Perinatal SSRI exposure disrupts G protein-coupled receptor BAI3 in developing dentate gyrus and adult emotional behavior: relevance to psychiatric disorders

Abstract

Selective serotonin reuptake inhibitor (SSRI) antidepressants are widely prescribed to pregnant women suffering with depression, although the long-term impact of these medications on exposed offspring are poorly understood. Perinatal SSRI exposure alters human offspring’s neurodevelopment and increases risk for psychiatric illness in later life. Rodent studies suggest that perinatal SSRI-induced behavioral abnormalities are driven by changes in the serotonin system as well as epigenetic and transcriptomic changes in the developing hippocampus. A major gene altered by perinatal SSRI exposure is the G-protein coupled receptor Brain Angiogenesis Inhibitor 3 (BAI3). Our present study shows that perinatal exposure to the SSRI citalopram increases mRNA expression of Bai3 and related molecules (including its C1ql ligands) in the early postnatal dentate gyrus of male and female offspring. Transient Bai3 mRNA knockdown in perinatal SSRI-exposed dentate gyrus lessened behavioral consequences of perinatal SSRI exposure, leading to increased active stress coping. To determine translational implications of this work, we examined expression of BAI3 and related molecules in hippocampus and prefrontal cortex from patients that suffered with depression or schizophrenia relative to healthy control subjects. We found sex- and region-specific changes in mRNA expression of BAI3 and its ligands C1QL2 and C1QL3 in men and women with a history of psychiatric disorders compared to healthy controls. Together these results suggest that abnormal BAI3 signaling may contribute to molecular mechanisms that drive adverse effects of perinatal SSRI exposure, and show evidence for alterations of BAI3 signaling in the hippocampus of patients that suffer depression and schizophrenia.

Introduction

Millions of pregnant women worldwide suffer from depression and are prescribed selective serotonin reuptake inhibitor (SSRI) antidepressants (Bennett HA et al., 2004;Borri C et al., 2008;Dayan J et al., 2002;Marcus SM, 2009;Yonkers KA et al., 2009), but the long-term impact of SSRIs on children exposed in utero is poorly understood. It is essential to treat depression occurring in pregnancy to protect the well-being of women and their children (Sawyer KM et al., 2018), but also crucial to understand how early life SSRI exposure affects offspring’s neurodevelopment and behavior. Growing evidence shows that perinatal SSRI exposure in humans alters infant limbic brain development (Lugo-Candelas C et al., 2018) and increases risk for depression, internalizing behavior, and abnormal social behavior in later life (Liu X et al., 2017;Malm H BA, Gissler M, Gyllenberg D, Hinkka-Yli-Salomäki S, McKeague IW, Weissman M, Wickramaratne P, Artama M, Gingrich JA, Sourander A, 2016). Identifying neurobiological mechanisms whereby perinatal SSRI exposure disrupts brain development and emotional behavior is necessary to develop therapeutic interventions for affected individuals.

Preclinical studies show that perinatal SSRI exposure in rodents triggers behavioral abnormalities that are relevant to depression (see (Glover ME and Clinton SM, 2016) for review) and disrupts offspring’s serotonin system function (Kinney GG et al., 1997;Maciag D et al., 2006;Weaver KJ et al., 2010). Our laboratory showed that perinatal SSRI exposure in rats causes widespread disruptions in gene expression (Glover ME et al., 2015) and the epigenetic process DNA methylation (Glover ME et al., 2019) in the developing limbic brain, particularly the hippocampus. Several functional classes of genes are modified, including molecules involved in neurodevelopment, neurogenesis, synaptogenesis, and synaptic plasticity (Glover ME,Pugh PC,Jackson NL,Cohen JL,Fant AD,Akil H and Clinton SM, 2015). These findings support the notion that perinatal SSRI exposure disrupts brain development, but molecular mechanisms that directly link perinatal SSRI exposure and later depression-related behaviors remain essentially unknown.

Prior studies of neurobiological consequences of perinatal SSRI exposure largely focused on changes in the adult brain (Bourke CH et al., 2013;Darling RD et al., 2011;Sarkar A et al., 2014;Simpson KL et al., 2011;Zhang J et al., 2011), with few studies apart from ours examining the early neurodevelopmental period when SSRI exposure occurs. This knowledge gap has created a barrier to identifying specific molecules that mediate disruptive effects of perinatal SSRI exposure on brain development and behavior. To begin to address this, we cross-referenced our developing hippocampus transcriptome and methylome datasets to find genes with altered epigenetic regulation and expression following perinatal SSRI exposure (Glover ME,McCoy CR,Shupe EA,Unroe KA,Jackson NL and Clinton SM, 2019). We then refined the gene list, focusing on G-protein coupled receptors (GPCRs) altered by perinatal SSRI exposure. GPCRs represent valuable drug discovery targets since one third of current FDA-approved drugs and 100% of central nervous system-regulating drugs approved since 2016 act on GPCRs (Hauser AS et al., 2017;Hauser AS et al., 2018). We reason that GPCRs disrupted by perinatal SSRI exposure could serve as plausible ‘druggable targets’ to treat perinatal SSRI-related behavioral abnormalities or offer a novel therapeutic for depression more broadly. The adhesion GPCR BAI3 (Brain-specific angiogenesis inhibitor 3; ADGRB3) emerged as a top candidate from our genome-wide molecular assessments of perinatal SSRI-exposed brain. The studies described in this report aimed to examine the role of BAI3 in mediating ill effects of perinatal SSRI exposure on brain development and behavior.

In rodents, BAI3 is highly expressed in the central nervous system where these receptors often localize at the postsynaptic density of glutamatergic synapses (Bolliger MF et al., 2011;Martinelli DC et al., 2016). A major function of BAI3 is to regulate dendritic arborization and activity-dependent synapse formation (Bolliger MF,Martinelli DC and Sudhof TC, 2011;Lanoue V et al., 2013). Levels of Bai3 mRNA are particularly high in the developing hippocampus (Kee HJ et al., 2004;Kim MY et al., 2004). The receptor is activated by four major ligands, C1q-like (C1ql) factors 1–4, that trigger Rho-GTPase signaling through interactions with the intracellular protein Elmo2 (Duman JG et al., 2016). In humans, genome-wide association studies have linked BAI3 with predisposition for schizophrenia and addiction (DeRosse P et al., 2008;Liu QR et al., 2006). However, no studies to-date have investigated BAI3 expression in healthy human brain, nor have prior studies reported changes in brain tissue of psychiatric patients.

Based on previous observations of perinatal SSRI-induced alterations of limbic neurodevelopment (including abnormal expression of synapse-related markers), the present experiments addressed the working hypothesis that early life SSRI exposure perturbs brain development and behavior through altered BAI3 receptor signaling. Work from our lab and others shows that perinatal exposure to SSRI antidepressants elicits similar behavioral anomalies in rodents (Glover ME and Clinton SM, 2016) and humans (Liu X,Agerbo E,Ingstrup KG,Musliner K,Meltzer-Brody S,Bergink V and Munk-Olsen T, 2017), although much of this work has focused on male offspring. Our first experiment examined emotional behavior in male and female offspring that were perinatally exposed to the SSRI citalopram, one of the most commonly prescribed SSRIs in pregnancy (Jimenez-Solem E et al., 2013). Next, based on our prior finding that perinatal SSRI exposure disrupts expression of Bai3 and associated molecules (ligands C1ql2–3 and intracellular interacting partner Elmo2) in the early postnatal dentate gyrus of male offspring (Glover ME,McCoy CR,Shupe EA,Unroe KA,Jackson NL and Clinton SM, 2019), we sought to determine: a) whether these Bai3 expression changes extend to other limbic brain regions; and b) whether effects occur in female as well as male offspring. A third experiment used a small interfering RNA (siRNA) approach to test whether disrupting perinatal SSRI-induced Bai3 expression changes in the early postnatal dentate gyrus would lessen its detrimental impacts on behavior. To begin translating our rodent work to humans, we used an in silico approach to assess expression of BAI3 and related molecules in the developing and adult human brain. Finally, to affirm clinical relevance of our studies to psychiatric disorders, our last experiment examined mRNA expression of BAI3 and related transcripts in hippocampus and dorsolateral prefrontal cortex (dlPFC) of subjects that suffered from Major Depressive Disorder (MDD) or schizophrenia. Overall, these studies aim to better characterize molecular alterations triggered in the brain by early life SSRI exposure, which can potentially point to novel therapeutic targets for associated emotional disturbances or perhaps mood disorders more broadly.

Experimental Procedures

Rodent experimental procedures were approved by the Virginia Tech Institutional Animal Care and Use Committee and performed in accordance with the National Institutes of Health (USA, 2011) guidelines on animal research. Male and female breeder rats were purchased from Charles River Laboratories (Kingston, NY, USA). Animals were housed in a temperature- and humidity-controlled facility (21–23°C, 50%−55% humidity) with 12/12 h light–dark cycle (lights on 07:00). Food and water were available ad libitum.

Perinatal SSRI exposure

Adult female Sprague Dawley rats (n=18) received the SSRI citalopram (10mg/kg/day p.o.) via drinking water for one week prior to conception, throughout pregnancy, and until offspring were weaned on postnatal day (P)21. Control rats for Experiments 1 and 2 received normal drinking water during this timeframe (n=12). One week after citalopram (or control) treatment began; adult male rats were introduced to the females’ cages for mating. Evidence of successful mating was checked daily in the first two h of the light cycle by vaginal lavage and visual evaluation for a sperm plug. Male breeders were removed once mating was confirmed. Throughout the drug treatment period, water intake and animal weight were monitored daily, and drug concentrations were adjusted to maintain a steady dose. We previously used this approach with the SSRI paroxetine and found that this dose and administration method are sufficient to achieve detectable drug levels in pups’ blood and brain tissue (Glover ME,Pugh PC,Jackson NL,Cohen JL,Fant AD,Akil H and Clinton SM, 2015). At birth, litters were culled to six male and six female pups. To control for litter effects, no more than 1–2 pups per sex per litter was used for a particular assay. Pups remained with dams undisturbed except for routine cage changes until weaning on P21. Rats were pair housed throughout behavioral testing with a cage mate that experienced the same control or SSRI exposure condition in early life.

siRNA infusion to transiently suppress Bai3 in developing dentate gyrus

To determine if overexpression of Bai3 in perinatal SSRI-exposed developing dentate gyrus contributes to later emotional behavior changes, we utilized an siRNA approach to transiently reduce Bai3 expression in the developing dentate gyrus of perinatal SSRI-exposed offspring. Adult female rats (n=8) were treated with the SSRI citalopram in drinking water and mated as described above. At birth, litters were culled to six male and six female pups. On P10, using a within-litter design, SSRI-exposed pups were randomly assigned to receive either (1) a stereotaxic injection of an Accell SMARTpool siRNA targeting Bai3 (GE Dharmacon, Lafayette, CO, USA; E-082554–00-0005), or (2) an injection of Accell non-targeting siRNA control (K-005000-G1-02). Injections of control or Bai3 siRNA (0.5μl at 0.1μl/min) bilaterally targeted the dorsal portion of the dentate gyrus (AP: −2.3, ML: +/−2, DV:− 2.5). Offspring were monitored daily to assess potential negative outcomes following surgery; all dams resumed sufficient maternal care of pups following surgeries, and no offspring were rejected by the dam. Dams and pups remained undisturbed other than standard cage changes and drug treatment procedure. A subset of offspring (n=3/condition) were decapitated at P14, and brains were removed, flash frozen, and later sectioned to collect dentate gyrus tissue punches to confirm Bai3 knockdown with qRT-PCR. A second cohort of animals (n=12/sex/condition) matured to adulthood (P60) when they were subjected to the emotional behavior test battery described above.

Behavioral test battery

A subset of perinatal citalopram-exposed (and control) male and female offspring were raised to adulthood (n=15 per sex per condition) and subjected to a behavioral test battery at P60 to examine emotional behaviors. In a separate cohort, perinatal citalopram-exposed male and female offspring that underwent neonatal surgeries were raised to adulthood (n=12 per sex per condition) and subjected to the same battery of tests. The test battery included the Elevated Plus Maze (EPM), a social interaction test, the Forced Swim Test (FST), and sucrose preference test. Females and males were tested on separate days for each test to minimize olfactory cues that could impact behavior. All tests were performed in a dimly lit test room (approximately 30 lux). For each of these tests, animals were transferred to a holding room to be housed overnight prior to behavioral testing on the following morning with the exception of FST, for which animals were transferred to the holding room one h prior to testing. All behavior testing with the exception of sucrose preference occurred between 08:30 to 12:30. Ethovision XT 8.0 video tracking software (Noldus, Wageningen, The Netherlands) was used to video record and analyze behavior for all tests. Testing arenas were disinfected between trials with 70% isopropyl alcohol.

Elevated Plus Maze.

The EPM apparatus was constructed of black Plexiglas, with four elevated arms (70 cm from the floor, 45 cm long and 12 cm wide). The arms are arranged in a cross, with two opposite arms being enclosed by 45-cm high walls. The two other arms are open, having at their intersection a central 12×12-cm square platform giving access to all arms. Rats were placed in the central square facing a closed arm, and time spent with the four paws in every arm was recorded over the course of five min.

Social Interaction.

Social behavior was assessed in a three chambered apparatus (24” × 36” × 12” Plexiglas box divided into three chambers with two partitions). Each divider had an opening that allowed rats to freely move between chambers using a previously published protocol (McCoy CR et al., 2016). On the first test day (Habituation), rats were habituated to the test area by being placed in the middle chamber and allowed to explore freely for five min. On the second test day (Sociability), a wire cup-like container (“interaction cage”; 6-in diameter) was placed in each of the two side chambers. A novel, same-sex conspecific was placed inside one of the interaction cages and an inanimate object was placed in the other interaction cage. The position of the wire cages containing an inanimate object or a novel rat was systematically changed between trials to reduce a side preference. At the start of the day 2 session, the test subject was placed into the middle chamber and allowed to freely explore the apparatus for 10 min. Sociability was calculated as [time spent in social stimulus chamber] ÷ [combined time spent in social stimulus and object chambers]. During the test sessions, time spent in each chamber was evaluated by an observer blinded to experimental conditions.

Forced Swim Test.

Porsolt’s FST was performed as previously described across two sessions using Plexiglas cylindrical containers (45 cm height × 20 cm diameter) (McCoy CR et al., 2018;Nam H et al., 2014). On both days, water depth was approximately 30 centimeters and the temperature was 25°C. Immobility in this test has been classically regarded as an indicator of passive coping and depression-like behavior (Porsolt RD et al., 1977). On test Day 1, rats were singly placed into the water-filled cylinder for 15 min. On Day 2 (24 h later), the rats were returned to the water-filled cylinder for five min. Water was changed after every session. Time spent immobile on Day 2 was scored by a trained observer blinded to experimental conditions.

Sucrose Preference.

The sucrose preference test was administered at the onset of the dark cycle following Day 2 of the FST. All animals were individually housed for the night of testing. Each rat cage was provided with two bottles, with one containing tap water and the other a 0.5% sucrose solution. The weight of each bottle was measured at the start of the test (19:00) and the following morning (09:00) to ascertain preference for sucrose drinking.

Rat tissue collection and quantitative Real-Time PCR

Perinatal SSRI-exposed and control male and female rats were sacrificed by rapid decapitation to harvest brains at two time points: P14 and P60 (n=6/time point/sex/treatment group). Whole brains were flash frozen in isopentane cooled to −35–40°C on dry ice and stored at −80°C until further use. Brains were later cryostat-sectioned at −10 to −12°C with alternating section thickness of 30 μm and 300 μm. The 30 μm sections were stained with cresyl violet to identify target anatomical regions to be dissected in the 300 μm sections. Brain microdissections were collected from the dentate gyrus, dorsal CA hippocampus, amygdala, prefrontal cortex, and nucleus accumbens.

RNA was extracted from each tissue sample for quantitative Real Time PCR (qRT-PCR) using a QuantStudio 6 (Applied Biosystems, USA) with TaqMan detection chemistry as previously described (Cohen JL et al., 2015). The following genes were examined (with the corresponding Taqman assay probe): Bai3 (Rn01511743_m1), C1ql1 (Rn01773925_m1), C1ql2 (Rn01774081_g1), C1ql3 (Rn04244468_m1), Elmo2 (Rn01423444_m1), and the housekeeping genes Abl1 (Rn01436238_g1) and Gapdh (Rn01775763_g1). Relative fold changes between experimental groups were compared for a given gene at a particular time point using the ΔΔCT method.

Bioinformatic analysis of BAI3 and related genes in human brain tissue

Little is known about BAI3 related gene expression in the human brain, whether in a healthy, typically developing brain or in neuropsychiatric conditions. To investigate the expression of BAI3 in the human brain, we performed an in silico analysis of publicly available mRNA expression in postmortem human brain tissue utilizing the R shiny application Kaleidoscope (https://kalganem.shinyapps.io/BrainDatabases/), a database compiled of expression data and visualization tools (Alganem K et al., 2020). Our analysis focused on the BAI family of receptors (BAI1, BAI2, and BAI3), their ligands C1QL1–4, and two intracellular protein interactors, ELMO1 and ELMO2. We then visualized expression patterns of these molecules across developmental phases using data from BrainCloud (https://www.libd.org/brain-cloud/), a freely available resource described by Colantuoni et al. (Colantuoni C et al., 2011). These data were derived from postmortem human prefrontal cortex collected at different time points across the lifespan, from fetal development through middle and advanced adulthood. A total of 269 subjects without known neuropathological or neuropsychiatric disorders were included in this dataset.

To determine specific cell types that express BAI3 and related molecules in human brain tissue, we used the BrainRNA-seq database that was created in a previous study by Zhang et al. (Zhang Y et al., 2016). As described in that publication, human fetal, juvenile, and adult postmortem cortical tissue was processed using cell-type specific antibodies to separate neurons (anti-Thy1 CD90), astrocytes (anti-hepaCAM), oligodendrocytes (anti-GALC hybridoma supernatant, microglia/macrophages (anti-CD45), and endothelial cells (BSL-1). Transcriptomes of these separated cell populations were then determined using RNA-seq.

Similarly, we utilized BrainAtlas (http://portal.brain-map.org/) to survey mRNA expression in the human medial temporal gyrus (Hodge RD et al., 2019). Data here were derived from intact nuclei from frozen human postmortem brain specimens. In total, 15,928 nuclei from 8 human brain specimens ranging from 24 to 66 years of age were previously analyzed. As described in that publication, nuclei were isolated, sorted into neuronal (NeuN+) and non-neuronal (NeuN-) populations using FACS, and cell-types were identified following single-nucleus RNA-seq. This initial analysis revealed 75 transcriptionally distinct cell types that were subdivided into 45 inhibitory neurons, 24 excitatory neurons, and 6 non-neuronal cell types including astrocytes, microglia, oligodendrocytes, oligodendrocyte precursor cells (OPCs), and endothelial cells.

Expression of BAI3 and related molecules in hippocampus and dorsolateral prefrontal cortex of patients with Major Depression or Schizophrenia.

To investigate the relevance of BAI3 in Major Depressive Disorder (MDD) or schizophrenia (SCZ), we measured expression of BAI3, C1QL2, and C1QL3 in human hippocampus and dlPFC of male and female patients with MDD and SCZ, as well as healthy control subjects with no history of psychiatric disorder. RNA samples for this experiment were obtained from the National Institute of Mental Health (NIMH) Human Brain Collection Core (HBCC). Subject demographics are included in Table 3, and other characteristics of the samples can be found at dbGaP Study Accession phs000979.v3.p2. Taqman assay probes used for qRT-PCR included BAI3 (Hs00938873_m1), C1QL2 (Hs00704398_s1), C1QL3 (Hs00864244_m1), and housekeeping genes ACTB (Hs99999903_m1), B2M (Hs99999907_m1), and GUSB (Hs99999908_m1).

Table 3:

Subject demographics for human postmortem brain tissue study

Hippocampus							Dorsolateral Prefrontal Cortex
	Male Control	Male MDD	Male SCZ	Female Control	Female MDD	Female SCZ		Male Control	Male MDD	Male SCZ	Female Control	Female MDD	Female SCZ
N	115	36	67	59	33	37	N	151	78	111	64	57	65
pH	6.58 ± 0.27	6.28 ± 0.22	6.36 ± 0.27	6.49 ± 0.32	6.07 ± 0.11	6.32 ± 0.27	pH	6.56 ± 0.26	6.42 ± 0.25	6.42 ± 0.25	6.55 ± 0.29	6.25 ± 0.28	6.39 ± 0.27
PMI (hours)	30.68 ± 14.57	34.18 ± 28.18	33.37 ± 14.85	33.03 ± 14.65	34.71 ± 16.53	43.84 ± 24.61	PMI (hours)	30.66 ± 15.37	39.66 ± 29.07	38.47 ± 25.39	29.92 ± 12.34	35.55 ± 19.51	38.92 ± 21.96
RIN	8.07 ± 0.81	7.89 ± 1.17	7.47 ± 1.14	7.76 ± 0.89	7.62 ± 0.94	7.30 ± 1.09	RIN	8.32 ± 0.61	8.10 ± 0.84	7.92 ± 0.92	8.14 ± 0.87	7.94 ± 0.96	7.72 ± 1.03
Age	41.17 ± 15.67	44.71 ± 14.77	50.09 ± 15.71	45.25 ± 15.67	50.35 ± 15.14	56.65 ± 16.36	Age	42.90 ± 15.58	44.09 ± 13.38	47.01 ± 14.56	45.07 ± 16.49	47.85 ± 13.81	55.20 ± 14.55
Suicide	115NS	25S/11NS	14S/53NS	59NS	21S/12NS	4S/33NS	Suicide	151NS	49S/29NS	28S/83NS	64NS	37S/20NS	7S/58NS

Group demographics of human postmortem samples. Data presented as mean ± standard deviation. Abbreviations: postmortem interval (PMI); RNA integrity number (RIN); major depressive disorder (MDD); schizophrenia (SCZ); suicide (S); non-suicide (NS)

Statistical Analyses

For all behavioral measures data are presented following Z normalization, which is a dimensionless tool that compares experimental group means across behavioral measures both within and between behavioral tests (Glover ME et al., 2021;Guilloux JP et al., 2011). This type of analysis offers an estimate of an animal’s gestalt emotionality since it takes into account behavioral responses across a range of tests of emotional behavior. The Z value represents how many standard deviations (σ) a value (X) is above or below the group average (μ) and is represented by the equation: Z = X − μ/σ where μ and σ are the mean and standard deviation, respectively, of the control group. For each behavioral test, all groups were normalized to the control male (Fig. 1) or scrambled siRNA male (Fig. 2) cohort.

Figure 1. Behavioral effects of perinatal citalopram exposure in rats.

Experimental timeline illustrating treatment of adult female rats with the selective serotonin reuptake inhibitor (SSRI) citalopram (p.o. 10 mg/kg/day in drinking water) throughout gestation until offspring weaning at postnatal day (P) 21 (A). Z-score normalized behavioral data of adult male and female offspring in the Forced Swim Test (FST; B), sucrose preference test (C), Elevated Plus Maze (EPM; D), and social interaction with a novel, same-sex rat (E). * p<0.05; ** p<0.01; **** p<0.0001.

Figure 2. Behavioral effects of siRNA-mediated Bai3 knockdown in perinatal citalopram-exposed rats.

Experimental timeline illustrating timing of exposing pregnant/postpartum females to the selective serotonin reuptake inhibitor (SSRI) citalopram throughout gestation until offspring are weaned at postnatal day (P) 21. At P10, male and female offspring were stereotaxically injected with a short interfering RNA (siRNA) to knockdown Bai3 siRNA in the dorsal dentate gyrus (or with a scrambled siRNA control; A). Z-score normalized behavioral data of adult offspring in the Forced Swim Test (FST; B), sucrose preference test (C), Elevated Plus Maze (EPM; D), and social interaction (E). * p<0.05; *** p<0.001; **** p<0.0001.

Data from rodent studies were analyzed using GraphPad Prism Software (Version 9.1.0, GraphPad, La Jolla California USA). In each case, data were verified to be normally distributed before further analysis. Two-way ANOVA was used to examine the effect of perinatal SSRI exposure and sex on adult offspring emotional behavior (Experiment 1) and mRNA expression of Bai3 and related molecules in multiple limbic brain regions (Experiment 2). In Experiment 3, two-way ANOVA examined the effect of early life siRNA-mediated Bai3 knockdown and sex on adult behavior. Behavioral and molecular measurements were excluded from analysis if they were identified as outliers in the ROUT method of outlier identification in the GraphPad Prism software (Version 9.1.0, GraphPad, La Jolla California USA). Significance was set at p<0.05, and results are presented as mean ± SEM.

The human postmortem qRT-PCR expression data were normalized and a general linear model using multiple stepwise regression (both directions) was fitted for the relation between normalized gene expression and the psychiatric diagnosis. Covariates of suicidality, age of death, pH, and postmortem interval (PMI) were also considered (Durrenberger PF et al., 2010;Stan AD et al., 2006). Samples from fetal tissue or from children under the age of 17 were excluded from the analysis. Outliers (± three standard deviations from the mean) within each diagnosis group per gene of interest were removed. Each gene of interest was evaluated for data normality and homogeneity of variance. Data normality was checked via a QQ Plot while homogeneity of variance across diagnosis groups was checked via Levene’s test using the leveneTest() in R (Version 1.2.5019, R. RStudio, Inc., Boston, MA). When necessary, data were transformed using either the log10 or square method to reduce skewing. In these cases, transformation increased the Levene’s test p-value to p>0.05, with the exception of BAI3 expression in the male dlPFC (p=0.01715). To remove unnecessary, forced relationships within covariates and between the covariates and diagnosis groups, stepAIC() function in R was used to perform step-wise linear regression to generate a more simplified and accurate model of important interactions between these variables. Data were analyzed by analysis of covariance (ANCOVA) using the aov() function, and if there was a main effect of diagnosis on mRNA expression, Tukey’s post hoc test was performed. Data were plotted using GraphPad Prism and the median for each group is indicated by a horizontal line.

Results

Experiment 1: Perinatal exposure to the SSRI citalopram increases passive coping and anhedonia, but does not impact anxiety-like behavior and sociability in male and female rodent offspring

Past work by our group and others demonstrated myriad emotional behavior perturbations in offspring exposed perinatally to different types of SSRI antidepressants (Glover ME and Clinton SM, 2016). The majority of this past work focused on male offspring, so this first experiment examined a range of emotional behaviors in adult male and female offspring that were exposed perinatally to the SSRI citalopram (experimental timeline shown in Fig. 1A). Perinatal exposure to citalopram led to increased passive stress coping (immobility) in the FST in both male and female offspring compared to controls (main effect of SSRI exposure (F(1,53) = 9.48; p=0.0033); no effect of sex; no sex × drug interaction; Fig. 1B). Perinatal citalopram exposure also diminished sucrose preference in both male and female perinatal SSRI-exposed offspring versus controls (main effect of SSRI exposure (F(1,51) = 34.0; p<0.0001). There was no effect of sex, but the sex × drug interaction was significant (F(1,51) = 6.43; p=0.0143; Fig. 1C). Post hoc analysis revealed reduced sucrose preference in male (p=0.0196) and female (p<0.0001) offspring. Perinatal citalopram exposure did not affect anxiety-like behavior in the EPM, although there was a significant effect of sex (F(1,54) = 21.28; p<0.0001), with females spending more time in the open arms compared to males (Fig. 1D). Finally, perinatal citalopram exposure did not impact social behavior in male or female offspring. As in EPM, there was a significant effect of sex (F(1,55) = 19.94; p<0.0001), with males spending more time interacting with a novel conspecific compared to females (Fig. 1E).

Experiment 2: Perinatal SSRI exposure alters expression of Bai3 and related molecules in the developing rodent dentate gyrus, nucleus accumbens, and prefrontal cortex with limited impact in adulthood

We next sought to determine whether perinatal exposure to citalopram changes expression of Bai3 and related molecules in the dentate gyrus, dorsal CA hippocampus, amygdala, prefrontal cortex, and nucleus accumbens of male and female offspring. We focused on two time points: P14 (Table 1), when pronounced SSRI-related molecular changes occur, and P60 (Table 2), when behavioral abnormalities are present. Tables 1 and 2 show statistical output for each time point.

Table 1:

Perinatal SSRI exposure impacts expression of Bai3 and related molecules in select brain regions at postnatal day 14.

Brain region	Gene	Mean ± SEM				Effect of perinatal SSRI exposure	ANOVA output
		Female		Male
		Control	PNSSRI	Control	PNSSRI		Exposure	Sex	Interaction	Post Hoc
Dentate Gyrus	Bai3	0.907 ± 0.101	1.068 ± 0.044	1.065 ± 0.208	1.543 ± 0.084		F(1,22) = 9.516; p=0.0054	F(1,22) = 9.300; p=0.0059	F(1,22) = 2.330; p=0.1412
	C1ql1	-	-	-	-	-	-	-	-
	C1ql2	1.092 ± 0.086	1.184 ± 0.086	1.012 ± 0.092	1.484 ± 0.131		F(1,21) = 5.372; p=0.0306	F(1,21) = 0.8142; p=0.3771	F(1,21) = 2.450; p=0.1325
	C1ql3	0.997 ± 0.094	1.612 ± 0.372	1.053 ± 0.180	1.284 ± 0.112	-	F(1,22) = 3.185; p=0.0881	F(1,22) = 0.3307; p=0.5711	F(1,22) = 0.6570; p=0.4263
	Elmo2	0.868 ± 0.080	1.115 ± 0.182	1.012 ± 0.085	1.349 ± 0.111		F(1,22) = 4.349; p=0.0488	F(1,22) = 1.821; p=0.1909	F(1,22) = 0.1043; p=0.7498
Dorsal Cornu Ammonis (CA) hippocampus	Bai3	0.803 ± 0.139	0.790 ± 0.161	1.005 ± 0.049	0.973 ± 0.125	-	F(1, 16) = 0.4395; p=0.5168	F(1, 16) = 1.213; p=0.2870	F(1, 16) = 0.3127; p=0.5838
	C1ql1	0.82 2 ± 0.089	0.77 0 ± 0.087	1.01 0 ± 0.071	0.92 0 ± 0.146	-	F(1, 16) = 0.4755; p=0.5003	F(1, 16) = 2.739; p=0.1174	F(1, 16) = 0.03365; p=0.8568
	C1ql2	0.96 1 ± 0.163	1.40 7 ± 0.384	1.10 5 ± 0.210	0.93 8 ± 0.196	-	F(1, 16) = 0.3038; p=0.5891	F(1, 16) = 0.4113; p=0.5304	F(1, 16) = 1.470; p=0.2429
	C1ql3	0.795 ± 0.063	0.947 ± 0.149	1.002 ± 0.034	0.705 ± 0.098	-	F(1, 14) = 0.5056; p=0.5056	F(1, 14) = 0.02721; p=0.8713	F(1, 14) = 4.498; p=0.0523
	Elmo2	0.743 ± 0.023	0.741 ± 0.039	1.000 ± 0.017	0.973 ± 0.125	-	F(1, 14) = 0.03856; p=0.8471	F(1, 14) = 10.59; p=0.0058	F(1, 14) = 0.02714; p=0.8715
Amygdala	Bai3	0.893 ± 0.053	0.686 ± 0.174	1.038 ± 0.142	0.767 ± 0.100	-	F(1, 15) = 3.232; p=0.0923	F(1, 15) = 0.7288; p=0.4067	F(1, 15) = 0.05837; p=0.8124
	C1ql1	0.993 ± 0.179	0.817 ± 0.118	1.033 ± 0.132	0.758 ± 0.082	-	F(1, 16) = 2.890; p=0.1085	F(1, 16) = 0.005135; p=0.9438	F(1, 16) = 0.1400; p=0.7132
	C1ql2	1.711 ± 0.259	1.237 ± 0.235	1.029 ± 0.124	0.752 ± 0.108	-	F(1, 16) = 3.776; p=0.0698	F(1, 16) = 9.110; p=0.0.0082	F(1, 16) = 0.2612; p=0.6163
	C1ql3	0.712 ± 0.123	0.529 ± 0.137	1.016 ± 0.090	0.610 ± 0.214	-	F(1, 16) = 3.942; p=0.0645	F(1, 16) = 1.695; p=0.2114	F(1, 16) = 0.5628; p=0.4640
	Elmo2	0.980 ± 0.123	0.677 ± 0.043	1.015 ± 0.086	0.848 ± 0.077	↓	F(1, 15) = 6.654; p=0.0209	F(1, 15) = 1.280; p=0.2757	F(1, 15) = 0.5522; p=0.4689
Nucleus Accumbens	Bai3	0.710 ± 0.080	0.479 ± 0.046	1.046 ± 0.140	0.596 ± 0.080	↓	F(1, 16) = 13.47; p=0.0021	F(1, 16) = 5.965; p=0.0266	F(1, 16) = 1.391; p=0.2555
	C1ql1	0.803 ± 0.077	0.673 ± 0.048	1.033 ± 0.133	0.748 ± 0.078	↓	F(1, 16) = 5.379; p=0.0339	F(1, 16) = 2.888; p=0.1086	F(1, 16) = 0.7526; p=0.3985
	C1ql2	0.745 ± 0.050	0.333 ± 0.045	1.053 ± 0.161	0.409 ± 0.034	↓	F(1, 16) = 35.41; p<0.0001	F(1, 16) = 4.685; p=0.0459	F(1, 16) = 1.713; p=0.2090
	C1ql3	0.508 ± 0.120	1.346 ± 0.383	1.069 ± 0.212	1.602 ± 0.330	↓	F(1, 14) = 5.028; p=0.0417	F(1, 14) = 1.787; p=0.2026	F(1, 14) = 0.2476; p=0.6265
	Elmo2	0.855 ± 0.062	0.763 ± 0.079	1.044 ± 0.189	0.955 ± 0.073	-	F(1, 15) = 0.7838; p=0.3900	F(1, 15) = 3.485; p=0.0816	F(1, 15) = 0.0002007; p=0.9889
Prefrontal cortex	Bai3	0.908 ± 0.246	2.575 ± 0.453	1.056 ± 0.188	0.970 ± 0.061		F(1, 16) = 8.207; p=0.0112	F(1, 16) = 6.978; p=0.0178	F(1, 16) = 1.552; p=0.230
	C1ql1	1.113 ± 0.056	1.101 ± 0.111	1.013 ± 0.083	0.884 ± 0.084	-	F(1, 15) = 0.6235; p=0.4420	F(1, 15) = 3.151; p=0.0962	F(1, 15) = 0.4351; p=0.5195
	C1ql2	0.435 ± 0.078	0.268 ± 0.060	1.181 ± 0.324	1.291 ± 0.280	-	F(1, 15) = 0.01499; p=0.9042	F(1, 15) = 14.52; p=0.0017	F(1, 15) = 0.3567; p=0.5592
	C1ql3	0.918 ± 0.222	1.866 ± 0.282	1.038 ± 0.151	0.994 ± 0.069		F(1, 16) = 5.237; p=0.0361	F(1, 16) = 3.628; p=0.0749	F(1, 16) = 6.300; p=0.0232	Female Control vs female SSRI p=0.0175 Fe male SSRI vs male control p=0.0409 Female SSRI vs male SSRI p=0.0301
	Elmo2	0.772 ± 0.155	0.927 ± 0.150	1.002 ± 0.032	0.850 ± 0.114	-	F(1, 16) = 0.0001174; p=0.9915	F(1, 16) = 0.3888; p=0.5417	F(1, 16) = 1.552; p=0.2308

We assessed mRNA expression of the Bai3 receptor, its ligands (C1ql1–3), and intracellular interacting protein Elmo2 in several limbic brain regions at postnatal day 14 of male and female offspring perintally exposed to the SSRI citalopram (or control condition). Transcript levels (fold change relative to control males) are presented as group mean ± SEM. Statistical output from two-way ANOVA indicates effects of perinatal SSRI exposure, sex, and exposure by sex interaction. Significant effects (p<0.05) are shaded in gray.

Table 2:

Perinatal SSRI exposure impacts expression of Bai3 and related molecules in select brain regions in adulthood

Brain region	Gene	Mean ± SEM				Effect of perinatal SSRI exposure	ANOVA output
		Female		Male
		Control	PNSSRI	Control	PNSSRI		Exposure	Sex	Interaction	Post Hoc
Dentate Gyrus	Bai3	0.827 ± 0.150	0.742 ± 0.124	1.031 ± 0.127	0.783 ± 0.194	-	F(1, 16) = 1.211; p=0.2873	F(1, 16) = 0.6569; p=0.4295	F(1, 16) = 0.2886; p=0.5985
	C1ql1	0.994 ± 0.214	0.627 ± 0.058	1.022 ± 0.109	0.883 ± 0.128	-	F(1, 16) = 3.285; p=0.0887	F(1, 16) = 1.039; p=0.3233	F(1, 16) = 0.6702; p=0.4250
	C1ql2	1.050 ± 0.212	0.782 ± 0.089	1.095 ± 0.247	0.926 ± 0.225	-	F(1, 16) = 1.160; p=0.2975	F(1, 16) = 0.2146; p=0.6494	F(1, 16) = 0.06029; p=0.8092
	C1ql3	0.826 ± 0.113	0.821 ± 0.118	1.023 ± 0.115	0.884 ± 0.217	-	F(1, 15) = 0.2267; p=0.6408	F(1, 15) = 0.7282; p=0.4069	F(1, 15) = 0.1932; p=0.6665
	Elmo2	0.868 ± 0.118	0.654 ± 0.059	1.014 ± 0.080	0.828 ± 0.162	-	F(1, 16) = 3.181; p=0.0935	F(1, 16) = 2.038; p=0.1727	F(1, 16) = 0.01648; p=0.8994
Dorsal Cornu Ammonis (CA) hippocampus	Bai3	1.013 ± 0.084	0.963 ± 0.059	1.034 ± 0.060	0.829 ± 0.049	-	F(1, 15) = 3.808; p=0.0699	F(1, 15) = 0.7463; p=0.4012	F(1, 15) = 1.405; p=0.2543
	C1ql1	0.886 ± 0.910	0.820 ± 0.087	1.015 ± 0.088	0.955 ± 0.068	-	F(1, 15) = 0.5670; p=0.4631	F(1, 15) = 2.466; p=0.1372	F(1, 15) = 0.001521; p=0.9694
	C1ql2	2.747 ± 0.910	1.140 ± 0.325	1.171 ± 0.330	0.908 ± 0.157	-	F(1, 14) = 3.929; p=0.0674	F(1, 14) = 3.675; p=0.0759	F(1, 14) = 2.027; p=0.1764
	C1ql3	2.49 2 ± 0.059	0.97 5 ± 0.168	1.31 1 ± 0.395	1.54 9 ± 0.277	↓females only	F(1, 14) = 4.686; p=0.0482	F(1, 14) = 1.058; p=0.3212	F(1, 14) = 8.823; p=0.0101	Female Control vs SSRI p=0.0200
	Elmo2	1.02 9 ± 0.116	1.06 4 ± 0.113	0.98 5 ± 0.107	0.87 3 ± 0.107	-	F(1, 15) = 0.1194; p=0.7345	F(1, 15) = 1.087; p=0.3137	F(1, 15) = 0.4283; p=0.5228
Amygdala	Bai3	1.053 ± 0.052	1.049 ± 0.112	1.014 ± 0.081	1.086 ± 0.032	-	F(1, 16) = 0.2034; p=0.6580	F(1, 16) = 0.0002348; p=0.9880	F(1, 16) = 0.2555; p=0.6201
	C1ql1	0.885 ± 0.057	1.152 ± 0.192	1.003 ± 0.046	0.905 ± 0.024	-	F(1, 15) = 0.6007; p=0.4504	F(1, 15) = 0.3481; p=0.5640	F(1, 15) = 2.810; p=0.1144
	C1ql2	1.041 ± 0.060	1.110 ± 0.146	1.014 ± 0.089	0.797 ± 0.072	-	F(1, 16) = 0.5727; p=0.4602	F(1, 16) = 3.032; p=0.1008	F(1, 16) = 2.145; p=0.1624
	C1ql3	1.772 ± 0.916	0.510 ± 0.035	1.431 ± 0.575	2.565 ± 0.260	-	F(1, 14) = 0.01044; p=0.9201	F(1, 14) = 1.890; p=0.1908	F(1, 14) = 3.686; p=0.0755
	Elmo2	1.269 ± 0.085	1.254 ± 0.127	1.014 ± 0.096	1.023 ± 0.067	-	F(1, 15) = 0.001088; p=0.9741	F(1, 15) = 6.248; p=0.0245	F(1, 15) = 0.01566; p=0.9021
Nucleus Accumbens	Bai3	1.117 ± 0.068	1.117 ± 0.066	1.008 ± 0.065	1.132 ± 0.079	-	F(1, 16) = 0.8044; p=0.3831	F(1, 16) = 0.4471; p=0.5132	F(1, 16) = 0.7846; p=0.3889
	C1ql1	1.197 ± 0.073	1.139 ± 0.050	1.033 ± 0.138	1.330 ± 0.117	-	F(1, 16) = 1.412; p=0.2521	F(1, 16) = 0.01719; p=0.8971	F(1, 16) = 3.118; p=0.0965
	C1ql2	1.365 ± 0.119	1.253 ± 0.152	1.005 ± 0.051	1.011 ± 0.054	-	F(1, 15) = 0.2406; p=0.6309	F(1, 15) = 7.650; p=0.0144	F(1, 15) = 0.2964; p=0.5941
	C1ql3	1.282 ± 0.220	0.793 ± 0.035	1.136 ± 0.271	1.042 ± 0.170	-	F(1, 15) = 1.984; p=0.1794	F(1, 15) = 0.06187; p=0.8069	F(1, 15) = 0.9102; p=0.3552
	Elmo2	1.256 ± 0.091	1.172 ± 0.099	1.013 ± 0.079	1.179 ± 0.123	-	F(1, 16) = 0.1663; p=0.6888	F(1, 16) = 1.406; p=0.2530	F(1, 16) = 1.588; p=0.2256
Prefrontal cortex	Bai3	1.40 5 ± 0.165	1.06 5 ± 0.192	1.02 4 ± 0.118	0.62 5 ± 0.062	↓	F(1, 16) = 6.669; p=0.0200	F(1, 16) = 8.230; p=0.0111	F(1, 16) = 0.04402; p=0.8365
	C1ql1	1.144 ± 0.098	1.016 ± 0.221	1.021 ± 0.101	0.634 ± 0.065	-	F(1, 16) = 3.633; p=0.0748	F(1, 16) = 3.492; p=0.0801	F(1, 16) = 0.9130; p=0.3535
	C1ql2	2.177 ± 0.234	1.429 ± 0.258	1.018 ± 0.095	0.783 ± 0.121	↓	F(1, 16) = 6.645; p=0.0202	F(1, 16) = 22.45; p=0.0002	F(1, 16) = 1.809; p=0.1973
	C1ql3	1.451 ± 0.184	1.229 ± 0.280	1.029 ± 0.131	0.740 ± 0.065	-	F(1, 16) = 1.950; p=0.1817	F(1, 16) = 6.185; p=0.0243	F(1, 16) = 0.03356; p=0.8570
	Elmo2	1.422 ± 0.167	1.090 ± 0.181	1.044 ± 0.160	0.729 ± 0.029	↓	F(1, 16) = 4.795; p=0.0437	F(1, 16) = 6.285; p=0.0233	F(1, 16) = 0.003483; p=0.9537

We assessed mRNA expression of the Bai3 receptor, its ligands (C1ql1–3), and intracellular interacting protein Elmo2 in several limbic brain regions of adult male and female offspring perintally exposed to the SSRI citalopram (or control condition). Transcript levels (fold change relative to control males) are presented as group mean ± SEM. Statistical output from two-way ANOVA indicates effects of perinatal SSRI exposure, sex, and exposure by sex interaction. Significant effects (p<0.05) are shaded in gray.

In the P14 brain, perinatal SSRI exposure had the most prominent effects on the expression of Bai3 and related molecules in dentate gyrus, nucleus accumbens, and prefrontal cortex (Table 1). Perinatal citalopram exposure increased expression of Bai3, C1ql2, and Elmo2 in the dentate gyrus (main effect of exposure) and Bai3 (main effect of exposure) and C1ql3 (increase in females only) in the prefrontal cortex. By contrast, early exposure to citalopram decreased expression of Bai3, C1ql1, C1lq2, and C1ql3 in the nucleus accumbens (main effect of exposure). There were multiple incidences of sex differences in expression of Bai3 or related molecules. Primarily, we observed increased expression of these mRNAs in male versus female brain (e.g., Bai3 in the dentate gyrus; Elmo2 in CA hippocampus; Bai3 and C1ql2 in the nucleus accumbens).

In the adult (P60) brain, perinatal SSRI exposure had far less effect on expression of Bai3 and related molecules (Table 2). Perinatal SSRI exposure had the greatest effect on the prefrontal cortex, decreasing Bai3, C1ql2, and Elmo2 relative to controls (main effect of exposure). In the CA region, there was a decrease in C1ql3 expression as a result of perinatal SSRI exposure (in females only). In the adult brain, there were also fewer sex differences in expression of Bai3 or related molecules. We primarily observed increased expression of Bai3-related mRNAs in female versus male brain (e.g., Bai3, C1ql2, C1ql3, and Elmo2 in the prefrontal cortex).

Experiment 3: Blocking perinatal SSRI-induced Bai3 expression increase in developing dentate gyrus lessens impact on rodent offspring emotional behavior

Because Bai3 expression is increased in the developing dentate gyrus of perinatal SSRI exposed offspring compared to controls, we next investigated whether blocking this change through an siRNA approach would lessen behavioral effects of early life SSRI exposure (Fig. 2A). We found that the Bai3 siRNA treatment lead to a 32.44% reduction of Bai3 levels compared to control siRNA-treated animals (control siRNA fold change: 1.033 ± 0.1079, Bai3 siRNA fold change: 0.7086 ± 0.1220; one-tail unpaired t-test: t(10)=1.993, p=0.0371). Bai3 knockdown significantly reduced passive stress coping (immobility) in the FST (effect of Bai3 siRNA; F (1, 50) = 14.88; p=0.0003). There was also an effect of sex (F (1, 50) = 6.56; p=0.0135), with males showing overall greater immobility compared to females, but no treatment × sex interaction (Fig. 2B). In the sucrose preference test, there were main effects of Bai3 knockdown F (1, 48) = 4.814; p=0.0331; Fig. 2C) and sex (F (1, 48) = 8.377; p=0.0057), as well as a sex × Bai3 siRNA interaction (F (1, 48) = 11.63; p=0.0013). Post hoc analysis revealed SSRI exposed females treated with scrambled siRNA were significantly different from all other groups. There was no impact of the Bai3 siRNA knockdown on anxiety-like behavior in the EPM, although females spent more time in the open arms compared to males (main effect of sex; F (1, 48) = 22.46; p<0.0001; Fig. 2D). In the social interaction test, Bai3 knockdown increased time spent with the interactor rodent relative to an object in males, as indicated by a sex × Bai3 siRNA interaction (F (1, 42) = 5.437; p=0.026; Fig 2E), but we did not find a main effect of Bai3 siRNA or sex.

Experiment 4: Expression of Bai3 and related molecules in the human brain

To follow up our assessment of Bai3 mRNA expression in the developing and adult rat brain, we utilized the BrainCloud database (https://www.libd.org/brain-cloud/) to assess expression of BAI3 and associated ligands from publicly available transcriptome data from human prefrontal cortex. Expression levels were assessed at four developmental stages including fetal (gestational weeks 14–20), infant (0–6 months), childhood (1–10 years), and adolescence through adulthood (until ~80 years of age). BAI3 expression is notably higher than BAI1 and BAI2 in during gestation and in the adolescent and adult brain (Fig. 3A). ELMO2 tends to be more highly expressed than ELMO1 in the infant and adult brain (Fig. 3B). Early in gestation, C1QL1 and C1QL4 are more highly expressed than C1QL2–3. However, by late gestation, C1QL3 expression levels increase, remain relatively high compared to the other C1QL ligands throughout infancy and childhood, and are notably higher in the adolescent and adult brain (Fig. 3C).

Figure 3. Expression of BAI network genes in human brain across the lifespan.

Using publicly available RNA-seq data from human prefrontal cortex, expression of BAI network genes including BAI receptors BAI1–3 (A), ligands C1QL1–4 (B), and intracellular interacting proteins ELMO1–2 (C) were assessed across four developmental stages: fetal (gestational weeks 14–20), infant (0–6 months), childhood (1–10 years), and adolescence and adulthood (until ~80 years of age).

We next investigated the cell-type specific expression of these genes using RNA-seq data from the human cortex (Fig. 4A, 4B) and medial temporal gyrus (Fig. 4C). Similar to the BrainCloud dataset (Fig. 3), adult BAI3 expression is higher than BAI1–2 expression, and this is evident across multiple cell-types, with the highest level of BAI3 occurring in neurons and astrocytes. The high expression of C1QL3 compared to the other C1QL ligands is largely driven by neuronal expression. ELMO1–2 have similar expression across the measured cell-types. However, ELMO2 expression is noticeably higher in mature astrocytes and lower in microglia and macrophages than ELMO1 (Fig. 4A, 4B).

Figure 4. BAI network gene cell-type-specific expression in human brain tissue.

Cell-type specific expression (A) and cell proportion (B) of BAI network genes in multiple cell types were determined from human cortical tissue, including endothelial cells, mature astrocytes, fetal astrocytes, oligodendrocytes, microglia/macrophages, and neurons. Single nuclei RNA-seq identified 75 distinct cell types in which BAI network genes were expressed (C).

Several of these BAI network genes are predominantly expressed in fetal astrocytes, which comprise ~40–60% of the expression of BAI1, BAI2, C1QL2, and C1QL4, and to a lesser extent, BAI3, C1QL1, and ELMO1–2. Approximately 60% of C1QL3 expression can be attributed to neurons (Fig. 4B). Figure 4C is a heatmap of mRNA expression of the BAI network across 75 cell types in the human medial temporal gyrus. Similar to its expression in cortical grey matter, BAI3 has the highest expression in this network across most cell types, with the exception of endothelial cells, microglia, and oligodendrocytes. The C1QL ligands have the most cell-type specific expression patterns when compared to the BAI receptors and ELMO intracellular proteins. C1QL3 expression is predominantly limited to excitatory and inhibitory neurons, with some expression observed in microglia and macrophages, similar to the findings in Fig. 4A. In astrocytes, C1QL2 expression is higher in protoplasmic astrocytes that reside in layers L2–6 when compared to primate-specific intralaminar astrocytes residing in L1–2. In neurons, C1QL2 expression is largely restricted to a subset of inhibitory and excitatory neurons (Fig. 4C).

Experiment 5: BAI3 expression is altered in human postmortem brain tissue from patients suffering with Major Depression or Schizophrenia

Relative mRNA expression of BAI3, C1QL2, and C1QL3 were measured in postmortem hippocampus and dlPFC of men and women diagnosed with MDD, SCZ, and nonpsychiatric controls. In the male hippocampus, there was a main effect of diagnosis on BAI3 expression (F(2, 187) = 10.913; p=3.29×10⁻⁵). Post hoc analysis found that BAI3 was significantly increased in MDD compared to controls (p=0.0344) and SCZ (p=0.0001) and decreased in SCZ versus controls (p=0.0330) (Fig. 5A, left panel). In females, there was a main effect of diagnosis on BAI3 expression (F(2, 101) = 3.214; p=0.04432) (Fig. 5A, right panel). For C1QL2, there was no effect of diagnosis in the male hippocampus (F(2, 177) = 0.946; p=0.3903) or female hippocampus (F(2, 94) = 2.712; p=0.0716; Fig. 5B). For C1QL3, there was a main effect of diagnosis for male hippocampus (F(2, 185) = 4.362; p=0.0141). Post hoc tests indicate a significant decrease in expression in SCZ vs controls (p=0.0234) (Fig. 5C). In female hippocampus, effect of diagnosis was not significant (F(2, 125) = 2.686; p=0.0721).

Figure 5. BAI3, C1QL2, and C1QL3 expression in the hippocampus of patients with major depression or schizophrenia.

In males (left) and females (right), normalized mRNA expression of BAI3 (A), C1QL2 (B), and C1QL3 (C) was measured in the hippocampus of patients with major depressive disorder (MDD), schizophrenia (SCZ), and non-psychiatric control subjects. The median and 95% confidence intervals are indicated for each group. * p<0.05; *** p<0.001.

In the dlPFC, there was no effect of diagnosis on BAI3 expression in males (F(2, 248) = 0.017; p=0.98361) or females (F(2, 115) = 1.040; p=0.3569; Fig. 6A). For C1QL2, there was a main effect of diagnosis in males (F(2, 298) = 4.003; p=0.01926), with post hoc comparisons indicating a significant increase in SCZ vs control (p=0.0376) (Fig. 6B). In females, there was also a main effect of diagnosis (F(2, 167) = 3.971; p=0.02065); post hoc comparisons indicated a significant increase in SCZ vs controls (p=0.0360; Fig. 6B). For C1QL3 in dlPFC, there was no effect of diagnosis in males (F(2, 298) = 0.174; p=0.84033) or females (F(2, 174) = 0.744; p=0.47688; Fig. 6C).

Figure 6. BAI3, C1QL2, and C1QL3 expression in the dorsolateral prefrontal cortex of patients with major depression or schizophrenia.

In males (right) and females (left), normalized mRNA expression of BAI3 (A), C1QL2 (B), and C1QL3 (C) was measured in the dorsolateral prefrontal cortex of patients with major depressive disorder (MDD), schizophrenia (SCZ), and non-psychiatric control subjects. The median and 95% confidence intervals are indicated for each group. * p<0.05.

Discussion

Several studies demonstrate that SSRI use in pregnancy alters human offspring’s brain development (Lugo-Candelas C,Cha J,Hong S,Bastidas V,Weissman M,Fifer WP,Myers M,Talati A,Bansal R,Peterson BS,Monk C,Gingrich JA and Posner J, 2018) and increases risk for emotional disorders (Liu X,Agerbo E,Ingstrup KG,Musliner K,Meltzer-Brody S,Bergink V and Munk-Olsen T, 2017;Malm H BA, Gissler M, Gyllenberg D, Hinkka-Yli-Salomäki S, McKeague IW, Weissman M, Wickramaratne P, Artama M, Gingrich JA, Sourander A, 2016). Parallel work in rodent models shows that perinatal SSRI exposure (a) alters behavioral domains relevant to depression (Glover ME and Clinton SM, 2016); (b) disrupts serotonin signaling (Kinney GG,Vogel GW and Feng P, 1997;Maciag D,Simpson KL,Coppinger D,Lu Y,Wang Y,Lin RC and Paul IA, 2006;Weaver KJ,Paul IA,Lin RC and Simpson KL, 2010); and (c) perturbs epigenetic and transcriptomic patterns in the developing hippocampus (Glover ME,McCoy CR,Shupe EA,Unroe KA,Jackson NL and Clinton SM, 2019;Glover ME,Pugh PC,Jackson NL,Cohen JL,Fant AD,Akil H and Clinton SM, 2015). One of the top molecules we found to be altered in the developing hippocampus following perinatal SSRI exposure was the adhesion GCPR BAI3, which plays several key roles in dendrite formation, maturation and function. We hypothesized that early life SSRI exposure in rats perturbs brain development and behavior through changes in BAI3 receptor signaling. Our present experiments began by demonstrating perinatal SSRI-induced behavioral and BAI3-related molecular changes in male and female offspring since most prior work in the field has focused on males. We then used an siRNA approach to test whether suppressing perinatal SSRI-induced increases in Bai3 expression in the developing dentate gyrus lessened its impact on emotional behavior. Finally, to examine translational implications of this work to humans, we report expression patterns of BAI3 and related molecules in human brain as well as alterations of BAI3 mRNA expression in the hippocampus and dlPFC of patients that suffered with depression and schizophrenia.

Perinatal SSRI exposure altered expression of Bai3 and related molecules in developing rodent limbic brain

Our prior studies found that perinatal exposure to the SSRI paroxetine altered DNA methylation and increased gene expression of Bai3 in the P14 rat hippocampus (Glover ME,McCoy CR,Shupe EA,Unroe KA,Jackson NL and Clinton SM, 2019). Our current data show that perinatal exposure to the SSRI citalopram elicits similar changes in Bai3 expression in P14 hippocampus (primarily P14 dentate gyrus), and shows that it occurs in both male and female offspring. The present study also examined possible perinatal SSRI-induced changes in Bai3 and related molecules across limbic regions, showing decreased expression in the P14 nucleus accumbens, and increased Bai3 in P14 prefrontal cortex.

Expression of BAI3 and related molecules in human brain

Since most prior work on BAI3 has been conducted in rodents, we sought to expand the understanding of BAI3 and related genes in human brain tissue. To do so, we used in silico analyses of publicly available single cell RNA-seq data from human cortex collected from psychiatrically healthy subjects across a range of developmental ages (gestation through late adulthood). Bai3 expression in mouse brain peaks at P1 while Bai1 and Bai2 peak around P10 (Kee HJ,Ahn KY,Choi KC,Won Song J,Heo T,Jung S,Kim JK,Bae CS and Kim KK, 2004). Human data followed a similar trajectory, with BAI3 levels consistently higher than BAI1 and BAI2 in fetal brain. BAI3 was also elevated relative to BAI1 and BAI2 during adolescence and adulthood, a time when mood and psychiatric disorders tend to arise. The BAI receptors are thought to participate in a variety of developmental processes including establishment of cerebellar primary climbing fibers as well as activity-dependent synapse formation and stabilization and dendritic arborization in various brain areas (Kakegawa W et al., 2015;Lanoue V,Usardi A,Sigoillot SM,Talleur M,Iyer K,Mariani J,Isope P,Vodjdani G,Heintz N and Selimi F, 2013;Martinelli DC,Chew KS,Rohlmann A,Lum MY,Ressl S,Hattar S,Brunger AT,Missler M and Sudhof TC, 2016). Studies to date suggest that BAI receptors are predominantly expressed in neurons at the post synaptic density (Kakegawa W,Mitakidis N,Miura E,Abe M,Matsuda K,Takeo YH,Kohda K,Motohashi J,Takahashi A,Nagao S,Muramatsu S,Watanabe M,Sakimura K,Aricescu AR and Yuzaki M, 2015;Sigoillot SM et al., 2015). Our analysis of publicly available cell-type specific RNA-seq data suggests that BAI1–3 may be more ubiquitously expressed than originally thought, with notable expression in astrocytes and oligodendrocytes. Due to its role in establishment of the tripartite synapse, it is possible that BAI3 plays a role in connecting astrocytic processes to the synapse, although future work will be required to investigate this possibility.

While BAI1–3 are thought to be predominantly expressed on the postsynaptic density of neurons, their C1QL ligands are traditionally thought to be expressed in the presynaptic terminal. C1QL1–4 expression has previously been studied in the developing and adult mouse brain (Iijima T et al., 2010), showing that C1ql1–2 expression peaks around embryonic day (E)18 and remains high through adulthood. C1QL1 expression patterns in the human cortex appear less dynamic while C1QL2 expression increased from early to late infancy. C1ql3 peaks in mice around P0, declines during the first postnatal week, and then significantly increases from P10 onward. In human cortex, C1QL3 levels peaked late in fetal development, showed variable levels infancy and childhood. C1QL3 was then the most highly expressed of all the C1QLs evaluated in the adolescent and adult brain. Finally, C1ql4 expression in mice peaks around E18-P0 before declining in later life. In human brain, C1QL4 levels were relatively stable across developmental periods. In terms of cell types where C1QLs are present, most evidence points towards expression in neurons, although some data suggest C1QLs are expressed in glia (Iijima T,Miura E,Watanabe M and Yuzaki M, 2010). For instance, one report found that C1QL1 was released from activated microglia following injury (Glanzer JG et al., 2007). The cell-type specific RNA-seq data presented here are generally consistent with findings in rodents. C1QL3 was highly expressed in human cortical tissue and appears to be particularly enriched in excitatory neurons. Cortical expression of C1QL2 appears to be in glia cells (astrocytes and oligodendrocytes) as well as a small fraction of both inhibitory and excitatory neurons.

In the human cortex, ELMO2 levels were relatively stable across developmental periods measured here, while ELMO1 expression dropped from fetal development to infancy before gradually rising to adult levels. In adolescent and adult brain tissue, ELMO2 levels were higher than ELMO1 levels, which is generally consistent with previous studies in mouse brain (Katoh H et al., 2006). Regarding cell-types where ELMO1–2 are expressed, past work in rodent brain found Elmo1 and Elmo2 are expressed in both neuronal and non-neuronal cell-types (Katoh H,Fujimoto S,Ishida C,Ishikawa Y and Negishi M, 2006;Sato Y et al., 2019). Our present analysis of single-cell RNA-seq human data are largely consistent with past rodent findings. Overall, BAI receptors and ELMO1–2 intracellular binding partners are well distributed throughout the human and rodent brain across neuronal and non-neuronal cell types. The C1QL1–4 ligands show greater cell-type specificity, suggesting that the role of BAI network molecules may depend on the region- and cell-type specific expression of C1QL ligands. Future studies can begin to interrogate specific cell types where BAI3 network molecules are perturbed by perinatal SSRI exposure as well as depression in general.

Expression of BAI3 and related molecules is altered in hippocampus and prefrontal cortex of human subjects that suffered with major depression or schizophrenia

To begin to explore potential clinical significance of our rodent findings, we sought to determine whether expression of BAI3 and related molecules is disrupted in brain tissue from patients suffering with major psychiatric disorders (depression or schizophrenia). We found that BAI3 levels were increased in the hippocampus of male patients suffering with depression relative to healthy controls; in females, BAI3 levels were reduced in patients with depression and schizophrenia relative to controls. For the C1QL ligands, we found decreased C1QL3 levels in the hippocampus of male subjects with schizophrenia compared to controls. In the prefrontal cortex, C1QL2 levels were increased in both male and female subjects suffering with schizophrenia compared to non-psychiatric controls.

Two single nucleotide polymorphisms (SNPs) in BAI3 have been associated with disorganized symptoms in schizophrenia (DeRosse P,Lencz T,Burdick KE,Siris SG,Kane JM and Malhotra AK, 2008), while SNPs in BAI3 and ELMO1 were linked to addiction (Liu QR,Drgon T,Johnson C,Walther D,Hess J and Uhl GR, 2006). The present study is the first to measure expression of BAI3 and ligands C1QL2–3 in the hippocampus and prefrontal cortex of patients with depression. Future studies will need to determine if altered BAI3 expression in patients with depression is due to the disease state, or an effect of antidepressant treatment. Due to the role of BAI3 signaling in synapse maturation and stabilization, as well as dendritic arborization, alterations of BAI3 in depression could explain some of the changes in expression of synaptic genes (Holmes SE et al., 2019;Liu Y et al., 2019;Yoshino Y et al., 2021;Zhang G et al., 2020) as well as loss of hippocampal volume (Mikolas P et al., 2019) associated with depression. Similarly, schizophrenia may be characterized by a loss of synapses in the cortex (Feinberg I, 1982;Jacob S et al., 2011) and reduced cortical thickness as psychosis progresses (Cannon TD et al., 2015). Such effects could be related to activated microglial-driven changes in synaptic plasticity, which have been shown to release C1QL1 (Glanzer JG,Enose Y,Wang T,Kadiu I,Gong N,Rozek W,Liu J,Schlautman JD,Ciborowski PS,Thomas MP and Gendelman HE, 2007). Future research is needed to understand how the BAI3 network is involved in and interacts with neuronal and glial populations as well as their functions that may contribute to psychiatric disorders.

Perinatal SSRI exposure-induced increases in passive stress coping and anhedonia in male and female offspring.

Previous work shows that perinatal exposure to multiple types of SSRI antidepressants (i.e., citalopram, paroxetine, and fluoxetine) increases a number of depression-relevant behaviors in rodents (Glover ME and Clinton SM, 2016) as well as humans (Liu X,Agerbo E,Ingstrup KG,Musliner K,Meltzer-Brody S,Bergink V and Munk-Olsen T, 2017). The present study showed that perinatal exposure to the SSRI citalopram increased passive stress coping in the FST and reduced sucrose preference in both male and female offspring, but did not substantively affect anxiety-like behavior or social measures examined. This is one of only a few studies to examine behavioral effects of perinatal SSRI exposure in females (Glover ME and Clinton SM, 2016). Overall, our current results together with past findings suggest that perinatal SSRI elicits similar behavioral consequences in both male and female offspring (Lisboa SF et al., 2007;Zohar I et al., 2016).

When considering the ways in which perinatal SSRI exposure impacts offspring behavior, an important consideration is whether neurodevelopmental and behavioral changes in offspring are driven by SSRI exposure itself, or if effects occur indirectly due to SSRI-induced changes in the mother’s behavior with her pups. We examined this question in an earlier study and found that treatment with the SSRI paroxetine did not substantially impact maternal behavior, which suggests that observed effects of perinatal SSRI treatment on offspring are likely not driven by mothers’ behavioral changes (Glover ME,Pugh PC,Jackson NL,Cohen JL,Fant AD,Akil H and Clinton SM, 2015).

Another interesting point to consider with the present study and others that examine behavioral consequences of perinatal SSRI exposure is the fact that most studies treat rat mothers that are ‘naïve’/normal (as opposed to a rodent model of depression or other type of emotional dysregulation). One of our past studies began to address this question by studying the effects of perinatal SSRI exposure in two selectively-bred strains of Sprague Dawley rats that exhibited stark differences in emotional reactivity (high behavioral response to novelty rats [HRs] versus low behavioral response to novelty rats [LR]) (Glover ME,Pugh PC,Jackson NL,Cohen JL,Fant AD,Akil H and Clinton SM, 2015). LR rats exhibit serotonin system differences as well as a behavioral phenotype that includes anhedonia, passive stress coping, and psychomotor retardation compared to HR rats. We therefore hypothesized that LR offspring would be more susceptible to negative effects of perinatal SSRI treatment. Our results showed that, indeed, LR offspring exposed to perinatal SSRI exposure showed an exacerbated ‘depressive like’ phenotype in adulthood (vs. LR controls) while HR offspring were unaffected by perinatal SSRI exposure.

Blocking perinatal SSRI-induced Bai3 increase in developing dentate gyrus lessens behavioral effects of perinatal SSRI exposure.

Because perinatal SSRI exposure increased Bai3 expression in the developing dentate gyrus, we hypothesized that transiently blocking this increase using an siRNA approach would lessen the behavioral effects of perinatal SSRI exposure. Previous studies utilizing Accell siRNA have demonstrated a high affinity for uptake by neurons (Nakajima H et al., 2012) where Bai3 is predominantly expressed, and our prior work with Accell siRNAs found that they effectively knockdown target mRNAs for a 4–7 day period (Glover ME,McCoy CR,Shupe EA,Unroe KA,Jackson NL and Clinton SM, 2019;McCoy CR et al., 2019). On postnatal day 10, rats that were going through perinatal SSRI exposure either (a) received an siRNA injection into the dentate gyrus on postnatal day 10 to transiently knockdown Bai3; or (b) were injected with a scrambled siRNA control that would not impact gene expression. Male and female offspring that experienced perinatal SSRI exposure combined with early postnatal Bai3 knockdown showed less passive coping in the FST than those receiving scrambled siRNA. The transient Bai3 knockdown also increased sucrose preference in perinatal SSRI-exposed offspring, although only in female offspring. A caveat of this study was that we were unable to include a control group where perinatal SSRI-exposed offspring had no siRNA manipulation; inclusion of this group could have demonstrated that the control siRNA manipulation did not alter behavioral consequences of perinatal SSRI exposure. Although we were unable to show this group comparison directly, we can confirm that behavioral measures (e.g., FST immobility) for the scrambled control siRNA perinatal SSRI-exposed offspring were comparable to several past cohort of animals that received perinatal SSRI exposure (including those shown in Experiment 1 of the present manuscript).

The present experiment examined a limited set of behaviors and the Bai3 manipulation was conducted in animals that were simultaneously experiencing perinatal SSRI exposure. Future work could more fully explore effects of Bai3 manipulation in naïve animals. Moreover, Bai3-C1ql signaling likely plays disparate roles in distinct brain regions/circuits. For instance, Bai3-C1ql3 signaling in the amygdala has been linked to fear memory (Martinelli DC,Chew KS,Rohlmann A,Lum MY,Ressl S,Hattar S,Brunger AT,Missler M and Sudhof TC, 2016), while disruption of Bai3-C1ql3 in the olfactory bulb impairs acquisition of the social transmission food preference memory (Wang CY et al., 2020). Consequently, future studies could examine effects of manipulating Bai3 expression in different brain regions at different developmental timeframes to determine its impact on behavior.

In summary, perinatal SSRI exposure alters Bai3 network gene expression in the developing rat brain, leading to increased passive coping and anhedonia. Preventing this increase in Bai3 with siRNA mitigates some of the behavioral effects of perinatal SSRI exposure. In humans, the BAI receptors are highly expressed in neurons compared to other cell-types, making it an intriguing target for future CNS medications. Analysis of RNA collected from postmortem brain samples suggests the genes are dysregulated in MDD and SCZ subjects, providing further support linking BAI3 and the ligands C1QL2 and C1QL3 to psychiatric disorders.

Highlights.

Perinatal exposure to the selective serotonin reuptake inhibitor (SSRI) citalopram increased offspring passive stress coping and anhedonia.

Perinatal SSRI exposure increased mRNA expression of G-protein coupled receptor Bai3 and related molecules in the early postnatal dentate gyrus.

Transient Bai3 knockdown in perinatal SSRI-exposed dentate gyrus lessened behavioral consequences of perinatal SSRI exposure.

Human postmortem work showed BAI73 network dysregulation in limbic brain regions of individuals with depression or schizophrenia.

Abbreviations

BAI3

Brain Angiogenesis Inhibitor 3

C1ql

C1q-like

CA

Cornu Ammonis

dlPFC

dorsolateral prefrontal cortex

EPM

Elevated Plus Maze

ELMO

Engulfment and cell motility protein

FST

Forced Swim Test

GPCRs

G-protein coupled receptors

MDD

Major Depressive Disorder

PMI

Postmortem interval

P

Postnatal day

qRT-PCR

quantitative Real-Time PCR

SCZ

Schizophrenia

SSRI

selective serotonin reuptake inhibitor

SNP

single nucleotide polymorphism

siRNA

small interfering RNA