Gestational stress and perinatal SSRIs differentially impact the maternal and neonatal microbiome-gut-brain axis

Selective serotonin reuptake inhibitors (SSRIs) are the most popular antidepressant medications used to manage perinatal mood disturbances, yet our understanding of how they affect the microbiome-gut-brain axis of the mother and offspring is limited. The purpose of this study was to determine how peripartum SSRI treatment may prevent the effects of gestational stress on plasticity in the maternal hippocampus, plasticity in the neonatal brain and related changes in gut microbiota. To do this Sprague – Dawley female rats were left untreated or subjected to unpredictable stress during pregnancy. Half of the females were supplemented daily with fluoxetine. On postpartum day 2 brains were collected for measurement of plasticity (neurogenesis and microglia content) in the maternal hippocampus and in the neonatal brain. Glucocorticoid receptor density was also investigated in the maternal hippocampus. Microbiota composition was analyzed in fecal samples of dams during and after pregnancy, and colon tissue samples from offspring on postnatal day 2. Main findings show there are significant changes to the maternal microbiome-gut-brain axis that may be fundamental to mediating plasticity in the maternal hippocampus. In addition, there is significant impact of gestational stress on neonatal gut microbiota and brain microglia density, while the effects of SSRIs are limited. This is the first study to explore the impact of gestational stress and SSRIs on the microbiome-gut-brain axis in the mother and neonate. Findings from this study will help inform pathways to intervention strategies including stress reduction techniques and/or microbiota targeted nutritional approaches directed towards improving maternal gut health and outcomes for mother and neonate.


| INTRODUCTION
Selective serotonin reuptake inhibitors (SSRIs) are prescribed to 3%-10% of pregnant women globally [1][2][3] to treat perinatal mood and anxiety disorders. [4][5][6][7][8][9] SSRIs and their metabolites can cross the placenta [10][11][12] and are detected in breast milk 13 raising questions about the effect of using these serotonergic medications on fetal development. [14][15][16][17][18] Furthermore, the impacts of untreated perinatal mood and anxiety disorders are pervasive and can affect the parent and developing child, even with administration of treatments. 7,[19][20][21][22][23] While our understanding of how the maternal brain is affected by perinatal mental illness and SSRI treatment is limited, 7,8,24 it is important that a form of intervention, either pharmacotherapy or psychotherapy, is available for the 10% of mothers and parents struggling with these debilitating conditions. 24 Human studies of the impact of SSRI exposure during developmentally sensitive periods are constrained by "confounding by indication" and distinguishing the impact of SSRIs from the maternal mood disturbances that led to antidepressant treatment remains methodologically challenging. Even with stringent analytic approaches to "confounding by indication", 23,25 many questions remain about the use of SSRIs for both the mother and developing child. For example, unmeasured variables related to the nature of the maternal illness (i.e., stress, anxiety, depression), variations in symptom severity during pregnancy and differing indications for SSRI treatment have contributed to studies showing negative, positive or no altered effect of SSRIs on childhood outcomes. 19,22,[26][27][28][29] Understanding why the variance in effects of SSRI treatment exist on mother and developing neonate is needed. 30 Emerging research points to the impact of the microbiome-gutbrain axis on mental health and child development. [30][31][32][33][34][35] The gut microbiota is a collection of microorganisms within the bowel that are extensively linked to health and disease. 36 They are also capable of producing metabolites and neurotransmitters that allow communication between the gut and the brain, including via the immune system. 36 Colonization of the infant gut largely begins at birth and current research shows that early life factors, such as maternal stressrelated illnesses, impact the maternal brain, microbiome and the developing microbiome-gut-brain axis. [37][38][39][40][41][42][43][44][45] For example, stress during gestation reduces the quantity of a number of gastrointestinal microbial strains in mouse dams 46 and prenatal stress alters fetal intestinal transcriptome and the adult offspring gut and hypothalamic gene expression (in male rat offspring). 47 In addition, early life stress affects the richness and diversity of serotonin-regulating gut microbiota. [33][34][35] This suggests that the gut microbiome of the mother and developing neonate may offer a critical link between early adversity and brain plasticity.
The impact of SSRIs on the microbiome-gut-brain axis of the mother and offspring has only recently begun to be explored 48,49 with research investigating the impact of SSRIs on the maternal microbiome and/or neurodevelopment. 50,51 SSRI treatment with fluoxetine alters the maternal fecal microbial structure and relative abundance of a number of operational taxonomic units (OTUs) in the rat dam and these effects are strongest if the dam has been exposed to early life stress herself. 50 Furthermore, gestational SSRI exposure to nonstressed dams alters the fetal brain transcriptome in mice and these effects are related to maternal gut microbiome influences in response to SSRIs. 51 Perinatal SSRI exposure and perinatal mental illness also impact the neonatal brain. [52][53][54][55] Both prenatal exposure to maternal depression and SSRIs affect neonatal functional brain organization on postnatal day 6 56 and the impact of prenatal SSRI-exposure effects on the neonatal brain are sexually differentiated in some brain regions (i.e., the corpus-callosum microstructure). 52 This research highlights the important early impact that SSRIs and perinatal mood symptoms can have on neurobehavioral outcomes; adding to the growing literature of perinatal SSRI exposure on the developing offspring later in life. 15,16,26,[57][58][59][60] In turn, we are also seeing an impact of SSRIs on the maternal brain. 61,62 Understanding how the maternal gut bacteria may contribute to the effects of SSRIs on mother and offspring will aid in our understanding of the microbiome-gut-brain axis in perinatal mental health.
The present study aimed to understand the impact of perinatal SSRIs and gestational stress on key components of the maternal and neonatal microbiome-gut-brain axis. An animal model of gestational stress and SSRI treatment was used to disentangle the effects of stress-related disorders and SSRIs on mother and offspring. This resulted in four groups of mother-offspring dyads: (1) Ctr_Vehunstressed and treated with vehicle; (2) Ctr_Flx: unstressed and treated with fluoxetine; (3) GS_Veh-gestationally stressed and treated with vehicle; (4) GS_Flx-gestationally stressed and treated with fluoxetine. Both male and female offspring were used to determine if sex had an impact on any measures. We hypothesized that gestational stress exposure would impact the maternal microbiome-gut-brain axis during the peripartum period, that these alterations would be similar to effects seen in the neonate, and that SSRIs would prevent some of the effects of gestational stress.

| Animals
The present study used 54 Sprague-Dawley female rats (average weight 310 g) and 15 male rats for breeding from the Laboratoires Janvier (Le Genest-Saint-Isle, France). Animals were kept under standard laboratory conditions in a 12:12 h light/dark schedule (lights on at 7:00 am) with access to standard rat chow and tap water ad libitum. For breeding, two females and one male were paired. Daily vaginal smears were done to determine if copulation had taken place as evidenced by sperm in the vaginal smears (gestation Day 1; GD1). Once sperm was evident in a vaginal smear the female was removed from the cage with the male and singly housed. Fifteen females did not have sperm in the vagina or did not maintain a pregnancy (lack of weight gain) and were removed from the study. This left 39 pregnant females randomly divided into four groups: (1) Ctr_Veh-unstressed and treated with vehicle (n = 8); (2) Ctr_Flx: unstressed and treated with fluoxetine (n = 7); (3) GS_Veh-gestationally stressed and treated with vehicle (n = 11); (4) GS_Flx-gestationally stressed and treated with fluoxetine (n = 13). Weight and litter characteristics were recorded. For a timeline of the experimental design, see Figure 1. In brief, confirmed pregnant females were randomly assigned to the four treatment groups. Maternal fecal samples were collected prior to treatment on GD2, and then on GD22. Affective behaviors were assessed from GD16-GD9. Two days after giving birth (postpartum Day 2: PD2), offspring-retrieval tests were performed after which maternal and offspring brains, offspring colon tissue and maternal fecal samples were collected.

| Gestational stress
Repeated unpredictable stress during pregnancy was carried out between GD 5 and GD21 according to a previously described procedure. 63,64 The repeated stress consisted of 0-2 stressors per day.
Stressors included restraint under bright light for 1 h; overnight exposure to damp bedding; 12 h of food deprivation; or 5 min of forced swimming at 26 C.

| Fluoxetine administration
Treatment with vehicle (saline) or fluoxetine (10 mg/kg/day-Fagron, Belgium) was administered to the pregnant rats as previously described. [64][65][66] Females were fed 1/9th of a vanilla wafer biscuit (Crousti fondante, Delacre, Belgium) once per day in the morning between 8:00 and 10:00 am from GD5 until sacrifice on postnatal day 2 (PD2). Biscuit portions were injected with vehicle (saline) or fluoxetine dissolved in vehicle. After administration of the biscuit to the home cage, females were monitored to ensure that biscuits were eaten. Biscuits were consistently eaten by all females in less than 5 min.

Gestational Stress + Fluoxetine
F I G U R E 1 Experimental design. After pregnancy was determined, female rats were randomly assigned to four treatment groups: (1) Ctr_Veh-unstressed and treated with vehicle (n = 8); (2) Ctr_Flx-unstressed and treated with fluoxetine (n = 7); (3) GS_Veh-gestationally stressed and treated with vehicle (n = 11); (4) GS_Flx-gestationally stressed and treated with fluoxetine (n = 13). Maternal fecal samples were collected prior to treatment on gestation day 2 (GD2). Affective behaviors were assessed from GD16-GD19 with the sucrose preference test, open field test and elevated plus maze. Feces from the mothers were collected on GD22. On postnatal day 2 (PD2), pup-retrieval tests were performed after which maternal and offspring brains, offspring colon tissue and maternal fecal samples were collected. For further details of the methodology refer to the Materials and Methods section.

| Offspring retrieval test
In order to investigate aspects of maternal behavior a test of offspring retrieval was performed on PD2. In brief, offspring were sep- GS_Flx (n = 13).

| Mother fecal collection
In addition to standard cage changing, the dams were transferred to new cages with clean bedding weekly and the following day feces were collected with sterilized tweezers. Feces were stored in sterile 1.5-mL tubes and frozen at À80 C until analysis. Two feces were taken each week and stored separately. At the time of terminal anesthesia on PD2, feces were collected from the colon to ensure no contamination of fecal matter from offspring.

| Bioinformatic sequence analysis
Bioinformatic sequence analysis was performed as previously described. 69 Briefly, paired-end sequences were assembled using FLASH 70 and analyzed using QIIME version 1.8.0 (Quantitative Insights into Microbial Ecology). 71 Sequences were quality checked and the remaining sequences were clustered into OTUs using USEARCH (version 7-64bit). 72

| Maternal fecal microbiota composition of Bacteroidetes, Firmicutes and Actinobacteria
Based on the 16S RNA data which found significant effects of gestational stress or fluoxetine on Firmicutes and Bacteroidetes ( Figure 3A), we chose to focus on these two phyla with Actinobacteria as a control.
Fecal samples from GD2, GD 22, PD2 were used to investigate any changes in Bacteroidetes, Firmicutes, and Actinobacteria in dams during pregnancy and the early postpartum period as a function of gestational stress or fluoxetine treatment.
Macherey-Nagel bacterial DNA extraction kits were used for DNA extraction from the feces (60-80 mg per sample). The amounts of extracted DNA were quantified by Nanodrop 8000 spectrophotometer (ThermoFisher). Each sample was then diluted to 1 ng/μL. Primers for the bacterial families were used based on a previous study 74 (Table S1). The amounts of phyla Actinobac-

| Maternal brain histology
On PD2, approximately 90 min after the offspring retrieval tests, the dams were deeply anesthetized via an intraperitoneal injection.
Within minutes dams were subjected to intracardiac perfusion with saline and then 4% paraformaldehyde with picric acid. Brains of dams were extracted and post-fixed in 4% paraformaldehyde for 24 h.
Brains were then cryoprotected up to 1 week into a phosphate-saline solution containing 30% sucrose, frozen on dry ice and kept at À80 C. Brain tissue was sliced using a cryostat (Leica) in 40 μm sections in series of 10 and were stored in antifreeze solution and maintained at À15 C.
To investigate effects on immature neurons and microglia cells in the dentate gyrus of the hippocampus the endogenous markers for doublecortin (to label immature neurons) and Iba1 (to label microglia) were used. Glucocorticoid density was also assessed in the CA1, CA3 and dentate gyrus (DG) regions of the hippocampus.  2.9 | Neonatal offspring brain and colon collection on PD2 On PD2, at the time the dams were perfused, a maximum of two males and two females from each litter of each dam were rapidly decapitated, brains and colons were dissected, frozen on dry ice and stored at À80 C until further analysis. In total 106 pups were used.  3.2 | Gestational stress and fluoxetine administration significantly impact the maternal microbiome in late pregnancy

| Alpha and beta diversity
We detected on average 150 bacterial genera of eight different phyla in the fecal material of each experimental group. There were no differences in the alpha-diversity (Chao1 and Shannon indices) of microbiota communities across groups ( Figure S1). Principal component analysis did not show significant separation of gut microbiota across groups as well ( Figure S2).

| Phylum level
On the phylum level, the two dominant phyla-Bacteroidetes and Firmicutes-comprised more than 95% of all bacterial taxa in all experimental groups ( Figure S3). For better visualization of changes occurring within the minor phyla, we constructed a heatmap depicting Log2 change ratios in the Ctr_Flx, GS_Veh and GS_Flx groups relative to the "reference" Ctr_Veh group ( Figure 3A). Gestational stress induced a significant increase in the abundance of the Bacteroidetes phylum (56.6% vs. 47.4% in Ctr_Veh group), administration of fluoxetine during the gestational stress ameliorated the impact of stress, significantly decreasing Bacteroidetes and increasing Firmicutes bacterial species in the GS_Flx group to the control levels ( Figure 3A). Yet, fluoxetine administration alone did not affect the microbiome (Ctr_Flx group, Figure 3A).

| Genus level
We also observed differential changes in various bacterial genera in all experimental groups (depicted on Figure 3B) 3.3 | Maternal gut bacteria are altered from late pregnancy to the early postpartum On GD2, prior to stress or fluoxetine treatment there were no baseline differences in fecal abundance of Bacteroidetes, Firmicutes or Actinobacteria of the dams randomly assigned to each of the four groups ( p > .08). There was a significant decrease in the proportion of Bacteroidetes in maternal feces from late gestation (GD22) to PD2 (F [1,36] = 4.6208, p = .03838; Figure 3C) and a significant increase in the abundance of Firmicutes from late gestation to PD2 (F[1, 36] = 4.5644, p = .03951), regardless of fluoxetine treatment or stress ( Figure 3D).
This led to a significant decrease in the ratio of Bacteroidetes to

| Gestational stress increases offspring retrieval time on PD2
GS_Veh dams took significantly less time retrieving all offspring than 3.6 | Maternal glucocorticoid receptor (GR) density in the hippocampus is decreased with fluoxetine treatment (Ctr-Flx) and is associated with changes in the maternal gut microbiome DG: r = .3430, p = .033. Figure 4B).

| Maternal Iba1 density in the hippocampus is correlated with abundance of bacteria in maternal feces
There were significant negative correlations between Iba1 density in the dorsal hippocampus of the dam and the abundance of Actinobacteria in maternal feces on PD2 (r = À.3724, p = .023; Figure 4C) and Iba1 density in the hippocampus of the dam and the ratio of Bacteroidetes to Firmicutes (r = À.3482, p = .035; Figure 4D). GS_Flx offspring (p = .034) ( Figure 5B). There was a significant positive correlation between Bacteroidetes content in offspring colon and in maternal feces on PD2 (r = .2277, p = .035; Figure 5E).
3.11 | Offspring brain immature neuron content is negatively correlated with offspring gut microbiota on PD2 There was a significant negative correlation between offspring colon Bacteroidetes content and DCX/actin in the forebrain (r = À.2877, p = .014; Figure 6C). The main findings are summarized in Table S2. F I G U R E 4 (A) Mean (±SEM) glucocorticoid receptor (GR) density in the dentate gyrus of rat dams. Ctr_Veh dams had significantly greater GR density than Ctr_Flx dams ( p = .027) and the GR density in the dorsal dentate gyrus was significantly greater than in the ventral dentate gyrus ( p < .018; 3-way ANOVA). (B) There was a significant positive correlation between GR optical density in the ventral CA3 of the dam and the ratio of Bacteroidetes to Firmicutes in maternal feces on PD2 (CA3: r = .38720, p = .012). There were significant negative correlations between Iba1 density in the dorsal hippocampus of the dam and the (C) abundance of Actinobacteria in maternal feces on PD2 (r = À.3724, p = .023) and (D) the ratio of Bacteroidetes to Firmicutes (r = À.3482, p = .035). (n = 7-12/ group) *p < .05.
T A B L E 1 Mean (± SEM) for DCX-ir cell counts and Iba1-ir optical density (OD) in the dentate gyrus of the dam (n = 7-13 per group; 3-way ANOVA).

| DISCUSSION
The maternal and neonatal gut microbiome and brain undergo a number of dynamic changes during the perinatal period as a consequence of stress. 50,79 The main findings from our study show that gestational stress and fluoxetine treatment can differentially impact the maternal microbiome-gut-brain axis. These effects may play a fundamental role in mediating plasticity in the maternal hippocampus and possibly maternal mental health. In addition, we found effects of both gestational stress and fluoxetine exposure on neonatal gut microbiota and brain plasticity.

| Gestational stress effects on the microbiomegut-brain axis of the mother
Our work shows that the maternal gut microbiome is significantly impacted by gestational stress, fluoxetine and reproductive state (pregnant vs. postpartum) and can impact plasticity in the hippocampus of the mother. We found that gestational stress significantly alters the gut microbiome during late pregnancy with increases and decreases in abundance of a number of bacterial genera ( Figure 3A,B).
These findings are in line with previous research showing that gestational stress reduces the bacterial abundance in the feces of rat dams during late pregnancy 46 and shifts gut bacterial abundance during pregnancy in mice. 80 In women, high levels of psychosocial stress also result in a shift in the abundance of fecal microbiota during pregnancy further demonstrating the impact of maternal mental health on gut microbial composition during reproduction. 81 In addition to the effects of gestational stress on the maternal gut microbiome in late pregnancy, we found that the abundance of prominent gut microbiota in the rat dam was altered with the transition to motherhood such that there was a significant decrease in the proportion of Bacteroidetes and an increase in the proportion of Firmicutes in the maternal gut from late pregnancy to the early postpartum period. Given these findings, and the well known relationship between the microbiome and the brain, it is not surprising that the microbiome-gut-brain axis has begun to be a focus for intervention in perinatal mental health research. 31 In the current study we found that maternal gut microbiota abundance was associated with changes in the hippocampus of the dam such that the ratio of Bacteroidetes to Firmicutes was positively associated with glucocorticoid receptor density in the CA3 region of the maternal hippocampus.
This suggests an important role between the maternal gut bacteria

| Gestational stress impacts microglia density in neonatal offspring
When looking at the neonatal brain, we found that gestational stress decreases the density of microglia on postnatal day 2. We know that microglia have receptors that respond to stress hormones in early development 86 4.4 | SSRI effects on the microbiome-gut-brain axis of mother and neonate SSRIs are prescribed to 3.0% of pregnant women globally and are often used as a first line of treatment to manage perinatal mental illness. 1 One of the main questions of the present study was how SSRIs may alter the effects of gestational stress on the microbiomegut-brain axis and how these effects compare to the effects of gestational stress alone on both the mother and neonate. When looking at the maternal gut microbiome, we found that fluoxetine treatment during gestation prevented the stress-induced increase of Bacteroidetes abundance and prevented the stress-induced changes in bacterial taxa (Bifidobacterium, Bacteroides, Erysipelotrichaceae uncultured species) during late gestation suggesting a protective effect of fluoxetine against the effects of stress on certain microbes of the maternal microbiome. These findings are in line with previous research showing that dams exposed to early life stress and treated with fluoxetine during pregnancy have a decrease in the relative abundance of Bacteroides. 50 A recent study in male Sprague-Dawley rats shows that long-term treatment with fluoxetine after weeks of chronic unpredictable stress significantly enhanced Bacteroidetes and reduced Firmicutes relative abundance in the gut. 34 This is opposite of what has been reported in pregnant female rats. These differences may be due to the length of the treatment and stress exposure but also the significant impact that pregnancy can have on the gut microbiome, 50,91,92 as well as the impact of gonadal hormones on gut microbiota composition. 93 In neonatal offspring we found minimal effects of SSRIs on measures of the microbiome-gut-brain axis. However, when looking at brain microglial content we found that fluoxetine exposure, regardless of gestational stress, increased the density of microglia in the neonatal brain. We know that perinatal SSRIs and stress have an enduring impact on offspring brain plasticity later in life with the majority of research focusing on neural and not glial plasticity. [94][95][96] Research has found that SSRI administration to stressed adult male mice increases the spacing and cell body areas of microglia and decreases the arborization of these cells 97 but further research is needed to understand how SSRIs can alter the effects of gestational stress on the developing brain.

| SSRI effects in the absence of stress
It should be noted that treatment with SSRIs such as fluoxetine, in the absence of stress, often results in a different pattern of effects on the brain and gut as found in the present study and previous research. 15 SSRIs and stress affect a number of physiological systems 97,98 and, at times, in opposite directions or at least in significantly different ways (i.e., the 5-HT system 99 ). Therefore, it is important that translational research use appropriate models of stress-related illnesses in order to determine the full impact of SSRI medications on mother and offspring.

| Limitations and future directions
The present study offers a basis for a number of future studies to further investigate the mechanisms and outcomes of the effects reported here. For example, future studies would benefit from understanding the link between the maternal microbiome-gut-brain axis and postpartum maternal affective state as well as maternal offspring-directed care. Understanding in more detail the link between the maternal and neonatal gut would be beneficial, particularly with regards to other forms of stress (i.e., pregestational) as well as other antidepressant medications commonly prescribed during the perinatal period (i.e., sertraline). One key finding in the present study was the impact of gestational stress on the density of microglia in the neonatal brain. Future studies detailing the changes in microglia content specific to different brain areas in the developing brain would aid to our understanding of how early life adversity affects these cells that are essential to brain function.

| CONCLUSIONS
There is a growing body of research investigating how gestational stress and SSRIs can affect offspring outcomes, 26,57,95 more recently pointing to the important impact of perinatal mental health on these outcomes. 19 We now know that untreated depression and stressrelated disorders can have a more significant impact on the mother and offspring than perinatal SSRI medication. 8,19 The present study further demonstrated these findings and is the first study to begin to understand key components of the maternal-neonatal microbiomegut-brain axis, potentially reflecting a nonplacental gestational stress transfer mechanism. Exploring the dyadic relationship of this axis in the mother and neonate in the context of perinatal mental illness and medication exposure will better elucidate the extent to which this biological system can be used as a biomarker and treatment target to improve parent and offspring development.