC1 correlated with higher levels of deprivation, indicating a potential role in the biological response to environmental adversity. While the exact mechanisms remain unclear, these findings underscore the complex interplay between genetic predisposition and environmental stressors in shaping mental health outcomes. Although life adversities undeniably contribute to the development of psychiatric symptoms, our study did not find evidence for a direct mediating or moderating effect of gene expression on this relationship. This suggests that the influence of these genes on psychopathology may operate through independent pathways or interact with environmental factors in more nuanced ways. Importantly, our findings highlight the potential of specific gene expression patterns as early indicators of psychiatric vulnerability. The associations observed for NR3C1, HSPBP1, SIN3A, SMAD4, CRLF3, FAR1 and USP38 with internalizing and externalizing symptoms across multiple time points suggest their potential utility in predicting future mental health trajectories. Similarly, the association between DENND11 and PRRC1 expression and deprivation underscores the importance of considering both genetic and environmental factors in understanding and addressing mental health challenges.