Developmental disruption of amygdala transcriptome and socioemotional behavior in rats exposed to valproic acid prenatally

The etiology of autism spectrum disorder (ASD) likely involves complex interplay between genetic and environmental factors. As over 800 genes are implicated in the etiology of ASD (Simons SFARI database), any one knockout will not entirely represent the complexity of core abnormalities. Moreover, although genetic heritability plays an important role in the etiology of ASD, a number of early environmental exposures have also been linked to ASD risk [1]. The antiepileptic drug and mood stabilizer, valproic acid (VPA), has a well-documented history of increasing the susceptibility to ASD. Children exposed to VPA during the first trimester of pregnancy are at increased risk of developing ASD, with estimates varying between 2.9 and 18-fold greater risk compared to the general population [2‐4].

Rats exposed to VPA prenatally around embryonic day (E) 12.5 display abnormalities in neurological and behavioral development. Notably, the VPA rat model recapitulates many of the core symptoms of ASD, including impaired social behaviors, increased repetitive behaviors, and communication impairments, as well as hyperserotonemia, heightened dopamine levels, increased ratio of excitatory to inhibitory neurotransmission, elevated physiological and behavioral measures of anxiety, and enhanced responsivity to sensory stimulation [5‐9].

VPA alters neural circuits in brain regions implicated in ASD, such as the amygdala, and thus is a useful tool to investigate how the disruption of these circuits can lead to emotional and behavioral abnormalities. Abnormal functioning of the amygdala has long been implicated in the etiology of ASD [10]. Amygdala abnormalities in structure [11‐13], neuronal density [14, 15], and functional MRI activity during social tasks [16‐22] have been reported in patients with ASD. Prenatal VPA treatment induces hyper-excitability, enhanced long-term potentiation, and hyper-plasticity of neurons in the amygdala, a reduction in inhibitory synaptic transmission [23, 24], and local hyper-, but distal hypo-connectivity of neural microcircuits [25]. Local amygdala hyperactivity contributes to enhanced fear memories, over-generalized fear, reduced fear extinction, and enhanced anxiety in rats prenatally exposed to VPA [23, 26, 27]. The impact of VPA on local hyperactivity and hyperplasticity is in line with the “intense world” theory of ASD, which postulates that excessive neuronal processing leads to a hyper-functionality underlying ASD symptomatology [28]. Impairment in GABAergic circuitry [29] and under-connectivity between brain regions [30, 31] are also similarly implicated ASD etiology.

Premature amygdala development is commonly reported in youth with ASD [32‐34], potentially contributing to the observation of clinically significant anxiety in 40% of children and adolescents with ASD, twice that seen in typically developing children [35‐37]. The basolateral amygdala (BLA) is an important center for multimodal sensory information processing in the control of emotional arousal and social behavior [38]. We have previously shown that the rat amygdala undergoes phenomenal changes in the first few weeks of life in GABA and glutamatergic signaling, structure, and function [39‐41]. At postnatal (P) day 10, the ability to learn fear association emerges, along with functional, molecular, and physiological changes in the amygdala [39, 42‐44]. Between P7 and 28, extensive changes occur in molecular and electrophysiological properties of BLA principal neurons [39, 41] and in GABAergic fiber and cell body density [45]. Behaviorally, amygdala-mediated fear learning and extinction undergo dynamic changes between juvenile and adolescent stages [46]. Plasticity in BLA neurons is bidirectional and under very tight control by second messenger systems like cAMP-Protein Kinase A (PKA) [47, 48]. However, surprisingly little is known about the cellular, molecular, or genetic changes that occur in the amygdala over development in ASD individuals or rodent models. In order to develop appropriate early interventions and treatments for neurodevelopmental disorders, it is important to investigate changes across development. Appropriate treatments may differ greatly across time, depending on the natural biological state of neurons in the amygdala. Here, we investigated the gene expression in the BLA across development in prenatally VPA-exposed rats. This is the first comprehensive analysis of genetic changes in this region over development.

Prenatal exposure to VPA disrupted later social behavior, fear expression, anxiety-like behavior, and amygdala gene expression in infant and juvenile animals, with many alterations corresponding to ASD-like symptomatology (Table 2). Maternal behavior was largely unaltered in dams given VPA, suggesting changes in offspring are not due to indirect maternal effects. A reduction in time spent in social proximity and enhancement in isolation were detected in the VPA-treated males. This reduction in sociality is in accordance with a number of previous reports of the VPA model [8, 23, 57‐61]. Prenatal VPA also lead to enhancements in basal anxiety with males displaying greater startle amplitudes and females investigating the center of the open field less than saline counterparts. Enhanced baseline acoustic startle responses are correlated with anxiety disorders like PTSD [62], as well as with ASD in adolescents [63].

Notably, behavioral effects were found primarily in male rats, paralleling the male predominance of ASD and neurodevelopmental disorders resulting from prenatal VPA [3, 64, 65]. Sex differences have been detected in some [57, 60, 66, 67] but not all [59] previous studies of the VPA animal model in which sexes were compared. Impairments in social behavior that are enhanced or exclusive to male VPA-exposed offspring have been reported in rats and mice [32, 57, 60]. Reduced pain sensitivity, enhanced anxiety-like behavior in an elevated plus maze, and increased cortical and hippocampal post-synaptic marker proteins have also been detected exclusively in males [57, 60]. Raza et al. [67] found that prenatal VPA exposure impaired motor behavior and decreased closed-arm time exclusively in female rats. Stefánik et al. [68] reported increased social interaction in females prenatally exposed to VPA. However, Roullet et al. [59] reported male and female mice exposed prenatally to VPA were similarly impaired in sociability scores and nest-seeking responses. An increase in social investigation was reported by Cohen et al. [69]; however, a lower 350 mg/kg dose was used and sexes were combined in the analyses. In other studies, sexes were combined in analyses with no reports of sex differences [23, 70] or only male behavioral deficits were examined [8, 58, 61, 71, 72]. Due to enhanced behavioral deficits observed in males in our sample and previous reports [32, 57, 60], genomic analyses were run in male amygdala. However, future investigation into potential protective pathways in females that accounts for reduced deficits warrants future investigation.

The amygdala undergoes significant functional, morphological, and physiological maturation during the first 2 weeks of development, and insults to this region can dramatically alter later socioemotional behavior. By postnatal day 10, the ability to learn fear associations emerges, which corresponds to changes in amygdala including differential responsivity to shock [42], enhanced synaptic plasticity [44], an emergence of dendritic spines [43], and a switch in the postsynaptic GABA response from excitation to inhibition [39]. Significantly, learned fear in young rats (P17) is no longer expressed after 10 days, but early maternal separations induce adult fear retention [73]. Prenatal VPA exposure may similarly have led to early maturation of fear responses, which is supported by the observed enhancements in fear expression after infant fear conditioning in our data. A hyperactive, precocious amygdala may lead to an early termination of the attachment-learning period and the beginning of fear learning [74].

Early in development, neonates experience a sensitive period for attachment learning and impaired fear learning, thought in part to be mediated via maternal suppression of pup corticosterone levels [75]. VPA-treated males displayed reduced approaches to maternal bedding and increased startle amplitudes in the presence of the maternal odor, and both sexes displayed reduction in distress ultrasonic vocalizations upon separation from the mother. Reduced vocalizations and altered call structures (reduced length, increased frequency) are suggestive of impaired social communication and imply a functional difference in these calls, which may further exacerbate social impairments through development. The number and length of calls increase after an isolated pup is briefly reunited with the mother, known as maternal potentiation [76, 77]. Furthermore, anxiolytics and antidepressants reduce call length in neonates during isolation [78], suggesting the decreased duration and rate of calls in VPA pups may indicate reduced distress from maternal separation. Increased infant call frequencies were also observed in VPA pups, a pattern previously associated with reduced fitness and increased risk for ASD [79, 80]. Olfactory learning is particularly important for attachment during this time period, and the quality of the infant-mother bond and early olfactory learning is a salient predictor of later social relationships [81]. Learned maternal odors serve as safety signals later in life, reducing depressive-like behavior, attenuating fear conditioning, and enhancing social behavior in rats exposed to early stress [82, 83]. Rather than attenuating startle, the presence of the maternal odor (acetophenone) enhanced baseline startle responses, while not impacting responses in control animals. It should be noted that female startle amplitudes in response to maternal odor were not measured and warrant future investigation. The observed alterations in prenatal VPA-treated pup behavior may reflect abnormal amygdala functioning and maternal-infant bonding.

VPA, a nonselective histone deacetylase inhibitor, may transiently increase histone acetylation in the developing embryonic brain when exposed in utero [66], leading to widespread changes in the amygdala transcriptome. This is the first analysis of transcriptomic and proteomic changes across development in the amygdala in both normative and VPA-exposed conditions. Many pathways involved in development, nervous system function, and the immune system displayed predicted activations in VPA amygdala samples at P10, whereas at P21 pathways involved in cell death and developmental disorders predominated (Fig. 5a). Across time, VPA appears to be stunting the normal developmental alterations in amygdala gene expression from P10 to P21. A number of pathways involved in synaptic plasticity, neurotransmission, cellular growth, immune function, and metabolism are reduced in VPA-exposed animals across time relative to saline-treated animals. Conversely, pathways involved in cell death and neurological and developmental disorders are increased across time exclusively in the VPA group. Previous microarray analyses of adult amygdala gene expression from prenatally exposed VPA animals pointed to alterations in similar pathways, such as neuronal projection, cell-cell signaling, synaptic transmission, vesicle, and calcium signaling pathways [69]. Moreover, a microarray study from 35-day-old rats exposed to VPA on E12 reported that the amygdala displayed alterations in many similar IPA gene pathways, including cell death, cell signaling, development, proliferation, movement, inflammatory disease, molecular transport, neurological disease, developmental, psychological disorder, and tissue development [84]. It should be noted that although changes in gene expression across development within each group survived multiple comparisons, differences between the groups did not, and thus these results should be considered exploratory analyses to identify targets for future investigation. Furthermore, the BLA was specifically targeted at P21, but P10 samples may include surrounding amygdala regions.

These data are remarkably in line with transcriptomic analysis of the temporal and frontal cortex showing alterations in genes involved in immune function and synaptic plasticity between individuals with and without ASD [85]. A gene co-expression module involved in synaptic function, vesicular transport, and neuronal projection was downregulated, whereas a module associated with astrocyte markers, activated microglia, immune and inflammatory responses was upregulated in ASD adult samples. The authors also observed an attenuation of gene expression differences between frontal and temporal cortices. Previous microarray-based analyses have revealed similar alterations in developmental or immune pathways in ASD [86‐88]. Chow et al. [87] examined dorsolateral prefrontal cortex gene expression in both young and adult brains of individuals with ASD and reported alterations in pathways involved in neurogenesis, neurodevelopment, cell cycle, DNA damage response, apoptosis, cytoskeleton remodeling, and the immune response. ASD susceptibility genes involved neurodevelopmental pathways such as neuritogenesis, axogenesis, quantity of axons, and formation of neural tube display prenatal enrichment in the dorsolateral prefrontal cortex [88]. Transcriptomic analyses have also found enrichment of ASD genes in cortical projection neurons during embryonic development [89, 90], suggesting future analyses of the cortex are warranted at P10, the rat equivalent of the third trimester in humans [91]. A previous microarray analyses in rats revealed differing effects of VPA on OFC, as compared to anterior amygdala and cerebellar vermis gene expression, which were more similar [69]. However, in a study of primate neuro-development using laser-capture microdissection and microarray profiling from early gestation (E50) through the first 48 months, the amygdala had both an earlier onset of genes related to synaptogenesis and myelination, and a quicker downregulation of those same processes compared to basal ganglia, hippocampus, and neocortex [92]. Genes related to ASD were also enriched in early developing regions like the amygdala, as opposed to genes related to schizophrenia which saw increased expression post birth. Together, these data would suggest that gene disruption in response to fetal VPA exposure will be regulated not only by the process of normative cellular development but also by the time at which individual components of circuits involved in affective regulation come “on-line”.

At P10, pathways involved in the development, morphology, proliferation, and differentiation of cells displayed predicted increases in the amygdala of males exposed to VPA in utero. Increases in pathways involved in neuronal signaling and immune function, and decreases in those involved in developmental disorders were predicted in VPA animals. An early enhancement in cellular differentiation and proliferation is in line with an early overgrowth of the amygdala in ASD [32‐34] and may underlie the observed enhancement in fear learning in infants exposed to VPA. Indeed, VPA promotes neurogenesis and cortical neuronal growth in primary cultures [93].

A number of pathways were similarly altered across P21 transcriptome and proteome samples (Figs. 4 and 5b). Although more immediate effects of VPA in early development may appear beneficial, abnormalities appeared later in the development. In juvenile P21 males exposed to VPA prenatally, a number of amygdala gene pathways involved in neurotransmission, cellular development, nervous system development, cell death, and immune function displayed alterations. Both proteomic and transcriptomic analyses of contralateral amygdala from the same animals at P21 revealed reductions in pathways involved in cellular development and proliferation, which is in contrast to the enhancement in these pathways found at P10. Both transcriptomic and proteomic analyses revealed reduced activity of the ephrin receptor signaling pathway, which regulates axonal guidance and cell migration early in development. Over time from P10 to 21, the ephrin pathway was reduced exclusively in the VPA animals. Knockouts in the ephrin pathway cause ASD-like symptoms [94]. Interestingly, although early premature overgrowth of the amygdala is described in ASD [32‐34], reductions in amygdala volume [11, 12], neuron number [15], and functional activity [16‐19] are observed later in life.

A predicted reduction in activity of immune function pathways was detected in P21 proteomic and transcriptomic samples, as well as in transcriptomic samples from P10 to 21, in VPA animals, which is in line with previously reported immune alterations in VPA rats [57]. Interleukin function has been correlated to social behavior and ASD [95]. Enhanced autoimmunity, allergies, asthma, and reduced immune function have been reported in individuals with ASD [96]. Enhancements in gastrointestinal tract and colorectal cancer were predicted at P21, which is in line with commonly reported gastrointestinal problems in ASD [97]. Reductions in natural killer cell function [98], increases in chemokines, and enhancements in pro-inflammatory cytokines have been associated with ASD [99, 100]. As cytokines serve dual functions in both the immune system and in fetal brain development [101], immune regulators may underlie impairments in both.

The calcium signaling pathway was significantly altered in proteomic and transcriptomic analyses, with reduced activity in VPA animals predicted by mRNA. Between P10 and 21, both saline and VPA amygdala displayed predicted increases in the calcium signaling pathway, however, to a greater extent in the saline group. Ryanodine receptor 2 (Ryr2) and the Calcium Voltage-Gated Channel Auxiliary Subunit α2δ 1 (Cacna2d1) were two of the three genes altered in both mRNA and protein analyses and both displayed reduced expression across analyses. Notably, RYR2, which mediates sarcoplasmic calcium release, is a potential ASD risk gene in humans [102, 103]. RYR2 mRNA is differentially expressed between the frontal and temporal cortices in postmortem tissue from control individuals, but not those with ASD [85]. Cacna2d1 promotes calcium-induced neurotransmitter exocytosis through regulating voltage-gated calcium channel trafficking and increased efficiency of exocytosis [104]. Significantly, functional mutations in several genes encoding voltage-gated calcium channels have also been linked to ASD (see [105]). Alteration in calcium signaling, and thus neuronal communication, may contribute to aberrant amygdala functioning and thus of behavior in VPA rats and individuals with ASD.

Pathways altered across proteomic and transcriptomic analyses at both developmental time points were protein kinase A (PKA) signaling and signaling by Rho family GTPases. A reduction in PKA signaling at P10 and an increase at P21 in VPA were predicted relative to saline amygdala samples. Overtime, both groups displayed predicted increases in the PKA pathway. PKA signaling regulates emotionality and social anxiety, and dysfunction in this pathway is implicated in ASD [106, 107]. Alterations in PKA may underlie changes in synaptic plasticity and activity seen in the amygdala of VPA-exposed animals. We have previously shown that PKA signaling in the amygdala is critical for dopamine D1 receptor facilitation of long-term potentiation [108], enhances membrane potential oscillations in BLA principal neurons [48], and bidirectionally controls synaptic strength [47]. Membrane potential oscillations enhance spike-timing precision and coordinated firing of BLA principal neurons [48], thus VPA-induced changes in PKA activity may significantly alter functional activity of the amygdala.

An increase in signaling by Rho family GTPases at P10, but a reduction at P21, was predicted in VPA amygdala. Over time, this pathway displayed a predicted increased in saline animals but a decrease in VPA exposed males. RhoGTPases are essential regulators of neuronal motility and morphology, particularly in the development and maturation of dendritic spines, and alterations in these systems have been linked to ASD [109]. GTP bound Rho activates Rho-kinase 2 (ROCK2), which ultimately inhibits neurite outgrowth and promotes cell contraction (for review see [110, 111]). Down-regulation of Rho pathways at P21 may contribute for amygdala hyperplasticity in the VPA model [23, 24]. However, knockout of Rock2 expression reduces dendritic spine density and synaptic transmission [112]. Thus, functional outcomes of altered Rho signaling in juvenile animals warrant future investigation. Interestingly, the Rho-Guanine nucleotide dissociation inhibitor (GDI) signaling pathway, which inhibits Rho family GTPases, displayed a corresponding predicted enhancement in activity at P21 in both transcriptomic and proteomic analyses. Additionally, the RhoGDI pathway displayed an exclusive increase in activity in VPA animals.

It should be noted that previous work has found very modest overlap between transcriptomic and proteomic samples, which may be due to posttranscriptional processing such as alternative splicing, modifications, translational efficacy, and degradation [113]. In a comprehensive genomic analysis of mouse liver tissue, an average correlation of only 0.27 was obtained between levels of transcripts and proteins [114]. Similarly, an analysis of human brain tissue yielded only 0.25 mRNA-protein correlation [115]. When controlling for neural cell type, a maximum of 0.47 correlation was reached [116]. We are also limited in mRNA-protein comparisons in that transcriptomic analyses mapped reads to ~17,000 genes, whereas significantly less proteins were assessed (~3500). Future targeted quantification of differentially expressed gene proteins, especially ones highlighted in Fig. 5b that overlapped in transcriptomic and proteomic assays, should be performed, as well as cell-type specific or single cell genomic approaches.

These exploratory genomic analyses provide potential targets of interest in the amygdala in the study of molecular underpinnings of ASD. Furthermore, behavioral and genetic alterations observed support the use of prenatal VPA exposure as an effective tool for the study of pathways underlying social dysfunction relevant to ASD. Future studies should investigate gene expression and physiology of individual BLA neuronal populations throughout development in animals prenatally exposed to VPA. An understanding of the genetic alterations occurring in BLA neurons could identify potential drug targets and critical windows for treatment interventions for anxiety in children with ASD.