Maternal antioxidant treatment prevents behavioural and neural changes in offspring exposed to prenatal social stress

Maternal exposure to social stress during pregnancy is associated with an increased risk of psychiatric disorders in the offspring in later life. However, the mechanism through which the effects of maternal stress are transmitted to the foetus is unclear. Using a rat model, we explored the mechanisms by which maternal social stress is conveyed to the foetus and the potential for targeted treatment to prevent disease in the offspring. Maternal stress increased circulating corticosterone in the mother, but not the foetuses. Maternal stress also induced oxidative stress in the placenta, but not in the foetal brain, and this was prevented by administration of a nanoparticle-bound antioxidant. Moreover, antioxidant treatment prevented prenatal stress-induced anxiety-like behaviour in the adult male offspring, along with several stress-induced neuroanatomical, neurochemical and gene expression changes in the offspring brain. Importantly, many of these neural effects were mimicked in neuronal cultures by application of placental-conditioned medium or foetal plasma from stressed pregnancies. Both placental-conditioned medium and foetal plasma contained differentially abundant extracellular microRNAs following prenatal stress. The present study highlights the crucial role of the placenta, and the molecules it secretes, in foetal brain development and provides evidence of the potential for treatment that can prevent maternal stress-induced foetal programming of neurological disease.


Introduction
Stress and anxiety experienced during pregnancy are associated with an increased risk of psychiatric disorders, such as anxiety and depression, attention deficit-hyperactivity disorder, autism spectrum disorder and schizophrenia [1][2][3] , in the offspring in later life. In a rodent model, maternal social stress in pregnancy gives rise to physiological and behavioural changes in the offspring, including low birthweight, abnormal glucose homeostasis, heightened anxiety behaviour, enhanced stress responses, cognitive deficits and abnormal social behaviours [4][5][6] . Despite much research on the consequences of maternal stress on the offspring, it is still unclear how the effects of psychological stress are transmitted from mother to foetus. Most rodent studies have focussed on a role for maternal glucocorticoids as a mediating factor, with mixed results 7,8,9 . Moreover, in women maternal stress or anxiety during pregnancy is only weakly associated with raised maternal cortisol, if at all 7 .
It is not known if the placenta plays a role in mediating the effects of maternal social stress on foetal neurodevelopment and adult behavioural phenotypes. Moreover, it is not known if in utero intervention through treatment of placental changes can rescue altered neurodevelopment in the offspring and hence prevent abnormal behaviour in later life. Here, we used a rat model to examine the role of the placenta in mediating the effects of maternal social stress on the foetus. After ascertaining a role for maternal social stress in inducing oxidative stress in the placenta, we tested if maternal treatment with a nanoparticle-bound antioxidant is able to prevent both the short-and long-term neurological and behavioural effects of prenatal social stress in the offspring. Given dysfunction of the GABAergic, glutamatergic and dopaminergic systems have been implicated in the pathophysiology of psychiatric disorders for which prenatal stress increases risk [32][33][34][35] , we hypothesised that maternal exposure to social stress during pregnancy may lead to imbalances in the expression of GABA and glutamate receptors, along with neuroanatomical changes, in particular of tyrosine hydroxylase-immunoreactive dopaminergic neurons and parvalbumin-positive GABAergic interneurons.

Maternal social stress increases circulating corticosterone in the mother but not in the foetus
Stress-induced elevations in maternal glucocorticoids have been proposed to mediate the effects of prenatal stress on the offspring; however, there is limited evidence to support this assertion 8 .
Circulating corticosterone concentrations were significantly greater in mothers exposed to repeated social stress during late pregnancy, compared with non-stressed mothers, irrespective of drug treatment (Figure 1a). In contrast, neither maternal social stress nor drug treatment had any significant effect on plasma corticosterone concentrations in the male (Figure 1b) or female foetuses (Figure 1c).
These data suggest that the programming effects of prenatal social stress are unlikely to be mediated directly by changes in foetal corticosterone levels and indicate other factors may be involved.

Maternal social stress increases oxidative stress in the placenta but not in the foetal brain
There is growing evidence that excessive oxidative stress is associated with adverse outcomes of pregnancy, which in turn are believed to contribute to foetal programming of health and disease in later life 8 . Therefore, we examined whether exposure to psychosocial stress in pregnancy induces oxidative stress in the placenta, and in the maternal and foetal tissues. Following maternal social stress, there was a significant increase in reactive oxygen species in the junctional and labyrinth zones of the placenta (Figure 1d,g), maternal liver and cerebral cortex (Figure 1e,g), and foetal liver ( Figure   1f). In contrast, oxidative stress levels in the foetal cerebral cortex and cerebellum were not affected by maternal social stress (Figure 1i,j).
To assess if oxidative stress in the placenta could be prevented or reduced, the pregnant mother was  (Figure 1f). There was no effect of MitoQ-NP administration on the oxidative stress marker in the foetal brain (Figure 1i,j).
The results suggest that prenatal social stress induces placental oxidative stress, which can be prevented by MitoQ-NP administration to the mother.
3. Maternal antioxidant treatment rescues anxiety-like behaviour in male offspring exposed to prenatal social stress Adult male rats born to mothers exposed to social stress showed increased anxiety-like behaviour, There was no significant effect of maternal social stress exposure on depression-like behaviour in either the male or female adult offspring (Figure 2e,h). However, maternal MitoQ-NP administration evidently had an antidepressant action in the forced swim test, independent of prenatal experience ( Figure 2e,h), as reflected by reduced passive coping behaviours, (i.e. decreased floating behaviour in male rats), and increased active coping behaviours (i.e. increased swimming in males and increased climbing in females). Neither prenatal stress nor MitoQ-NP treatment affected sucrose preference in either the male or the female offspring (Supplementary Figure S1).
In addition, female offspring exposed to prenatal stress showed reduced social odour memory (Supplementary Figure S1); they were less able to distinguish between novel and familiar social odours compared to controls, consistent with previous findings 5 . This effect was not significantly altered by maternal MitoQ-NP treatment (Supplementary Figure S1).
In summary, administration of an antioxidant to the mother prior to social stress exposure prevented stress-induced changes in anxiety-like behaviour in adult male offspring and seemingly had antidepressant effects in both sexes, regardless of prenatal experience. 4. Prenatal social stress leads to differential expression of serotonin and noradrenaline-related genes in the foetal frontal cortex To establish the impact of prenatal stress exposure on central gene expression, RNA sequencing was performed. Following prenatal stress, 143 genes were found to be differentially expressed in the foetal frontal cortex most of which were downregulated (12 upregulated, 131 downregulated) (Supplementary Data S1). Enriched pathways corresponded to serotonin and adrenaline/noradrenaline biosynthesis ( Figure 3a). MitoQ-NP treatment reduced the prenatal stress-induced up or downregulation of cortical genes, as demonstrated by a negative correlation between fold change differences in gene expression induced by prenatal stress (compared to non-stressed controls) and fold change differences induced by prenatal stress with MitoQ-NP treatment (compared to prenatally stressed rats that had received saline treatment) (Spearman correlation = -0.798, p < 2.2E-16). Genes related to serotonin and noradrenaline biosynthesis, transport or differentiation were generally found to be downregulated and maternal MitoQ-NP treatment significantly prevented the downregulation of three of these genes in the foetal brain ( Figure 3b). Western blot analysis of the protein products in foetal frontal cortices confirmed the significant downregulation of tyrosine hydroxylase in offspring exposed to prenatal stress compared to control offspring (Figure 3c). In contrast, in the juvenile offspring (P30) tyrosine hydroxylase expression was significantly greater in the frontal cortex, following prenatal stress (Figure 3d). Protein levels of dopa decarboxylase tended to be reduced in the cortex of foetal offspring in response to prenatal stress (Figure 3e), however this effect was not significant (p = 0.095). In the juvenile offspring, prenatal stress exposure did not affect dopa decarboxylase levels in the frontal cortex ( Figure 3f). As observed by RNA sequencing, maternal MitoQ-NP treatment did not significantly affect the levels of tyrosine hydroxylase or dopa decarboxylase in the frontal cortex of the foetuses (Figure 3c,e), or the juveniles (Figure 3d,f).
The results indicate that prenatal stress alters the expression of serotonin and noradrenaline-related genes in the foetal brain; these changes were transient and recovered by P30. Nonetheless, some of these changes could be prevented by administration of MitoQ-NP to the mother. Taken together, the results indicate that prenatal social stress exposure causes long-term neuroanatomical changes in the juvenile offspring brain, and support a role for dysregulated GABAergic and glutamatergic signalling in prenatally stressed offspring. Many of these neuroanatomical and neurochemical changes are prevented by maternal MitoQ-NP administration prior to the stress exposure.
6. Exposing cortical cultures to foetal plasma or placenta-conditioned medium from stressed mothers mimics the neurological effects of prenatal stress Due to the observed lack of increased corticosterone in the foetal circulation in response to maternal social stress, we hypothesised that the placenta, as the interface between mother and foetus, might secrete factors into the foetal circulation that alter foetal brain development. To test this hypothesis, neuronal cell cultures from dissociated cortical tissue were exposed to foetal plasma collected from pregnant rats that were either undisturbed or exposed to social stress, with and without MitoQ-NP pretreatment. Additionally, placental tissue from the same pregnancies was cultured ex vivo and the resulting placenta-conditioned medium was applied to cortical cultures. A significant reduction in dendritic length was observed in cortical cultures following exposure to placenta-conditioned medium or foetal plasma collected from prenatally stressed rats (Figure 4i ). This effect of maternal stress on TH+ process length was not observed in the cortical cultures exposed to foetal plasma or placenta-conditioned medium collected from stressed mothers treated with MitoQ-NP ( Figure 4j). As observed in vivo, the total density of neurons was not altered in cortical cultures exposed to conditioned medium or foetal plasma from any treatment groups (Supplementary Figure S3). Because of the paucity of TH+ and PV+ cells in cortical cultures, no meaningful counts for these cell types could be obtained.
Immunoreactivity for GABA receptor subunits Aα1, Aα2 and B1, as well as glutamate receptor subunit GluN1 was significantly lower in cortical cultures exposed to either conditioned medium or foetal plasma from stressed rats (Figure 5i-l), replicating the loss of immunoreactivity for these receptors seen across multiple brain regions in both the male and female prenatally stressed offspring In summary, placenta-conditioned medium and foetal plasma induced morphological and neurochemical changes in cortical neurons in vitro, similar to those observed in distinct brain regions of juvenile offspring exposed to prenatal stress, suggesting in response to stress the placenta may secrete an as yet unidentified factor which accesses the foetal blood. Prenatal treatment of the mother with MitoQ-NP prevented many of these changes. 7. microRNAs are differentially secreted from the placenta in response to prenatal social stress The evidence suggests the presence of factors in the culture medium conditioned by placental tissue and in the foetal plasma in response to maternal stress, that induce similar changes in dissociated cortical neurons to those changes observed in the brains of prenatally stressed juvenile offspring.
Hence we conducted an initial search for potential mediators of these changes. microRNAs are small post-transcriptional regulators that pay a role in neurodevelopment and in neurodevelopmental disorders, including anxiety and depression 38,39 . We detected differentially abundant microRNAs in response to prenatal stress in both conditioned medium and foetal plasma ( Figure 6a; Supplementary Data S2). Ten microRNAs were similarly up or downregulated in both placenta-conditioned medium and foetal plasma from stressed rats and may therefore be of particular interest for further study. microRNAs significantly altered by prenatal stress consisted of a higher proportion of hypoxamiRs, hypoxia-regulated microRNAs and this enrichment was significant for Predicted targets of those microRNAs that showed significantly different abundance in placentaconditioned medium were specifically enriched among the differentially expressed genes in the foetal cortex detected by RNA-sequencing (Figure 6d). Correlation analysis of miR-455 with the four differentially expressed genes that had been shown to be affected by both prenatal stress and MitoQ-NP treatment (Dmbx1, Gata3, Slc6a2 and Slc6a4), suggested a negative correlation between gene expression in the foetal frontal cortex and levels of mir-455 secreted from the placenta, i.e. transcript levels of the genes were downregulated by prenatal stress compared to non-stressed rats, while miR-455 levels in conditioned medium were upregulated (Figure 6e). This correlation was significant for the transcription factor Dmbx1.
The results suggest that the pattern of microRNAs released from the placenta is altered following maternal social stress. Some of these changes can be prevented by maternal MitoQ-NP treatment and may be correlated with gene expression changes in the foetal brain.

Discussion
Psychosocial stress during pregnancy has been associated with an increased risk of psychiatric and neurodevelopmental disorders in the offspring 1-3 . In the present study, we observed long-term neuroanatomical, neurochemical, behavioural and gene expression changes in the offspring of rats that experienced social stress during pregnancy.
The mechanism by which the effects of maternal social stress are conveyed to the foetus is unclear.
Previous studies have focussed, almost exclusively, on a role for maternal glucocorticoids 7,8 . In the present study however, despite a marked stress-induced increase in corticosterone secretion in the mother, we did not observe any such increase in corticosterone concentrations in the foetal circulation at gestational day 20, indicating that maternal corticosterone is unlikely to directly program the foetal brain. Whether maternal glucocorticoids contribute to foetal programming indirectly by initiating a cascade of downstream events, for example by interacting at the placenta, remains to be tested.
Previous research has suggested that the placenta, as the interface between mother and foetus, plays a key role in neurodevelopment and in foetal programming of psychological disorders that result from adverse events during pregnancy 31,40,41 . In keeping with this concept, the results presented here demonstrate for the first time that maternal exposure to social stress during pregnancy induces oxidative stress in the placenta, highlighting the potential significance of the placenta in mediating the effects of social stress on the offspring. This observation is in line with previous studies showing that placental oxidative stress is increased in response to other maternal insults during pregnancy, such as hypoxia 31 .
Maternal administration of a nanoparticle-bound antioxidant, MitoQ-NP, prior to the chronic social stress exposure was able to reduce placental and maternal oxidative stress, as seen previously in pregnant rats exposed to hypoxia 31 . Moreover, a single MitoQ-NP injection to the mother was able to prevent anxiety-like behaviour in male offspring exposed to prenatal stress and had an overall antidepressant effect in males and females, regardless of maternal stress exposure. This observation further suggests a potential role for oxidative stress in the placenta in the foetal programming of long-term disease and suggests that these long-term effects can be prevented through treatment of the mother during pregnancy. The advantage of using nanoparticle-bound MitoQ lies in the observation that it does not enter the foetal brain, as previously shown 31 , thereby reducing the risk of potential side effects to the developing foetus. MitoQ is a mitochondrial targeted antioxidant that is repeatedly recycled back to its active form by the respiratory chain after fulfilling its antioxidant role 42,43 and can thus produce long-lasting effects. In light of the evident need for preventative treatment, the observation that maternal MitoQ-NP treatment prevents behavioural changes in the offspring in later life warrants further assessment.
A comparison was made of the neurological changes between male and female adult offspring in view of the specific vulnerability of the male offspring to increased anxiety-like behaviour. Particularly marked neurological changes were observed in the basolateral amygdala, hippocampus and the thalamic reticular nucleus and there was a general tendency for male offspring to show more pronounced neurological changes compared to females, consistent with other studies that report sexspecific foetal programming in response to maternal insults during pregnancy [44][45][46][47] . For example, here the finding that GABAA receptor subunit expression was lower in the BLA of prenatally stressed males, but not females, may be important particularly given only the males displayed an anxious phenotype and changes in both receptor expression and anxiety behaviour were prevented by maternal antioxidant treatment. Indeed, decreased GABAergic transmission in the BLA is strongly linked to heightened anxiety behaviour in rodents and reduced GABAA receptor binding has been reported in the brains of patients with generalised anxiety disorder 32,48-50 . Moreover, the observed reduction in the number of PV+ neurons was more widespread in prenatally stressed males than in females, which may be particularly relevant to anxiety behaviour, as impaired inhibitory drive from PV+ interneurons is consistently reported in animal models of psychiatric disorders and an abundance of evidence supports disrupted inhibitory control in several psychiatric disorders 33,34 .
Some prenatal stress-induced neurological changes were only observed in the male offspring. These included a shortening of the dendrites in the basolateral amygdala, hippocampal CA3 region and thalamic reticular nucleus; a decrease in PV+ cells in basolateral amygdala and CA3, and a reduction in TH+ cells in the basolateral amygdala. The basolateral amygdala, hippocampus and thalamus are thought to play key roles regulating anxiety-related behavioural responses 35,51,52 . In particular, a role for PV+ neurons in the basolateral amygdala behaviours has been reported in the regulation of anxiety-related behaviours [53][54][55] . Furthermore, dendrite remodelling in hippocampus and amygdala has been observed previously in response to chronic stress 56,57 . The selective impact on the CA3 region supports the hypothesis that the hippocampus is not a unitary structure but consists of subfields with specific functions. Through region-specific knockouts the CA3 has been shown to play an essential role in anxiety-like behaviours, as opposed to fear responses 58 .
Why these brain regions are particularly susceptible to stress-induced changes in male, but not female rats is unclear. Differences in sex steroids may render certain brain regions more vulnerable to prenatal stress in males versus females; however, differences in gene expression or epigenetic profiles are also likely to play a role 59,60 . In the basolateral amygdala differential expression of CRH receptor subtypes has been observed in male rats exposed to prenatal stress, but not in females 61 . Furthermore, sex-specific differences in the placenta, specifically differences in gene expression 20,22,26,62,63 , epigenetic profiles 24,28 , and metabolism 64 , could render male and female placentae differentially responsive to maternal insults 45,65 .
Application of culture medium conditioned by placentae, or foetal plasma, collected from mothers exposed to social stress to neuronal cultures, replicated many of the neurological effects observed in the adult offspring, such as reductions in dendrite length, TH+ process lengths and levels of GABA and glutamate receptor subunits. These effects were prevented when using conditioned medium or plasma from mothers who had been treated with MitoQ-NP. Due to the low volume of foetal plasma that could be collected from each foetus, samples were pooled across litters, meaning that no comparisons could be made between males and females. However, the above suggests that in response to maternal social stress the placenta secretes molecules into the foetal circulation that can alter the foetal brain. We have previously proposed that the placental barrier may respond to a variety of physiological insults by secreting factors that can damage cells distal to the barrier, both in vitro and in vivo, thereby having the potential to play a role in foetal development 31,40,41,66,67 . The current work further supports the hypothesis that factors released from the placenta may mediate the effects of maternal insults during pregnancy on the developing foetal brain.
Preliminary exploration of potential factors that were secreted from the placenta showed that microRNAs were differentially abundant both in placenta conditioned medium and foetal plasma from pregnant rats exposed to social stress, compared to unstressed rats. Some of these changes were prevented by maternal administration of MitoQ-NP. Furthermore, a potential correlation between microRNA released from the placenta and gene expression changes in the foetal brain was observed.
Extracellular microRNAs can act as long-distance messengers. They can be taken up by cells, including neurons, where they fulfil their regulatory function 68 . Furthermore, microRNAs have the potential to cross the blood-brain barrier and enter the brain, when encapsulated within extracellular vesicles 69 . We hypothesise that following chronic maternal social stress, extracellular microRNAs released into the foetal circulation from the oxidatively stressed placenta may cross the blood-brain barrier, where they can be taken up by neurons and transiently affect the expression of crucial developmental genes. In the foetal cortex of offspring exposed to prenatal stress, we observed a reduction in the expression of genes associated with serotonin and noradrenaline biosynthesis and transport. However in the prenatally stressed juveniles, protein levels in the frontal cortex had returned to levels equivalent to those observed in non-stressed offspring. The serotonin system plays a critical role during neurodevelopment, in particular in the development of the limbic system (including the amygdala, hippocampus and prefrontal cortex), which is important for processing emotional behaviours 70 . Therefore, this transient dysregulation of gene expression in prenatally stressed foetal brains could temporarily affect crucial neurodevelopmental processes at a period of high vulnerability, in line with the concept of the foetal origins of adulthood disease 71,72 , potentially leading to long-term alterations in the offspring brains. Sex-specific differences at any level of this process (e.g. differences in the levels of secreted factors, uptake into the brain, or a specific vulnerability of particular brain regions) may lead to differences in neurological and behavioural phenotypes in adulthood in males and female offspring, although this remains to be tested.

Conclusion
The prenatal period has been shown to be critical for healthy development. However, very little in the way of preventative treatment exists for cases when the prenatal environment is sub-optimal (e.g. during periods of chronic maternal stress). The present study highlights for the first time the importance of oxidative stress in the placentaand factors released from the placenta that likely access the foetal circulationin mediating the effects of maternal psychosocial stress exposure on the offspring's brain and behaviour. While further assessment of maternal antioxidant treatment in pregnancy will be required, especially in light of the observed differences in male and female offspring, the present study highlights the potential for drugs such as MitoQ-NP to be used as preventative treatment with long-term implications for mental health.

Experimental design and animals
All animal procedures were conducted in accordance with the UK Animals (Scientific Procedures) Act Maternal social stress paradigm and drug treatment On GD16, pregnant rats were weighed and 4% (w/w) lidocaine cream was applied topically to the tail prior to administration of either 0.9% sterile saline (vehicle; 0.34 ml/kg) or 125 µM nanoparticle-bound mitoquinone (MitoQ-NP; 0.34 ml/kg) i.v. into the tail vein. Immediately after the injection, each rat was transferred into a clean cage. Approximately 1 h later, half of the vehicle and MitoQ-NP-treated rats were exposed to 10 min social stress, while the remainder served as non-stressed controls and remained undisturbed in their home cage. Social stress was induced using a modified resident-intruder paradigm as previously described 4 . Briefly, pregnant rats ('intruder') were placed in the home cage of an unfamiliar aggressive lactating rat ('resident' between days 1-7 of lactation) for 10 min/day for 5 consecutive days from GD16-20. A different resident rat was used on each day. The number of attacks and latency to the first attack were scored manually. All rats were weighed daily from GD 16-20.
Two sets of in vivo experiments were performed:

Experiment 1: Effects of prenatal stress on the mother, foetus and placenta at GD20
Pregnant rats underwent drug treatment and social stress exposure (as above). Immediately after the final social stress exposure on GD20, pregnant rats and foetuses were killed by conscious decapitation.
Maternal trunk blood was collected into chilled tubes containing 5% (w/v) EDTA. Foetal trunk blood was collected into EDTA-coated microvette tubes (Sarstedt, Germany), pooled by sex and litter. In both cases, plasma was separated by centrifugation at 4°C and then frozen on dry ice and stored at -20°C until analysis of corticosterone and microRNA levels along with application to cortical cultures (see below). Placental explants (n = 3/litter) were collected and cultured as described below to generate conditioned media. Additional placentae, together with maternal and foetal samples of brain and liver were also collected and frozen on dry ice before storage at -80°C for oxidative stress, transcriptome and protein analyses (see below).

Experiment 2: Effects of prenatal stress on offspring brain and behaviour
An additional cohort of pregnant rats underwent the same drug treatment and social stress exposure (as above) but were allowed to give birth (typically on GD22.5-23). Pups remained with their mothers until weaning on postnatal day (P) 23. Litter weights were measured on the day of birth, on P8 and at weaning (P23). Following weaning, offspring were housed in same sex groups by litter (4-6 females, 3-5 males) and were left undisturbed except for handling during routine husbandry. At P30, a cohort of rats (n = 6-7 males and n = 6-7 females per treatment group each from separate litters) were deeply anaesthetised with 3% isoflurane in 1000 ml/min O2, before cardiac perfusion fixation. Following fixation, brains were removed and transferred into 4% (w/v) paraformaldehyde in 15% (w/v) sucrose solution and stored overnight at 4°C, after which time they were transferred to 30% (w/v) sucrose solution at 4°C for ca. 48 h, after which time they were frozen on dry ice and stored at -80°C until processing for immunohistochemistry (see below). Another cohort of male offspring (n = 5/treatment group each from different dams) were killed by CO2 overdose and the prefrontal cortex was collected for protein analysis (see below). The remaining littermates were group housed (by litter and sex) until 9 weeks of age before behavioural experiments commenced (see below). One week after behavioural testing was completed, rats were killed by decapitation and their brains were collected and frozen on dry ice, then stored at -

Cortical cultures
Cortical cultures were prepared from dissociated rat embryonic day 18 cortical tissue and grown on glass coverslips as described previously 40 , in neuronal culture medium (see above). Placental conditioned medium or foetal plasma (n = 3, each replicate originating from a different dam; collected in Experiment 1) was applied to the cortical cultures for 6 days starting from day 12 in vitro.
Exposures were performed at least in triplicate with 3 different sets of cortical cultures. show the thalamic reticular nucleus, hippocampal regions (CA1, CA2 and CA3), the basolateral amygdala and, as control areas, cortical regions (primary somatosensory, primary auditory and retrosplenial granular cortex). For each region, in each brain, a minimum of 5 fields of view were examined for each of 3 sections in both hemispheres using a LASX (Leica) widefield microscope or a SP5II (Leica) confocal microscope at 40x magnification using oil. This resulted in a minimum of 30 fields of view being analysed for each brain region. The analysis of the intensity of the immunostaining of GluN1 and GABA receptor subunits was performed using a similar approach to that previously described 31 . Briefly, nine sections per treatment were imaged and 10-12 images were collected for every brain region using the same microscopy settings of intensity and magnification to allow for a meaningful comparison. Images were converted to greyscale and the total pixel number was determined using a corrected macro for ImageJ. The total pixel number was subsequently subtracted from the background.

In situ hybridisation (ISH)
Adult offspring brains collected after behavioural testing in Experiment 2 were sectioned through the amygdaloid complex at 15 µm using a cryostat and thaw-mounted onto Polysine slides. Sections were then fixed, dehydrated and processed for ISH as previously described 4,61 . 35 S-UTP radio-labelled (Perkin Elmer UK Ltd., Beaconsfield, Buckinghamshire, UK) cRNA sense and antisense probes were synthesised from the linearised pBluescript vector expressing a 518 bp cDNA fragment encoding rat Crh and used to detect Crh mRNA expression, as previously described 4,61 . Following overnight hybridisation and subsequent washes, slides were dipped in autoradiographic emulsion (Illford K5, Knutsford, Cheshire, UK), exposed at 4°C for 4 weeks, before being developed, counterstained with haematoxylin and eosin, and coverslipped. Sections hybridised with sense probes served as negative controls and showed no signal above background. For quantitative analysis, the region of interest (central amygdala) was localised under a light microscope and the number of cells expressing the mRNA of interest were manually counted bilaterally. A positive cell was defined as one with <5x the number of overlying silver grains than background.

RNA sequencing
The transcriptome of the foetal frontal cortex (n = 4/treatment, originating from different litters; collected in Experiment 1) was analysed as previously described 31 . Briefly, RNA was extracted from 50 mg of foetal frontal cortex using the RNeasy Mini kit (Qiagen) and subjected on-column incubation with RNase-free DNase according to manufacturer's protocol (Qiagen). RNA quality and integrity was measured on the 2100 Bioanalyzer (Agilent Technologies). mRNA sequencing was performed by Edinburgh Genomics, UK. Libraries were prepared from total RNA samples using the Illumina TruSeq stranded mRNA Sample Preparation Kit and sequenced by 75 bases paired-end sequencing across two lanes of an Illumina HiSeq 4000. This level of sequencing produced greater than 42 million paired reads per sample. The FASTQ files were generated using the standard Illumina pipeline for bcl2fastq.

Bioinformatic analyses
Tophat 74 was used to align RNA sequencing reads to the rat reference genome Rnor_6.0 (GenBank Assembly ID GCA_000001895.4). Total alignment rates were 83-85% on average, after removing multiple alignments and discordant pairs. Read counts were generated from the resulting BAM files using HTSeq 75 . Differential expression analysis was performed using DESeq. Genes were classed as significant differentially expressed genes for p < 0.05. Pathway analysis was performed using PANTHER 11.0 76,77 .
NanoString nCounter data was analysed using a pipeline described previously 31 . RUVSeq 78 was used to adjust the counts to account for unwanted variation and then edgeR 79 to predict differentially abundant microRNAs from the adjusted counts. miRNAs were classed as significant differentially secreted miRNAs if p < 0.05. Predicted targets of differentially abundant microRNAs were derived from TargetScanHuman v7.0 80 (Total Context Score < -0.2)and subjected to gene ontology analysis using DAVID 6.8 81,82 . Enrichment analyses were performed in R/Bioconductor using the Fisher's exact test and investigating enrichment only. Correlation of abundance changes of significant miRNAs with abundance changes of significant mRNAs was analysed with the miRComb package 83 for R/Bioconductor, using Spearman correlation and Benjamini-Hochberg adjustment for multiple comparisons. Correlation analysis between log2 fold change values of prenatal stress exposure compared to control and prenatal stress exposure + MitoQ compared to controls was performed using Pearson's correlation in R/Bioconductor. Venn diagrams were produced using Venny 2.1 (http://bioinfogp.cnb.csic.es/tools/venny/index.html).

Behavioural tests on adult offspring
Starting between 9-10 weeks of age, offspring (one male and one female per litter; n = 8/group/sex; generated in Experiment 2) were subjected to a battery of behavioural tests. Anxiety tests were carried out first, starting with the light-dark box 84 and followed by the elevated plus maze 85 4 days later. In the subsequent 2 weeks, the rats were tested for the presence of two hallmark traits of depression: anhedonia, using the sucrose preference test, and behavioural despair using the forced swim test 86 . A separate cohort of female offspring (aged 16-17 weeks; littermates of those used in the anxiety/depression tests) was tested in a social olfactory memory paradigm, as it was shown previously that prenatally stressed females, but not males, showed deficits in social odour memory 5 .
In all tasks, the mazes were thoroughly cleaned with 70% ethanol and allowed to air-dry between each trial and males and females were tested on separate days.

Light-dark box
The light-dark box consisted of a transparent box and lidded opaque black box connected by an opening, as described previously 87 . Rats were placed into the dark chamber and allowed to explore freely for 5 min. The trials were recorded with an infrared camera and the time spent in each compartment and the latency to enter the light box was automatically scored using EthoVision XT software v12 (Noldus, Wagenigen, The Netherlands).

Elevated plus maze
The

Sucrose preference test
Rats were singly housed in clean IVCs with two drinking bottles per cage. Fluid consumption was measured by weighing the drinking bottles every 24 h over 3 days. On the first day, both bottles contained filtered tap water. On the second day, one water bottle was substituted for a bottle containing 2% (w/v) sucrose solution. The following day, the positions of the water and sucrose bottles were reversed. Preference for sucrose was expressed as a percentage, calculated by the volume of sucrose solution consumed, divided by the total volume of fluids consumed within the 48 h choice period.

Forced swim test
The forced swim test consisted of 2 swimming bouts spaced 24 h apart. On the first day, rats were forced to swim for 10 min in a 50 cm tall cylinder filled with water (23-25°C) to a depth of 30 cm.
The following day, the rats were placed in the same cylinder, forced to swim for 5 min and their behaviour was digitally recorded. Videos were analysed using a pharmacologically validated timesampling technique where the 5-min bout was divided into 60 x 5 s bins, and the predominant behaviour in each bin was recorded. Behaviours analysed included climbing (with vertical movement of forepaws against the walls of the cylinder), swimming (horizontal movement in the cylinder) and floating (assuming a passive position without struggling, making only small movements necessary to keep their heads above water).

Social olfactory memory
Wooden beads (25 mm in diameter) were impregnated with social odour by being kept for 1 week in a cage housing unrelated and unfamiliar female conspecifics. Rats were allowed to habituate to the test arena, a clean open top cage without bedding, for 2 h. In the exposure session, rats were first introduced to a wooden bead (designated as 'familiar') and were allowed to freely interact with it for 4 min. After a 3-h consolidation period, spent in the test arena cleaned of odours, the rats were exposed to 2 beads simultaneously for 4 min, a fresh familiar bead with the same odour as before and a bead with a novel odour. The time spent interacting with each bead (which included sniffing, biting or holding bead in mouth) was scored manually using Observer XT software (Noldus). The preference score was calculated as time spent in investigating the novel bead divided by the total time spent investigating both beads during the test session.

Statistics
Analysis of behaviour and microscopy images was performed with the experimenter blind to the group allocation. Data are presented as group means ± s.e.m. As indicated in the figure legends, oneway or two-way ANOVA was performed in Prism 6.0 (GraphPad, USA) or SPSS 21.0 (IBM Corp., USA) with post-hoc analysis using Bonferroni correction, Tukey's test or Student Newman Keuls test for multiple comparisons. Two-way ANOVA was used to test for main effects of drug and of stress conditions and for interaction effects.

Data availability
Raw and processed data files generated from NanoString analysis and RNA sequencing have been deposited in NCBI's Gene Expression Omnibus 88      and protein levels (c-f) were analysed in cerebral cortex tissue collected from foetal (a-c,e) and P30 juvenile (d,f) offspring of non-stressed mothers and those exposed to social stress with either vehicle or MitoQ-NP pretreatment. Significant differentially expressed genes, identified by RNA sequencing, from prenatally stressed rats compared to control rats were subjected to pathway analysis (a). Differentially expressed genes involved in serotonin and noradrenaline-related processes are shown with orange arrows indicating differential expression under prenatal stress conditions and green arrows representing differential expression following prenatal stress combined with maternal MitoQ-NP treatment. Values represent log-2 fold changes (b). Relative protein levels of tyrosine hydroxylase (TH; c,d) and dopa decarboxylase (DDC; e,f) were quantified by Western blot. One-way ANOVA showed a significant difference between experimental groups of tyrosine hydroxylase expression in the frontal cortex of foetal (F2,11 = 5.07, p = 0.034) and juvenile offspring (F2,11 = 6.53, p = 0.018). *p < 0.05. Group numbers: n = 4/group for foetal and juvenile tissue.

Figure 4 | Neuroanatomical effects of maternal social stress and antioxidant treatment in
juvenile offspring in vivo and on cortical cultures in vitro. Brain tissue from juvenile rats, born to stressed or non-stressed mothers administered either vehicle or MitoQ-NP during pregnancy, were processed using immunohistochemistry (a-h). Dendrite lengths (a,e), lengths of TH+ cell processes (b,f), TH+ cell numbers (c,g) and PV+ cell numbers (d,h) were assessed in the CA1, CA2, CA3 hippocampal subregions and in the basolateral amygdala (BLA) in male and female offspring. Foetal plasma (FP) and culture medium conditioned by placentae collected on gestational day 20 from dams of the four different treatment groups were applied to embryonic cortical culture (i,j). Dendrite lengths (i) and TH+ process lengths (j) are shown. *p < 0.05, **p < 0.01, ***p < 0.001. Group numbers: n = 6/group for males and females.

Figure 5 | Neurochemical effects of maternal social stress and antioxidant treatment in juvenile
offspring in vivo and on cortical cultures in vitro. Immunoreactivity for GABA and glutamate receptor subunits was assessed in the brains of juvenile rats, born to mothers either exposed to social stress or undisturbed during pregnancy, and pre-treated with either vehicle or MitoQ-NP. GABA Aα1 (a,e), GABA Aα2 (b,f), GABA B1 (c,g) and GluN1 (d,h) subunit expression was quantified in the hippocampal subregions CA1, CA2, CA3 and in the basolateral amygdala (BLA) in both male and female offspring. Embryonic cortical cultures were exposed to foetal plasma or culture medium conditioned by placentae, collected at gestational day 20 from pregnancies exposed to social stress and/or MitoQ-NP treatment (i-l). Expression of GABA Aα1 (i), GABA Aα2 (j), GABA B1 (k) and GluN1 (l) subunits was quantified in cortical cell cultures. *p < 0.05, **p < 0.01, ***p < 0.001. Group numbers are: n = 6/group for males and females. (yellow), collected from pregnant rats exposed to prenatal social stress and treated with vehicle, was assessed for levels of microRNAs. Differentially abundant microRNAs following prenatal stress were compared between foetal plasma and placental conditioned medium (a). Significant differentially abundant microRNAs were enriched for hypoxamiRs (b), both in foetal plasma (top) and placenta conditioned medium (bottom). Gene ontology analysis was performed on the predicted targets of differentially abundant microRNAs (c). Predicted targets of differentially abundant microRNAs were analysed for enrichment among differentially expressed genes in cortex of foetuses from the same pregnancies (d). Correlation between levels of genes Dmbx1, Gata3, Slc6a2 and Slc6a4 in the foetal cortex with levels of miR-455 in placental conditioned medium is shown (e). The red box signifies a significant correlation (p < 0.05). Group numbers: n = 3/group for plasma and conditioned medium.