Serotonin-related rodent models of early-life exposure relevant for neurodevelopmental vulnerability to psychiatric disorders

Mental disorders including depression and anxiety are continuously rising their prevalence across the globe. Early-life experience of individuals emerges as a main risk factor contributing to the developmental vulnerability to psychiatric disorders. That is, perturbing environmental conditions during neurodevelopmental stages can have detrimental effects on adult mood and emotional responses. However, the possible maladaptive neural mechanisms contributing to such psychopathological phenomenon still remain poorly understood. In this review, we explore preclinical rodent models of developmental vulnerability to psychiatric disorders, focusing on the impact of early-life environmental perturbations on behavioral aspects relevant to stress-related and psychiatric disorders. We limit our analysis to well-established models in which alterations in the serotonin (5-HT) system appear to have a crucial role in the pathophysiological mechanisms. We analyze long-term behavioral outcomes produced by early-life exposures to stress and psychotropic drugs such as the selective 5-HT reuptake inhibitor (SSRI) antidepressants or the anticonvulsant valproic acid (VPA). We perform a comparative analysis, identifying differences and commonalities in the behavioral effects produced in these models. Furthermore, this review discusses recent advances on neurodevelopmental substrates engaged in these behavioral effects, emphasizing the possible existence of maladaptive mechanisms that could be shared by the different models.


Introduction
Mental disorders including depression and anxiety are devastating and disabling conditions for the individual's life, with an extremely concerning high prevalence worldwide. Currently, around 4.4% and 3.6% of the global population suffers from depressive and anxiety disorders, respectively, while the burden of mental illnesses is continuously rising 1,2 .
Psychiatric disorders arise from a complex combination of genetic, biological, environmental, and psychosocial factors. However, one of the main risk factors contributing to psychopathology corresponds to early-life adverse experiences during childhood, especially those involving traumatic or stressful events, such as maltreatment, abuse, poor or neglectful parenting, and loss of a family member, among others [3][4][5][6][7] . Furthermore, adverse experiences not only could contribute to pathological mental states, but may also affect the effectiveness of prospective therapeutic treatments 7,8 .
Brain imaging studies have shown that childhood maltreatment results in persistent alterations in a wide repertoire of neurocognitive systems involved in threat processing, reward, emotions, and executive control 6,8,9 . Interestingly, such alterations can often be found even in the absence of psychiatric conditions, indicating that maladaptive mechanisms produced by adverse environments in the early-life could predispose to latent vulnerabilities to future psychiatric disorders 8,9 . Early-life exposure to certain drugs that interfere with the normal neurodevelopmental trajectories could also increase the predisposition of individuals to develop psychiatric disorders [10][11][12][13] . In particular, chemical substances that can interfere with crucial molecular and cellular neurodevelopmental events (e.g., cell proliferation, migration, differentiation, axon growth, synaptic connectivity, etc…) during the so-called critical periods [14][15][16] . These critical periods could represent unique developmental windows of brain plasticity, particularly relevant for preventive or therapeutic interventions [16][17][18] .
One example of this, is the use of antidepressant drugs that act as selective serotonin (5-HT) reuptake inhibitors (SSRIs) in pregnant and lactating depressed mothers 19,20 . These molecules have as a main molecular target the 5-HT transporter (SERT) typically present at the axon terminals of 5-HT neurons. Accumulating clinical evidence indicates that perinatal exposure to SSRIs could have a detrimental impact on infant's neurodevelopment, with long-lasting consequences on cognitive and emotional capabilities [19][20][21][22][23][24][25][26] . In addition, other lines of evidence indicate that early exposure to SSRIs during pregnancy could increase the risk of Autism Spectrum Disorders (ASD) [27][28][29][30][31][32] , which in turn present a robust comorbidity with other psychiatric disorders including anxiety and depression 33,34 . However, there is a general agreement that adverse consequences observed after antenatal SSRI treatments should be always contrasted against the highly detrimental consequences of the unmedicated maternal mental illness.
To investigate the impact of early-life adverse experiences and exposure to drugs that could interfere with the normal neurodevelopment, several preclinical rodent models were developed, in which many of the emotional, social and cognitive aspects of human psychopathology can be recapitulated 10,12,[35][36][37][38][39][40] . These models represent very useful tools to interrogate different aspects of the highly complex human pathophysiology, likely implying dysregulation of neurotransmitter systems, hormones and neuromodulators, growth factors, immune and inflammatory molecules, etc... 3,6,7,35,36 . In addition, brain structures such as prefrontal and sensory cortical regions, as well as other limbic regions including the amygdala, hippocampus and brainstem nuclei, have been increasingly driving the attention of clinicians and researchers as plausible neurodevelopmental substrates contributing to the vulnerability to mental illness 3,6,35,36,41 .
In this review, we explore well-established rodent models applied to study the impact of early-life environmental perturbations on adult behaviors of relevance to psychiatric disorders. We limited our analysis to models in which alterations in the 5-HT neurotransmission system appear to play a crucial role in the pathophysiological mechanisms [42][43][44] . Thus, we perform a comparative analysis across models of early-life exposures to stress and psychotropic drugs such as SSRIs or valproic acid (VPA), identifying differences and commonalities in their emotional and social behavioral outcomes, but also considering the impact of the perinatal period of exposure and treatment schedule on these effects. Importantly, this review discusses recent advances on developmental neural substrates engaged in such behavioral outcomes, also identifying possible maladaptive mechanisms likely shared by these models. To this aim, we carry out an exhaustive search in the PubMed database until February 2021, using a combination of the following key words: SSRI, fluoxetine, citalopram, serotonin, stress, early-life adversity, maternal separation, restrain, limited bedding/nesting, valproate, VPA, prenatal, gestational, pregnancy, postnatal, lactation, critical period, rodent, rat, mouse, emotional behavior, anxiety, depression, and mood. The results of this search were further refined and narrowed down to obtain only those studies that had sufficient methodological details to allow comparative analyses across the different models, as well as studies investigating the possible neurodevelopmental mechanisms contributing to the emotional and social behavioral effects.

Early-life stress models: behavioral outcomes
In consonance with the primordial role of the early-life adverse experiences in the etiology of psychiatric disorders, several rodent models were established in which the dams and/or the offspring are exposed to different stressful stimuli during different neurodevelopmental stages. Importantly, in these models many of the emotional and cognitive impairments present in the human psychopathology can be consistently replicated 5,35,36 .
Gestational stress: restraint, unpredictable stress Two of the more frequently applied stressors to dams during the gestation are the restraint and unpredictable stress protocols. While the first one implies the physical body restraint of the dams in a closed compartment for time periods ranging from 1.5 to 6 h per day, in the second one, dams are daily subjected to a variety of different stressful stimuli including exposure to anxiogenic environments (social defeat paradigm), foot shocks (learned helplessness paradigm), forced swim, among others. Besides, these models vary according to whether the stress protocol is applied during the entire gestational period (i.e., until the offspring birth) or selectively during defined gestational windows, often resulting in distinctive behavioral outcomes.
Daily application of restraint or unpredictable stress protocols during almost the entire gestational period, in various rat and mouse strains, produces several stressrelated and emotional alterations in the offspring later in life ( Table 1). One of the most consistently reproduced phenotypes observed in these models, is the enhancement of anxiety behaviors, often accompanied by reduced locomotor and exploratory activities [45][46][47][48][49] . Additionally, these anxiety effects can predispose the offspring to other compulsive addictive behaviors like alcohol consumption 48 (Table 1). Importantly, stress exposure in the late phase of the pregnancy, somewhere within the period from gestational day (G) 11 to postnatal day (P) 0, in mice and rats, has also produced robust anxiety phenotypes [50][51][52][53][54][55][56][57] (Table 1). Depressive-like symptoms were also reported after gestational exposure to stress in rats 49,58 , though the evidence is more limited ( Table 1).
Examination of other behavioral components associated with psychiatric conditions, such as the social interaction to conspecifics, showed a marked reduction after prenatal stress 57 . In contrast, cognitive functions do not seem to be substantially affected in these models 55 , though moderate improvements in spatial learning 52,59 and object recognition memory 59 , were observed in the offspring (Table 1).

Postnatal stress: maternal separation, limited bedding and nesting material
Other models directly expose the offspring to the stressful stimuli. The repeated maternal separation of the pups from the dams has been consistently used as an efficient stressful condition, especially when it happens during the lactation period 60 . This model sometimes also includes other stressors like the early weaning of the pups or a limited access to the bedding and nesting materials.
Protocols of maternal separation vary according to the time the pups spend isolated from their dams, but also to during which postnatal period the protocol is applied. Thus, while the time of separation typically ranges from 1 to 8 h per day, the postnatal period usually covers the first 2 or 3 postnatal weeks. In addition, in these models, distinctive behavioral effects can be observed when considering the mouse/rat strain and the sex of the offspring.
Daily separation from the dams during the first 2 or 3 postnatal weeks, in various mouse/rat strains, have consistently produced a reduction in the exploratory activity of the offspring [61][62][63][64][65][66][67] . Besides, these effects appeared to be more commonly detected in males than in female littermates 65 . However, other studies using slightly different conditions, could not reproduce these findings [68][69][70][71][72] . A combination of maternal separation with an early weaning of the pups also produced a decreased exploration 61 . On the other hand, stress protocols of limited access to bedding and nesting material, were shown to be inefficient per se to reproduce the exploratory defects [73][74][75] (Table 1).
Anxiety phenotypes were consistently observed in various mouse/rat strains after maternal separation during the first 2 62,63,65,67,71,76-78 or 3 postnatal weeks 61,64,66,79,80 (Table 1). However, milder protocols (1 h/day) appeared to be insufficient to produce anxiety effects, after a chronic social distress 72 . On the other hand, anxiety phenotypes were reported to be more evident in females than in male descendants 65 . Lastly, less robust anxiety phenotypes were observed when both the dams and pups had a limited access to the bedding and nesting material during early postnatal life 74,75,81 (Table 1).
Depressive-like behaviors have been repeatedly observed after maternal separation protocols in various mouse strains 64,67,69 , though others failed to reproduce these effects 79 . Similar depressive-like symptoms were reported when the maternal separation was followed by an early weaning in mice 61,82 . Interestingly, a mouse study suggested that the maternal separation from P10 to P20 would be sufficient to induce depressive-like symptoms in the offspring 64 . In rats, depressive-like phenotypes produced by maternal separation protocols are less robust. Thus, depressive-like symptoms were reported in Sprague-Dawley and Wistar rats 66,77 , while other studies using briefer protocols (1.2 h/day) in Wistar-Kyoto rats, have described the opposite effects 71 (Table 1).
The impact of early maternal separation upon cognitive behaviors was consistently evidenced (Table 1). In general, postnatal exposures to stress appeared to have more profound consequences on cognitive abilities than prenatal exposures. Studies carried out in various mouse strains have shown deficits in behavioral flexibility 83 and short-term memory 67 , after different maternal separation protocols. In contrast, no apparent effects were observed on spatial learning and working memory in mice 79 . However, in rats, improvements in spatial learning and attention were recently described 78 . Detrimental consequences of the limited bedding and nesting material have been observed on the novel object exploration and location memory 73,81 , accompanied by deficits in spatial learning and working memory tasks 73,74,81 ( Table 1). Consequences of maternal separation on social behaviors were reported in mice and rats (Table 1). Thus, a delayed latency to initiate social contacts, and a decreased engagement in social interactions and activities, have been found 64,71,78,84 . However, other studies could not reproduce some of these effects 65 . Interestingly, distinct behavioral features associated with social behaviors, like aggression or dominance, were also found to be enhanced by maternal separation 79,85 . Consistent with this, a reduction of freezing responses to conspecifics, or after a punishment, were also observed 86 (Table 1).

Pharmacological models: behavioral outcomes
Psychotropic drugs can interfere directly or indirectly with the brain's developmental trajectory. Importantly, the age period when such perturbations occur will define the emergence of long-lasting detrimental consequences on the individual's brain architecture and function. Rodent models have been very useful to understand how such dysregulation of neural mechanisms during developmental critical periods can impact adult behaviors. Here, we analyze the behavioral outcomes of two pharmacological models: the exposure to SSRI antidepressants, and to the anticonvulsant and mood stabilizer, VPA. Both drugs have a high capacity of crossing the placenta to reach the fetus 10,87,88 . While SSRIs inhibit the reuptake of 5-HT from the extracellular space by blocking the 5-HT transporter, SERT, the precise mechanism of action of VPA is less known. In fact, it is thought that one of the main actions of VPA is to modulate the synthesis and metabolism of gammaaminobutyric acid (GABA), resulting in an enhancement of GABA neurotransmission 87 . Besides, VPA was also shown to have a direct inhibitory action on histone deacetylases (HDACs), modulating the epigenetic control of gene expression 89 .

Gestational exposure to SSRIs
One of the most widely prescribed SSRI antidepressants is fluoxetine, and its exposure during the gestational period has a neurodevelopmental impact in the offspring (Table 2), as indicated by a recent meta-analysis by Ramsteijn et al. 90 . Rat pups prenatally exposed to fluoxetine showed a mild delayed neurological maturation 91 , followed by a general reduction in exploratory behaviors, accompanied by anxiety and depressive-like phenotypes 92,93 . On the other hand, studies applying a much broader schedule of fluoxetine treatment, covering from the last phase of the gestation to somewhere until the third postnatal week, could not detect such exploratory and emotional deficits [94][95][96][97][98] , while including the full gestational period produced emotional modifications 99 ( Table 2).
Regarding cognitive and social behaviors, only moderate effects of gestational exposures to SSRIs were observed ( Table 2). Thus, no major changes in the exploration or memory of novel objects were detected in mice and rats 91,100 . However, these studies did observe increments in conspecific exploration and aggression 100 . Consistently, an increased proportion of "attacker" individuals was found after fluoxetine exposure at later perinatal periods (G15 to P12) 95,96 . In the same conditions, no major effects on passive avoidance, fear conditioning, spatial learning and prepulse inhibition (PPI) were detected in mice 94,96 . On the other hand, in rats, the freezing behavior was shown to be enhanced by prenatal fluoxetine treatment, and several sex-specific social outcomes were also described 97,101,102 (Table 2).
Other behavioral effects of relevance for psychiatric conditions were also reported after postnatal exposure to SSRIs. That is, several rat studies have reported reductions in ultrasonic vocalizations and sensorimotor responses, accompanied by alterations in social interactions to conspecifics 104,105,115,117,118 . On the other hand, spatial learning and memory abilities seemed not to be affected by the fluoxetine treatment at lower doses 116,119 (Table 2), consistent with a recent meta-analysis 90 .
Postnatal exposure to other SSRIs like citalopram or escitalopram, during the same postnatal period, also produces anxiety and depressive-like phenotypes in mice 120,121 . However, other studies have only partially replicated some of these emotional effects [122][123][124][125] . In addition, citalopram exposure in the same period, was shown to reduce the exploration of novel objects and the     (Table 2). Since SSRIs are often prescribed to depressed pregnant women, preclinical investigations on the possible interaction of these treatments with the maternal stress become highly relevant for obvious direct translational reasons. Several studies have shown that early postnatal fluoxetine treatment in pups exposed to either prenatal stress or maternal separation, can alleviate anxiety and depressive-like symptoms produced in these models 110,[129][130][131] . However, this does not seem to be the case when using another SSRI (i.e., citalopram) 132 .

Gestational and postnatal exposures to VPA
Acute treatment with a high dose of VPA during pregnancy causes a wide repertoire of emotional, social and cognitive alterations in the offspring (Table 3). Although the exact molecular mechanism of action of this drug remains unknown, VPA has been consistently applied during neurodevelopment in an attempt to replicate common phenotypic features present in ASD patients 39,40 .
Anxiety phenotypes induced by VPA exposure during the pregnancy were more consistently found in rats than in mice 135,[137][138][139]142,146,147,[152][153][154][155][156][157] . However, other studies could not replicate some of these effects 144 . On the other hand, the emergence of depressive-like symptoms was also reported in a mouse study 147 (Table 3).   Social behaviors appeared to be markedly affected by prenatal VPA in various mouse/rat strains, indicating a highly consistent reduction in the number of ultrasonic vocalizations and social interactions (e.g., nose pokes, approaches, etc…), accompanied by a diminished juvenile social play, and limited interest for novel conspecifics 134,135,139,140,142,143,[145][146][147]149,150,153,154,[157][158][159][160][161][162][163][164] . However, several studies failed to replicate some of these In the studies where both sexes were analyzed, the sex-specific effects observed are indicated. OF Open field, EPM Elevated plus maze, PPI Prepulse inhibition, MWM Morris water maze, NSF Novelty-suppressed feeding test, TST Tail suspension test, FST Forcedswim test, NOR Novel object recognition. effects 136,137,151 . On the other hand, and related to the social deficits observed, a decreased olfactory motivation and sensorimotor capacity were found in VPA-exposed mice 134,139,142,145,155 (Table 3). Additionally, sex-specific aggressive and defensive strategies in social settings appeared to be affected by the VPA exposure, switching from a defensive tactic to a more aggressive one 161 . Furthermore, enhanced freezing and startle responses were described in different fear-conditioning paradigms after VPA treatment 142,153,154,157,160,165,166 (Table 3).
Certain cognitive aspects have been described to be altered in VPA gestational models, though the behavioral outcomes were less consistent. Thus, several studies reported either deficits 133,135,137 or improvements in spatial learning 146 , while others could not reproduce these effects 142,143 . Similarly, evaluation of exploration time of a novel object showed that VPA treatment increased the interaction to familiar objects 144,148 , though other researchers could not replicate these findings 147,152,154 . Additionally, deficits in working memory 146,147 , temporal accuracy 167 , attentional, and negative reinforcement learning tasks 148,164 were also reported ( Table 3).
There are fewer studies applying postnatal exposures to VPA, having in general, less robust behavioral effects (Table  3). VPA treatment somewhere during the first 2 postnatal weeks produces a delayed neurological maturation 133,168 , without any apparent locomotor effects 133,169 , though others observed an enhanced locomotion 170,171 . On the other hand, changes on anxiety behaviors were described in rats 171 , but not in mice 169 (Table 3).
Social behaviors were shown to be affected by postnatal exposure to VPA in rats and mice 168,170,171 , though others could not replicate some of these findings 169 . In the same conditions, higher levels of social aggression were found 169 (Table 3).
Several cognitive aspects appeared to be affected in these models, including deficits in spatial 133 and reversal learning 169 , accompanied by enhanced cue-dependent reward learning 168 . Additionally, other studies described longer latencies in a passive avoidance paradigm in mice 133 , while others failed to reproduce some of these effects in rats 171 (Table 3).

Neural mechanisms engaged in the rodent models of early-life exposure
In the last decade, many preclinical studies have contributed to our understanding of the possible molecular, cellular and circuit mechanisms implicated in the neurodevelopmental vulnerability to psychiatric conditions. A main concept emerging from those studies is that the ontogenetic occurrence of unique developmental events defines critical periods of plasticity with a maximal sensitivity to environmental functional demands [14][15][16]18 . Thus, depending on the period when the environmental challenge/perturbation takes place, the long-lasting impact that it will have on the mature brain. In this section, we explore maladaptive neural mechanisms that have been implicated in the behavioral outcomes of the analyzed rodent models, paying special attention to brain regions and neural mechanisms that could be similarly altered across the different models.

Developmental role of 5-HT and Prefrontal circuits
Perturbation of the 5-HT neural signaling during earlylife has been long associated with developmental origins of several psychiatric conditions, including anxiety, depression, and ASD 43,44,[172][173][174][175][176] . In rodents, activation of 5-HT receptors in different brain regions has been implicated in long-term emotional alterations. Thus, in the early-life stress model of maternal separation, an enhanced adult 5-HT2A/C-mediated prefrontal function was found 62 , while the early postnatal pharmacological blockade of these receptors prevented the emergence of the anxiety phenotype in the same model 63 . Interestingly, the expression of 5-HT2A/C receptors appeared to be substantially modified by prenatal stress but not after maternal separation 55,63 . Emotional alterations produced by the postnatal exposure to SSRIs also appear to be mediated, at least in part, by 5-HT2A/C receptors. That is, the exposure to 5-HT2A and 5-HT2C antagonists during the early postnatal period prevents the anxiety and depressive-like phenotypes induced by fluoxetine, while the treatment in the same period with agonists of the same receptors produces anxiety 104 .
Other 5-HT receptors such as the 5-HT1A and 5-HT7 were also implicated in the behavioral consequences of early postnatal SSRIs. That is, some of the adult emotional effects induced by postnatal fluoxetine were found to be enhanced by the co-treatment with a 5-HT1A receptor agonist 116 . More recently, the prefrontal 5-HT7 receptors have been shown to have a crucial developmental role in the emergence of anxiety and depressive-like symptoms in the model of postnatal fluoxetine 112 . Additionally, increased stimulation of 5-HT1A receptors during early postnatal life was also shown to mimic deficits produced by early SSRIs on adult social interactions of relevance to ASD 128 . On the other hand, in the VPA model, activation of 5-HT1A receptors improves some of the social and cognitive deficits produced in that model 166 .
Accumulating evidence indicates a primordial role of the prefrontal cortex (PFC) in these neurodevelopmental mechanisms. Thus, disruptions of developmental processes in the PFC such as circuit formation/refinement, synaptic connectivity, and oligodendrogenesis/myelinization, have been directly linked to the early vulnerability to stress-related and emotional alterations (Fig. 1). Prenatal stress has been reported to produce a decrease in spine density in mPFC pyramidal neurons 48 . A similar reduction was described in layer 2-3 mPFC pyramidal cells after postnatal exposure to fluoxetine, accompanied by altered excitability of this neuronal population and exuberant dendritic branching 106,115 . Functional imaging analysis in trumpet-tailed rats subjected to maternal separation stress has shown a global decrease in brain activity in PFC circuits likely engaged in prefrontal-limbic control 177 . Besides, decreases in the excitability of PFC pyramidal neurons have been recently described after maternal separation 178 . Consistent findings were described when analyzing the expression of activity-related immediate early genes in the PFC in the same model, accompanied by a precocious oligodendrocyte differentiation, and hypo-myelination 67,179,180 . In agreement with this, postnatal exposure to the SSRI citalopram, but not the prenatal treatment, was shown to alter the oligodendrocyte morphology and the callosal connectivity 126 .
Recent studies have shown a synaptic hyperconnectivity of corticolimbic circuits such as the PFCto-dorsal raphe nucleus (DRN) one, after postnatal exposure to fluoxetine during the early postnatal period 111 . Furthermore, these changes were found to be mediated by the developmental 5-HT signaling through the 5-HT7 receptors in the PFC 112 . Importantly, reciprocal circuits connect the PFC to DRN 5-HT neurons, and these pathways have a crucial role in controlling stress-coping strategies and emotional responses throughout life 111,181,182 .
Direct inhibition of the activity of DRN 5-HT neurons using chemogenetic tools has been shown to prevent the emergence of the emotional alterations produced by Fig. 1 Schematic overview of some of the main effects reported after either stress, SSRI or VPA exposures during gestation and/or early postnatal life, on different brain regions and neurodevelopmental processes. Different rodent models of exposure to stress (forced swim, helplessness, social defeat, restraint, maternal separation) and to chemical substances (SSRIs, VPA) are shown. Main brain regions affected in these models are indicated (Olfactory bulb OB, Prefrontal cortex PFC, Lateral habenula LHb, Amygdala AMY, Hippocampus HIP, Ventral tegmental area VTA, Dorsal raphe nucleus DRN, Cerebellum CBL), together with summarized effects on neurodevelopmental molecular, cellular, and circuit mechanisms. These include hyperexcitability of glutamate neurons, exuberant glutamate synaptogenesis, decreases in dendrite and spine remodeling, decreased neurogenesis, reduction of the balance of glutamate/GABA transmission, reduced gliogenesis and myelination, and multiple changes in epigenetic control of gene expression. postnatal fluoxetine 107 . On the other hand, the early-life treatment with another SSRI (i.e., citalopram), was shown to reduce the 5-HT synthesis and its forebrain release 123 . Additionally, maternal separation reduces the firing activity of DRN 5-HT neurons 183 . Interestingly, gestational exposure to VPA was reported to delay the migration and differentiation of developing DRN 5-HT neurons 184 .
Altogether, this evidence indicates a complex role of the developing 5-HT system and its targeted neural circuits, in the detrimental emotional effects produced by early-life exposures.

Other neural circuits implicated
Other studies have implicated other brain structures such as the ventral tegmental area (VTA) and lateral habenula (LHb), in the emotional vulnerability during early-life 64,82 (Fig. 1). There, genetic invalidation of an early transcriptional program in the VTA was sufficient to replicate the life-long susceptibility to stress observed after maternal separation 64 . Furthermore, a recent study using the same model showed enhancements in the excitability of VTA dopamine neurons and their excitatory synaptic inputs 185 . In another study, researchers showed that limiting the hyperactivity of LHb output neurons relieves the depressive-like symptoms caused by maternal separation 82 . Interestingly, both VTA and LHb neurons are interconnected, and receive dense 5-HT afferents 186 , and in turn, send projections back to DRN neurons 187,188 .
The hippocampus appeared to be more robustly affected in preclinical models using either gestational or precocious early postnatal exposures. Thus, alterations in neurogenesis were found among the different preclinical models 52,59,81,110,189 (Fig. 1). Specifically, changes in the survival of newborn cells in the dentate gyrus of the hippocampus 52,59,81,189 and alterations in the expression of neurogenic modulators such as brain-derived neurotrophic factor (BDNF) 52,110 , have been consistently reported after either prenatal or early postnatal stress protocols. Similarly, effects on the hippocampal neurogenesis and BDNF expression levels were also described after maternal separation, SSRI and VPA postnatal exposures 108,110,113,[189][190][191] . Interestingly, postnatal SSRI exposure reverses alterations on hippocampal neurogenesis produced by prenatal maternal stress 189 .
Alterations in the synaptic morphology and function of the hippocampus were implicated in the models of prenatal exposures to VPA and stress 53,135,159,192 . Interestingly, in the stress model, postnatal treatment with SSRIs restores the normal density of dendritic spines and glutamate synapses 119 . In this last model, altered formation and maintenance of the hippocampal perineuronal net has been recently shown after SSRI exposure 193 .
Furthermore, an altered synaptic development of glutamate/GABA transmission systems has been also described in the PFC, after gestational exposures to SSRIs or VPA 48,99,135,144,159,192 (Fig. 1). In addition, these changes were accompanied by a delayed neuronal migration followed by a precocious neuronal maturation, and longstanding modifications in cortical thickness 137,144,168 . Interestingly, cortical modifications on dendritic spines and neuronal morphology were also observed after postnatal exposure to VPA 169 (Fig. 1), likely reflecting the protracted developmental period of frontocortical circuits. Lastly, at the system level, the early-life stress experience of maternal separation was shown to perturb the communication within the network engaging hippocampal and PFC circuits 194 .
Other studies point out to the amygdala as a key neural substrate involved in the effects of maternal separation and in utero VPA exposure (Fig. 1). These investigations showed a reduced functional connectivity between the basolateral amygdaloid nucleus (BLA) and the PFC after maternal separation 195 , in agreement with previous imaging findings in depressive patients 196,197 . On the other hand, enhancements of the neuronal excitability and glutamate transmission in the amygdala have been reported in the VPA model 142,153 . Besides, similar findings were described in the same model in frontocortical circuits 158,165,198,199 (Fig. 1). Interestingly, treatment with a metabotropic glutamate receptor 5 (mGluR5) antagonist rescued some of the social behavioral deficits observed in the VPA model 138 . Additionally, changes on GABAergic neurochemical markers and modifications on the neuronal/glial cytoarchitecture, have been also reported in the amygdala and cerebellum, after prenatal exposure to VPA 135,156,200 .

Epigenetic mechanisms
Other lines of evidence indicate an important role for epigenetic control of gene expression in the neurodevelopmental mechanisms at play in these preclinical models (Fig. 1). For instance, a transcriptional disruption of genes involved in developmental and immune gene networks has been described in the amygdala after VPA exposure 157 . In other studies, researchers found that acetylation regulation of hippocampal gene expression by the histone deacetylase (HDAC) 4 is crucially involved in the adult emotional alterations caused by postnatal SSRIs 105 . Consistent with its relevance, a decreased expression of several members of the HDAC family was reported in frontocortical regions including the PFC of maternallyseparated rats and mice 180,201 . Furthermore, in one of these studies the authors suggested that this mechanism could be a major contributor to the susceptibility/resilience to early-life stress, and the subsequent efficacy to antidepressant treatments 201 . In utero effects of VPA have been directly linked to its actions on the activity of HDACs. Thus, exposure to the HDAC inhibitor trichostatin A, can phenocopy many of the social behavioral deficits observed in the VPA model 145 . Consistently, the prenatal treatment with valpromide, a VPA analog lacking the HDAC's inhibitory activity, failed to reproduce the social interaction deficits 137 .
Methylation of promoter regions has been also implicated in the effects of SSRIs during pregnancy and lactation 98,109 . These investigations showed long-lasting changes in methylation levels of multiple genes in the hippocampus and cortex, including the Bdnf gene 98,109 . Interestingly, chromatin remodeling and increased methylation levels in the PFC were reported, after either gestational stress 48 or VPA exposure 202 . Lastly, the transgenerational epigenetic inheritance of VPA-induced imbalance in excitatory/inhibitory transmission in the frontal cortex has been recently demonstrated 203 , illustrating the rather complex repertoire of neurodevelopmental mechanisms that could be engaged in the different preclinical models of exposure.

Translational Aspects and Concluding Remarks
Rodent models have been very useful to investigate the mechanisms implicated in the developmental vulnerability to psychiatric disorders. Thus, preclinical research surveyed in this review allows the identification of neural substrates and neurobiological mechanisms impacted by early-life environmental exposures. Importantly, many of these findings may directly relate to what occurs in the complex human psychopathology. However, other biological processes linked to gene regulation and its possible interaction to environmental conditions, is growingly emerging as crucial actors involved in neuropsychiatric disorders. For instance, investigations in rodents with reduced 5-HT synthesis have shown a differential impact on the DRN 5-HT system of maternal separation 204 , while reduction of the 5-HT transporter, SERT, leads to an enhanced anhedonia under similar stress conditions 205 . Furthermore, combination of stress exposure and postnatal SSRI treatment in dams alters affective susceptibility of the offspring in a SERTdependent manner 206 . Conversely, behavioral alterations produced by in utero VPA exposure were not affected by the presence of SERT 155 . On the other hand, rats with reduced SERT expression, and a previous history of maternal separation, showed improvements in stress-coping responses 207 . Interestingly, in utero exposure to citalopram mitigates the detrimental effects of maternal stress on the fetal forebrain development, and these changes are thought to be mediated by normalizing brain 5-HT levels 88 .
Human studies have shown that gestational exposure to SSRIs is associated with adverse neonatal outcomes that can be moderated by the SERT promoter polymorphism of the infants 20 . Moreover, methylation of the SERT promoter can influence the soothability of infants that had exposition to SSRIs during the gestation 208 . Additionally, the hyperserotonemia and several stereotypical behavioral outcomes present in ASD patients can be replicated in mice that express a hyperfunctional SERT variant 209,210 .
A critical point to be considered when analyzing the risk of exposure to SSRIs in pregnant women is the psychopathological state of the future mother. Evidence indicates that SSRI treatment in depressed mothers can prevent the modifications in brain's connectivity produced in newborns that are exposed to unmedicated depressive mothers 211 . In addition, other studies have shown that adverse effects produced by prenatal maternal depression on infant's problematic temperament can be amplified by a concurrent prenatal traumatic stress 212 .
In the case of VPA, human evidence supports only a poor interaction between the effects of prenatal exposure to VPA and maternal mental health. Thus, a very recent nationwide population-based analysis has shown that in utero exposure to VPA is associated with an increased risk of neurodevelopmental mental and behavioral disorders in exposed children 213 . Moreover, these effects were stronger when exposures occurred during the second and third trimesters of pregnancy. However, these findings were not affected by the mother's mental health 213 .
Future investigations need to consider neurodevelopmental regulatory mechanisms as well as other biological factors, such as the neuroinflammatory and hormonal contexts, that could influence the brain's homeostatic capacity to mitigate early-life environmental perturbations. Altogether, these further considerations will improve understanding of how developmental maladaptive mechanisms could increase the risk of vulnerability to mental disorders.