Introduction

Over the past decades, it has become clear that the perinatal environment in conjunction with genetic factors shapes an individuals’ predisposition to increased risk of disease1. Brain development is influenced by various intrauterine, prenatal and postnatal events interacting with the genotype to affect the neural systems related to emotions, cognitive and language abilities, which may also increase the risk for psychopathology later in life. Noxious stimuli include maternal infection yielding excess inflammation, maternal stress, obesity and metabolic disorders, many of them associated with intrauterine growth restriction (IUGR) and/or preterm birth. In addition to these foetal elements, the immature brain is at risk to be exposed to additional injurious stimuli during or early after birth, e.g. infection, foetal hypoxia, intrapartum oxygen deprivation, hyperoxia, stress from maternal separation and medical procedures, exposure to medication or anaesthesia further increasing the risk for the development of behavioural and neuropsychiatric brain diseases in later life.

Perinatal brain injury affects infants born at all gestational ages with a higher incidence in the most vulnerable very immature group of patients. Improved obstetric and neonatal care in conjunction with the use of preventive and/or therapeutic strategies such as antenatal steroids prior to preterm delivery and the introduction of hypothermia treatment for hypoxic–ischaemic encephalopathy (HIE) has led to increased survival rates and also reduction of severe neurological impairment. For instance, a recent meta-analysis of world-wide cohorts born after 2006 revealed that the pooled prevalence of cerebral palsy in infants born extremely preterm was with a rate of 6.8% (95% confidence interval 5.5–8.4) reduced compared to previous studies2. In the majority of cases suffering from an adverse perinatal environment, physiological brain maturational processes are affected leading to a variety of neurodevelopmental disorders in later life such as cognitive impairment, autism spectrum disorders (ASD), attention deficit (hyperactivity) disorder (ADHD) and psychiatric disease like schizophrenia (SZ)3.

While these sequelae are frequently related to perinatal brain injury of various origins, investigations of the underlying anatomical and pathophysiological processes remain a major goal for research in neonatal neurology. In this review, we give an overview on multiple risk factors for the development of neuropsychiatric diseases focussing on recent findings derived from experimental models (Table 1). We further summarise shared cellular and molecular mechanisms across a variety of neurodevelopmental disorders. Finally, we discuss challenges for preclinical research focussing on the assessment of neuropsychiatric symptoms in experimental models.

Table 1 Modelling adverse neuropsychiatric development following perinatal insults
Full size table

Risk factors for the development of neuropsychiatric diseases

Maternal and early postnatal stress

Human studies suggest a positive association between maternal stress and the risk of preterm delivery, foetal growth restriction, altered social behaviour, ADHD symptoms and impaired cognitive performance4,5. Furthermore, analysis in siblings demonstrated that exposure to high levels of cortisol in utero are associated with lower adult educational attainment and verbal cognition6. The underlying reasons for these effects are assumed to be stress-related alterations in brain structure. As such, it was shown that children born to mothers who experienced high levels of anxiety in the early second trimester of pregnancy or chronic perceived stress had region-specific reductions in grey matter volume in the cortex, hippocampus and cerebellum7,8. Nevertheless, information about the detailed cellular and molecular mechanisms is mainly limited to animal studies4. In experimental models, prenatal stress can be induced either by restraint under bright light, nest manipulation or predator stimuli, which alters the offsprings’ anxiety-related behaviour, learning and memory and stress regulation4,9,10,11,12. Neuropathology includes alterations in hippocampal gene expression, cortical inhibitory neuron subtypes and hypothalamic–pituitary–adrenal axis responses or cortisol reactivity13,14,15. Furthermore, basic neurodevelopmental processes like cellular proliferation, differentiation and migration are delayed16,17. Importantly, maternal stress is supposed to interact with multiple genes involved in neurogenesis, serotonergic genes and genes involved in GABAergic transmission, implicated in both SZ and ASD18. For example, interaction between the serotonergic gene 5-hydroxytryptamine transporter (5-HTT aka. sodium-dependent serotonin transporter and solute carrier family 6 member 4 (SLC6A4)) and stress was observed for ultrasonic vocalisation19 and ASD-like stereotypic repetitive grooming behaviour20. In addition to maternal stress, early postnatal stress models have been developed with maternal separation being one of the most prominent causes leading to long-lasting increases in anxiety, depressive behaviour and disruption of stress responses. These changes are accompanied by reduced neural activation in the amygdala and other areas implicated in social functioning, including the medial prefrontal cortex and nucleus accumbens21,22. Recently, it was shown that long-lasting stress (i.e. 3 h maternal separation daily for the first 3 postnatal weeks) results in increased anxiety and reduced social interaction accompanied by impaired oligodendrocyte differentiation and myelination in the prefrontal cortex23. Nevertheless, often only single regions and single cell types were analysed. Other molecular events such as alteration not only in synaptic strength and neuron connections but also inflammatory processes may contribute to early stress-induced behavioural alterations in later life. The relationship between synaptogenesis, synaptic pruning and inflammation has been demonstrated in a variety of studies24,25,26. However, whether these mechanisms are associated with early-life stress and later life changes remains to be investigated.

Maternal obesity

Causal links between maternal obesity and the development of severe perinatal brain injury, e.g. periventricular leukomalacia, HIE or intraventricular haemorrhage are rather scarce27,28. In contrast, high amount of human studies indicate that it is associated with an increased risk for various mental health disorders, such as not only ADHD, ASD, depression and SZ but also cognitive impairments (reviewed elsewhere29). This is supported by animal research30. For instance, in a rodent model of maternal high-fat diet (HFD) it was shown that female offspring revealed decreased sociability and males presented with increased hyperactivity31. However, this seems to contradict other reports demonstrating that HFD leads to signs of anxiety and depression assessed in the elevated plus maze (EPM), open field (OF) and forced swim test in male offspring32,33. In addition to rodents, a non-human primate model was used demonstrating that adiposity in pregnancy increases the risk of impulsive and disruptive behaviour34. Besides alteration of emotions, cognitive function is reduced with delayed spatial learning, deficits in reference memory and associative learning accompanied by attenuated levels of amino acids in the medial prefrontal cortex and hippocampus35,36. Preclinical studies in rodents and primates suggest abnormal development of the serotonergic and dopaminergic system and obesity-induced inflammation as underlying mechanisms29,34. This is supported by human studies demonstrating that maternal obesity is associated with changes in cytokines, hormones and acute-phase proteins in the second trimester and with an increased risk of maternal infection or childhood asthma37,38.

Placental insufficiency

IUGR often caused by placental insufficiency is a major contributor to neurodevelopmental deficits involving problems in motor skills, cognition and memory in school children39. It is assumed that deviant brain development and altered maturational processes may lead to increased chances of later cognitive, emotional and behavioural problems40,41,42. Neuroimaging studies revealed structural brain alterations and disturbed white matter organisation leading to impaired neurodevelopmental outcomes, which persist into young adulthood43,44,45.

Growth-restricted foetuses due to placental insufficiency are chronically hypoxemic and hypoglycaemic. In rodents, placental dysfunction can be caused by reduced utero placental perfusion leading to hypertension and placental ischaemia46,47 or genetic knockdown of the insulin receptor reflecting maternal diabetes48. Despite initial epidemiological evidence indicating a relationship between these maternal metabolic disorders (i.e. hypertension and diabetes) and the risk for the development of SZ49,50, animal research in terms of long-term neuropsychiatric outcomes in above described experimental models is still lacking.

Most commonly applied models for evaluation of neurodevelopment in IUGR offspring include umbilical/uterine artery ligation or chronic hypoxia in sheep, piglets or rodents39. Neuropathological consequences include reductions in total brain volume, myelination deficits and impaired brain connectivity39. Furthermore, unilateral ligation of the uterine artery at mid-gestation was associated with ventriculomegaly, reduced basal ganglia volume, impaired synaptogenesis and altered cerebellar Purkinje cell development in the guinea-pig brain at adolescence, resembling anatomical changes found in some individuals with SZ18,51,52,53,54. Of note, prepulse inhibition (PPI) of the acoustic startle response (ASR) was reduced53. In addition to the PPI deficit that is supposed to parallel sensorimotor gating deficits observed in SZ, cognitive deficits have been observed in rodent models of IUGR induced either by maternal caloric undernutrition or by meso-ovarian vessel cauterisation leading to learning disabilities and spatial memory deficits in adolescent offspring55,56.

Hypoxia–ischaemia (HI)

Another frequent insult in the perinatal period is term asphyxia with HIE affecting approximately 1.5–3 in 1000 live births57. Despite improvement of outcomes after introduction of hypothermia treatment mortality remains significant and severe motor and cognitive impairment may lead to cerebral palsy evident in the first 2 years of life. Beside the severe forms of brain injury children may also suffer from attention, memory and behaviour problems, requiring developmental surveillance beyond 2 years of age. Diversity of outcome across grades of HIE is reported but only few studies have addressed the milder consequences of HIE at school age both in the pre- and post-cooling period58,59,60. Furthermore, children with diagnosed stroke in the neonatal period, a disease with an incidence between 7.1 in 100.000 and 16 in 100.000 live births, may also present with poor neurological and psychosocial outcome, which is largely dependent on the lesion size61,62,63,64. Those deficits can be attributed to the pattern of injury and affected brain structures, such as hippocampus, cortex and striatum65.

The most commonly used neonatal asphyxia model has been established in rodents by Rice–Vannucci in 1981 with unilateral common carotid artery (CCA) occlusion followed by systemic hypoxia (HI) leading to excitotoxicity, apoptosis, inflammation and subacute white and grey matter injury resulting in long-term brain atrophy mainly affecting cortex, hippocampus and striatum66,67,68. This type of injury leads to long-term sensorimotor deficits while general locomotor activity is only transiently increased resolving at 3 weeks of age67,69,70. This transient increase in locomotor activity may reflect hyperactivity, a major hallmark of ADHD. Furthermore, adolescent and young adult animals reveal reduced anxiety, an increased risk-taking behaviour, deficits in sustained attention and increases in impulsivity and compulsivity67,69,71,72,73. Together, these results suggest that HI contributes to the development of characteristics related to ADHD. In addition to alterations in impulsive behaviour, several groups also demonstrated deficits in cognitive abilities either in recognition or working memory or learning capability67,69,70,71,74,75. With regard to symptoms associated with SZ, discrepant results have been described. While unilateral HI in 10-day-old mice did not modulate sensorimotor gating72, bilateral CCA ligation in 12-day-old rats revealed a significant deficit in this behavioural domain76. The variability of outcomes can be most likely explained by differences in experimental models, age and species.

In addition to postnatal HI in term-born equivalent rodents a prenatal hypoxic–ischaemic brain injury model has been recently developed. Offspring of dams exposed to transient intrauterine systemic HI have been shown to develop symptoms of ASD revealed by poor social interaction77. Subplate neuronal loss and impaired cortical maturation resulting in impaired thalamocortical connections have been discussed as potential underlying mechanisms77.

Hyperoxia

While HI causes an acute disruption of oxygen supply reflecting asphyxia, preterm birth is frequently associated with exposure to non-physiological high oxygen concentrations. This is caused by a sudden rise in oxygen tension compared to intrauterine conditions and is further increased by intensive care78,79. Although our understanding of neuropathological hallmarks of hyperoxia-triggered brain injury has increased within the past decade80, involving inflammatory responses and induction of reactive oxygen species associated with oligodendroglial cell death and hypomyelination triggering ultrastructural changes of the developing white matter81,82,83,84,85,86,87,88,89,90,91,92,93,94, the evaluation of long-lasting functional deficits has only recently begun88,95,96. Neonatal hyperoxia is associated with reduced motor-skill learning and memory deficits assessed in the complex running wheel, Barnes maze and novel object recognition that persist into late adulthood88,95,96. These findings resemble the difficulties observed in neurodevelopmental follow-up studies in ex-preterm infants, emphasising the importance of experimental preclinical research.

Inflammation

Early perinatal immune activation is associated with long-lasting consequences on cognitive and emotional health throughout life. Clinical studies have linked early- and late-onset infections to adverse outcome with development of cerebral palsy97. In addition, pattern of inflammatory markers have been related to neurological outcome in very immature infants as investigated up to 10 years of age98,99. Different models of immune activation have been established. The most common immunogens are the viral mimetic polyinosinic–polycytidylic acid (poly I:C) and the bacterial mimetic lipopolysaccharide (LPS). Poly I:C induces several symptoms of ASD and SZ, such as impaired cerebral communication, abnormal social and repetitive behaviour, decreased sensorimotor gating, working memory deficits and increased anxiety100,101,102. These behavioural abnormalities are associated with neuropathological alterations, e.g. not only reduced cortical thickness and decreased volumes of hippocampus, amygdala and striatum103 but also ASD-associated aberrations in Purkinje cells and altered levels of synaptic proteins, as well as deficits, in synaptic transmission102,104,105,106. Furthermore, neurochemical alterations involving serotonergic and dopaminergic signalling have been linked to SZ and ASD pathology101,107. While LPS and poly I:C induce a general rise in pro-inflammatory cytokines, prenatal administration of selective cytokines such as interleukin-6 (IL-6) or IL-17 are sufficient to elicit ASD- and SZ-like behaviours102,108. Besides these artificial immune stimuli, models of congenital viral infections have been used, such as cytomegalovirus infection leading to delayed acquisition of neurological milestones (e.g. righting and grasping reflexes, cliff aversion and negative geotaxis) and disturbances in social behaviour and exploratory activity in adolescent animals109. Clinical signs were accompanied by brain atrophy, cerebellar hypoplasia and neuronal loss109. In spite of the huge number of available models altogether underpinning the detrimental impact of perinatal inflammation on long-term neurodevelopment, it needs to be critically emphasised that most outcome measures focus on distinct and selective behavioural domains, single brain regions and/or single analysis time points in a single animal model. Comprehensive systematic analyses identifying common neurobiological principles underlying distinct pathologies should be a future goal in this research field.

In addition to immune stimuli, several different species have been studied with various outcomes, i.e. while maternal treatment with poly I:C in mice leads to long-lasting deficits in sensorimotor gating in both male and female102, Wistar rats seem to be less sensitive to maternal immune activation as sensorimotor deficits were only observed in males110. To overcome limitations regarding translation, non-human primate models of maternal immune activation have been developed111,112 also recapitulating symptoms of neurodevelopmental disorders, e.g. abnormal repetitive behaviour.

Timing of immune challenge seems to be critical not only for foetal brain pathology113 but also for long-term behavioural outcomes. For instance, maternal immune activation with poly I:C at late gestation is supposed to be more effective to induce working memory deficits compared to early gestation110,114. Similarly, maternal LPS injection at early gestation (i.e. embryonic day 12 (E12)) leads to impaired reward seeking behaviour while at late gestation (i.e. E16) impaired motor behaviour is induced115. Comparative analysis in the poly I:C model further demonstrated that prenatal immune activation at E9 but not at E17 reduces OF exploration, impairs the latent inhibition effect of associative learning and alters sensorimotor gating82,114,116,117. On the other hand, E17 but not E9 immune activation led to perseverative behaviour117. Considering the delay in rodent brain development compared to humans, with the first postnatal week in rodent brain development corresponding to the third trimester in humans118,119, several postnatal models of inflammation have been developed. A single low-dose LPS injection at postnatal day 3 (PND3) or repetitive injections of IL-1beta over the first 5 PNDs impairs recognition memory, which is accompanied by long-lasting alterations of the white matter microstructure120,121. Interestingly, locomotor and anxiety-related behaviour were not modulated120. In addition to cognitive dysfunction, pain and conditioned fear responses can be modulated by neonatal inflammation as previously shown after immune challenge with LPS at PND14122.

Multiple hit hypothesis

Epidemiological studies suggest that the probability to develop a neuropsychiatric disorder does not only rely on a single risk factor but also requires exposure to multiple “hits”. For instance maternal immune activation is supposed to be a “disease primer” increasing susceptibility to effects of genetic mutations and postnatal risk factors that synergise to trigger mental diseases in later life102,123. This is supported by clinical and experimental data, e.g. the effect of maternal infection in increasing the risk for SZ is higher in families with SZ history101. In preclinical models, it was shown that low-dose poly I:C synergises with SZ- and ASD-related genes like DISC1, NRG1, NR4A2 and TSC2 inducing increased ASD- and SZ-related abnormalities compared to single insults101,102,107.

In addition to the combination with genetic variations, animal models of prenatal inflammation have been combined with intrauterine or early postnatal HI124,125,126,127. However, the majority of studies mainly focussed either on analysis of acute inflammatory responses and/or pathological alterations involving for example myelination deficits126,127. Furthermore, correlations to behavioural changes have rarely been assessed and if so these analyses focussed on motor deficits that were similar in single intrauterine-induced hypoxic–ischaemic injury and in the combined setting125. Analysis of behaviours relevant to mental disorders have only recently been explored in more subtle models of white matter injury, e.g. by combination of foetal inflammation with postnatal hypoxia not involving severe lesions or gross anatomical changes. This combination led to long-term changes in oligodendrocyte maturation, which was associated with increased anxiety and signs of autism-like behaviour, i.e. reduced social play behaviour and increased repetitive grooming128. Besides episodes of hypoxia, exposure of preterm babies to high oxygen concentrations is critical in terms of brain development (see above). Combining an inflammatory stimulus (i.e. LPS) with hyperoxia (80% oxygen) results in long-lasting impairment of white matter microstructural integrity, which did not significantly differ compared to single insults. However, underlying mechanisms were different in so far that hyperoxia but not LPS predominantly induced apoptotic cell death while LPS rather impaired oligodendrocyte maturation82. Even though these pathophysiological processes are important to understand, analyses of behaviours potentially relevant to cognitive and neuropsychiatric disorders are lacking.

In addition to early prenatal and postnatal factors, events in later in life may also increase the risk for neurodevelopmental diseases. This was elegantly demonstrated in a translational mouse model that combined exposure to prenatal immune challenge and peripubertal stress resulting in synergistic effects on adult behavioural alterations associated with neuropsychiatric diseases129.

In spite of a whole body of clinical and experimental evidence (with only a few selected here) confirming the multiple hit hypothesis for the development of mental illness, the underlying mechanisms remain obscure. Further research is necessary combining diverse prenatal and postnatal risk factors to characterise phenotypes and illuminate the cellular and molecular pathways for each factor.

Sexual dimorphism

Compelling clinical and preclinical evidence demonstrates a pronounced sexual dimorphism across a huge variety of cognitive deficits and neuropsychiatric disorders. This is supported by experimental work since males display deficits in executive function and attention while females presented increased immobility in the forced swim test and reduced saccharin intake after immune activation in lactating dams130. These findings recapitulate the clinical situation where men have an increased risk for SZ and women for depression131,132. Of note, hippocampal and striatal volume reductions were similar in males and females. However, the inflammatory response, i.e. pro-inflammatory cytokine levels (i.e. IL-6, interferon-gamma) in the offsprings’ brain were increased in males supporting the concept of cytokine-mediated perinatal programming of neuropsychiatric disorders. Similar sex differences were observed in a model of intrauterine inflammation induced by LPS injection at E15 demonstrating increased cytokine responses in males133. However, behavioural phenotypes were similar in that study contradicting the results by Arad et al., which might be explained by different behaviour tests applied but more likely by the different timing of perinatal immune activation133. This is supported by a recent report where postnatal LPS challenge between P5 and P7 resulted in differential behavioural responses in males and females suggesting that the sexual dimorphism in behavioural domains is more prominent after late immune challenges134. Even though divergent behavioural responses dependent on sex were confirmed, the kind of behavioural deficits differed comparing studies of Custodio et al. and Arad et al., i.e. depressive-like behaviour was observed in female but not in male mice after immune challenge by lactating dams130 while direct immune challenge via intraperitoneal injection to neonatal animals at a similar time point resulted in the opposite with males but not females presenting depressive-like behaviour134. These different outcomes suggest that not only timing but also the mode of immune challenge may be decisive for final outcome and sexual dimorphism in adulthood. On the other hand, these differences might have been caused by alteration of maternal behaviour after injection of LPS to the lactating dam. Similar low levels of maternal care have been described when immune activation occurs in mid-gestation135,136. Besides long-lasting divergent outcomes in behaviour, emerging evidence suggests that inflammatory responses also differ between adult males and females. As such, males demonstrate increased myeloperoxidase activity in the hippocampus at P35 and P70134. Furthermore, cytokine responses to adult LPS challenge are increased in males at adulthood after intrauterine immune challenge133.

In addition to perinatal immune challenges, neonatal HI induces long-lasting immune responses, which differ between genders. For instance, in adolescent male mice circulating levels of tumour necrosis factor (TNF)alpha were increased and males revealed more lymphocyte infiltration in injured brain hemispheres accompanied by reduced neurogenesis137. Regarding HI-induced atrophy and neurobehavioural deficits, gender differences were observed after neonatal HI in P9 mice with reduced brain atrophy in females while deficits in OF behaviour were increased, likely representing injury not detectable by volume measurements alone138.

Sex differences have also been observed in models of maternal obesity with female offspring of HFD-fed dams revealing increased anxiety in the OF but males being more hyperactive31. Furthermore, latency in initial and reverse Barnes maze trials were decreased in male but not in female offspring of HFD-fed dams139.

The most obvious mechanism underlying sexually dimorphic outcomes seem to be sex hormones that is supported by epidemiological data revealing a higher incidence and earlier onset of SZ in males compared to females while incidence in females increases with age and outnumbers that of males after the age of 50 years140. The sex hormone hypothesis is in line with numerous reports on the neuroprotective effects of oestrogen84,141,142,143. Major target mechanisms of sex hormones are supposed to be regulation of neurotransmitter dysfunction and synaptic plasticity144,145. Nevertheless, preclinical reports demonstrating no sex differences in hormone levels but differences in outcome regarding early and late inflammatory responses and histological outcomes after neonatal HI137,146 imply that additional intrinsic mechanisms may play a role in divergent susceptibility to the development of neurological deficits. This was demonstrated in in vitro cultures of isolated cells or brain slices revealing differences in cell death pathways and differentiation of specific hippocampal neuron subpopulations independent of the presence of hormones147,148. In view of the huge amount of epidemiological evidences for more than two decades, preclinical in vivo research is now beginning to include stratified analyses for sex across a huge variety of perinatal injury models.

Mechanisms potentially relevant for the development of neurodevelopmental disorders

In spite of the well-known heterogeneity of cognitive and neuropsychiatric symptoms, only some of them overlapping between different diseases and despite the obvious diversity of risk factors, several common molecular and cellular mechanisms have been proposed to underlie the pathology of neuropsychiatric disorders. Here we will focus on the most prominent molecular and cellular components.

Cytokines

The major hypothesis how maternal immune activation impairs neurodevelopment and potential consecutive development of neuropsychiatric symptoms is an alteration of the balance between pro- and anti-inflammatory cytokines. As such, blockade of IL-6 and IL-17 during maternal immune activation via poly I:C reverses behavioural abnormalities resembling ASD-like behaviour while an increase of IL-10 is protective101,108. Whether maternal increases in pro-inflammatory cytokines mediate direct effects on the foetal brains or rather indirectly through induction of placental inflammation leading to downstream effects such as oxidative stress and nutrient deficiency is still not entirely clear149,150. Of note, elevated cerebral cytokines levels may persist into adulthood in a region- and age-specific manner114,117,151,152. Changes in cerebral major histocompatibility complex-I expression and associated intracellular mammalian target of rapamycin signalling are proposed mechanisms underlying altered synaptic plasticity and function associated with neuropsychiatric disorders100.

In addition to prenatal immune activation, hyperoxia also generates inflammatory responses in the developing brain, as demonstrated by a marked increase in mRNA and protein levels of caspase-1 and its downstream effectors IL-1beta, IL-18, and IL-18 receptor alpha (IL-18Ralpha) in PND6 rodents exposed to hyperoxia for 2–48 h88,153. Of note, both intraperitoneal injection of a specific inhibitor of IL-18 and genetic deficiency of IL-1 receptor-associated kinase 4, which is pivotal in both IL-1beta and IL-18 signal transduction, attenuated hyperoxic brain injury153. Even though a direct correlation to cognitive development and neuropsychiatric symptoms remains to be determined, these findings causally link inflammation triggered by pro-inflammatory cytokines such as IL-1beta and IL-18 to early hyperoxia-induced cell death resulting in long-lasting white matter structural abnormalities86,87,88,89,95.

Similarly to hyperoxia, maternal obesity and neonatal HI increase cytokine levels in the offsprings’ brain. This was demonstrated in rodent and non-human primate models of obesity34,154,155. Moreover, after neonatal hypoxic–ischaemic brain injury in rodents, cerebral levels pro-inflammatory proteins (e.g. IL-1beta and TNFalpha) were increased compared to sham-controls while anti-inflammatory cytokines (e.g. IL-10) were decreased 4 days after injury that was associated with long-term alteration of emotional and impulsive behaviour assessed in the EPM71. In addition to short-term regulation of brain cytokines, recent work also demonstrated a long-lasting increase in the pro-inflammatory cytokine TNFalpha up to 30 days after injury at least in males137.

Taken together, these findings strongly suggest that cytokines are a contributing mechanism underlying the pathology of developmental neuropsychiatric disorders. Nevertheless, determination of molecular targets and dynamic changes of a larger number of cytokines across brain regions, developmental stages and animal models to identify immune signatures associated with specific neuropsychiatric symptoms will be a major goal for future.

Microglia

Microglia as resident immune cells of the brain play an important role during physiological brain development. Microglia functions during development include not only phagocytosis of synaptic elements and living and dying cells but also stimulation of myelination, neurogenesis and cell survival24. The majority of studies in early brain injury models propose a detrimental activation of microglia mainly based on their amoeboid morphology associated with increased pro-inflammatory cytokine release similar as in adult neurodegenerative diseases. However, these simple conclusions might be misleading because immature microglia differ from adult microglia for example in gene expression in response to LPS activation156. Microglia development proceeds through three distinct stages—early, pre-adult and adult—with characteristic gene expression and functional states. Therefore, it is not surprising that perturbations of this maturational process during early brain development modulate microglia. For instance, maternal immune activation results in a transcriptional shift of pre-microglia towards the more advanced developmental stage157. Thus analysis by morphology alone might provide limited information. Assessment of microglia function by gene expression, phagocytic activity or altered ultrastructure is suggested to be more accurate24.

A major mechanism potentially relevant for the pathophysiology of developmental brain disorders is an alteration of microglia-mediated synaptic pruning. This is supported by a recent report using a model of prenatal inflammation by a single LPS stimulus at E15 resulting in symptoms reminiscent of ASD (e.g. increased stereotypic movements and reduced ultrasonic vocalisation, reduced social preference)158. Of note, these long-term behavioural alterations were accompanied by reduced hippocampal expression of the fractalkine receptor CX3CR1 at P15158, the time point and structure where effects would be most readily detected for microglia-mediated synaptic pruning159. Even though microglia-specific CX3CR1 deletion studies are still missing, a causal relationship between CX3CR1 signalling and impaired social behaviour was recently demonstrated by Zhan et al.160. Deficient synaptic pruning due to CX3CR1 deletion (CX3CXR1 knockout mice) coincided with weak synaptic transmission and altered functional connectivity in the prefrontal cortex leading to deficits in social interaction and increased grooming behaviour160. Interestingly, CX3CR1 mRNA expression is only reduced in males after prenatal immune activation158 suggesting that different microglia responses to perinatal insults might be a major cause for observed sex differences in behavioural outcomes as described above.

While the importance of CX3CR1 is largely acknowledged by previous reports159,160, recent work in a model of postnatal inflammation by repetitive IL-1beta injection very elegantly identified a potential new molecular player previously not linked to microglia function but rather considered as a marker of neuronal postsynaptic density161. Using complex gene network analysis of the microglial transcriptomic response to injury, it was demonstrated that discs large homologue 4 (DLG4, aka. postsynaptic density protein-95) is expressed in microglia of the immature mouse and human brain, which was developmentally regulated and modulated by inflammation161. Since DLG4 knockout mice reveal increased repetitive behaviour, abnormal communication and social behaviour162, this might be a new important mechanism how microglia contribute neurodevelopmental disorders after perinatal infection. Even though the underlying mechanism of DLG4 action in microglia remains unclear, its involvement in crosstalk between oligodendrocytes, astrocytes and microglia or regulation of glutamatergic and GABAergic signalling was suggested161. Nevertheless, cell-specific deletion would be important to dissect the functional role of DLG4 on microglia for long-term neurodevelopmental outcome.

In addition to immune activation, early-life stress has also been implicated in perturbation of microglia function in the hippocampus163. Furthermore, neonatal HI and hyperoxia are known to result in increased abundance of the microglial cell marker Iba-171,88,164. In-depth analysis in experimental neonatal HI revealed a change in microglia phenotypes even though no consistent classification could be made regarding the classical M1 and M2 polarisation states165. Furthermore, the classical view of detrimental microglia activation following perinatal insults is challenged by the growing number of evidence for a protective role of microglia under certain conditions, such as neonatal ischaemic brain injury25. It was shown that depletion of microglia enhances local inflammation induced by ischaemia–reperfusion and increases injury severity in the developing brain166. Taken together, the functional relevance of phenotypic and functional microglia differences associated with perinatal noxious insults for the development of cognitive and neuropsychiatric disorders in later life remains to be further investigated.

Oligodendrocytes

The development of the mammalian brain involves complex cellular processes, like migration, glial cell proliferation, axonal and dendritic outgrowth, synaptogenesis and myelination of axons167. Although neurogenesis is mostly completed around 24 weeks of gestation, i.e. at the border of viability of preterm infants, outgrowth and the formation of long-range connectivity supported by glia maturation and axonal myelination are still in progress168,169. Development of diffuse white matter alterations and a reduction of cortical and hippocampal grey matter volume are often associated with cognitive impairment, attention deficit disorder, behavioural problems, autism and development of psychiatric disease in later life170,171,172,173.

Experimental work detecting long-lasting subtle alterations of white matter integrity and brain connectivity following perinatal insults like HI, hyperoxia, inflammation or IUGR80,164,174,175 confirm reports about white matter abnormalities in psychiatric patients176,177,178. Oligodendrocyte development is characterised by different developmental stages, ranging from oligodendrocyte precursor cells to fully mature, myelinating oligodendrocytes. Whereas mature oligodendrocyte seem to be less sensitive to noxious stimuli, oligodendrocyte precursor cells and pre-myelinating oligodendrocytes, which are the most abundant oligodendrocyte populations around birth, are particularly vulnerable174. Hypoxia or hyperoxia-induced imbalances of the cellular redox system of developing, immature oligodendrocytes have been suggested to cause cell death84,179,180 while inflammation-triggered perinatal brain injury rather leads to maturational arrest of oligodendrocytes82,121. In spite of different underlying causes, both insults result in disturbed structural white matter development82,86,87,88,89,120,121.

While most of these findings remain at a descriptive level, a causal link between minor myelination deficits and psychiatric symptoms has recently been reported. Using mice heterozygous for the oligodendroglial gene MBP, Poggi et al. demonstrated that very subtle hypomyelination without gross lesions and/or striking differences in basic and cognitive behaviour results in persistent defects of PPI of the startle response, as a surrogate marker of gating defects in SZ patients181. A similar association between myelin deficits and behaviour was demonstrated in CNP−/− mice that show signs of catatonia, a psychomotor syndrome observed across neuropsychiatric diseases182. Interestingly, these mice have increased inflammatory responses predominantly in white matter tracts. Furthermore, signs of catatonia were reversed after microglia depletion. These data demonstrate that altered myelin gene expression and minor structural abnormalities of central nervous system myelin trigger white matter inflammation as underlying cause of catatonic signs182. Even though experimental data are sparse in the field of neonatology, these basic findings strongly support the concept that perinatal injurious stimuli inducing subtle myelination deficits, e.g. by inflammation and/or oxygen disturbances, contribute to long-lasting neuropsychiatric consequences.

Epigenetics

Epigenetic mechanisms, i.e. the enzymatic regulation of transcription activity and gene expression without altering DNA sequence via acetylation, methylation, ubiquitination, phosphorylation and sumoylation on histones, DNA or via microRNA-mediated regulation of translation plays a pivotal role in normal and disturbed brain development and may be associated with neurodevelopmental and/or neuropsychiatric diseases later in life183,184,185. Interestingly, perinatal brain injury either induced by inflammatory triggers or by stress has been linked to sustained epigenetic alterations associated with adverse neurodevelopmental outcome18,150. Furthermore, maternal HFD-associated alteration of cognitive performance and anxiety-related behaviour were associated with increased brain-derived neurotrophic factor and Grin2b methylation and increased expression of DNA methyltransferases186. Whereas the latter reviews and reports rather focus on prenatal insults, several studies have also provided a direct link between early postnatal insults, epigenetic modifications and neuropsychiatric outcome. For instance, early-life stress-induced cognitive impairment and anxiety is attenuated by silencing of miRNA124a in the hippocampus of adult male rats187. Furthermore, maternal separation was associated with persistent dysregulation of histone modifiers in the medial prefrontal cortex, one of the key structures regulating stress responses and mood-related behaviour188. For perinatal brain injury induced by postnatal inflammation, alterations in miRNA expression and histone acetylation have been described, which potentially contribute to disturbed oligodendrocyte maturation121,189,190 even though this hypothesis needs to be proven in cell- and region-specific analyses.

Challenges for preclinical research

Clinical and epidemiological studies are limited to establish a causal relationship between perinatal insults and the development of neuropsychiatric disorders. This may be attributed to impossible randomisation of handling or behaviour. Humans are ecologically and behaviourally heterogeneous limiting the associative power and thus necessitating large studies. Furthermore, symptoms often do not appear for many years after birth. The multiple hit scenario further impedes clear definitions of causal relationships. Therefore, animal models are essential to identify target mechanisms and therapeutics.

Challenges for experimental modelling and data analysis

Technical difficulties emerge from currently used animal models of perinatal brain injury that often induce severe cerebral lesions and associated motor deficits that may confound interpretation of results from behavioural analyses assessing emotional, social and cognitive tasks. Furthermore, owing to improvements in obstetric and neonatal care the number of infants suffering from rather subtle and late-onset cognitive and socio-emotional impairments in later life requires the development of additional more subtle injury models than the currently used severe injury models with high mortality and morbidity. First attempts have been made by combining prenatal immune challenge with postnatal hypoxia only, instead of combination with the more severe model of HI128.

Nevertheless, even with refinement of experimental models, which are indispensable to increase our abilities for the design of therapies, most likely no single animal model can reflect all neuropathological alterations and behavioural deficits of a complex human disorder. For instance, while maternal immune activation has been shown to induce behavioural abnormalities resembling symptoms of ASD and SZ100,101,102, neuropathological alterations do not always parallel observations in clinical studies. As such, while cortical thickness is reduced in the rat model of maternal immune activation, clinical studies demonstrated widespread increased cortical thickness in ASD patients191. Another example is neonatal encephalopathy related to HI that is modelled by induction of unilateral brain damage and thus is unlikely to fully model the global brain injury seen in humans. Nevertheless, work in this model has shown suppression of electroencephalogram power after unilateral HI, as is often seen in infants with HIE192. Thus, despite the unilateral nature of the injury, the HI model replicates at least some important aspects of the encephalopathy observed clinically. With regard to preterm-birth-related complications caused by high oxygen concentrations, the hyperoxia model has been developed. It is a postnatal model where experimental animals do not suffer from lung injury when exposed for 24–48 h to high oxygen concentrations in contrast to preterm infants. It also lacks the clinical fluctuation of oxygen concentrations observed in clinical practice of neonatal care85,193. However, it represents a very standardised and reproducible type of injury that can be combined with additional noxious stimuli, i.e. an inflammatory challenge can be easily added to closer mimic the clinical situation82.

In addition to these difficulties, one has to keep in mind for all models with maternal targeting that this involves whole-litter manipulations where littermates share similar in utero and postnatal environments, which can produce considerable litter effects. Therefore, in experimental designs with maternal targeting, littermates may not be treated as fully independent in statistical analyses that would artificially increase sample size. It is rather suggested to randomly select only one subject from each litter and to allocate animals of one litter to different outcome measures to avoid ethical issues of wasting animals194. An alternative would be the use of appropriate statistical models as previously described195. Potential effects by altered maternal care and/or pain due to injections/interventions can also be confounders in studies with maternal targeting.

Challenges for assessment of neuropsychiatric symptoms in rodents

Symptoms like hallucination, delusions and major thought disorders are difficult to measure in primates and impossible to assess in small animals, such as rodents. Therefore, the aim should not be to fully mimic the entire syndrome but rather focus on distinct behavioural physiological and neuroanatomical phenotypes. Nevertheless, multi-task approaches are needed to model multi-symptomatic diseases and several studies have shown that neuropsychopathological deficits and disorders can be reliably modelled in animals with the help of a battery of neurological and neuropsychological tests screening behavioural abnormalities in rodents196,197. This chapter briefly presents and describes paradigms that are commonly used in rodents to screen for behavioural deficits frequently associated with certain neuropsychiatric disorders or after experimentally induced developmental manipulations198,199. However, rodent behaviours have limitations when compared to the complexity of human behaviour. Thus any tests used for the assessment of behavioural patterns in general are crucial to provide established face validity, construct validity and predictive validity200.

Anxiety-related behaviours

The most commonly used unconditioned tests of anxiety comprise the OF test, the EPM, and the light/dark test (LD). All of them reveal the conflict between avoidance behaviour towards potentially dangerous spaces and the motivation to explore novel, unfamiliar areas201.

Elevated plus maze

One prominent paradigm assessing anxiety-related behaviour is the EPM test. The apparatus consists of a centre platform with four branching arms, two open and two opposing closed arms. By placing the animal on the centre platform always facing an open arm, behaviour is assessed over a certain time. The dependent measures are usually the number of open arm entries, time spent in open/closed arms, the distance covered on the open/closed arms and head dips (the frequency of the animal protruding its head over the ledge of an open arm and down towards the floor). An increase in anxiety-like behaviour is characterised by reduced number of entries into the open arms and the increased time spent on the closed arms of the maze compared to controls202,203,204,205. Likewise, the reduction in head dips is suggested to reflect deteriorated risk assessment-related behaviour206 and thus also supports the assumption that anxiety-like behaviour is apparent.

Rodents’ sickness behaviour is in part reflected by reduced exploration and motor activity207,208,209,210,211,212. Thus observed changes in EPM performance may be simply explainable due to impaired locomotor activity induced by experimental interventions. Such an assumption can, however, be ruled out easily by quantifying the total distance covered on the maze between experimental groups. Second, analysing the number of entries into the closed arms is also considered as good indicator of locomotion rather than anxiety213.

Open field

A simple rectangular acrylic box with black walls and a frosted floor with infrared backlighting can be used as an OF arena. For testing, animals are placed in the centre of the dimly lit OF arena214 and movements are assessed by a video tracking system over 5, 10, 20 or more minutes. The dependent measures are usually general motor activity, that is covered distance and velocity, as well as straighten up-postures for vertical activity. Anxiety-like behaviour in this test is reflected as less time that is spent in the centre of the arena and vice versa more time spent at the borders of the arena. Like in the EPM, quantifying the total distance covered in this test may check simple effects on locomotor activity.

LD test

Another simple but meaningful setting to investigate decreased exploratory behaviour as an index for anxiogenic effects in laboratory rodents is provided by the LD. This plexiglas-made arena consists of two compartments, while two-thirds of the arena is white coloured and illuminated, one-third was black coloured and darkened. The two compartments are separated by a black partition equipped with photocells across the opening, as well as in the overall arena, to measure locomotor activity. Animals are placed in the centre of the bright chamber facing the separating wall. The total numbers of transitions between the two compartments and total locomotor activity, as well as time spent in the bright side are recorded215. In this test which is based on rodents innate aversion to brightly illuminated areas216, less time exploring the lit compartment (risk assessments) and low latency to enter the lit compartment are usually interpreted as increased levels of anxiety217.

While the aforementioned paradigms are frequently used to assess anxiety in rodents, there are drawbacks that need to be taken into consideration. First of all, the ideal animal model of anxiety does not exist, and the tests available (EPM, OF and the LD) are characterised by their originality. Second, using solely one task to measure anxiety-like effects is biased, since rodent behaviours have limitations when compared to the complexity of human behaviour200. Thus, to gain broader understanding about underlying mechanisms and to increase validity of data, multiple behavioural tests should be used to characterise the anxiogenic/anxiolytic impact of experimentally induced interventions218,219, which also improve translation from animals to humans. Moreover, it is proposed that short-term, intra-individual variations in emotionality probably constitute an important factor for investigating anxiety-related behaviour that may differ between tests220.

Deficits in sensorimotor gating, break point and stereotypic behaviour

Sensorimotor gating

The startle response is a fast twitch of eye-lid-closure, facial, neck and skeletal muscles evoked by a sudden and intense tactile, visual or acoustic stimulus221,222. This characteristic response pattern, best described as ASR, is observed in a variety of species such as rodents and humans and is thought to be a protective function against injury from a predator or harm and of the preparation of a flight-or-fight response223. However, when a distinctive non-startling stimulus (prepulse) is presented 30–500 ms prior to the actual startling stimulus, the response magnitude is naturally reduced. This mechanism of inhibited contemporaneous sensory or motor events that would interfere with the ongoing processing of the prepulse reflects a fundamental principle of the neuronal control of behaviour, which in turn is necessary for stimulus recognition and sequential organisation of the appropriate behaviour224. This phenomenon also known as PPI is considered to be a pre-attentive filter mechanism, reflecting the ability of an organism to gate out irrelevant sensory information221,222,225,226,227. PPI can be measured with almost identical procedures in humans (eye blink reflex) and rodents (startle chamber)228,229. Since PPI is disrupted in schizophrenic patients, animal models of sensorimotor gating may offer the possibility to investigate the neural mechanisms related to some characteristical impairment in SZ230. While lesioning the medial prefrontal cortex198 or the ventral thalamus231 in adults are known to induce deficits in sensorimotor gating, neonatal lesions of the entorhinal cortex did not impact PPI measured during adulthood232. However, multiple studies involving different forms of maternal immune activation demonstrated reduced PPI responses105,108,117.

Break point

The so-called “break point” or progressive ratio (PR) test is an operant behaviour task that measures the instrumental effort a rat is willing to invest in order to obtain a reward233,234. In this paradigm, animals are initially trained to press a lever in order to obtain a reward. At the level of stable performance, the number of lever presses required for the reward is progressively increased235, and at a certain point of instrumental effort, animals cease to “work” for the reward and stop responding. The “break point” is considered as operational measure for a shift in motivation, indicating that the value of the reward is lower than the effort the animal is willing to invest236. PR schedules have been shown to provide a valuable method to measure the impact of experimental manipulations that might affect the perceived reinforcement value of gustatory stimuli233. Moreover, it was assumed that a reduced break point and hence a lower performance in the PR- test might serve as an animal model for anhedonia, a core negative symptom in SZ236. When adult rats were tested for motivation in the PR task following neonatal lesions of the entorhinal cortex, a reduced break point in operant responding was observed. These findings indicate that early-life brain injury reduced motivation during operant behaviour reflecting a state of anhedonia as also observed in SZ232.

Stereotypy

Stereotypy is not only a fundamental feature of the behavioural syndrome induced by psychomotor stimulant drugs but also a cardinal feature of a broad range of neuropsychiatric disorders237,238,239,240. Behaviours of patients that typify stereotypy in clinical disorders range from repetitions of single or multiple movements (motor stereotypies) to repetitive, inflexible patterns of attention, emotion, planning and cognition241. Most commonly, video recordings of animals for a certain amount of time are evaluated for specific patterns, such as focussed sniffing, repetitive head and limb movements and oral movements (chewing, licking and biting)242,243.

Cognitive deficits

Morris water maze

Neurodevelopmental disorders due to perinatal insults are often associated with cognitive deficits later in life244. A prominent task, which assesses spatial learning in rodents is the Morris water maze test245. This test is based upon the premise that animals have evolved an optimal strategy to explore their environment and escape from an unpleasant environment (the water tank) with a minimum amount of effort (swimming the shortest distance possible). The time the animals need to discover a hidden platform in the tank after previous exposure to the set-up with only external proximal and distal cues available is considered as an index for spatial memory246,247.

Barnes maze

Similarly, but not as aversive, however, is the Barnes maze task, which also assesses spatio-temporal memory248. Here animals are placed in the centre of a brightly illuminated round board equipped with 20 holes at the border. Animals are required to escape from the very unpleasant environment by seeking an escape box that is attached to one of the 20 holes at the border region. After training, the latency to find the escape box is assessed after a couple of training days reflecting the memory performance249.

Radial maze

Spatial memory abilities may also be assessed in an eight-arm radial maze using a reinforced alternation task. The maze consists of an octagonal central platform 48 cm in diameter connected to eight equally spaced arms projecting radially from the central platform with adjacent arms separated by 45°. At the distal end of each arm, food cups are located (2.5 cm high × 4 cm in diameter) that prevents the rodent from viewing the food reward. The test, usually in an experimental room equipped with several extra-maze visual cues, is initiated by placing the rodent into the centre of the maze with orientation varying from trial to trial towards distinct arms232. Four arms (alternating) of the maze are always baited with food rewards. The animals are placed in the centre of the maze, and arm choices are recorded when an animal collects the reward or reaches the end of an arm. Entering an arm that had never been baited reflects a reference memory error, while a re-entry into an arm that was baited indicates a working memory error250. Since a re-entry into a never-baited arm could be related to either process of working or reference memory, it is scored as a “perseveration error”251.

Novel object recognition

Another method to test cognitive functioning in rodents is provided by the novel object recognition test252,253,254. To verify that performance is less dependent on spatial information, an actual modified version of the task is performed in a Y-maze. Here no external cues are visible for the animal from inside the maze. Initially, three identical objects are placed at the end of the maze arms for familiarisation purposes. On testing days, one familiar object is replaced by a novel, differently shaped object, which is consistent in height material. As most reliable parameter for assessing recognition memory is the time spent in the distal half-arms during the first 2 min of the test. In healthy subjects, this should indeed be longer for the arm with the unfamiliar object255.

Social behaviour

Social interaction

In humans, early prefrontal–cortical damage has been shown to impair cooperative behaviour and social interaction, as well as social cognition256,257. The so-called “unsocial” behaviour is also a characteristic outcome of neurodevelopmental psychiatric disorders, such as SZ and autism. Hence, social withdrawal and social isolation are most commonly seen as negative symptoms in SZ, while the core symptoms of autism comprise specific impairments of reciprocal social relationships, affected cooperative play with peers and atypical social behaviour258,259.

Impairment of social behaviour in rodents can be measured using the social interaction test. This paradigm is based on the fact that active interaction reflects motivation to interact and does not necessarily reflect the behaviour of both animals, while passive interaction depends on the pair. Moreover, the evaluation of active and passive interaction separately and for each animal of the pair allow the investigation of social behaviours between animals of different strains or drug treatments, increasing the repertoire of information originated from this task260,261,262. Briefly, pairs of unfamiliar rats or mice are simultaneously but placed apart in an unfamiliar arena (e.g. OF). Over a certain time period, the time spent in active (sniffing and following) or passive behaviour (animals lie next to each other in distance) are scored263. Based on its predictive and face validities, a decrease in social interaction has been described in several animal models of SZ as a behavioural parameter that mirrors the negative symptoms of this disease199,264,265,266,267 but need to be evaluated following subtle perinatal insults in the future.

Three chamber sociability test

A different method to analyse sociability in rodents is provided by the three chamber sociability test. Here the animal is placed in a roofless apparatus comprising three compartments, a small centre compartment and two equally sized end compartments. During testing, the subject is placed in the middle compartment. One end compartment hosts a strange, unfamiliar rodent in an inverted grid cup, while the other end compartment contains an identical but only an empty grid cup, sometimes with an object268,269. Restraining the unfamiliar rodent in a grid cup limits the mobility of this animal but still allows visual, olfactory and tactile contact to the testing subject. Moreover, possible aggressive responses between test and stranger animal are prevented270. However, rodents normally prefer to spend more time with an unfamiliar counterpart. Thus valid measures for sociability in this test are the amount of time the test subject spends in each compartment and the amount of exploration time (sniffing), as well as the transitions between the compartments. Besides sociability, social novelty may be assessed in a subsequent test setting271. Here a second stranger animal is placed in the former empty grid cup. In this phase, the test subject may choose whether to explore the rodent that was already present during the sociability phase (now a familiar animal) or the newly introduced one. Measures for social novelty during this testing phase are the same ones as described above.

Impulsive behaviour

Although not yet experimentally assessed following insults to the developing brain, high levels of impulsive behaviour are evident in a number of psychiatric disorders, such as obsessive–compulsive disorder, ADHD, SZ, antisocial and addictive behaviour272,273. Owing to the range of behaviours that the term impulsivity encompasses, it is proposed that impulsivity not as a unitary construct but rather as a set of diverse and complex phenomena that may have independent underlying biological mechanisms274,275. Common aspects of impulsivity comprise decreased inhibitory control, intolerance of delay to rewards and quick decision making due to lack of consideration, as well as poor attention ability and hyperactivity276,277. Despite this very broad range of symptomatology, it is possible to devise different paradigms to measure the various forms of impulsive behaviour in both humans and laboratory animals278,279,280. In general, impulsivity includes two major categories: behaviour that results from diminished ability to inhibit actions, often referred to as impulsive action, and behaviour that reflects impulsive decision making, for example intolerance to delay of gratification also known as delay aversion277,281.

Impulsive action (inhibitory control)

Impulse control is often described as active inhibitory mechanism, modulating internally and externally driven pre-potent desires for primary reinforcers, such as food, sex or other highly desirable rewards. This inhibitory control mechanism may provide lower cognitive mechanisms to guide behaviour, while rapid conditioned responses and reflexes are transiently suppressed277,282. The most popular clinical measure of sustained attention, vigilance and response inhibition in humans is the continuous performance task281,283. The preclinical analogue of the continuous performance task designed for rodents is the five-choice serial reaction time task, an operant-based test originally developed to measure visuo-spatial attention. In this 30–40 min task, animals are required to be attentive and withhold responding (nose poke) while monitoring five apertures for brief light stimuli (e.g. ≤1 s) presented randomly therein284,285. Subsequently, at the beginning of a trial and prior to presentation of a light stimulus, there is a 5-s inter-trial interval during which the animals have to withhold a response at all. Any responses made during this time are described as premature responses. Low levels of premature responses require the ability to inhibit actions, whereas high levels are seen as a measure of motoric impulsivity reflected by disturbances in the inhibition of behaviour279,284.

Impulsive decision-making (delay aversion)

Owing to the fact that impulsive patients are not able to take time to carry out appropriate evaluations of incoming information in order to choose behavioural responses on a detailed analysis of a given situation286,287, the inability to tolerate delays of gratification or reward is also an important aspect of impulsive behaviour288,289,290. A well-established paradigm for testing sensitivity to delays of reward in laboratory animals is the delay-based decision-making task, carried out in random operant chambers with two retractable levers and a food pellet dispenser. Here the ability to wait for a food reward is taken as an operational measure of impulsive-like behaviour275,291. More specific, rats are given the choice between a small (one food pellet) and a large reinforcer (five food pellets). The programmed delay associated with the large reinforcer is increased stepwise from 0 to 60 s during the session. Trained rats begin each session by choosing the lever providing the larger reinforcer but then switch the preference to the smaller one as the delay increases. This switch from the large to the small reinforcer indicates an increase in impulsive choice275,292.

Conclusion

Owing to the increasing amount of clinical neurodevelopmental follow-up studies in large cohorts confirming the increased risk for cognitive, neurobehavioural and neuropsychiatric problems following perinatal insults, the focus of translational experimental research has shifted from the sole analysis of the underlying pathophysiology to assessment of long-lasting neurobehavioural deficits reminiscent to typical neurodevelopmental disorders. However, several hurdles remain. While the multiple hit hypothesis is increasingly implemented in experimental research, underlying cellular and molecular mechanisms of each factor and potential synergistic actions remain largely unclear. Furthermore, stratified analyses for sex across a huge variety of perinatal injury models needs to be included in future studies. Major evidence for a direct link between perinatal insults and long-term neurodevelopmental disorders is derived from prenatally or postnatally induced inflammation while the impact of prenatal stress, IUGR and oxygen imbalances on long-lasting neuropsychological symptoms is less well explored. Moreover, most outcome measures focus on distinct and selective behavioural domains, single brain regions and/or single analysis time points in a single animal model. Thus comprehensive systematic analyses including multi-task behavioural tests and the identification of common neurobiological principles underlying distinct pathologies should be a future goal in this research field. This will need multidisciplinary experimental approaches combining knowledge from the field of basic neurobiology, neonatology, paediatrics, psychology and psychiatry to explore the full spectrum of perinatal programming of neurodevelopmental disorders.