Most previous animal model studies have only assessed one developmental ‘hit’, either during early development (e.g. maternal deprivation or neonatal lesions) or during adolescence/young adulthood (e.g. cannabinoid treatment). Similarly, most of the studies using genetically-modified mice have previously assessed baseline behavior only without a second ‘hit’ [90, 91].

In recent years, more groups have begun to generate models that integrate multiple development disruptions. The first ‘hit’ is usually during early development (in the prenatal or early postnatal phase) while the second ‘hit’ is sometimes also during the early developmental phase and in other studies during adolescence/young adulthood. Some studies investigated behavior during the application of the second ‘hit’ but usually behavioral testing is done in adulthood some time after the ‘hits’ have been applied to test for long-term and not acute effects of the stressors. However, the time span between the end of the second ‘hit’ and beginning of behavioral testing can vary extensively among studies with some leaving only a few hours in between, and others weeks. An early study combining two different methods that are individually used to induce schizophrenia-like behavior was done by Hori and colleagues [92]. They used neonatal ventral hippocampus lesions and combined it with 14 days of repeated phencyclidine treatment starting at postnatal day 42 and showed that the phencyclidine-induced locomotor hyperactivity was more pronounced in lesioned rats [92]. However, behavioral testing was only assessed during the phencyclidine treatment phase and the study did not investigate the long-term effects of both ‘hits’ combined. In 2007, Schneider and Koch used neonatal medial prefrontal cortex lesions as their first ‘hit’ and then treated the animals with a cannabinoid receptor agonist during puberty [93]. Behavioral testing in adulthood revealed that both ‘hits’ individually impaired recognition memory but that the combination exacerbated this effect, showing that two ‘hits’ can be more detrimental than one individual ‘hit’ [93]. In a different study, social isolation was used as the first ‘hit’ and repeated MK-801 treatment during young adulthood represented the second ‘hit’ [94]. After a 1-week washout period, animals were tested for amphetamine-induced locomotor hyperactivity and while both treatments significantly enhanced the response to amphetamine, no synergistic or additive effects were found for the ‘two hit’ group [94].

A few studies have combined maternal deprivation/separation with a second stressor. The first study to show that the postweaning environment can influence the behavioral outcome of maternal deprivation was by Ellenbroek and Cools [95]. Animals were maternally-deprived for 24 h on pnd 9 and at weaning age they were either housed in groups or in social isolation. Paradoxically, in both single ‘hit’ groups, prepulse inhibition (PPI), a measure of sensorimotor gating, was decreased in adulthood while the combination of both insults brought PPI levels back to normal [95], although similar ‘hits’ in mice showed different effects [96]. This study looked at the effects of early separation combined with social isolation after weaning and found that early separation induced deficits in a variety of tasks including social interaction, Y-maze, novel-object recognition (NOR), fear conditioning and PPI and these effects were usually more pronounced in animals that were additionally single-housed after weaning [96]. This shows that the combination of two ‘hits’ can produce stronger effects than one ‘hit’ alone.

In previous studies from our laboratory, a ‘two hit’ rat model was developed that included maternal deprivation at pnd 9 and corticosterone (CORT) treatment during young adulthood from 8 to 10 weeks of age [97–100]. Behavioral testing from 12 weeks of age showed that the combination of the two ‘hits’ induced a marked deficit in short-term spatial memory in the Y-maze but no such changes were seen after one ‘hit’ only [98]. Baseline PPI was disrupted after maternal deprivation but CORT treatment had no additional effects [97]. Furthermore, acute apomorphine treatment disrupted PPI in all control groups but not in the ‘two hit’ group indicating differences in dopaminergic regulation of PPI even though no differences in dopamine receptor levels were observed [99, 100]. A more recent study used a similar method but replaced the chronic CORT treatment with a chronic unpredictable stress paradigm [101]. Both ‘hits’ induced a deficit in the NOR in both sexes but no additive or synergistic effects were found for the ‘two hit’ group [101].

Other studies have investigated the effects of chronic cannabinoid treatment in maternally deprived animals. Treatment with the cannabinoid receptor agonist, CP55,940, during adolescence significantly disrupted PPI in adult female animals but maternal deprivation did not further influence this [102]. Maternal deprivation impaired recognition memory in female rats but Δ9-tetrahydrocannabinol (THC) treatment paradoxically restored this deficit [103]. Both treatments were accompanied by changes in NMDA and dopamine receptors but none of these changes were exclusively seen in ‘two hit’ animals indicating that there were no synergistic effects after the combination of maternal deprivation and THC treatment [103].

We recently examined the combined effect in rats of a combination of maternal separation stress and chronic young-adult treatment with CP55,940 from 8 to 10 weeks of age [104]. Here, the combination of maternal separation and cannabinoid exposure induced anhedonia-like behavior in males, expressed as a significant decrease in sucrose preference. Additionally, in male rats both maternal separation and cannabinoid receptor stimulation induced an anxiety-like phenotype in the plus maze and expressed as center time in the open field. These effects were additive and most pronounced after a combination of these ‘hits’. In both males and females, PPI was reduced by maternal separation but there was no effect of CP55,940 treatment. Moreover, memory performance in the Y-maze and novel object recognition test was not affected by either of the two ‘hits’ [104].

Studies utilizing genetically-modified mice and exposing them to an environmental stressor differ in their protocols in regard to stressors used, as well as timing of these stressors. While some apply the second ‘hit’ in the early developmental phase, others use the adolescent phase. NRG1 mutant mouse models have been used to study schizophrenia-like behavior but have yielded mixed results. A recent study investigated the behavior of adult heterozygous transmembrane-domain NRG1 mutant mice that were exposed to chronic social defeat stress during adolescence [105]. NRG1 mice exhibited impaired PPI and a reduction in social novelty preference which were independent of the second ‘hit’. Chronic stress during adolescence reduced the hyperactivity seen in unstressed NRG1 but induced a memory deficit in the mutant but not wild-type mice. Furthermore, BDNF levels in the striatum were significantly decreased in the ‘two hit’ animals only [105]. A few studies have investigated the effects of cannabinoids in NRG1 mice but most investigated the acute effects of cannabinoids immediately after a chronic cannabinoid treatment schedule [106–108]. Only one study additionally investigated the effects of cannabinoid treatment after withdrawal [109]. Animals were tested for locomotor activity, PPI, and anxiety at 48 h after the last injection of a 21-day long chronic THC treatment regimen. NRG1 mice displayed higher locomotor activity than wild-type controls but this was more pronounced in vehicle-treated animals. PPI was not affected by either genotype or cannabinoid treatment and this was similar for anxiety-related behavior [109]. Overall, these results indicate that THC treatment does not significantly alter behavior in NRG1 mice, although more studies are necessary to investigate a larger battery of behavioral tasks. Similar studies have investigated the effects of cannabinoid treatment in COMT mice. Animals were treated with a cannabinoid agonist for 20 consecutive days and behavioral testing started 21 days after treatment had ceased. Cannabinoid treatment impaired PPI, enhanced locomotor activity, impaired spatial memory, and decreased anxiety-like behavior in male COMT knock-out mice [110, 111]. Interestingly, some of these behavioral deficits were only seen after adolescence and not adult cannabinoid treatment [110].

Other studies have investigated the effects of two ‘hits’ during the early developmental phase. For example, one study looked at the effects of maternal separation in heterozygous reeler mice [112]. Maternal separation decreased levels of BDNF but this was less marked in heterozygous reeler mice. Similar effects were seen in the social novelty task indicating that heterozygous reeler mice are less vulnerable toward an environmental stressor [112].

BDNF belongs to the family of neurotrophins which also includes nerve growth factor and neurotrophin 3 and 4 [113]. Neurotrophins are necessary for the normal development of the central nervous system and have important roles in the regulation of neural development, maintenance, survival, and activity-dependent synaptic plasticity [113–115].

The BDNF gene contains nine promoters, and through alternative splicing, a number of different transcripts can be produced which all encode the BDNF protein [116]. BDNF is initially synthesized as the precursor protein pre-pro-BDNF which is then cleaved into pro-BDNF. Pro-BDNF is either secreted and then extracellularly cleaved to mature BDNF (mBDNF) or less commonly converted intracellularly and then secreted as mBDNF [117]. mBDNF binds to the high-affinity tropomyosin-related kinase B (TrkB) receptor which is associated with promoting cell survival and long-term potentiation (LTP) [113, 118]. Major pathways that can be activated through BDNF-TrkB signaling include the phospholipase C-γ pathway, important in synaptic plasticity, which leads to an increase in intracellular levels of Ca²⁺; the mitogen-activated protein kinase pathway, which leads to the activation of several transcription factors resulting in neuronal differentiation and growth; and the phosphatidylinositol 3-kinase pathway, which leads to the activation of the protein kinase Akt which is involved in cell survival and protein translation [113, 119].

In addition to its role during development, BDNF also plays an important role in learning and memory in adulthood. LTP is considered to be the cellular mechanism of learning and memory and is defined as a stimulation-induced persistent increase in synaptic strength [120]. Neuronal activity that leads to LTP increases the transcription of BDNF [121] and BDNF seems to be necessary for early phase LTP as well as late phase LTP (L-LTP) [122]. l-LTP has been shown to be impaired in mice with a deletion of the BDNF gene [123] and these impairments could be rescued by exogenous BDNF administration [124, 125]. BDNF mRNA in the hippocampus was found to be increased after acquisition of memory tasks such as the radial arm maze [126] and the water maze [127]. Furthermore, administration of exogenous BDNF improved performance in spatial memory tasks [128] while inhibition of BDNF mRNA expression in the hippocampus impaired performance [129]. These studies all show that BDNF is strongly implicated in the process of learning and memory and that loss or low levels of BDNF can have functional consequences for cognition. In addition to its implication in cognitive deficits, altered BDNF expression has been associated with a range of diseases including depression, addiction, eating disorders and schizophrenia [130].

Due to its role in development, decreased BDNF functionality could alter normal neurodevelopment leading to dysfunctional neural networks, thereby contributing to the development of schizophrenia. Indeed, post-mortem studies comparing tissue from schizophrenic patients and healthy controls have revealed changes in BDNF levels in schizophrenia. However, while some studies have found decreased protein levels of BDNF in the hippocampus of patients diagnosed with schizophrenia [131], others reported an increase of BDNF but a decrease of TrkB [132]. Protein as well as mRNA levels of BDNF were shown to be significantly decreased in the prefrontal cortex of schizophrenic patients and this was accompanied by a significant reduction of TrkB expression [133–135]. Most of the data derived from these studies came from the brains of schizophrenic patients treated with various medications. If is therefore not entirely clear whether the changes seen are indeed a result of the pathology of schizophrenia or whether they are induced by antipsychotic treatment. BDNF can cross the blood-brain-barrier [136] and several studies have therefore assessed BDNF serum levels in patients with schizophrenia with or without medication. In 2011, Green et al. conducted a meta-analysis of these studies and concluded that there is a moderate reduction of BDNF levels in both medicated and drug-naïve patients with schizophrenia [137].

In addition to altered levels of BDNF, a single nucleotide polymorphism in the BDNF gene which leads to a substitution from valine to methionine at codon 66 (Val66Met) has been implicated in schizophrenia [89, 138–141]. The Val66Met polymorphism leads to altered trafficking of BDNF and subsequently to impaired activity-dependent secretion of BDNF [142, 143]. Healthy Met allele carriers have been shown to perform worse in hippocampus-dependent tasks [143, 144] and to have smaller hippocampal volume compared with homozygous Val/Val controls [145]. Data regarding schizophrenic patients seem to be more complex. While some studies found that the Val allele was associated with schizophrenia [139] and psychosis [140], a meta-analysis from 2007 reported that Met/Met carriers had a higher risk of developing schizophrenia [138] while further studies found no interactions [146, 147]. Age of onset seems to be associated with the Val66Met polymorphism with Met carriers showing an earlier AOP [89, 141, 147]. However, not all studies were able to replicate these findings and a recent study suggested that cannabis use might contribute to the genotype-dependent age of onset [89]. Age of onset was significantly associated with cannabis use in male subjects independent of genotype while in female subjects earlier age of onset as a result of cannabis use was only seen in Met carriers [89]. Overall, despite some conflicting results, it seems that BDNF is associated with schizophrenia and further studies are needed to clarify its role in the disease process. Animal models investigating the role of BDNF can help elucidate some of the findings.

Because of its important role during development, complete deletion of BDNF from birth leads to severe brain abnormalities and BDNF knockout mice die soon after they are born [148]. Therefore, other models have been produced which include BDNF heterozygous mice, conditional and region-specific deletion of BDNF, as well as BDNF Val66Met knock-in mice.

BDNF heterozygous mice (BDNF HET) have about 50 % reduction of normal BDNF expression [149] which is similar to what has been observed in schizophrenic patients [134]. In previous studies, BDNF HET mice showed enhanced amphetamine-induced locomotor hyperactivity [150, 151] and while one study showed enhanced baseline activity [152] this was not replicated by other studies [151, 153, 154]. Studies investigating spatial learning and memory in the water maze have provided conflicting results with one reporting significant impairment in BDNF HET mice [155], although this was not replicated in another study [156]. BDNF HET mice display impaired contextual fear conditioning and this deficit could partially be restored by infusion of BDNF into the hippocampus [157]. No differences between wild-type and BDNF HET mice were found for sucrose preference indicating that animals were normal when tested for anhedonia-like behavior [154]. Overall, the phenotype of BDNF HETs seems to be very subtle. It may be that partial loss of BDNF does not in itself lead to schizophrenia but that it is rather a risk factor and that further ‘hits’ are needed to induce behavioral impairments.

Mice with region-specific deletions of BDNF are used to assess the role of BDNF in specific brain regions that have implications in schizophrenia. Forebrain-specific deletion of BDNF during early development resulted in impaired spatial memory in the water maze but did not alter sensorimotor gating [158]. A similar mouse model with forebrain-specific deletion of BDNF during early development resulted in hyperactivity and severe deficits in fear conditioning while loss of BDNF during adulthood resulted in a less profound outcome [159]. A specific deletion of BDNF only in the dorsal hippocampus during adult age produced impairments in the water maze and novel object recognition task but had no effect on baseline locomotor activity [160]

To study the role of the Val66Met polymorphism in animals, knock-in mice carrying the human Met allele were generated by Chen and colleagues [161]. These mice showed smaller hippocampal volume and demonstrated deficits in context-dependent memory compared to wild-type mice [161]. The study was conducted in male animals and when female mice were used, results were less clear as impairments in novel object recognition were dependent on the phase of the oestrous cycle, indicating that BDNF genotype might interact with oestradiol to regulate memory function [162].

As discussed above, previous preclinical studies show that the combination of two adverse events (either genetic and/or environmental) may lead to additive or synergistic effects, however, the results strongly depend on the stressors used and differences between protocols regarding timing of stressors and time span between application and behavioral testing. We recently conducted a series of studies to assess the involvement of BDNF in the combined effects of early-life stress and a young-adult second ‘hit’.

As mentioned above, our initial studies included a ‘two hit’ model with maternal deprivation at postnatal day 9 and CORT treatment during young adulthood from 8 to 10 weeks of age [97–100]. Behavioral testing from 12 weeks of age showed that the combination of the two ‘hits’ induced a marked deficit in the Y-maze which was accompanied by a decrease of BDNF mRNA in the hippocampus. No such changes were seen after one ‘hit’ only [98]. Baseline PPI was disrupted after maternal deprivation but CORT treatment had no additional effects [97]. Furthermore, acute apomorphine treatment disrupted PPI in all control groups but not in the ‘two hit’ group indicating differences in dopaminergic regulation of PPI even though no differences in dopamine receptor levels were observed [99, 100].

In subsequent studies, we used BDNF heterozygous mice treated from 6 to 9 weeks of age either with CORT to simulate chronic stress, or with the cannabinoid receptor agonist, CP55,940. The stress hormone, CORT, was administered to female and male BDNF heterozygous mice and their wild-type controls through their drinking water [163]. When the animals reached 11 weeks of age, we observed a profound memory deficit in the Y-maze in male, but not female BDNF heterozygous mice treated with CORT. The groups displayed no differences in baseline PPI or its disruption by the NMDA receptor antagonist, MK-801. In addition to the expected reduced levels of BDNF in the mutant mice, there were no additional changes in the expression of this neurotrophin after CORT treatment. Protein levels of the NR2B subunit of the NMDA receptor were markedly increased in the dorsal, but not ventral hippocampus of male BDNF heterozygous mice treated with CORT, an effect which could be related to the spatial memory deficits in these mice, respectively. No significant changes in the levels of subunits NR1, NR2A, and NR2C were observed in males and there were no changes in any of the female groups [163].

We were then interested to see if the role of BDNF in this ‘two hit’ mouse model would generalize to other second ‘hits’. Thus, we investigated whether a BDNF deficit would interact with chronic cannabis intake, a well-described risk factor for schizophrenia development. As with the previous study, BDNF heterozygous mice and wild-type controls were chronically treated during weeks 6–9 of life, this time with the cannabinoid receptor agonist, CP55,940 [164]. Behavioral testing again commenced at 11 weeks of age and revealed no CP55,940-induced deficits in short-term spatial memory in the Y-maze and no changes in novel object recognition memory either. In this study, baseline PPI was found to be reduced in BDNF heterozygous mice and chronic CP55,940 treatment did not alter this. However, acute CP55,940 administration caused a marked increase in PPI particularly in male BDNF heterozygous mice pre-treated with this same drug but not in any of the other male groups. All female groups showed small increases of PPI after acute CP55,940 administration. We then analyzed the levels of [³H]CP55,940 binding by autoradiography and found a significant increase in the nucleus accumbens, but not caudate nucleus of male BDNF HET mice previously treated with this drug. There were no changes in binding in any of the other groups [164]. These results contrast with the effect of CORT treatment in the BDNF HET mice in that BDNF deficiency and chronic young-adult cannabinoid receptor stimulation did not interact on learning and memory later in life. Again in contrast to CORT, cannabinoid receptor stimulation elicited a hypersensitivity to the effect of acute CP55,940 on PPI, which could be related to up-regulation of cannabinoid receptor density in this region.

Clearly, ‘two hit’ effects in BDNF animal models appear to depend markedly on the nature of the second ‘hit’. To investigate if this was unique to the BDNF heterozygous mouse model, we followed this up with a series of studies in rats that were subjected to a maternal-separation protocol as the first ‘hit’ and either CORT treatment or CP55,940 treatment as the second ‘hit’.

Wistar rats were exposed to neonatal maternal separation for 3 h per day on postnatal day 2–14 and/or received CORT in their drinking water during 8–10 weeks of age [165]. Male, but not female ‘two hit’ rats showed marked disruptions in short-term spatial memory in the Y-maze. However, female ‘two hit’ rats showed signs of anhedonia in a sucrose preference test, which were not observed in males. Novel object recognition and anxiety measures in an elevated plus maze task were unchanged by either of the two ‘hits’. We then obtained dorsal and ventral hippocampus regions and used quantitative polymerase chain reaction (qPCR) to assess exon-specific BDNF gene expression or Western blot to assess BDNF protein expression and downstream signaling. In the dorsal hippocampus, maternal separation caused a male-specific increase in BDNF exons I, II, IV, VII, and IX mRNA but a decrease in mBDNF and phosphorylated TrkB (pTrkB) protein expression in adulthood. These effects were not seen in the male ventral hippocampus. However, in female rats only, maternal separation caused a significant decrease in mBDNF and pTrkB protein expression in the ventral hippocampus in adulthood. Thus, in this maternal separation model, long-lasting, region-specific, and sex-specific effects on BDNF expression and signaling were observed, which could be involved in the sex-specific qualitative differences in the behavioral profile of these animals, particularly with respect to spatial memory and anhedonic behaviors.

Together with the above-mentioned studies on the effects of CORT or cannabinoid receptor stimulation in BDNF HET mice [163, 164], these results in rats confirm that early developmental disruptions and young-adult stress or cannabis use [104] on their own or in combination can differentially affect behaviors related to neuropsychiatric disorders and that BDNF is likely to play a central role in this interaction. However, different ‘two hit’ combinations produce markedly and qualitatively different behavioral phenotypes in male vs. female animals.

There is ample evidence that the etiology of schizophrenia and other neuropsychiatric disorders involves complex gene-environment interactions where altered expression of brain factors relevant to plasticity and development produces a vulnerability to other ‘hits’ later in life, such as adolescent young-adult stress or drug abuse. Adding to substantial, but incomplete and often inconsistent clinical literature, animal model studies are beginning to unravel the complex multifactorial mechanisms in the brain which mediate these interactions. Our work has focused on BDNF but it is clear that several other early neuromodulators, including immune factors and neuregulin, COMT and DISC1, could be involved and may synergize with different second ‘hits’ to induce their own unique profile of behavioral effects in adulthood. Further pre-clinical studies will be of paramount value to elucidate the brain mechanisms involved in these various combinations. This may be relevant to recognize targets for early intervention which appear more promising to reduce the burden of complex psychiatric illnesses like schizophrenia than symptomatic pharmacotherapy which is often associated with severe side-effects.