Brain-derived Neurotrophic Factor (BDNF) is a protein involved in several biological pathways – from neurogenesis, promotion of neuronal survival and differentiation, to modulation of synaptic plasticity – playing a central role in both the developing and adult nervous system (Hempstead, 2014). Acting through its high-affinity tyrosine receptor kinase B (TrkB) receptor, it mediates neurite and spine outgrowth (Binder and Scharfman, 2004; Ji et al., 2005), and this signaling is also important for synaptic plasticity (Ji et al., 2005; Lu et al., 2014), a phenomenon that enables the organism to change according to environmental stimuli, and makes possible learning and memory. Also, it controls short and long-lasting synaptic interactions in the hippocampus, and its expression mediates working memory processes in the prefrontal cortex (Gold et al., 2003; Xing et al., 2012; Kowiański et al., 2018). BDNF is expressed in almost all brain regions, but the highest levels are found in the frontal cortex, hippocampus, and amygdala (West et al., 2014). Several studies indicate altered BDNF expression in brain structures like the prefrontal cortex (PFC), hippocampus, and striatum in post-mortem human brains of patients that suffered psychiatric illnesses. There are decreased levels of BDNF mRNA and protein expression in the hippocampus of suicide victims (Banerjee et al., 2013), and significant differences in BDNF transcripts allow to distinguish schizophrenia, bipolar disorder, and major depressive disorder patients from healthy subjects, suggesting that the BDNF system is implicated in several physiological aspects of brain development (Molendijk et al., 2012; Banerjee et al., 2013; Reinhart et al., 2015).

Prenatal exposure to stress, maternal depression/anxiety, low social support, and poor access to prenatal health services have long-term effects on child cognitive development that are well documented (Monk et al., 2012; O’Donnell et al., 2014a; Silveira et al., 2017). Brain plasticity and maturation are affected by positive and negative environmental exposures during sensitive periods of development (Nelson and Gabard-Durnam, 2020). The brain matures in a hierarchical manner, meaning that the quality of maturation of early-developing regions will affect the subsequent development of other regions (Tottenham, 2019). Gene expression in different brain regions at different developmental stages indicates that timing is an important factor at the transcriptome level (Somel et al., 2009; Haeussler et al., 2017). This makes complex cerebral regions, for instance, the PFC, particularly sensitive to environmental conditions. The PCF receives several inputs from all other cortical areas, playing a key role in planning and performance of higher thinking, cognitive, affective, and social behaviors throughout development (Kolb et al., 2012); such interconnectivity results in a longer period needed for maturation (Fuster, 2015).

Expression of BDNF, and TrkB receptors begins early during brain development, especially in the cortical plate, both in rodents and primates (for a review, see Bartkowska et al., 2010). Therefore, it is not surprising that disturbances in its function early in life have remarkable effects upon neuronal structure and function. For example, transgenic mice with a functional reduction in BDNF or TrkB genes have a curtailment of dendritic arborization in cortical neurons in the prepubertal period (Xu et al., 2000; Gorski et al., 2003), and impairments in memory (Gorski et al., 2003). In this scenario, studies using animal models of prenatal stress have reported altered BDNF signaling during post-natal development (Badihian et al., 2020; Sobolewski et al., 2020). Stressors such as maternal immune activation during gestation, repeated restraint, or variable stress during pregnancy, cause altered BDNF expression in the PFC at different ages during development of the offspring (Matrisciano et al., 2012; Hemmerle et al., 2015; Niu et al., 2020). Accordingly, prefrontal TrkB and glucocorticoid receptor (GR) activities are known to be modulated by exposure to stressors (reviewed in Barfield and Gourley, 2018), and TrkB-GR interaction has been suggested (Numakawa et al., 2009). Therefore, prolonged variations in glucocorticoids could affect both GR and BDNF-TrkB function in the PFC (Barfield and Gourley, 2018), contributing to stress-induced cognitive alterations. In addition, abnormal signaling in the BDNF/TrkB pathway was reported to lead to abnormalities in the GABAergic and glutamatergic activities in the PFC (Sakata et al., 2009; Zhang et al., 2013).

Negative exposures during the prenatal and early postnatal period have been associated with cognitive and brain development in different ways. Behaviorally, the attainment of cognitive skills is understood as a developmental cascade, characterized by a cumulative process in which functioning at a lower level of behavior (e.g., visuomotor integration, fine motor skills, habituation) affects higher-level functions that develop later (e.g., IQ, language and executive functions) (Almas et al., 2016; Choi et al., 2018; Camerota and Willoughby, 2019). In terms of neurodevelopment, experiments with infant rats exposed to caretakers that displayed abusive behaviors show increased levels of methylation of BDNF DNA throughout the life span, and reduced BDNF gene expression in the adult PFC (Roth et al., 2009). Prenatal exposure to stress was also associated with high methylation and lower expression of the BDNF gene in the PFC and hippocampus (Roth et al., 2011; Badihian et al., 2020). In humans, brain structure is also impacted by early life stressors, resulting in several morphological and functional alterations (Buss et al., 2010; Hair et al., 2015; Noble et al., 2015). The mentioned interrelated pathways affect the developing individual resulting in a predisposition for disease and poorer developmental outcomes later in life. Adolescence is also a sensitive period for PFC development. The PFC is one of the last brain regions to mature (Fuster, 2015; Hoops and Flores, 2017), and it is known to undergo significant structural remodeling, with dendritic and synaptic pruning during adolescence (Bourgeois et al., 1994; Shaw et al., 2020). This period of synaptic remodeling is believed to generate a refinement of connections (Barfield and Gourley, 2018). Therefore, exposure to adversities during this period can impact on PFC circuitry, and on adult behavior (Shaw et al., 2020). However, before this period of pruning, there is an initial phase of neuronal differentiation, dendritic spine and synapse overproduction, that occurs during prenatal and early childhood periods that will influence future development, stressing the relance of this sensitive window (Bourgeois et al., 1994; Lotfipour et al., 2009).

In summary, a large body of evidence indicates that early exposure to environmental adversity affects cognitive development, and some individuals are more susceptible than others to this long-term effect. Individual differences likely affect the impact of environmental exposure on several child developmental outcomes (Belsky, 2013; Silveira et al., 2017). It was shown that genetic variation in the BDNF gene (the Val66Met polymorphism), which decreases BDNF function (Egan et al., 2003), can lead to lower memory levels (Egan et al., 2003), and is associated with impairment in executive functioning (Nagata et al., 2012). This is particularly significant in individuals with high levels of early life adversity (Gabrys et al., 2017), in which this variation was associated with difficulties in attentional flexibility, a PFC-based function. Also, previous studies involving Val66Met polymorphisms suggested a role of the BDNF gene in moderating the effects of early adversity on attention problems and child behavior (Drury et al., 2012; Gunnar et al., 2012; O’Donnell et al., 2014b). However, it is known that the action of a gene is not isolated, but correlated in concert with other genes in functional networks (Gaiteri et al., 2014).

Genome-wide association studies (GWAS) are an important technological advance for the understanding of human health and disease but are still not able to inform the underlying tissue-specific mechanisms that explain phenotypic variation (Silveira et al., 2017; Hari Dass et al., 2019). GWAS considers only the highly significant genetic variants associated with a disease, thus are not enlightening of the several manifestations or endophenotypes that may precede the phenotype (Dalle Molle et al., 2017; Hari Dass et al., 2019). We propose a novel genomics approach, using a biologically informed genetic score based on genes co-expressed with the BDNF gene specifically in the PFC (BDNF ePRS) during the prenatal and early life periods to investigate the association with child cognitive development from 6 to 36 months of age. For a sub-sample of participants that we were able to follow up and collect structural Magnetic resonance images at age 9 we analyzed the multivariate association between the single nucleotide polymorphisms (SNPs) from the BDNF ePRS and gray matter density in order to uncover the mechanism of the interaction between prenatal environment and genotype and its association with brain development.

This study aimed at examining the hypothesis that the effects of prenatal exposure to adversity on cognitive trajectories are moderated by the prefrontal BDNF gene network. Differential response to prenatal exposure was captured using a novel bioinformatics approach that provides a biologically-informed genetic score, based on genes co-expressed with the BDNF in the PFC. Significant associations between SNPs weighted by gene expression and gray matter density at 8 to 10 years of age were located mainly in distinct portions of the occipital, frontal, and parietal cortex.

Longitudinally, high BNDF ePRS levels at the PFC were associated with higher environmental susceptibility in predicting the cognitive growth trajectory. Our data support the differential susceptibility model that postulates that individuals that are more likely to be affected by adverse environmental conditions are also most likely to benefit from positive conditions (Belsky, 2013; Belsky et al., 2018). The highest differences were observed in later development (36 months). This result might be related to the delayed messenger RNA expression in the PFC (Somel et al., 2009) and corroborates the enrichment analysis done with FUMA that shows different patterns of gene expression especially at beginning of infancy (Figure 4). Previous research found constancy on BDNF mRNA levels through development in the hippocampus, and variability at the temporal cortex with the highest expression in neonates that decreased with age (Webster et al., 2006).

The neurobiological processes enriched in the BDNF ePRS network were mostly associated with biosynthesis of complex macromolecules, regulation of gene expression and RNA transcription, maintenance of neuronal stem cells, neurogenesis, and neuron development. This is in line with previous animal research that proposes that BDNF system is critically involved in neuron development (Jones et al., 1994), regulation of genes that are associated with synaptic function (Mariga et al., 2015), dendritic growth of cortical neurons (Martin and Finsterwald, 2011), and formation of the neural networks being secreted locally by activity-dependent manner (Hayashi et al., 2007).

Going beyond the analysis of polymorphisms in G × E studies, we integrated information about the gene network of the BDNF with its function at the PFC in a specific developmental period, taking advantage of a cumulative measure of prenatal adversity that reflects a more global level of environmental influence (Silveira et al., 2017; Camerota and Willoughby, 2019). Several mechanisms might be involved in the relation of prenatal adversity and cognitive development, and the observed moderation by the BDNF ePRS. For example, activity-dependent transcription of BDNF is controlled by at least 9 distinct promoters, partially mediated by dynamic changes in DNA methylation (Martinowich et al., 2003; Sakata et al., 2009). Boersma et al. (2013) have found evidence of reduced BDNF expression in response to increased methylation of BDNF at the exon IV in both the amygdala and the hippocampus of prenatally stressed rat’s offspring. In humans, prenatal depressive symptoms were also associated with the BDNF promoter IV region along with NR3C1 1F (Braithwaite et al., 2015), suggesting that this epigenetic marker is developmentally sensitive to the quality of the early environmental exposure (Romens et al., 2015).

Another mechanism that may be involved, is the variability of gene expression in different brain regions at different developmental stages. In order to verify if the genes that composed the PFC BDNF ePRS are co-expressed in infancy and if the list of co-expressed genes is maintained in adulthood, we used databases that included gene expression levels in the human cortex. The heatmap obtained (Figure 2) demonstrates that a high proportion of these genes have different expression levels from early prenatal to late infancy, but others maintained a similar pattern. The pattern of the specific genes that varied from early prenatal to late infancy points to a special role of the network during development, which may be implicated in the relation of early life adversity effect on cognitive development (Gold et al., 2003; Braithwaite et al., 2015).

To further explore the interaction between cumulative prenatal adversity exposure with our genetic scores, we analyzed the association of the BDNF ePRS weighted SNPs in relation and brain matter density considering groups of high x low adversity. The strongest associations were observed at Brodmann areas 19, 6, 18, and 7 that contributed bilaterally and both negatively and positively to the overall pattern. The areas are related to visual association processes, including visual-motor integration, feature extracting, interpretation of images, attentional and multimodal integrating functions, as well as planning of complex and coordinated movements and convergence between vision and proprioception (Gentile et al., 2011; Bear et al., 2016). Significant associations were also found at the anterior PFC (Brodmann area 10). This area is involved in higher-order cognitive functions, for instance, the processing of internal states, strategic processes in memory recall, reallocation of attention, and more broadly the integration of information from across the supramodal cortex (Ramnani and Owen, 2004; Baird et al., 2013), all of which may be associated with the differential responsiveness to environmental adversity, as reflected in our main interaction finding.

Previous research indicates that prenatal exposure to tobacco correlates with a decrease in cortical thickness in the orbitofrontal cortex, in addition to the reduction in BDNF mRNA and protein levels (Lotfipour et al., 2009; Yochum et al., 2014). Experimental and prospective studies have shown that high pregnancy anxiety is negatively associated with gray matter volume, spine density, and dendritic complexity in the PFC (Murmu et al., 2006; Buss et al., 2010) supporting the idea that prenatal adversity has implications at the neurobiological and structural level.

The integration of genotype and gray matter data using p-ICA analysis suggests that environmental conditions have an especial impact on important neurodevelopmental processes. The G11 component is implicated in neural growth, DNA replication, regulation of mRNA processing, and commissural neuron axon guidance. The aforementioned processes are highly susceptible to environmental influences via epigenetic factors including DNA methylation and histone acetylation changes (Martinowich et al., 2003; Boulle et al., 2012; Braithwaite et al., 2015). The central role that the BDNF plays in neural development, learning, and memory processes suggests that prenatal exposure to unfavorable intrauterine conditions may compromise proper cognitive function via dysfunction of the BDNF system (Jones et al., 1994; Gomez-Pinilla and Vaynman, 2005; Boersma et al., 2013). The disruption of the BDNF network could be even more critical to the more susceptible individuals identified in our study since BDNF function have been repeatedly related to learning and memory, as well as the somatosensory and visual cortices (Gold et al., 2003; Lotfipour et al., 2009; Chiang et al., 2011; Xing et al., 2012). Thus, the observed environmental groups’ differences in common components of gray matter density and the weighted SNPs appear to play a role in a complex phenotype such as cognitive development.

Although long-lasting effects of prenatal adversity exposure were observed in cognitive behavior and gray matter density, we acknowledge that a continued influence of prenatal maternal adversity during the postnatal period is mediated through the quality of mother-infant interactions and the environmental conditions (Monk et al., 2012). The quality of interactions between caregivers and infants during the postnatal period can have a profound impact on several developmental domains predicting neuronal excitability and synaptic plasticity via epigenetic pathways (Meaney, 2010; Nguyen et al., 2015; Ohta et al., 2017). It is important to highlight that prenatal negative exposure is not determinant of a negative outcome, but rather offers possible optimistic opportunities for intervention during postnatal development (Bos et al., 2009; Silva et al., 2020).

With this work we expect to contribute with the understanding of how prenatal adversity and the BDNF gene network shape neural and cognitive development, aiming at ultimately inform and improve both prevention and intervention endeavors, yet a few limitations should be addressed. This study would benefit from replication in a different longitudinal cohort specific to the age bands that comprised our sample since during this period children go through several important sensitive periods of development. The smaller sample size of our neuroimaging study is also an aspect that suggests a need for replication using a falsification approach to avoid Type-I errors. Also, PFC subregions have been reported to develop following temporally different trajectories (Shapiro et al., 2017). Therefore, depending on the time when the stressor is applied, distinct effects could be expected in these different subregions, leading to later effects on specific aspects of cognitive behavior. Distinct PFC regions have also been shown to interact differently with the HPA axis: in rodents, GR gene knockdown in the IL cortex potentiated CORT response to a novel stressor in animals previously subjected to chronic stress, while GR knockdown in the PL cortex did not result in the same effect (McKlveen et al., 2013). In addition, functions such as attentional flexibility, reversal learning, and working memory, for example, are dependent on distinct PFC regions (Birrell and Brown, 2000; Manes et al., 2002; McAlonan and Brown, 2003; Gisquet-Verrier and Delatour, 2006). Although exposure to post-natal stress can have opposing effects on dendrite structure and spine density in distinct PFC regions, such as mPFC and OFC (Liston et al., 2006), specific effects of prenatal stress on neuronal structure according to different PFC regions are less studied. Unfortunately, the database (GTEx) used to calculate our polygenic score did not have expression data available from distinct PFC regions. We believe that future studies approaching this point considering specific PFC regions are warranted.

The broader literature on G x E contains few reports of a network approach specific to a determined brain region, use of psychometric modeling to obtain cognitive development trajectories, and the integration of genotype data with neuroimage. We demonstrated that the PFC BDNF gene network moderates the association between exposure to cumulative prenatal adversity and cognitive growth. Our results provide support for the developmental origins of health and disease (DOHaD), along with prenatal fetal programing of biological mechanisms, and differential susceptibility hypotheses (Silveira et al., 2007; Belsky, 2013; Barth et al., 2019). The focus on genes co-expressed with the BDNF allowed us to identify different patterns of enrichment throughout developmental stages that are in line with the multiple sensitive periods of brain development (Knudsen, 2004). It also made it possible to inspect specific pathways more comprehensively than the candidate-gene approach (Silveira et al., 2017). Thus, we expect to contribute to the understanding of neurobiological processes of cognitive development, and how prenatal adversity exerts a long-term influence on this complex phenotype.

The data are not publicly available due to information that could compromise the privacy of research participants. Requests to access the datasets should be directed to PS.

The studies involving human participants were reviewed and approved by Douglas Hospital Research Centre, Montreal, and St. Joseph Healthcare, Hamilton (protocol number IUSMD-03-45/IUSMD-06-09). Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.

EM was involved in data analysis, preparation, and review of the manuscript. BB was involved in parallel independent component analysis and preparation of the manuscript. DB was involved in the review of the cognitive development trajectories measure and manuscript. RL, DA, and CD were involved in enrichment analysis and review of the manuscript. IP was involved in preparation, data analysis interpretation, and review of the manuscript. RS and GH were involved in MRI acquisition and processing. MM and PS were involved in the design of study, preparation, and review of the manuscript. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.