Major depressive disorder, also known as major depression, carries the heaviest burden amongst all mental and behavioral disorders and is globally the largest contributor to years lived with disability (Ferrari et al., 2013). At any given time, over 4 percent of the global population suffers from major depression, with females being 1.7 times more likely than males to experience a depressive episode (Vos et al., 2012). Depression is diagnosed when signs and symptoms persist for at least 2 weeks and include daily feelings of depression, anxiousness or hopelessness and despair, insomnia, and anhedonia, i.e., loss of interest in activities that were once pleasurable (DSM-5; American Psychiatric Association, 2013). It has been widely acknowledged that both genetic and environmental factors contribute to the psychopathology of major depression, most likely by interacting in a complex and interdependent manner. For instance, aversive early life events such as childhood maltreatment contribute substantially to the risk of developing depression (Heim et al., 2004). However, the responsivity to such aversive early life events is significantly affected by the individual's genetic background.

A well-studied example of such gene × environment interactions is the influence of serotonin transporter gene (SERT, 5-HTT, or SLC6A4) variation on individual stress susceptibility (Caspi et al., 2003, 2010). The serotonin transporter is of major importance in regulating synaptic serotonin (5-HT) concentrations and signaling, and 5-HT synapses play a central role in the neural circuitry controlling mood and temperament. Disturbances in the serotonin system are known to contribute to the psychopathology of many psychiatric disorders (reviewed in Andrews et al., 2015). Various gene variants of SERT may interact to generate up to 20-fold differences in serotonin transporter gene expression and functional levels in vitro (Murphy et al., 2008). Of these SERT gene variants, the best-studied is the repeat length polymorphism in the promoter region of the serotonin transporter gene (5-HTTLPR). It is assumed that the human 5-HTTLPR drives allele-specific SERT promoter activity leading to differences in variations in transcriptional activity, and functional serotonin uptake. In the 5-HTTLPR, different lengths of the repetitive sequence containing GC-rich, 20–23-bp-long repeat elements in the upstream regulatory region of the gene have been identified. A deletion or insertion in the 5-HTTLPR are referred to as the 14-repeat short (S, low expressing) and the 16-repeat long (L, high expressing) alleles. However, 17 up to 22 repeat alleles (XL) also occur (Goldman et al., 2011). Furthermore, a single nucleotide polymorphism within or immediately outside the SERT gene results in two forms of the L-allele (L-A and L-G) and the S-allele (S-A and S-G) leading to transcriptional differences of the SERT gene. More specific, the L-G variant results in transcriptional activity similar to the short allele (Hu et al., 2006). Compared to the long variant, individuals carrying at least one allele with the short repeat in the promotor region have 60–70% lower SERT mRNA expression levels (Murphy et al., 2008; Wankerl et al., 2014) leading to reduced 5-HT uptake in lymphoblast cells (Lesch et al., 1996) and blood platelets (Greenberg et al., 1999; Nobile et al., 1999; Anderson et al., 2002). However, results on blood 5-HT levels are inconsistent. Whereas, an earlier study reported increased 5-HT levels in whole blood samples of patients with one (S/L) or two (L/L) long alleles (Hanna et al., 1998), other studies did not find differences in 5-HT storage of blood platelets (Greenberg et al., 1999; Anderson et al., 2002), or whole blood 5-HT levels (Betancur et al., 2002) among genotypes. Likewise, human studies on binding of SERT in both blood platelets and the brain show variable results, ranging from reduced SERT binding in S-allele carriers (Lesch et al., 1996; Stoltenberg et al., 2002) to no differences at all (Greenberg et al., 1999; Nobile et al., 1999; Anderson et al., 2002). Furthermore, cerebrospinal fluid levels of 5-HIAA, the main metabolite of 5-HT, was reported to be lower in healthy short allele carriers compared to long allele carriers (Williams et al., 2001). However, 5-HIAA cerebrospinal fluid levels did not differ between genotypes of healthy people (Jönsson et al., 1998) and depressed patients (Zalsman et al., 2006). Moreover, 5-HIAA cerebrospinal fluid levels seem to be dependent on gender and ethnicity (Williams et al., 2003). Remarkably, most studies only differentiate between the L and S variant and did not include XL alleles (17 up to 22 repeat alleles) or other allelic variants resulting from single nucleotide polymorphisms, possibly explaining the discrepancies in demonstrating a causal link between 5-HTTLPR genotype and in vivo SERT levels or function among the various studies. In addition, other variations such as in methodology, sampling size, or selection bias may have played an important role, but remain largely unclear.

At the behavioral level, human S-allele carriers show little evidence for changes in general behavioral functions compared to L-allele carriers. Large scale studies on neuroticism, a personality trait involved in the propensity to anxiety and depression, are inconsistent in finding an association with 5-HTTLPR variation (Sen et al., 2004; Terracciano et al., 2009). However, it was suggested that after a history of early life stress, such as childhood maltreatment, human S-allele carriers appear to be more stress sensitive, and are more likely to develop major depression (Caspi et al., 2003). Over the years, this link between the 5-HTTLPR genotype and depressive symptoms following stressful life events has been rather controversial. Multiple studies showing results against, but mostly in favor of this relationship have been performed (Munafò et al., 2009; Risch et al., 2009; Karg et al., 2011; McGuffin et al., 2011; Sharpley et al., 2014; Vrijsen et al., 2015). Given that about 70 percent of Caucasians have at least one copy of the short allele, the effects of these gene variants in the response to an environment with significant stressors may have a potential high-societal impact (Haberstick et al., 2015). The underlying mechanisms of these serotonin transporter gene variation × ELS interactions have gained increasing attention over the years as more mechanisms are being unraveled using animal models of SERT gene variation. In this review, we will (1) discuss the behavioral effects of ELS in heterozygous SERT knockout rodent and non-human primate models, (2) provide an overview of the underlying mechanisms by addressing (epi)genetic effects of ELS, lower SERT gene expression (heterozygous animals) and their interaction in the brain, (3) critically discuss the use of the heterozygous knockout rodent as a translational model to study early life stress induced psychopathology and (4) suggest relevant future studies. We selected literature focusing on both the SERT heterozygous genotype and some form of early life stress (we limitedly used literature about stressors during adulthood). Understanding the exact mechanisms underlying these interactions can be beneficial and is essential for the development of personalized treatment strategies in many mental disorders, including major depressive disorder.

Similar to humans, rhesus macaques carry an orthologue of the 5-HTTLPR, making it an excellent animal model to study the SERT gene polymorphism variants (Figure 1a; Lesch et al., 1997; Bennett et al., 2002). Promotor activity of the rhesus 5-HTTLPR long variant in an in vitro SERT promoter/luciferase gene construct, is almost twice as high as the short variant (Bennett et al., 2002). However, mRNA levels of SERT, hypothesized to be lowered as a result of lower transcriptional efficiency of the 5-HTTLPR short variant, were unaltered in peripheral blood cells compared to the long variant (Yu et al., 2010; Singh et al., 2012). Within the same genotype mRNA levels are highly variable indicating that other factors unrelated to the 5-HTTLPR genotype may influence SERT mRNA synthesis, stability, and degradation (Singh et al., 2012). Similar to human studies in lymphoblasts and blood platelets, 5-HT uptake in rhesus lymphocytes and peripheral blood cells (PBCs) is decreased in S-allele carriers (Singh et al., 2010, 2012). Furthermore, SERT binding was decreased and 5-HIAA concentrations were lower in peripheral blood cells of rhesus macaques carrying the short variant of the polymorphism (Singh et al., 2012). When taking into account the early rearing situation of the rhesus macaques, no differences in cerebrospinal fluid 5-HIAA levels were found among different genotypes of parent-reared rhesus macaques. However, when macaques were removed from their mother at birth and peer-reared, a form of early life social stress, cerebrospinal fluid 5-HIAA levels were lower in macaques with one short allele (S/L) compared to those with two long alleles (L/L) (Bennett et al., 2002). Furthermore, young rhesus monkey short allele carriers perform worse on orientation, attention and affective capacities but only in those macaques that were early deprived of their mother (Champoux et al., 2002). Peer-reared short allele carriers also display higher levels of aggression (Schwandt et al., 2010) and have increased sensitivity to alcohol consumption (Barr et al., 2003, 2004). Upon acute separation from either mother or peers, peer-reared S/L-allele carriers showed the highest levels of stereotypic behaviors compared to mother-reared S/L, mother-reared L/L, and peer-reared L/L carriers, whereas peer-reared L/L-allele carriers showed higher despair levels compared to mother-reared S/L, mother-reared L/L, and peer-reared S/L carriers (Spinelli et al., 2007). Overall, the rhesus 5-HTTLPR and early life stress interact to affect the behavioral outcome in rhesus macaques.

Since rodents do not carry an orthologue of the human 5-HTTLPR, genetic variation in the human polymorphism can be simulated by creating serotonin transporter knockout (SERT^−/−) rodents. For instance, using ENU-target driven mutagenesis, a premature stop codon has been induced in the SERT gene of rats resulting in a non-functional protein product (Figure 1c; Smits et al., 2004, 2006). In mice, a more common method is targeted disruption of the SERT gene by replacing exon 2 with a PGK-neo gene cassette using homologous recombination (Figure 1b; Bengel et al., 1998). Consequently, SERT knockout mice and rats, particularly the heterozygous ones, may demonstrate a similar loss-of-function in SERT activity as seen in the human and rhesus 5-HTTLPR genotype.

Animal models of early life adverse experiences consist of exposing animals to stressful procedures either pre- or post-natally, leading to considerable changes in behavioral and physiological responding of the offspring at later stages in life. Prenatal stress exposure, which is usually performed during the last week of pregnancy (corresponding to the late second trimester in human pregnancy; Homberg et al., 2010), can affect the development of offspring by transferring stress mediators from the pregnant mother to the fetus. Most likely this occurs by transport of maternal stress hormones across the placenta and induced release of placental hormones into the fetal blood circulation (Huizink et al., 2004). Postnatal stress, on the other hand, usually interferes with mother-pup interactions, which are essential for optimal brain development. At birth, the rat pup brain is not fully matured and continuous to develop. Approximately, postnatal days 1–10 of rat brain neurodevelopment are equal to the third trimester in humans (Dobbing and Sands, 1979; Andrews and Fitzgerald, 1997), and maturation of the rat cerebral cortex at postnatal day 12 and 13 is equivalent to that of the human neocortex at birth (Homberg et al., 2010). Disruption of mother-pup interactions can affect hypothalamic-pituitary-adrenal (HPA) axis regulation which may lead to considerable changes in neurobiology and behavior that persist into adulthood (Lajud and Torner, 2015). Interactions between SERT genotype and prenatal stress have so far only been studied in SERT knockout mice (van den Hove et al., 2011). A common method used as prenatal stress in mice is the maternal restraint-stress paradigm, which usually consists of placing individual pregnant dams in a small transparent cylindrical restrainer for multiple times per day. A study using this prenatal stress paradigm found that SERT^+/− mouse offspring had intact memory function in the novel object recognition test after a 3-h retention interval, while SERT^+/+ mice showed impaired memory performance (van den Hove et al., 2011). In addition, prenatally stressed SERT^+/− mouse offspring appeared less anxious in the elevated zero-maze compared to SERT^+/+ offspring. Interestingly, under control conditions both male and female SERT^+/− offspring show increased mobility compared to SERT^+/+ mice, whereas only prenatally stressed female SERT^+/− offspring tended to show lower mobility in the forced swim test compared to non-stressed controls (van den Hove et al., 2011).

Other studies choose to use both pre- and postnatal stressors (Heiming, 2009; Heiming et al., 2011; Bodden et al., 2015). In particular, exposing pregnant and lactating mouse dams to bedding of unfamiliar male mice has been used as an early life stressor. Putting unfamiliar male bedding every 2–3 days in the home cage of the dams is claimed to simulate a threatening environment to the mother and her pups, as unfamiliar male olfactory cues can signal infanticide (Weber and Olsson, 2008). Mouse offspring growing up in this so-called dangerous environment showed overall increased levels of anxiety-like behavior in the light/dark box test, but this was not evident in the elevated plus-maze (Heiming, 2009). In addition, reduced locomotor activity was found in both the light/dark box test and in the open field. However, SERT^+/− mice behavior was not different from SERT^+/+ in these behavioral measures. Since there was habituation of the stress response to the treatment procedure, it was suggested that the observed effects in the offspring result from a change in the amount of maternal care in response to the stressor. The maternal care provided is a crucial factor influencing offspring development. Indeed, in a follow-up study from the same authors it was shown that these mothers living in dangerous environments had increased levels of fecal corticosterone metabolites and displayed a decrease in various maternal care behaviors (Heiming et al., 2011). Surprisingly, in this study no clear effects of either housing condition or SERT genotype on offspring behavior were found. A subsequent study by these authors again used unfamiliar male bedding as an aversive stimulus, but only during the postnatal period (Kloke et al., 2013). In addition to control mouse dams that were given neutral bedding, they also studied communal nesting. This is considered a positive environment as pups growing up in a communal nest experience increased levels of maternal care and engage in higher levels of peer interactions (Branchi et al., 2013). SERT^+/− offspring showed no differences in anxiety-like and explorative behavior when compared to SERT^+/+ offspring. Furthermore, no effects of housing condition on anxiety were observed in mouse offspring during various behavioral tests (Kloke et al., 2013). Reasons for these discrepancies between results of the different studies remain unclear, but might be possibly due to age differences in the offspring between the studies. In addition, the authors suggested that there might have been an unknown extra stressor in the first study, as the mortality rate in mothers was high. Also, in the last study, the stressor was only given postnatal, compared to both pre- and postnatal in the previous studies, which shows the importance of the developmental period during which a stressor is given. Regardless of genotype, when the offspring was given the opportunity to freely choose between novelty and familiarity during free exploration, offspring from the adverse environment tended to show reduced anxiety, while communal nesting tended to increase anxiety-like behavior in this test. These results were contradictory to the expected findings and the authors proposed that both completely safe and stable as well as a maternal environment with significant stressors could result in higher levels of maternal corticosterone levels and increased HPA axis reactivity and fearfulness in offspring. However, they did not measure corticosterone levels in the mothers. Minimal interaction effects between housing condition and genotype were found for anxiety-like behaviors.

The same early life stress paradigm of Heiming (2009) and Heiming et al. (2011) was used and extended by Bodden et al. (2015) who added social defeat as an additional stressor during adolescence in the mice that were pre-and postnatally exposed to unfamiliar male bedding (dangerous environment). Offspring from the neutral bedding group also received a receptive female mouse during adolescence as a beneficial stimulus (beneficial environment). During adulthood, offspring from both environments were either exposed to escapable social defeat or a receptive female mouse. In line with the previous studies, SERT^+/− offspring did not show altered anxiety-like and exploratory behavior compared to SERT^+/+ mice. However, mice that grew up in the beneficial environment and were confronted with escapable social defeat in adulthood displayed overall lower levels of anxiety-like behavior compared to both groups of mice that grew up in a dangerous environment. No interaction effects were found between their life experiences and genotype.

In another study using a postnatal stressor mouse offspring received daily mild electric foot-shocks from postnatal day 7 through 13 (Carroll et al., 2007). Both before and after the stressor, SERT^+/−, and SERT^+/+ mice did not differ in anxiety-like behavior when subjected to the light/dark exploration test, elevated plus-maze, or open field. Also, no differences were found between these genotypes for depressive-like behavior during the second exposure to the forced swim test (Carroll et al., 2007). This indicates that SERT^+/− mice were not vulnerable to the effects of postnatal repeated foot-shock stress under the conditions met in this study. Possibly, repeated foot-shock is not lastingly acting as a stressor (Koolhaas et al., 2011) to resemble early life adversity.

Another postnatal stressor that is considered to be an aversive environmental stimulus is low maternal care. Genetically identical mice can differ in the amount of maternal care they give to their pups during development, either low or high as seen by differences in grooming and licking the pups (Carola et al., 2008). SERT^+/− offspring receiving low maternal care displayed enhanced anxiety-like behavior in the open field and elevated plus maze compared to high maternal care receivers, whereas SERT^+/+ mice were not affected by this early life stressor. Furthermore, SERT^+/− mice receiving low maternal care showed increased depressive-like behavior in the tail suspension test as measured by latency to reach immobility (Carola et al., 2008).

Next to low maternal care, maternal separation is also considered to be an adverse early life event (Lajud and Torner, 2015), as pups are separated from their mother for multiple hours per day, usually during the first 2 weeks after birth. SERT^+/− rats that were separated for 3 h per day on postnatal day two to 14 showed lower escape latencies following an escapable foot-shock, compared to control SERT^+/− rats (van der Doelen et al., 2013). This early stress effect was not present in SERT^+/+ rats. As SERT^+/− rats showed increased stress coping behavior after being exposed to maternal separation early in life, it was suggested that early life stress does not have to always result in negative consequences later in life, but may be a beneficial outcome to early life stress, especially when the stressor early in life was not too intensive (van der Doelen et al., 2013).

Overall, particularly after early life conditions that lower the level of maternal care, SERT^+/− mice and rats can be more sensitive compared to SERT^+/+ mice and rats, to develop anxiety- and depressive-like behavior or increased stress coping (Carola et al., 2008; van den Hove et al., 2011; van der Doelen et al., 2013), whereas in other studies using different stressors no such interaction effects were found (e.g., Kloke et al., 2013; Bodden et al., 2015). Even so, the results so far do not show profound behavioral effects as a result of SERT^+/− genotype × ELS interaction.

The observed behavioral manifestations in SERT^+/− rodents as a result of early life stress are assumed to be attributed to underlying molecular alterations throughout the body and brain. Most likely, one would expect to see adaptations at the level of serotonin homeostasis. Whereas, the normal stress response results in rapid changes in 5-HT levels and turnover (Tjurmina et al., 2002; Bartolomucci et al., 2010), early life stress did not induce long lasting alterations in 5-HT turnover (Carola et al., 2008). Nevertheless, maternal care level and 5-HT levels in the hippocampus interacted in 10-day old mouse pups. SERT^+/+ and SERT^+/− mice experiencing low maternal care did not differ in their 5-HT levels, while SERT^+/− high maternal care receivers showed higher 5-HT levels and 5-HT turnover compared to SERT^+/+ mice that received high maternal care (Carola et al., 2011). Besides the 5-HT homeostasis some other adaptations taking place at the stress regulating HPA axis and extra hypothalamic brain regions regulating HPA axis activity have been reported.

In this review, an overview of SERT gene variation × early life stress interaction and the currently known underlying molecular mechanisms have been provided, specifically by comparing SERT^+/+ with SERT^+/− animals. However, few early life stressors interact with SERT gene variants resulting in both behavioral and molecular alterations. Similar to human S-allele carriers, SERT^+/− rodents can display increased anxiety- and depressive-like behavior in response to early life stressors. Studies using unfamiliar bedding to induce low maternal care, or maternal restraint-stress as a prenatal stressor show little or no effects at all on anxiety levels in SERT^+/− mice (Heiming, 2009; Heiming et al., 2011; van den Hove et al., 2011; Kloke et al., 2013; Bodden et al., 2015). Similarly, daily foot-shocks on postnatal day 7–13 are not sufficient to elicit anxiety- and depressive-like behavior in SERT^+/− mice (Carroll et al., 2007). However, prenatal restraint-stress did increase depressive-like behavior in SERT^+/− offspring, but these mice were strikingly less depressive-like than SERT^+/+ before ELS (van den Hove et al., 2011). The most profound and convincing behavioral effects were found when SERT^+/− mice were subjected to low maternal care early in life, resulting in increased anxiety and depressive-like behaviors (Carola et al., 2008). Even so, this early-life stress method always compares low vs. high maternal care, and does not include normal maternal care levels. Similar studies also including normal maternal care should investigate whether the observed interaction effects are still present.

Furthermore, early life stress can even have a positive outcome on rodent behavior. Stress in the form of chronic maternal separation resulted in increased stress coping behavior in SERT^+/− rats, as seen by shorter escape latencies after an escapable foots-hock (van der Doelen et al., 2013). Nonetheless, it was proposed that SERT^+/− animals are not simply more susceptible to adverse early life experiences, but also seem to be more sensitive for beneficial experiences. For instance, the positive experience of cohabitation with a female mouse reduced anxiety-like behavior in the open field in SERT^+/− but not in wildtype mice (Kästner et al., 2015). Similarly, human S-allele carriers are more sensitive to positive emotions of their partner (Schoebi et al., 2012) and have a strong bias toward both negative and positive affective pictures (Fox et al., 2011) compared to non S-allele carriers. This concept of “for better and worse” suggests that genes “sensitive to stress” actually are plasticity genes and that individuals vary in their plasticity to environmental conditions (Belsky et al., 2009). Furthermore, the outcome of SERT gene variation × early life stress interactions can be of adaptive value. In particular, when the early life environment matches the adult environment, the observed behavioral responses, such as stress coping, can have high adaptive value. According to the match-mismatch hypothesis, stress coping responses are adaptive when they match current stress conditions, but are maladaptive when the conditions mismatch (Schmidt, 2011).

Overall, there is potential that the SERT genotype can determine an animals' stress response to their environment. However, it remains to be established what underlying mechanisms regulate SERT variation × ELS interaction and in particular the corresponding behavioral manifestations. One of the mechanisms through which SERT variation × maternal separation can interact is by inducing adaptations in HPA axis regulation. Overall, it seems that stress can induce a stronger increase in plasma ACTH levels in SERT^+/− rodents than in controls suggesting increased stress sensitivity. Some other HPA axis regulators were also differently affected in SERT^+/− compared to SERT^+/+ rodents in response to stress (Table 2). In humans HPA axis hyperactivity has been found in depressed patients, seen by increased salivary cortisol levels (Juruena et al., 2006; Cowen, 2010).

Furthermore, early life stress did not result in altered mRNA levels of the HPA axis key components or an interaction in SERT^+/− rodents. In the PVN, pituitary and adrenal cortex, only SERT^+/+ rodents were in some cases affected by early life stress. Underlying mechanisms in areas that regulate the HPA axis, such as the PVN and pituitary have been most extensively studied in male rats that underwent the maternal separation stress for 3 h per day from postnatal day 1–14 (van der Doelen et al., 2014a,b, 2015, 2017; Calabrese et al., 2015). Hypothalamic gene expression levels of various key HPA axis activity components such as CRF, MR, GR, and FKBP5 were not affected by SERT genotype × ELS interactions (van der Doelen et al., 2014a). However, in various extra-hypothalamic brain regions SERT genotype and ELS did induce adaptations of MR, GR, and FKBP5 mRNA levels in a brain area-specific way (van der Doelen et al., 2014b). In humans, major depression has been associated with an imbalanced response of glucocorticoid receptors (de Kloet et al., 2007). However, the exact direction of altered glucocorticoid mRNA levels is not always similar between humans and rodents, and might depend on the specific brain region. For example, MR mRNA levels are similarly decreased in the anterior hippocampus of depressed patients (Medina et al., 2013) and in the corresponding ventral hippocampus of SERT^−/− rats (van der Doelen et al., 2014b). However, the interaction of SERT genotype and ELS resulted in increased MR mRNA levels in the ventral hippocampus of SERT^+/− rats (van der Doelen et al., 2014b). Furthermore, SERT variation × ELS interactions are associated with adaptations in gene expression levels of neurotrophic factor BDNF (Calabrese et al., 2015). Overall, SERT variation and maternal separation resulted in a decrease in ventral hippocampus BDNF mRNA levels, although in the ventromedial PFC BDNF mRNA levels were only lowered in the SERT^+/+ and not in SERT^+/− rats. Interestingly, maternal separation stress resulted in increased BDNF mRNA levels in the dorsal hippocampus and dorsomedial PFC of the SERT^+/− rat (Calabrese et al., 2015). These increased bdnf expression levels might contribute to the observed enhanced stress coping behavior in SERT^+/− rats that have been exposed to maternal separation stress in an earlier study (van der Doelen et al., 2013). In addition, CRF mRNA levels negatively correlated with the escape latencies of these animals, thus providing a neural correlate for stress coping behavior (van der Doelen et al., 2015). Furthermore, SERT genotype and maternal separation stress interacted to alter DNA methylation of the crf promotor region in the central amygdala, but not CRF mRNA levels (van der Doelen et al., 2015). This suggests that additional mechanisms are involved in regulating crf expression.

Unfortunately, most of these studies combining SERT genotype and ELS only investigated mRNA levels of genes of interest, while not taking into account the protein levels. Protein levels do not always correlate with mRNA levels and thus protein levels should be verified (Vogel and Marcotte, 2013). In addition, most studies on ELS in SERT variants do not scrutinize the molecular effects at the level of serotonin homeostasis, while the serotonin transporter is obviously of great importance in SERT heterozygous knockout rodents. Also noteworthy, epigenetic repercussions of ELS in SERT heterozygous knockout rodents are currently limited and more studies investigating methylation of the serotonin system-related genes and HPA axis associated genes would be of significant importance. In humans, ELS in the form of childhood abuse is associated with increased methylation of the SERT promotor (Beach et al., 2010, 2011). In rhesus monkeys only a genotype effect was discovered for SERT promotor methylation, with S-allele carriers showing higher cytosine-phosphate-guanine methylation levels (Kinnally et al., 2010). In the van der Doelen study (2017) it was shown that a similar increase in DNA methylation patterns was found in the Ucn promoter of SERT^+/− rats compared to SERT^+/+ when exposed to early life stress, however these increasing levels did not result in different protein expression levels, confirming again that mRNA or epigenetic alterations do not always result into altered protein levels. More research is necessary to understand the functionality of the epigenetic processes.

Importantly, not only early life stressors such as childhood maltreatment, but also stressful events later in life can interact with the 5-HTTLPR and increase the risk for developing depression (Caspi et al., 2003). Likewise, acute or chronic stress later in life can affect behavior of the SERT knockout rodents [reviewed in van den Hove et al. (2011) and Homberg, 2012]. For example, repeated social defeat results in increased social avoidance in adult SERT^+/− mice compared to SERT^+/+ mice (Bartolomucci et al., 2010). Most rodent studies use environmental stressors during a single period in life and thus do not take into account possible interactions with the environment later in life. Therefore, using multiple stressors in both the early and adult life might enhance SERT genotype × stress interaction effects on brain and behavior. An interesting way to study these environmental influences both early and later in life is for example by using a semi-natural environment, where rodents are able to express their full repertoire of behaviors (Snoeren et al., 2015).

In addition, most studies investigating SERT variation × ELS interactions show differences in methodology. The timing of stressors (prenatal, postnatal), the number of stressors per day (how often, how long) as well as how many days the stressor is given differ among studies. Furthermore, different species (mouse or rat) or even different strains might result in discrepancies between studies. Another important notion is the fact that nearly all studies use only male rodents to minimize variation due to hormonal fluctuation. Human females are 1.7 times more likely to develop depression (Vos et al., 2012) and SERT variation × ELS interactions may differ between sexes. Therefore, further studies should include both males and females. Furthermore, most studies focusing on underlying molecular mechanisms came from the same research group using rat SERT variation × maternal separation stress. ELS-induced effects in extra hypothalamic brain regions regulating HPA axis activity are not supported by effects on depressive-like behavior and even show a positive effect for SERT^+/− rats compared to SERT^+/+ rats (van der Doelen et al., 2014a, 2015). Three hours of maternal separation may be too mild to produce adverse long term behavioral alterations in both SERT^+/− and SERT^−/− rats, and in line with the Belsky theory, longer maternal separation periods each day and applied at irregular and thus more unpredictable intervals may be more effective. When maternal separation occurs at fixed time intervals mothers can anticipate the absence of their pups and provide extra maternal care before and after separation, preventing adverse long term behavioral effects in the offspring (Franklin et al., 2010). Indeed, unpredictable maternal separation can induce depressive-like behavior in adult mouse offspring across generations (Franklin et al., 2010).

In conclusion, the exact molecular mechanisms underlying SERT genotype × ELS interactions appear to be highly complex, and diverse molecular alterations at the level of translation and transcription are substantially contributing to it. ELS induced alterations in SERT^−/− rodents may be less suitable to do research on SERT mechanisms, since total absence of the SERT does not exist in humans, although it might generate possible ideas about the importance of the SERT and serotonin homeostasis in behavior. Nevertheless, SERT^+/− rodents might be more translational to study SERT genotype × ELS interactions. Exposing the heterozygous SERT knockout rodent to an adverse early rearing environment may be of high translational value to the more stress sensitive human S-allele carrier, considering the similarities in neurochemical alterations. So far, most studies fail to show solid evidence for increased vulnerability to develop anxiety or depressive-like behavior after ELS in heterozygous knockout rodents. In addition, most molecular findings are obtained in studies that do not confirm depressive-like behavioral manifestations. Hence, further research is warranted as current studies might be insufficient. Main reasons may be: (1) stressors used might not optimal to induce maladaptations (e.g., not severe enough); (2) effects in females are largely absent due to a lack of studies, (3) few studies include both behavioral outcomes and molecular correlates of ELS-induced effects in SERT knockout rodents. The question arises whether the heterozygous SERT^+/− rodent creates a serotonergic subfunctional model. Although it is generally assumed that the approximately 50% remaining SERT molecules lead to differences in various processes in an adult animal, the supporting data is suggesting the opposite. Serotonergic functioning (Table 1) looks comparable to wildtypes. To find out either of these possibilities it is essential to try to elucidate the underlying mechanisms of SERT × ELS interactions, in particular in the heterozygous SERT knockout rodent. By extensively studying and including both long term behavioral and (epi) genetic aspects in both sexes, underlying mechanisms can be directly correlated. Moreover, more severe and realistic stressors should be applied to increase the chance that a not-optimally functioning serotonergic system (the SERT^+/−) might be influenced leading to lasting changes in stress sensitivity. The rodent SERT^+/− × ELS model has potentially high potential, but future research is indispensable before a solid conclusion can be drawn whether this is a suitable translational animal model for ELS-induced psychopathology in psychiatric disorders.

Drafting and/or revising the paper, including final approval to publish: DH, BB, EvdZ, SdB, and JO.

This work was financially supported by the Marie Sklodowska-Curie IF (grant SEP-210188707) and the Swedish Society of Medicine (grant SLS-411161).