The role of glia in the dysregulation of neuronal spinogenesis in Ube3a-dependent ASD

Abstract

Overexpression of the Ube3a gene and the resulting increase in Ube3a protein are linked to autism spectrum disorder (ASD). However, the cellular and molecular processes underlying Ube3a-dependent ASD remain unclear. Using both male and female mice, we find that neurons in the somatosensory cortex of the Ube3a 2X Tg ASD mouse model display reduced dendritic spine density and increased immature filopodia density. Importantly, the increased gene dosage of Ube3a in astrocytes alone is sufficient to confer alterations in neurons as immature dendritic protrusions, as observed in primary hippocampal neuron cultures. We show that Ube3a overexpression in astrocytes leads to a loss of astrocyte-derived spinogenic protein, thrombospondin-2 (TSP2), due to a suppression of TSP2 gene transcription. By neonatal intraventricular injection of astrocyte-specific virus, we demonstrate that Ube3a overexpression in astrocytes in vivo results in a reduction in dendritic spine maturation in prelimbic cortical neurons, accompanied with autistic-like behaviors in mice. These findings reveal an astrocytic dominance in initiating ASD pathobiology at the neuronal and behavior levels.

Introduction

Autism spectrum disorders (ASDs) are a complex, heritable, and diverse class of developmental disabilities present in 1:44 children at age 8 in the United States (Maenner et al., 2020; Parellada et al., 2014). Comorbidities include affective disorders, learning disabilities, and gastrointestinal issues (Lai et al., 2014). Hallmark ASD behaviors include impaired socialization, increased restrictive/repetitive behaviors, and altered communicative vocalization (Masi et al., 2017).

Extreme genetic heterogeneity engenders ASD diversity. Simons Foundation Autism Research Initiative (SFARI) implicates over one thousand genes in ASD. Heterogeneity partially stems from copy number variation, occurring in 5–40% of ASD patients, particularly of ubiquitin-associated, neuronal, and glia-related genes (J. A. Chen et al., 2015; Edmonson et al., 2014; Glessner et al., 2009).

Various cellular morphologies and functional perturbations exist in ASD, with monogenic studies indicating altered dendritic spine density, morphology, and maturation (Gilbert and Man, 2017; Pardo et al., 2005). Reduced dendritic spine density is observed in MeCP2^−/− Rett Syndrome mice (Landi et al., 2011), haploinsufficient reelin mice (Costa et al., 2001), and Angelman Syndrome mice (Ube3a^m-/p+) (Kim et al., 2016). Reduced dendritic protrusions were also observed in NEXMIF KO mice (Gilbert et al., 2020). Fragile X syndrome (FXS) Fmr1 KO mice display increased dendritic spine density (Dolan et al., 2013). These studies indicate improper dendritic spine maturation is a prominent cellular phenotype in ASD.

The mouse model used in this study, Ube3a 2X Tg, contains two extra copies of the Ube3a transgene (Smith et al., 2011). These mice have been shown to display hallmark autism-like behaviors with coincident reductions in glutamatergic signaling (Smith et al., 2011). The human UBE3A gene is located on chromosome 15q11–13, and maternal duplication (dup15) and triplication (isodicentric extranumerary chromosome, idic15) of this region are strongly implicated in human ASD, responsible for 1–3% of the ASD population (Abrahams and Geschwind, 2008; Cook et al., 1997; Noor et al., 2015). Human UBE3A yields UBE3A (previously referred to as E6-AP), a large, multidomain, multifunctional protein (Khatri and Man, 2019). Henceforth, we refer to the murine gene and protein as Ube3a and Ube3a, respectively. As an E3 ubiquitin ligase, Ube3a is crucial for proteostasis (Scheffner et al., 1993). Additionally, Ube3a complexes with transcription factors, influencing transcriptional networks (Furumai et al., 2019; Nawaz et al., 1999b).

Ube3a is paternally silenced by cis-acting antisense lncRNA, thereby expressed solely from the maternal allele (Albrecht et al., 1997; Meng et al., 2012; Rougeulle et al., 1997). Imprinting is restricted to neurons, whereas Ube3a is expressed biallelically in astrocytes (Vu and Hoffman, 1997; Yamasaki et al., 2003). Ube3a localizes to cytosolic and nuclear compartments in neurons and glia (Burette et al., 2017). Ube3a expression is higher in neurons than glia, with peak expression in CNS cells around postnatal day 5 (P5) and no spatially differential expression throughout the cortex (Burette et al., 2018; Judson et al., 2014; Khatri et al., 2018). Loss of maternal Ube3a expression or function ties Ube3a to the neurodevelopmental disorder, Angelman Syndrome (AS) (Kishino et al., 1997). Conversely, maternal Ube3a overexpression or hyperactivity is tied to ASD (Cook et al., 1997; Moreno-De-Luca et al., 2013; Urraca et al., 2013; Yi et al., 2015).

Glia, specifically astrocytes, regulate synaptic connectivity and are heavily implicated in ASD (Zeidán-Chuliá et al., 2014). Astrocytes encourage synapse and spine maturation, elimination, and stabilization (Allen and Eroglu, 2017). These studies agree with documented observations wherein brains of ASD individuals display atypical glial function coincident with hypoconnectivity (Petrelli et al., 2016; Rodriguez and Kern, 2011).

We found that Ube3a overexpression yields immature dendritic spines due to astrocyte dysregulation. We demonstrate Ube3a-dependent immature spines coincide with decreased levels of spinogenic astrocytic glycoprotein, thrombospondin-2 (TSP2), resulting from hindered transcriptional control upon increased astrocytic Ube3a gene dosage. Supplementing TSP2 in vitro rescues spine maturation to control levels. In vivo, astrocytic Ube3a overexpression in control mice impairs spine maturation similarly to 2X Tg mice, shown by reduced dendritic density and increased filopodia density, simultaneously recapitulating autistic-like behaviors including repetitive grooming, impaired communicative vocalization, and social aversion. These findings indicate a crucial role for astrocytes in autistic cellular and behavioral phenotypes, suggesting astrocyte-targeted therapeutic avenues for Ube3a-dependent ASD.

Results

Increased Ube3a gene dosage confers an immature dendritic spine phenotype

To investigate the cellular phenotypes of Ube3a-dependent ASD, we used the Ube3a 2X Tg transgenic mouse model (henceforth referred to as 2X Tg), which contains additional two copies of the Ube3a gene (Smith et al., 2011). These mice were created using BAC recombineering to insert a 162 Kb segment of mouse chromosome 7 (syntenic to human chromosome 15) containing the full exon-intron sequence of Ube3a into FVB-WT mouse embryos to generate Ube3a 1X Tg mice, which were then mated with each other to generate Ube3a 2X Tg mice (Smith et al., 2011). ASD model 2X Tg mice display all three hallmark ASD behaviors: an aversion to socialization, decreased communicative vocalizations, and increased restrictive and repetitive behaviors as modeled by grooming. This mouse also shows impaired glutamatergic excitatory neurotransmission (Smith et al., 2011).

Ube3a-dependent ASD is associated with several morphological abnormalities at the level of the neuron including a reduction in the complexity of dendritic arborization and a reduction in the density of dendritic spines (Khatri et al., 2018). To assess morphological differences between Ube3a 2X Tg mice and their wild-type (WT) littermates, we subjected mice to the Golgi-Cox staining at postnatal day 25 (P25) and measured neurons in the somatosensory cortex (Fig. 1A). Although studies on ASD often examine cellular changes in cortical layer 2/3 (Antoine et al., 2019; Fontes-Dutra et al., 2023; Iguchi et al., 2020; Isshiki et al., 2014; Wang et al., 2017), many studies have also highlighted the implication of layer 5 in the pathogenesis of autism (Brumback et al., 2017; Tang et al., 2014; Willsey et al., 2013; Yamamuro et al., 2020). Layer 4 was not analyzed since there is no clear evidence indicating sensory deficits in 2X Tg mice, and in other ASD models with sensory deficits, the pathology is primarily found in the spinal sensory pathway (Orefice, 2020; Orefice et al., 2019, 2016; Yong et al., 2009; Zimmerman et al., 2019). We therefore measured layer 5 neurons of the somatosensory cortex. We found that in layer 5, the density of overall dendritic protrusions including spines and filipodia was reduced in 2X Tg mice compared to WT mice (Fig. 1B). However, neurons of the 2X Tg mice display a reduction in dendritic spine density when compared to their WT littermates (Fig. 1C). Furthermore, we detected that the density of filopodia-type protrusions along the dendrite was significantly increased in 2X Tg mice (Fig. 1D). When the ratio of filopodia to spine was quantified, we found that the ratio was approximately three times higher in 2X Tg mice (Fig. 1E). This shift from mature dendritic spines to immature filopodia protrusions along the dendrites indicates that the maturation of the dendritic spines in the Ube3a 2X Tg mice is impaired.

Figure 1. Increased astrocytic, but not neuronal, Ube3a gene dosage impairs dendritic spine maturation.

(A) P30 Wild-type (WT) and Ube3a 2X Tg (2X Tg) brains were subjected to Golgi staining. Representative dendritic shafts were visualized in layer 5 neurons in the somatosensory cortex (n=9 mice for each WT and 2X Tg). Scale bar = 6 μm. The white arrow points to a filopodia, and the black arrow points to a spine in each image (B) Total dendritic shaft protrusions were significantly reduced in 2X Tg mice (WT: 5.29 ± 0.54 per 5 μm; 2X Tg: 4.44 ± 0.37 per 5 μm). (C) Dendritic spine density was significantly decreased in the 2X Tg mice (WT: 4.11 ± 0.41 per 5 μm; 2X Tg: 3.44 ± 0.88 per 5 μm). (D) Filopodia density was increased in the brains of the 2X Tg mice (WT: 0.82 ± 0.17 per 5 μm; 2X Tg: 1.63 ± 0.15). (E) The ratio of filopodia to dendritic spines in WT and 2X Tg neurons (normalized to WT) (WT: 1.00 ± 0.24; 2X Tg: 2.78 ± 0.44). (F) Primary neurons co-transfected with mGFP/pcDNA 3.1 or mGFP/Ube3a-MYC at DIV 10 were imaged at DIV 15. Scale bar = 45 μm. (G) Dendritic shafts with protrusions from both pcDNA and Ube3a transfected primary neurons (cells n=15 from 3 independent experiments per condition; protrusions n=201). Scale bar = 5 μm. (H-I) Neither filopodia or dendritic spines displayed significant changes (pcDNA filopodia: 0.36 ± 0.06 per 5 μm; Ube3a filopodia: 0.36 ± 0.04 per 5 μm; pcDNA spines: 5.01 ± 0.39 per 5 μm; Ube3a spines: 5.06 ± 0.38 per 5 μm). (J) Primary rat neurons were infected with control, AAV-Ube3a, or AAV-gfapUbe3a at DIV 1, then transfected with mGFP at DIV 10, and imaged at DIV 15 (cells n=19–22 per condition; protrusions n=539–732 per condition). Scale bar = 8 μm. The white arrow points to a filopodia, and the yellow arrow points to a spine in each image. (K) Dendritic spine density was decreased in the Ube3a and gfapUbe3a infected conditions (F=6.81; ANOVA; control: 4.29 ± 0.22 spines per 5 μm; Ube3a: 3.16 ± 0.23 spines per 5 μm; gfapUbe3a: 3.68 ± 0.19 spines per 5 μm). (L) Filopodia density was increased upon infection with AAV-Ube3a and AAV-gfapUbe3a (F=17.05; ANOVA; control: 0.99 ± 0.17 filopodia per 5 μm, Ube3a: 1.67 ± 0.23 filopodia per 5 μm; gfapUbe3a: 2.49 ± 0.16 filopodia per 5 μm). (M) Filopodia:spine ratios show a significant increase in the Ube3a and gfapUbe3a conditions compared to the control (F=11.23; ANOVA; control: 1.00 ± 0.22 filopodia/spines; Ube3a: 2.46 ± 0.33 filopodia/spines; gfapUbe3a: 2.78 ± 0.27 filopodia/spines). (N) The density of protrusions was only observed to be decreased in the Ube3a condition (F=4.58; ANOVA; control: 5.27 ± 0.17 per 5 μm; Ube3a: 4.48 ± 0.27 per 5 μm; gfapUbe3a: 5.44 ± 0.22 per 5 μm). Significance for (B-E, H, I) was determined via Student’s t-test; significance for (K-M) was determined via ANOVA within respective time points. *P<0.05. **P<0.01. ***P<0.001. ****P<0.0001. n.s., not significant, P>0.05. All error bars indicate standard error.

To investigate the dendritic spine maturation impairments in vitro, we cultured primary neurons from embryonic day 18 (E18) rat embryos and transfected neurons at 11 days in vitro (DIV 11) with either an empty vector (pcDNA) or a Ube3a-containing plasmid (pCMV-Ube3a-MYC), which was co-transfected with membrane-bound GFP (mGFP). Neurons were fixed at DIV 15 and subjected to immunostaining of the Ube3a C-terminal MYC tag (Fig. 1F, 1G). We then analyzed the density of filopodia (Fig. 1H) and dendritic spines (Fig. 1I). Unexpectedly, no significant difference was detected between neurons with Ube3a overexpression and the control.

Following the negative findings in neurons, we re-evaluated the experimental methods and noted that the transfection only affected sparsely distributed neurons in the mixed culture containing both neurons and glia. This does not mimic the state of the Ube3a 2X Tg mouse brain, where Ube3a is ubiquitously overexpressed in all cell types. We wondered whether increased gene dosage of Ube3a must exist in both neurons and glia to induce neuronal changes. Indeed, astrocytes are known to play a number of roles critical to neural connectivity, particularly in dendritic spine growth, developmental refinement, and synaptic maturation (Allen and Eroglu, 2017; Baldwin and Eroglu, 2017; Christopherson et al., 2005). To this end, we infected rat primary neuron and astrocyte co-cultures at DIV 1 with either a control AAV2 virus (AAV-scrambled), or an AAV2 virus expressing Ube3a (AAV-Ube3a), or an astrocyte-specific AAV2 Ube3a virus (AAV-gfapUbe3a). Cultures were maintained to DIV 11 at which time they were transfected with mGFP for spine visualization upon collection at DIV 15 (Fig. 1J). Differing from plasmid transfection, AAV-Ube3a was overexpressed in both neurons and astrocytes, whereas only astrocytes overexpressed Ube3a when infected with AAV-gfapUbe3a. We found that dendritic spine density was significantly reduced while the density of immature dendritic filopodia was increased in AAV-Ube3a infected co-cultures as compared to the control cultures (Fig. 1K, 1L). Consistently, the filopodia:spine ratio was increased in the AAV-Ube3a neurons (Fig. 1M). Surprisingly, we also detected changes in spine and filopodia in cultures infected with AAV-gfapUbe3a, where only astrocytes overexpressed Ube3a, to levels comparable to those found in neurons of AAV-Ube3a infected co-cultures (Fig. 1K, 1L). It was noted that the overall density of dendritic protrusions was markedly decreased in AAV-Ube3a cultures, but not in AAV-gfapUbe3a cultures when compared to the control (Fig. 1N). Because significant changes in dendritic spine and filopodia densities were induced upon both astrocyte-neuron and astrocyte-specific overexpression of Ube3a, but not upon neuron-only overexpression of Ube3a, these data suggest that increased Ube3a gene dosage in astrocytes is the prime contributing factor to impaired dendritic spine maturation.

Thrombospondin-2 is implicated in hindered dendritic spine maturation in Ube3a 2X Tg mice

We found that the defects in dendritic spine maturation could be induced by overexpression of Ube3a (AAV-gfapUbe3a) in co-cultured astrocytes, indicating an involvement of cross-talk between astrocytes and neurons. We therefore decided to further investigate astrocytic role in the impaired dendritic spine maturation with Ube3a overexpression.

Known mechanisms of astrocyte-mediated dendritic spine and excitatory synapse maturation may rely on contact-dependent mechanisms such as adaptor protein Hevin facilitating the formation of the transsynaptic bridge between pre-synaptic neurexin-1α and post-synaptic neuroligin-1 (Singh et al., 2016). Conversely, contact-independent mechanisms of synapse and spine maturation rely on functional synaptic strengthening via secreted proteins glypican-4 (Farhy-Tselnicker et al., 2017) and structural maturation-mediator proteins known as thrombospondins (Christopherson et al., 2005; Eroglu et al., 2009; Risher et al., 2018).

To determine if the Ube3a-dependent changes in dendritic spine maturation is due to contact-dependent or contact independent mechanisms upon astrocytic overexpression of Ube3a, we prepared neuronal cultures and astrocytic cultures separately (see Methods). Primary astrocytes were plated from postnatal day 1 (P1) wild-type (WT) or Ube3a 2X Tg mice. At DIV 13, fresh culture medium was added to the astrocytes to be conditioned for 48 hours. This astrocyte conditioned media from WT and 2X Tg mice (WT ACM and 2X Tg ACM, respectively) was then collected to feed the primary neurons. Primary neurons cultured from E18 rat embryos were incubated with the conditioned astrocyte medium at DIV1, DIV12, or DIV 13 until DIV15 for imaging. All neurons were transfected at DIV 10 with membrane-GFP for structural examination. Interestingly, we found that neurons treated with 2X Tg ACM displayed a reduced density in dendritic spines when compared to neurons grown in WT ACM (Fig. 2A, 2B). A significant increase in the density of filopodia and an increase in the filopodia:spine ratio was also observed in DIV 15 neurons grown in 2X Tg ACM when compared to WT ACM-grown neurons (Fig. 2C-D). These results strongly indicate that astrocytes overexpressing Ube3a lead to perturbation of dendritic spine maturation via a contact-independent mechanism.

Figure 2. TSP2 is reduced in 2X Tg brains and media of 2X Tg astrocytes is sufficient to cause impaired spine maturation.

(A) Primary rat neurons were treated with either WT or 2X Tg astrocyte conditioned media (ACM) for indicated times, and imaged at DIV15 to show the dendrites and spines. The white arrow points to a filopodia, and the yellow arrow points to a spine in each image (n=16 neurons from 3 independent experiments per condition; control N: 582 protrusions; 48 hr N: 400 protrusions; 72 hr N: 504 protrusions; 15 d N: 765 protrusions). Scale bar = 8 μm. (B) Dendritic spine density was decreased in the 2X Tg ACM treated conditions (WT 48 hr: 4.71 ± 0.29 spines per 5 μm; WT 72 hr: 4.55 ± 0.28 spines per 5 μm; WT 15 d: 5.41 ± 0.26 spines per 5 μm; 2X Tg 48 hr: 2.74 ± 0.28 spines per 5 μm; 2X Tg 72 hr: 2.58 ± 0.20 spines per 5 μm; 2X Tg 15 d: 3.61 ± 0.21 spines per 5 μm). (C) Filopodia density was significantly increased in the 2X Tg ACM treated neurons when compared to the WT ACM treated neurons (WT 48 hr: 1.51 ± 0.10 filopodia per 5 μm; WT 72 hr: 1.32 ± 0.08 filopodia per 5 μm; WT 15 d: 1.19 ± 0.09 filopodia per 5 μm; 2X Tg 48 hr: 2.20 ± 0.28 filopodia per 5 μm; 2X Tg 72 hr: 2.88 ± 0.20 filopodia per 5 μm; 2X Tg 15 d: 2.70 ± 0.15 filopodia per 5 μm). (D) Filopodia:spine ratios show a significant increase in the 2X Tg ACM treatment conditions (WT 48 hr: 0.32 ± 0.02 filopodia/spines; WT 72 hr: 0.41 ± 0.06 filopodia/spines; WT 15 d: 0.26 ± 0.04 filopodia/spines; 2X Tg 48 hr: 1.20 ± 0.08 filopodia/spines; 2X Tg 72 hr: 1.08 ± 0.08 filopodia/spines; 2X Tg 15 d: 0.85 ± 0.06 filopodia/spines). Significance for (B-D) was determined via two-way ANOVA. (E) Whole cortex homogenates of P30 WT or 2X Tg mouse brains were subjected to western blot analysis for Rac1, Glypican-4 (GPC4), α2δ−1, Thrombospondin-1 (TSP1), and Thrombospondin-2 (TSP2). GAPDH was used as a loading control. (F-J) Quantification of protein intensities of glypican-4 (WT: 0.37 ± 0.04 A.U.; 2X Tg 0.35 ± 0.02 A.U.), Rac1 (WT: 0.71 ± 0.09 A.U.; 2X Tg 0.86 ± 0.08 A.U.), α2-δ1 (WT: 0.50 ± 0.03 A.U.; 2X Tg 0.52 ± 0.02 A.U.), TSP1 (WT: 0.64 ± 0.04 A.U.; 2X Tg 0.68 ± 0.03 A.U.), and TSP2 (WT: 0.19 ± 0.02 A.U.; 2X Tg 0.03 ± 0.01 A.U.). n=9 brains per condition. A.U.= arbitrary units. Significance for (F-J) was determined with a Student’s T-test. *P<0.05. **P<0.01. ***P<0.001. ****P<0.0001. Not significant (n.s.) P>0.05. All error bars indicate standard error.

The observed neuronal effects resulting from the 2X Tg conditioned astrocyte medium suggests improper production, secretion, and insufficiency of spinogenic factors derived from the astrocytes. We thus sought to identify dysregulated proteins known to be secreted from astrocytes and implicated in structural maturation of dendritic spines. To address this, we collected whole cortical tissue from P30 WT and 2X Tg mice and subjected the homogenated lysates to western blot analysis (Fig. 2E). We saw no significant changes in glypican-4 (GPC4), a molecule involved in functional synapse maturation (Allen et al., 2012; Farhy-Tselnicker et al., 2017), nor thrombospondin-1 (TSP1), known to be implicated in synapse structural maturation (Fig. 2F, 2I) (Allen and Eroglu, 2017; Risher et al., 2018). However, when probed for thrombospondin-2 (TSP2), the only other thrombospondin expressed in the central nervous system (Stenina-Adognravi, 2014), we detected a significant reduction in the 2X Tg mice (Fig. 2J). No significant changes were found when other TSP2 signaling components were probed, including TSP2 receptor, α2-δ1 (Fig. 2H), or the downstream effector protein for actin reorganization, Rac1 (Fig. 2G).

Increased Ube3a gene dosage in astrocytes results in a decrease in cytosolic and secreted TSP2

In our search for cues released from 2X Tg astrocytes, we found a marked reduction in TSP2 and considered it as a prime candidate factor due to its astrocyte-specific expression, extracellular secretion, and known function in facilitating formation of excitatory synapses and spines (Christopherson et al., 2005; Risher et al., 2018). To further investigate the involvement of TSP2, we prepared astrocyte cultures from P1 WT and 2X Tg mice, and collected cells at DIV 11 for western analysis. We found that TSP2 protein was significantly reduced in 2X Tg astrocytes (Fig. 3A, 3B). To ascertain if TSP2 secretion was impaired due to Ube3a overexpression, we collected ACM from DIV 15 WT and 2X Tg astrocytes and performed a media immunoprecipitation of TSP2. We found that TSP2 precipitated from the media of 2X Tg astrocytes was significantly less than TSP2 from WT ACM (Fig. 3C, 3D). In addition, we prepared primary astrocytes cultured from E18 rat brains and virally infected the cells with either AAV-GFP or AAV-Ube3a at DIV 1. At DIV 15 the astrocyte medium was collected for immunoprecipitation of TSP2 (Fig. 3E). Similar to the findings from 2X Tg astrocytes, we found that Ube3a overexpression in rat astrocytes led to a significant reduction in cell lysate TSP2 and the secreted TSP2 isolated from the media (Fig. 3D, 3F).

Figure 3. Astrocytic Ube3a overexpression results in a reduction in TSP2 via transcriptional suppression.

(A) Lysates of DIV 11 primary astrocytes from WT or 2X Tg mice were subjected to Western blot to assess expression levels of Ube3a and TSP2. Tubulin was used as a loading control. (B) Quantification of TSP2 content in astrocyte lysates (WT: 0.52 ± 0.09 A.U.; 2X Tg: 0.35 ± 0.03 A.U.; n=6 mice per condition). (C) Media immunoprecipitation of TSP2 from DIV 15 primary mouse astrocyte cultures. TSP2 was reduced in the media of cultured astrocytes from 2X Tg mice (C) or media of WT astrocytes with viral overexpression of Ube3a. (D) Protein intensity quantification of (C) (WT IP: 4.28 ± 0.45; WT input: 0.62 ± 0.01; 2X Tg IP: 1.33 ± 0.12; 2X Tg input: 0.29 ± 0.02; n=3 mice per condition). (E) Media immunoprecipitation of TSP2 from DIV 15 primary neurons cultured from E18 rat embryos infected with AAV-GFP or AAV-Ube3a at DIV 1. (F) Protein intensity quantification of (D) (GFP IP: 3.65 ± 0.47; GFP input: 1.14 ± 0.02; Ube3a IP: 0.14 ± 0.00; Ube3a input: 0.37 ± 0.01; n=3 IPs per condition). (G) Ubiquitination assays on TSP2. p27 was used as a ubiquitination positive control (n=3). (H) Quantification of TSP2 mRNA levels by qRT-PCR indicates TSP2 transcription was significantly reduced upon Ube3a overexpression (F=2047.85; ANOVA; AAV-GFP: 1.00 ± 0.00 A.U.; AAV-gfapGFP: 1.00 ± 0.02 A.U.; AAV-Ube3a: 0.30 ± 0.00 A.U.; AAV-gfapUbe3a 0.46 ± 0.00 A.U.; n=3 replicates per condition). (I) Expression of the Ube3a C820A ligase-dead mutant in primary astrocytes significantly decreases TSP2 transcription to a level comparable to that observed with wild-type Ube3a (F=126.8; ANOVA; AAV-gfapGFP: n=4; AAV-gfapUbe3a: n=5; AAV-gfapUbe3a Mutant: n=5). Significance for (B) was determined with a Student’s T-test. Significance for (D) and (F) was determined with two-way ANOVAs. Significance for (H) and (I) was determined using one-way ANOVAs. All error bars indicate standard error. *P<0.05. **P<0.01. ***P<0.001. ****P<0.0001. Not significant (n.s.) P>0.05. A.U. = arbitrary units.

Transcriptional suppression of Thrombospondin-2 by Ube3a overexpression

The amount of TSP2 protein is negatively correlated with Ube3a expression, indicating that Ube3a may regulate TSP2 synthesis or stability. As an E3 ubiquitin-ligase, Ube3a is known to facilitate ubiquitination and degradation of specific target proteins, including p27, p53, XIAP and itself (Khatri and Man, 2019; Mishra et al., 2009). We thus performed ubiquitination assays to determine if TSP2 is a target for Ube3a-meditated ubiquitination and degradation. Astrocytes were infected with either AAV2 GFP or AAV2 Ube3a at DIV 1 and were treated with either vehicle or proteasome inhibitor MG-132 for 12 hrs on DIV 14, before cell lysate collection. Following immunoprecipitation, we found that cells overexpressing Ube3a showed strong ubiquitination accompanied with a reduction in protein content of p27, a known target of Ube3a ubiquitination (Mishra et al., 2009). Expectedly, p27 protein reduction was blocked by MG-132. However, TSP2 failed to show significant ubiquitination by Ube3a expression and MG-132 treatment (Fig. 3G), and its reduction in protein content by Ube3a was not blocked by MG-132 treatment. Therefore, Ube3a doesn’t appear to modulate TSP2 protein content via ubiquitination and subsequent proteasomal degradation.

Ube3a localizes to both the cytosolic and nuclear compartments (Burette et al., 2018, 2017; Dindot et al., 2008). When localized to the nucleus, Ube3a acts as a transcriptional coregulator of genes involved in immunity and synaptic maturation in mammalian and Drosophila models (Dindot et al., 2008; Ferdousy et al., 2011; Furumai et al., 2019; Nawaz et al., 1999b; Reiter et al., 2006). We therefore investigated the possible transcriptional regulation of TSP2 upon Ube3a overexpression. To this end, we performed qRT-PCR using RNA collected from primary astrocyte cultures following viral infection with AAV-Ube3a or astrocyte-specific AAV-gfapUbe3a. Primers against TSP2 and β-actin (as a control) were used for amplification. Quantification revealed a significant decrease in TSP2 transcription upon overexpression of Ube3a (Fig. 3H). Additionally, overexpressing the E3 ligase dead mutant Ube3a-C820A in primary astrocytes induced a similar reduction in TSP2 transcription, suggesting that the regulation of TSP2 was independent of Ube3a’s E3 ligase function (Fig. 3I). These findings indicate that the Ube3a-induced TSP2 reduction results from a down-regulation of TSP2 gene transcription.

TSP2 mediates the Ube3a-induced defects in dendritic spine maturation in vitro

With the findings on TSP2 reduction as a result of Ube3a overexpression, we next sought to validate that TSP2 is responsible for Ube3a-induced alterations in spine structural maturation. First, we immunodepleted TSP2 from WT mouse ACM using a TSP2-specific antibody. Mock immunodepletion in the same ACM was also performed using IgG as a control (Fig. 4A). DIV12 neuron cultures were then incubated with TSP2 immunodepleted media for 3d until structural examination at DIV15. Incubation with the mock- and TSP2-immunodepleted ACM showed no significant change in dendritic protrusion density (Fig. 4B). However, neurons displayed a reduction in dendritic spine density with a concomitant increase in filopodia density when incubated in ACM immunodepleted of TSP2 (Fig. 4C, 4D). In line with these changes, analysis showed a significant increase in the filopodia:spine ratio in neurons treated with TSP2 immunodepleted ACM (Fig. 4E).

Figure 4. TSP2 is responsible for Ube3a-induced defects in dendritic spine structural maturation in vitro.

(A) Astrocyte conditioned media collected from WT primary astrocyte cultures was immunodepleted with either a non-specific IgG or a monoclonal TSP2 antibody. Immunodepleted media was then used to incubate DIV 12 primary rat neurons transfected with mGFP. Neurons were imaged at DIV 15 to show the dendritic spines. Scale bar = 8 μm. White arrows point to filopodia, and the yellow arrows point to spines. (B-E) Quantifications for total dendritic protrusion density (Mock ID: 5.76 ± 0.45 per 5 μm; TSP2 ID: 5.55 ± 0.32 per 5 μm), dendritic spine density (5.37 ± 0.51 per 5 μm; TSP2 ID: 3.22 ± 0.24 per 5 μm), filopodia density (Mock ID: 1.68 ± 0.24 per 5 μm), and normalized filopodia:spine ratios (Mock ID: 1.00 ± 0.27; TSP2 ID: 3.68 ± 0.64) (n=15 cells from 3 independent experiments per condition; N=220 protrusions). (F) Astrocyte conditioned media from WT primary astrocytes (control) or 2X Tg astrocytes was used to treat primary rat neurons at DIV 12 over the course of 72 hours. 2X Tg ACM treated cultures were simultaneously supplemented with vehicle or recombinant TSP2 (rTSP2) protein at 10 nM. Representative images are from DIV 15 primary neurons transfected with mGFP. Scale bar = 6 μm. White arrows point to filopodia. Yellow arrows point to spines. (G-J) Quantifications for total protrusion density (F=0.70; ANOVA; control: 5.63 ± 0.25 per 5 μm; 2X Tg ACM: 5.28 ± 0.25 per 5 μm; 2X Tg + rTSP2: 5.77 ± 0.39 per 5 μm), dendritic spine density (F=4.02; ANOVA; control: 4.78 ± 0.38 per 5 μm; 2X Tg ACM: 3.63 ± 0.21 per 5 μm; 2X Tg + rTSP2: 4.99 ± 0.45 per 5 μm), filopodia density (F=33.35; ANOVA; control: 0.45 ± 0.06 per 5 μm; 2X Tg ACM: 1.11 ± 0.09 per 5 μm; 2X Tg + rTSP2: 0.39 ± 0.05 per 5 μm), and normalized filopodia:spine ratios (F=30.81; ANOVA; control: 1.00 ± 0.14; 2X Tg ACM: 3.79 ± 0.38; 2X Tg + rTSP2: 1.14 ± 0.28; n=20 neurons from 3 independent experiments per condition; N=1300 protrusions). Significance or lack thereof for (B-E) was determined using a Student’s T-test. Not significant (n.s.). Significance or lack thereof for (G-J) was determined with a one-way ANOVA. Not significant (n.s.) P>0.05. *P<0.05. **P<0.01. ***P<0.001. ****P<0.0001. All error bars indicate standard error.

We then decided to assess whether TSP2 is sufficient to rescue the Ube3a-induced spine defects. Primary astrocytes cultured from P1 WT or 2X Tg mice were given fresh media to conditioned for 48 hrs. The conditioned media were then used alone or supplemented with 10 nM TSP2 protein to treat DIV12 primary neurons for 3 days (Fig. 4F). We found that the overall protrusion density was maintained the same between the WT control, 2X Tg, and 2X Tg + TSP2 groups (Fig. 4G). Additionally, when compared to control neurons treated with WT ACM, the 2X Tg ACM cells showed a reduction in spine density (Fig. 4H) and an increase in filopodia density (Fig. 4I). However, we found that neurons with 2X Tg ACM with TSP2 supplement showed no change in dendritic spine density and filopodia density compared to the WT control, indicating that the TSP2 supplement blocked the effects of 2X Tg ACM (Fig. 4H, 4I). The increased density of filopodia observed upon 2X Tg ACM treatment was abolished upon treatment with TSP2. In agreement with these findings, the increase in dendritic filopodia:spine ratio observed in 2X Tg ACM treated neurons was similarly ablated, indicating that TSP2-supplemented ACM from Ube3a overexpressing astrocytes was sufficient to block the structural impairments of dendritic spines (Fig. 4J).

Viral-mediated astrocyte-specific expression of Ube3a in mice impairs dendritic spine maturation in vivo

Our findings indicate that selective overexpression of Ube3a in astrocytes leads to TSP2 down-regulation, which is sufficient to hinder spine maturation in vitro. However, whether the astrocyte-specific effects hold true in vivo in the brain remains unknown. To answer this question, we performed intraventricular brain injection in P1 WT mice with virus of AAV-gfapGFP as a control, or virus of AAV-gfapUbe3a. Intraventricular injection of AAV-gfapGFP in P1 2X Tg mice was also performed as a positive control. At P30, mouse brains were collected for examination. Immunostaining of astrocyte marker GFAP revealed a high level of colocalization of GFAP-positive and the infected GFP-positive cells in prefrontal cortical layers including layer 2/3 and layer 5 (Fig. 5A, 5B). The infection was of high efficiency, as GFP expression was present in over 90% of GFAP+ cells in each condition (Fig. 5C). Additionally, cortical brain tissue of the WT mice infected with AAV-gfapUbe3a showed a significant reduction in TSP2 protein content (Fig. 5D, 5E).

Figure 5. Astrocyte-specific expression of Ube3a impairs dendritic spine maturation in neurons.

(A) P1 mice were ventricularly injected with viruses and brains were collected for immunohistochemistry at P30. GFP (green) was detected in >90% of GFAP immunolabeled astrocytes (red) in all prelimbic cortices of injected mouse brains. WT + AAV-gfapGFP n=10 mice; 2X Tg + AAV-gfapGFP n=7 mice; WT + AAV-gfapUbe3a-GFP n=11 mice. Scale bar = 50 μm. (B) Representative astrocytes from (A), showing coexpression of GFP and GFAP (red). (C) Virus efficacy in each condition represented as a ratio of GFP+ cells to GFAP+ cells. In each instance, >90% of GFAP positive cells (astrocytes) expressed GFP (F=1.01; ANOVA; WT + gfapGFP: 0.91 ± 0.01 GFP+/GFAP+ cells; 2X Tg + gfapGFP: 0.92 ± 0.02 GFP+/GFAP+ cells; WT + gfapUbe3a: 0.91 ± 0.01 GFP+/GFAP+ cells). (D) Western analysis using P30 whole cortical lysates from mice with ventricular AAV injection at P1. GAPDH was used as a loading control. (E) Quantification of normalized TSP2 intensity from indicated mice (F=20.83; ANOVA; WT + gfapGFP: 1.00 ± 0.04 A.U.; 2X Tg + gfapGFP: 0.31 ± 0.09 A.U.; WT + gfapUbe3a: 0.43 ± 0.10 A.U.). n = 5 mice per condition. (F) Representative dendrites from Layer 2/3 prelimbic cortical neurons at P30 from the indicated conditions (n=13 images from individual neurons for each condition). Scale bar = 6 μm. The white arrow indicates filopodia, and the black arrow indicates spines. (G-J) Quantifications of dendritic protrusion density (F=8.48; ANOVA; WT + gfapGFP: 5.21 ± 0.32 per 5 μm; 2X Tg + gfapGFP: 3.91 ± 0.18 per 5 μm; WT + gfapUbe3a: 4.01 ± 0.22 per 5 μm), dendritic spine density (F=13.35; ANOVA; WT + gfapGFP: 4.79 ± 0.36 per 5 μm; 2X Tg + gfapGFP: 3.03 ± 0.20 per 5 μm; WT + gfapUbe3a: 3.14 ± 0.21 per 5 μm), filopodia density (F=3.37; ANOVA; WT + gfapGFP: 0.99 ± 0.05 per 5 μm; 2X Tg + gfapGFP: 1.38 ± 0.17 per 5 μm; WT + gfapUbe3a: 1.37 ± 0.12 per 5 μm), and normalized filopodia:spine ratio (F=6.71; ANOVA; WT + gfapGFP: 1.00 ± 0.10 filopodia/spine; 2X Tg + gfapGFP: 2.22 ± 0.33 filopodia/spine; WT + gfapUbe3a: 2.08 ± 0.29 filopodia/spine). WT + AAV-gfapGFP n=522 protrusions; 2X Tg + AAV-gfapGFP n=325 protrusions; WT + AAV-gfapUbe3a n=367 protrusions. Significance or lack thereof was determined using a one-way ANOVA. *P<0.05. **P<0.01. ***P<0.001. Not significant (n.s.) P>0.05. All error bars indicate standard error.

With viral expression of Ube3a in an astrocyte-specific manner in the brain confirmed, we then sought to determine whether an astrocyte-specific increase in Ube3a gene dosage affects dendritic spine maturation. WT mice were intraventricularly injected with either AAV-gfapGFP or AAV-gfapUbe3a at P1. As a positive control, 2X Tg littermate mice were injected with AAV-gfapGFP at P1. All mice were then collected at P30 for Golgi-Cox staining. We then imaged dendritic shafts of layer 2/3 prelimbic cortical neurons for dendritic spine analysis (Fig. 5F). We found that Ube3a overexpression of AAV-gfapUbe3a in astrocytes caused a reduction in dendritic protrusion density with a reduction in dendritic spine density and an increase in filopodia density when compared to WT + AAV-gfapGFP mice (Fig. 5G-I). These changes induced by AAV-gfapUbe3a were comparable to the 2X Tg mice (Fig. 5G-I). Consistently, the filopodia:spine ratio in the AAV-gfapUbe3a mice was increased to a level comparable to that observed in 2X Tg littermate mice (Fig. 5J). Together, these data suggest that increased astrocytic Ube3a gene dosage impairs dendritic spine maturation in vivo.

Astrocytic Ube3a overexpression in vivo results in autism-like behaviors similar to Ube3a 2X Tg mice

Ube3a 2X Tg mice display the hallmark ASD behaviors of impaired socialization, increased restrictive and repetitive behaviors, and decreased communicative vocalization (Smith et al., 2011). At the cellular level, we found that astrocyte-specific expression of AAV-gfapUbe3a in WT mice displayed an immature dendritic spine phenotype with decreased dendritic spine density and increased filopodia, which recapitulated the cellular defects in the Ube3a 2X Tg mice. If the cellular alterations constitute the pathobiological events in ASD, we wonder whether an astrocyte-specific increase in Ube3a gene dosage also confers autistic-like features at the behavioral level. To this end, AAV-gfapUbe3a or AAV-gfapGFP were intracerebroventricularly (ICV) injected in P1 WT mice, and AAV-gfapGFP was injected in 2X Tg littermate mice at P1. Mice were subjected to different behavioral tests at P30. We first evaluated sociability using a standard three-chamber method by analyzing time spent in each chamber when a stranger mouse, S1, was placed in a side chamber of the apparatus (Fig. 6A). We found that the WT+AAV-gfapUbe3a mice preferentially interacted with stranger mouse 1 (S1) to a degree similar to the WT+AAV-gfapGFP mice, whereas 2X Tg mice displayed diminished preference for socialization (Fig. 6C). After acclimating to S1 presence, a new stranger mouse, S2, was introduced on the opposite side chamber to assess social novelty (Fig. 6B). We found that the preference for social novelty was reduced in WT + AAV-gfapUbe3a mice as the 2X Tg mice when compared to WT + AAV-gfapGFP control mice (Fig. 6D). These findings indicate that astrocyte-specific expression of Ube3a results in abnormalities in social novelty with minimal effect on sociability.

Figure 6. Mice overexpressing Ube3a in astrocytes display autistic-like behaviors.

WT mice were ventricularly injected with viruses at P1, and behavior assays were performed at P30. (A) Representative path traces of P30 WT + AAV-gfapGFP, Ube3a 2X Tg +AAV-gfapGFP, and WT + AAV-gfapUbe3a in three-chamber test for socialization with stranger mouse (S1; denoted by “*”). (B) Path traces for the same mice with stranger mouse 2 (S2; “#”). (C-D) Social preference index for mice represented in path traces A (F=24.91; ANOVA; sociability: WT + gfapGFP: 1.69 ± 0.23 A.U.; 2X Tg + gfapGFP: 0.69 ± 0.07 A.U.; WT + gfapUbe3a: 2.07 ± 0.13 A.U.) and B (F=29.50 social novelty: WT + gfapGFP: 3.29 ± 0.39 A.U.; 2X Tg + gfapGFP: 1.72 ± 0.19 A.U.; WT + gfapUbe3a: 0.70 ± 0.04 A.U.). (E) Number of grooming episodes over the course of a 5 min observation (F=17.27; ANOVA; WT + gfapGFP: 2.40 ± 0.40; 2X Tg + gfapGFP: 7.33 ± 0.42; WT + gfapUbe3a: 6.50 ± 0.85). (F) Time spent on individual grooming episodes (F=25.20; ANOVA; WT + gfapGFP: 25.00 ± 4.46 s; 2X Tg + gfapGFP: 146.17 ± 12.66 s; WT + gfapUbe3a: 111.17 ± 22.91 s). (G) Representative vocalization recordings obtained for each condition depicted as spectrograms. Duration = 0.4 s per representative spectrogram (H-I) Quantification of the number of calls (F=8.21; ANOVA; WT + gfapGFP: 168.80 ± 14.10 calls; 2X Tg + gfapGFP: 131.60 ± 7.38 calls; WT + gfapUbe3a: 90.00 ± 18.18 calls) and the average duration of a call (F=0.01; ANOVA; WT + gfapGFP: 0.01 ± 0.00 s; 2X Tg + gfapGFP: 0.01 ± 0.00 s; WT + gfapUbe3a: 0.01 ± 0.00 s), respectively, from each condition. n= 5–7 mice per condition. All error bars indicate standard error. Statistical significance or lack thereof was determined with one-way ANOVA. *P<0.05. **P<0.01. ****P<0.0001. Not significant (n.s.) P>0.05.

Grooming is commonly used to test restrictive and repetitive behaviors in mice. Accordingly, ICV-injected mice were recorded in a familiar environment for detection of grooming behavior. Compared to the WT + AAV-gfapGFP littermate mice, WT + AAV-gfapUbe3a mice displayed an increased frequency of grooming behaviors, similar to 2X Tg + AAV-gfapGFP mice (Fig. 6E). WT + AAV-gfapUbe3a and 2X Tg mice also exhibited a longer duration of grooming episodes when compared to their control littermates (Fig. 6F). These results indicate that astrocyte-specific overexpression of Ube3a is able to cause restrictive and repetitive behaviors.

Abnormal communicative vocalization is another behavioral hallmark in ASD. We thus recorded ultrasonic vocalizations from the virus injected mice in the presence of a stranger mouse. Recordings were transformed to spectrograms for each mouse (Fig. 6G) which were then analyzed for the number of calls (Fig. 6H) and the duration of each call (Fig. 6I). No significant change in the average duration of calls was found when we compared WT+ AAV-gfapGFP, WT + AAV-gfapUbe3a, or 2X Tg + AAV-gfapGFP littermate mice (Fig. 6I). However, there was a significant reduction in the number of calls for both the 2X Tg mice and WT + AAVgfapUbe3a mice when compared to the WT + AAV-gfapGFP control littermate mice (Fig. 6H), indicating that astrocytic overexpression of Ube3a leads to a reduction in communicative vocalization in mice. Collectively, these data demonstrate that astrocyte-specific overexpression of the Ube3a gene in the brain is sufficient to trigger the neural dysregulation leading to the expression of the autistic-like behavioral phenotypes.

Discussion

In this study, our findings show a prominent astrocytic influence over neuronal and behavioral phenotypes present in a monogenic ASD caused by increased gene dosage of Ube3a. We find that selective neuronal overexpression of Ube3a has little effect on the structural maturity of dendritic spines, yet astrocyte-specific expression of Ube3a hinders dendritic spine maturation in vitro and in vivo, suggesting the involvement of astrocytic mechanisms. The loss of spine maturation was linked to astrocyte-derived spinogenic factor, thrombospondin-2. In astrocytes, Ube3a does not target TSP2 for ubiquitination and proteasomal degradation via E3 ubiquitin ligase activity, rather TSP2 gene transcription is suppressed, leading to a reduction in both cytosolic and released TSP2. Consistently, the defect in dendritic spine maturation is also observed when TSP2 is immunodepleted from astrocyte conditioned media. In support of TSP2’s necessity, we demonstrate that supplementing astrocyte conditioned media of 2X Tg with recombinant TSP2 protein blocked the impaired dendritic spine maturation phenotype. Importantly, mice infected with AAV-gfapUbe3a to selectively overexpress Ube3a in astrocytes displayed autistic-like behaviors to a similar degree as Ube3a 2X Tg mice. We noted that, while increased astrocytic gene dosage of Ube3a partially recapitulates autistic-like behaviors in vivo, not all of the behavioral phenotypes, namely defective sociability, are evident using this approach, which may suggest an involvement of multiple cell types (i.e., both glia and neurons) in the overall pathology in Ube3a-dependent ASD. For instance, neuronal Ube3a overexpression leads to simplification of dendritic arborization via caspase-3 mediated cleavage and subsequent destabilization of microtubules in vitro (Khatri et al., 2018). This is a phenotype that was not observed with astrocytic Ube3a overexpression, indicating its neuronal autonomy. Additionally, it was discovered that driving expression of Ube3a in excitatory neurons alone in mice did not appear to recapitulate autistic social, communicative, or restrictive behaviors, further suggesting that Ube3a-dependent ASD may not be neuron cell autonomous (Copping et al., 2017). CaMKII-promoted expression of Ube3a isoform II (camkiiUbe3a_iso2) in excitatory neurons of mice led to autism-associated comorbidity behaviors such as heightened anxiety, yet failed to recapitulate previously observed hallmark ASD-like behaviors, further indicating that the clinical phenotypes of ASD do not rely strictly on neuronal overexpression of Ube3a (Copping et al., 2017). In support of our findings, elevated expression of the Drosophila homologue of Ube3a, DUbe3a, in a glia-specific manner was found to be sufficient to confer a startle-induced seizure phenotype likely due to a loss of the N⁺/K⁺ ATPα (Hope et al., 2017).

When localized to the nucleus, Ube3a has been shown to function with a number of transcription factors such as the nuclear hormone receptors, thus influencing transcriptional networks as a transcriptional coregulator (Furumai et al., 2019; Nawaz et al., 1999b). Because we found no sign of direct ubiquitination of TSP2 by Ube3a, we suggest that Ube3a may act as a transcriptional cofactor, modulating TSP2 gene transcription. Ube3a binding partners implicated in TSP2 transcriptional control could be identified by future experiments including ChIP-seq, electrophoretic mobility shift assay (EMSA), or protein binding microarrays (PBM) assays.

This study suggests an important role for glia-released molecules such as TSP2 in mediating the expression of autistic cellular and behavioral phenotypes. In line with our findings, it has been shown that mice lacking the astrocytic protein glypican 4 (Gpc4) display lessened glutamatergic neurotransmission along with age-specific behavioral deficits akin to autistic-like behavior (Dowling and Allen, 2018). Single nucleotide polymorphisms (SNPs) identified in another glypican-family protein, GPC6, have been identified in an East-Asian genome-wide association study (GWAS) of ASD families (Liu et al., 2016). Furthermore, double knockout mice for astrocyte-derived ligands, ephrin-A2/A3, but not the neuronal Eph receptors, displayed hallmark autistic behaviors such as hampered social interaction and extensive, self-injurious grooming as a readout for restrictive and repetitive behavior (Wurzman et al., 2015). In addition to TSP2, there may be other glia-released proteins and peptides that influence the observed phenotypes in Ube3a 2X Tg mice. Future studies involving differential proteomic analysis on conditioned media of primary glia from WT vs. 2X Tg animals could provide a more systematic screening of the potential candidates. Furthermore, an astrocyte-specific knockout of Fragile X Syndrome gene, Fmr1, resulted in cellular phenotypes typical in FXS, including increased spine formation and upregulated protein synthesis (Higashimori et al., 2016), with the glial effect being mediated at least in part via astrocyte secreted protein TSP1 (Cheng et al., 2016). Additionally, reactive glia and extensive gliosis concomitant with glial-mediated neuroinflammation have been identified in human ASD patients (Edmonson et al., 2014; Morgan et al., 2012, 2010; Pardo et al., 2005; Rodriguez and Kern, 2011; Vargas et al., 2005). Coincident with aberrant spine/synapse formation, irregularities in glial cell density and glial functions have also been found in ASD (M. H. Chen et al., 2015; Falcone et al., 2021; Hodges et al., 2017; Wegiel et al., 2018).

Our findings demonstrate that a high dosage of Ube3a in astrocytes leads to a reduced availability of TSP2 in the extracellular space. The weakened signaling via the TSP2 receptor α2δ−1 and the downstream effector Rac1 is known to cause dysregulation in actin reorganization and hinders dendritic spine formation and structural maturation (Eroglu et al., 2009; Risher et al., 2018; Scott et al., 2003; Stenina-Adognravi, 2014; Tashiro et al., 2000). Consistently, we found that astrocytic overexpression of Ube3a resulted in an increase in immature filopodia density and decreased levels in mature spines and total protrusions overall. These changes indicate neural circuitries of reduced connectivity and weakened synaptic efficacy, as previously shown in 2X Tg mice (Smith et al., 2011). Thus, TSP2 may hold potential as a therapeutic agent for Ube3a-dependent ASD (Christopherson et al., 2005). Although the full coding sequence of TSP2 exceeds most therapeutic avenues, a previous study has demonstrated that three mid-sequence epidermal growth factor (EGF) like repeats of TSP2 induce synaptogenesis comparable to full length TSP2 (Eroglu et al., 2009). Further, as synaptopathy and spinopathy have been recognized as shared cellular bases in ASD, supplement of synaptogenic and spinogenic molecules such as TSP2 may offer promise as a general therapeutic approach (Dolan et al., 2013; Dong et al., 2020; Gilbert et al., 2020; Landi et al., 2011).

Limitations of the study

Intracerebroventricular injection of gfapUbe3a virus was performed in P1 mice, which limits the ability to evaluate the impact of astrocyte-selective Ube3a overexpression on prenatal brain development and its influence on autistic behaviors. Additionally, the virus expression may not be uniform throughout the entire brain, and the overexpression levels across brain regions may not accurately reflect the physiological pattern of Ube3a over-dosage in astrocytes. Therefore, transgenic mice with additional copies of the Ube3a gene that are specifically expressed in astrocytes will be a better model to study the role of astrocytic Ube3a overexpression in autism development.

The Ube3a 2X Tg mice express the C-terminal FLAG-tagged Ube3a protein in addition to the endogenous Ube3a. Recent studies have suggested that the extension of the C-terminus of the Ube3a protein can negatively impact its ubiquitin E3 ligase activity (Al-Maawali et al., 2013; Kühnle et al., 2013; Salvat et al., 2004). Therefore, the C-terminal FLAG-tagged Ube3a protein expressed in the Ube3a 2X Tg mice may be ligase defective. On the other hand, Ube3a is highly enriched in the nucleus of neurons, where its primary function involves gene regulation as a transcription coactivator. In fact, the most abundantly expressed Ube3a isoforms in mouse and human, which constitute approximately 80% of the total Ube3a protein in the brain, are mainly nuclear (Bossuyt et al., 2021). It has been shown that the coactivation function of Ube3a is independent of its E3 ubiquitin ligase activity, as both the E3-ligase-dead mutant of Ube3a and the C-terminal deletion mutant of Ube3a retain normal coactivation function (Nawaz et al., 1999a, 1999b; Ramamoorthy et al., 2010; Ramamoorthy and Nawaz, 2008). Therefore, the main function of Ube3a as a transcription coactivator in the nucleus should remain intact in the Ube3a 2X Tg mice. In our study, TSP2 is transcriptionally regulated by Ube3a in glia, and this regulation is expected to be independent of the E3 ligase function of Ube3a.

Materials and Methods

Antibodies

All antibodies used for Western blotting (WB) were diluted in 1X TBST + 5% milk and preserved with 0.1% sodium azide. All antibodies used for immunofluorescence (IF) were diluted in 1X PBS + 5% goat serum and preserved with 0.1% sodium azide. Antibodies used for immunoprecipitation (IP) were used in the indicated amounts. The following antibodies were used:

Mouse anti-Ube3a (WB: 1:1000; IF: 1:100; Sigma-Aldrich; #E8655). Mouse anti-GFAP (WB: 1:5000; IF: 1:100; NeuroMab/Antibodies Inc.; #75–240). Rabbit anti-GFAP (WB: 1:5000; IF: 1:100; Thermo Fisher Scientific; #PA5–16291). Rabbit anti-Rac1 (WB; 1:0000; Thermo Fisher Scientific; #PA1–091). Rabbit anti-Glypican-4 (WB: 1:10000; Thermo Fisher Scientific; #130481-AP). Mouse anti-p27 (WB: 1:500; IP: 1.0 μg; Santa Cruz Biotechnology; #sc-1641). Rabbit anti-Ubiquitin (WB: 1:1000; Abcam; #ab19247). Mouse anti-α2-δ1 (WB: 1:1000; Abcam; #ab2864) Rabbit anti-Thrombospondin-1 (WB: 1:1000; Abcam; #ab1823). Rabbit anti-Thrombospondin-2 (WB: 1:1000; IF: 1:100; Abcam; #ab84469). Mouse anti-Thrombospondin-2 (WB: 1:500; IP: 1.5 μg; Santa Cruz Biotechnology; #sc-136238). Mouse anti-GAPDH (WB: 1:5000; Millipore-Sigma; #MAB374). Mouse anti-α-tubulin (WB: 1:5000; Sigma-Aldrich; #T9026).

Plasmids and cloning

The following cDNA plasmids were obtained from Addgene: pE6-AP (pUbe3a; #5054) was cloned into pAAV-ReaChR-citrine (#50954) as previously described (Khatri et al., 2018). pAAV-gfapGFP (#50473). pAAV-gfapUbe3a was made in-house by cloning Ube3a out of pUbe3a #5054 using primers designed to include the HindIII and Xba1 restriction sites:

(Ube3a HindIII Reverse: 5’-GCAGGCATGCAAGCTTGTTTTACAGC-3’; Ube3a XbaI Forward: 5’-ACTTCCATCTAGATCAGATCCACCATGTACCC-3’). The isolated insert and pAAV-gfapGFP #50473 were subjected to the indicated restriction endonucleases. The insert was then ligated into the backbone vector and used for AAV preparation (see Adeno-associated virus (AAV) preparation and precipitation below).

Primary rat neuron and astrocyte culture

Hippocampal tissue dissected from embryonic day 18 (E18) rat embryo brains were used for all imaging experiments, whereas cortical tissue was used for all biochemical experiments. Tissues were placed in digestion solution containing 0.5 M ethylenediamine tertraacetic acid (EDTA) pH 7.0, papain (Millipore-Sigma; #P4762), and L-cysteine (Sigma-Aldrich; #C7352) in Dulbecco’s Modified Eagle Media (DMEM; Corning Cellgro; #10–017) for 20 min at 37°C. Tissues were gently dissociated in DMEM containing 1% bovine serum albumin (Sigma-Aldrich; #A2153) and 0.1% DNase (Life Technologies; #18068–015). Suspensions were gently spun down and moved to plating media [10% fetal bovine serum (Peak Serum; PS-FB2), heat inactivated horse serum (Atlanta Biologicals; #S12150), 260 μM L-cysteine, 1% penicillin/streptomycin (Corning Cellgro; #30–002-CI), 2 mM L-glutamine (Corning Cellgro; #25–005-CI) in DMEM]. Cells were counted and then plated on 18 mm circular coverslips (Carolina Biological Supply; #633013) in 60 mm dishes (5 coverslips/dish) that have been coated in 100 mg/ml poly-L-lysine (Sigma-Aldrich; P2636) borate buffer for at least 12 hours. Plating media was replaced with feeding media [horse serum (1%), penicillin streptomycin (1%), L-glutamine (1%), and 2% Neurocult SM1 (STEMCELL Technologies; #05711) in 500 ml Neurobasal media (Thermo Fisher Scientific; #21103–049)] the next day. Unless otherwise stated, glial growth was inhibited at DIV 7 by addition of 10 μM 5’-fluoro-2’-deoxyuridine (FDU; Sigma-Aldrich; #F0503).

Drugs and treatments

Recombinant thrombospondin-2 protein (R&D Systems; #1635-T2) was used at 10 nM (approximately 5 μg/mL) in sterile 1X PBS as established in a previous study (Christopherson et al., 2005). If culturing astrocytes alone, the above protocol was followed minus the use of poly-L-lysine for neuronal adherence. Microglial growth was ablated by supplementing with 10 μM PLX-3397/Pexidartinib (Active Biochem; #A-1330) at DIV 4 through DIV 15, every 2–3 days. Ablation of astrocytes was done by treating with 5 μM FDU at DIV 4 and continuing with 10 μM FDU treatment every 3–4 days.

Primary mouse astrocyte culture

Primary astrocytes were cultured from the brains of P1 mice. Digestion solution contained 10 μM EDTA, 33 μL papain, 16.6 μL L-cysteine, and 440 μL DMEM per brain. Trituration solution contained 62.5 μL BSA, 500 μL 1% DNase, and 437 μL DMEM per brain. Brains were digested and dissociated as described above. Dissociated cells were resuspended in 500 μL of plating media. Plating occurred as described above minus the use of poly-L-lysine in 6-well plates or 60 mm dishes. Plating media was replaced with glial media (feeding media + 10% horse serum) the next day. Microglial growth was ablated as described above.

Neuron transfection

Primary neurons were transfected at DIV 10 using Lipofectamine 2000 (Thermo Fisher Scientific; #11668019) at a 1:1 ratio with indicated plasmids at 1 mg/ml. Plasmids and lipofectamine were separately incubated in DMEM for 5 min at room temperature and then combined to incubate for 20 min at room temperature. The transfectant solution was then added to the appropriate coverslip and allowed to incubate for 4–6 hours at 37°C, after which transfected media was replaced with conditioned neuronal media. Neurons were given at least 48 hours to express the plasmid of interest.

Astrocyte conditioned media

Astrocyte conditioned media (ACM) was collected under sterile conditions. Briefly, astrocyte cultures of the indicated condition were brought to DIV 12. Culture media was replaced with fresh Feeding media (see “Primary rat neuron and astrocyte culture”), which was allowed to condition for 48–72 hours prior to collection. ACM was then used as a total replacement for neuronal culture feeding media.

Adeno-associated virus (AAV) preparation and precipitation

All viruses were prepared by transfection of HEK 293T cells at 70% confluency with the plasmid of interest (see “Plasmids”) along with the viral packing and envelop protein-coding plasmids pXX.680 and p50-Cap9 in polyethylenimine (PEI; Polysciences; #23966), as described in our previous study (Khatri et al., 2018). PEI was used at a 3:1 ratio (v/v) with plasmids. Plasmids (1 mg/ml) and PEI (1 mg/ml) were incubated in serum-free DMEM at room temperature for 5 min before being combined and incubating for 20 min at room temperature. The following day, 10 ml of fresh HEK media was added and supplemented with sodium citrate to reach a final concentration of 1 mM along with FDU at a final concentration of 10 nM. Media saturated with virus was sterilely collected 3 days after transfection and passed through a 0.45 μm filter. Cells were collected, lysed, and centrifuged, after which supernatant was passed through a 0.45 μm filter and combined with the viral media. Viral particles were precipitated using PEG-it (Systems Biosciences; #LV-825A) overnight at 4°C directly in the virus-containing media/homogenate. The mixture was subsequently centrifuged at 1500 g for 30 min. The viral pellet was diluted appropriately in sterile 1X PBS. Cultures were infected with desired viruses at DIV 2. At least 10 days was allowed for expression in vitro.

qRT-PCR

Astrocytes cultured for quantitative reverse transcriptase polymerase chain reaction (qRT-PCR) were collected and subjected to mRNA extraction using RNeasy Mini Kit (Qiagen; #74104) according to the manufacturer protocol. Concentration of mRNA was determined using a Nanodrop One/One^C Microvolume Spectrophotometer. mRNA (200 ng) was then used to synthesize cDNA using HiFiScript cDNA Synthesis Kit according to the manufacturer protocol (CoWin Biosciences; #2569) in a 25 μL reaction. qRT-PCR was performed using primers designed against TSP2 and β-actin. TSP2 forward: 5’- CCTCAACTACTGGGTAGAAGGC-3’; TSP2 reverse: 5’- TGACACTGTCGATAAGATCGCA-3’; β-actin forward: 5’AGTGTGACGTTGACATCCGTA-3’; β-actin reverse: 5’- GCCAGAGCAGTAATCTCCTTCT-3’. TSP2 and β-actin primers were validated in RT-PCR reactions prior to qRT-PCR, generating amplicons of 146 bp each. For qRT-PCR, 10 μL reactions were prepared in a 384-well plate in biological and technical triplicate using to CWBio UltraSYBR PCR Master Mix (CWBiosciences; #CW2602), cDNA, and nuclease free water. The plate was then loaded into an ABI 7900HT qPCR machine belonging to the Boston University Genomics Core. Threshold for detection was set to 0.85 for all samples. Following amplification, log₂ΔΔC_t was determined using standard quantifications in Excel.

Western blot

Cultured cells were collected in 0.1% SDS RIPA buffer [50 mM Tris HCl pH 7.4, 15 mM sodium chloride, 1% sodium deoxycholate, 1% NP-40, 0.1% sodium dodecyl sulfate]. Protease inhibitors in the form of Roche cOmplete EDTA-free Protease Inhibitor Cocktail tablets (1 tablet per 50 mL) were added to all lysis buffers (Sigma-Aldrich; #11873580001). Protein concentration was determined using a Pierce BCA Protein Assay Kit according to the manufacturer protocol (Thermo Fisher Scientific; #23225) and normalized by diluting appropriately with lysis buffer. Samples were diluted 1:1 with Laemmli 2X Sample Reducing Buffer [4% sodium dodecyl sulfate, 10% β-mercaptoethanol, 20% glycerol, 0.05% bromophenol blue, and 125 mM Tris-HCl]. Samples were boiled for 15 min at 95°C and then cooled on ice for 10 min. Samples were briefly vortexed before loading for SDS-PAGE.

Brains were collected immediately after sacrificing animals at the indicated time points. Brain tissue was homogenized in 1% SDS RIPA buffer [50 mM Tris HCl pH 7.4, 15 mM sodium chloride, 1% sodium deoxycholate, 1% NP-40, 1% sodium dodecyl sulfate + 1 cOmplete EDTA-free Protease Inhibitor Cocktail tablet]. Manual homogenization was performed using a pestle followed by sonication. Samples were then placed on a rotator arm for 1 hour at 4°C. homogenates were centrifuged for at 4°C for 30 min at 13,000 rpm in a microcentrifuge. Supernatants were then subjected to BCA assay for protein concentration, which was normalized across samples by diluting with lysis buffer. Samples were diluted 1:1 with 2X Laemmli Sample Reducing Buffer and boiled for 15 min at 95°C. Samples were cooled on ice for at least 10 min, briefly vortexed, and used for SDS-PAGE.

Proteins of interest were separated via SDS polyacrylamide gel electrophoresis using standard protocol. Proteins were transferred to methanol-activated Immun-Blot PVDF membranes (Bio-Rad; #1620177). Membranes were blocked in a solution of 5% milk in 1X TBS + 0.1% Tween-20 (TBST; Fisher Scientific; #BP-337). The indicated primary antibody (see “Antibodies”) was allowed to incubate on the membrane 1–2 hours at room temperature or overnight at 4°C. Membranes were washed 3 times in 1X TBST and then placed in a secondary antibody solution (5% milk, 1X TBST, 1:5000 secondary antibody; see “Antibodies”) for 1 hour at room temperature. Membranes were washed again 3 times with 1X TBST and developed using ECL Prime Western Blotting Detection Reagent (GE Healthcare; #RPN2236). Bands were detected on FUJI Super RX-N Full Speed Blue film (Radiology; #PPB5080) and analyzed for intensity using Image-J software.

Immunoprecipitation and ubiquitin assay

Protein samples were collected and normalized for protein content using a BCA assay (see “Western blotting”). Inputs from each sample were obtained before adding pull down antibody. Briefly, the desired sample was diluted with the appropriate lysis buffer (0.1% SDS RIPA for standard IP, 1% SDS RIPA for ubiquitin assay). The indicated pull-down antibody (1 μg) was then added and the sample was allowed to gently agitate for 1 hour at 4°C. Protein-A agarose beads (Santa Cruz Biotechnology; #sc-2001) were washed in the appropriate lysis buffer and then added to the antibody-containing sample. The sample was then gently agitated overnight at 4°C. Samples were washed 5 times before adding Laemmli 2X Sample Reducing buffer to the beads directly. Samples were boiled at 95°C for 15 min and cooled on ice for at least 10 min before being used for SDS-PAGE.

Immunodepletion and immunoprecipitation of conditioned media

All procedures, solutions and conditions were performed in a biosafety cabinet to maintain sterility of the conditioned media. Conditioned media was immunodepleted by sterilely collecting media and adding 1.5 μg of the desired antibody. The ACM was then gently agitated at 4°C for 1 hour. Protein-A agarose beads were washed in sterile 1X PBS and added to the ACM. The ACM was allowed to gently agitate overnight at 4°C before being centrifuged at very low speed to remove the agarose beads from the media. Immunodepleted media was then used to completely replace culture media and was used consistently for the duration of that time interval. Agarose beads were boiled in 50 μL of 2X Sample Reducing buffer for 15 min at 95°C and then subjected to SDS-PAGE and immunoblotting for media IPs.

Immunocytochemistry

Primary cultures were placed on ice and fixed in 4% paraformaldehyde/4% sucrose/1X PBS. Cells were then washed in cold 1X PBS. Cell membranes were permeabilized using a 0.3% Triton X-100 (Fisher Scientific; #bp151–100) in 1X PBS, rinsed twice more in PBS, then placed in a blocking solution of 5% goat serum (Fisher Scientific; #16–210-064) in 1X PBS for 1 hour. Cells were then incubated with primary antibodies (in 5% goat serum PBS) for 1–2 hours at room temperature or overnight at 4°C. Cells were then washed and fluorescently labeled using Alexa Fluor-conjugated fluorescent secondary antibodies (1:500, see “Antibodies”) in 5% goat serum in PBS for 1 hour at room temperature. Coverslips were then washed in cold 1X PBS and mounted on microscopy slides using Prolong Gold Antifade Reagent (Thermo Fisher Scientific; P36930). Slides were allowed to dry overnight at room temperature in the dark and subsequently stored at 4°C.

Immunohistochemistry of brain sections

Mice were collected at the indicated time points. Brains were removed after anesthetizing in 4% CO₂ as described previously. Brains were immersed in 4% paraformaldehyde for 4–6 hours at 4°C in the dark and subsequently cryoprotected and dehydrated in 30% sucrose in PBS at 4°C overnight. Brains were rinsed briefly in PBS and immersed in Optimal Cutting Temperature (O.C.T.) Compound (Tissue-Tek; #4583). Sections at 35 μm were obtained using a LEICA Biosystems CM1850 cryostat and mounted onto SuperFrost Plus and ColorFrost Plus microscope slides (Fisher Scientific; #12–550-15). Sections were rinsed in PBS twice for 30 min and then subjected to antigen retrieval using a 10 mM sodium citrate buffer [pH 6.0 + 0.05% Tween-20]. Sections were permeabilized and blocked using 5% goat serum/0.3% Triton X100/1X PBS buffer for 1 hour at room temperature. Primary antibodies were prepared at 1:100 dilutions in 5% goat serum/1X PBS and left to incubate overnight at 4°C in a humidity chamber. Sections were washed in PBS the next day, and fluorescently labeled with appropriate Alexa Fluor-conjugated secondary antibodies at 1:250 dilutions in 5% goat serum/1X PBS (see “Antibodies”). Sections were washed a single time in 1X PBS containing 1:5000 Hoechst 33352 (Thermo Fisher Scientific; #62249). Sections were then washed twice more in 1X PBS before being mounted using ProLong Gold Antifade Reagent. Slides were allowed to dry overnight at room temperature in the dark and subsequently stored at −20°C.

Golgi impregnation

Whole brains from test subject mice at the indicated ages were subjected to Golgi-Cox staining using FD GolgiStain Kit (FD Neurotechnologies; #PK401A) according to the provided manufacturers protocol. Briefly, mice were culled in a 4% CO₂ chamber. Brains were collected and rinsed in cold 1X PBS and then in MilliQ water. Brains were initially immersed in “A/B” solution [potassium dichromate, potassium chromate, and mercuric chloride], which was replaced with fresh A/B solution 12–24 hours after initial immersion. Brains were left at room temperature in the dark for 2–3 weeks. Brains were transferred to solution “C” for 24 hours, followed by fresh Solution C for 2–5 days. Cryosections were taken by rapidly freezing the tissue and sections of 100 μm were taken using a LEICA Biosystems CM1850 cryostat. Sections were mounted in Solution C on gelatin coated slides (FD NeuroTechnologies; #PO101) before allowing sections to dry overnight. Sections were rinsed with distilled water and stained using Solution D/E + distilled water (at a ratio of 1:1:2). Sections were rinsed again in distilled water post-staining. Sections were dehydrated in solutions of 50%, 75%, 95% and 100% ethanol. Sections were then cleared using xylene solution and mounted under coverslips using Fisher Chemical Permount Mounting Medium (Thermo Fisher Scientific; #SP15–100).

Neurons were imaged under 40X objective without oil as Z-stacks with steps of 2–4 μm to ensure acquisition of dendritic shaft images. Dendritic spines were measured using Image-J and classified by morphological parameters (Risher et al., 2014).

Microscopy

Primary neurons and dendritic spine images (fluorescent) were captured using an inverted fluorescent microscope (Carl Zeiss Axiovert) under a 40X oil immersion objective (numerical aperture 1.3) using AxioVision Release 4.5 software. Image analysis was performed using ImageJ (see “Image analysis”).

Golgi images were acquired using a Carl Zeiss Axiovert 200M microscope with AxioVision Release 4.5 software. Neurons were scanned under a 40X magnification air lens (numerical aperture 1.3) by varying depth of the Z-plane to capture fully intact cellular structures. Spines and protrusions were measured using ImageJ software.

Image analysis

Images were acquired by adjusting exposure time so all fluorescent signals for all images were in a full dynamic range using glow scale look-up tables or software histograms to prevent oversaturation. Images were analyzed using Image-J software. Images were converted to 8-bit gray scale images and used for raw data analysis. Fluorescence intensity was measured after applying an appropriate threshold to the image and subtracting background fluorescence.

Dendritic spines and filopodia from Golgi and fluorescent images were measured individually and then classified by type according to established morphological parameters (Risher et al., 2014).

Animal models

Model mice used in this study were created previously (Smith et al., 2011) and obtained from Jackson Laboratory (#019730). Briefly, the full intron-exon sequence Ube3a was inserted to mouse chromosome 7 using BAC recombineering to generate the FVB/Nj-Tg(Ube3a)1Mpan/J mouse, more commonly referred to as Ube3a 1X Tg. Wild type (WT) and coisogenic Ube3a 2X Tg were created by breeding Ube3a 1X Tg males and females. Littermates of the WT and 2X Tg genotypes were used for all experiments. Males and females were used for all experiments.

Animal care and use

Experiments and procedures requiring the use of animals were performed according to Boston University’s policies set forth by the Institutional Animal Care and Use Committee (IACUC). Mouse colonies were maintained on the Boston University Laboratory Animal Care Facility on the Charles River Campus. Both males and females were used for the experiments. All mice used for experiments were at the indicated ages.

Genotyping

Individual tail snips were collected from mice at appropriate time points for respective experiments (P21 unless otherwise indicated). Tissues were digested in 25 mM NaOH, 0.2 M EDTA, pH 12 and incubated at 95°C for 30 min for genomic DNA extraction. Neutralizing solution (40 mM Tris-HCl, pH 5.0) was then added at 1:1 with digestion solution. DNA concentrations were obtained using Nanodrop One/One^C Microvolume Spectrophotometer (Thermo Fisher Scientific; #ND-ONE-W). 200 ng of DNA was used per genotyping reaction using the PrimeStar HS DNA Polymerase system (Clonetech; #R010A). Touchdown PCR was performed via the following conditions: initial hold at 94°C for 2 min, 10 cycles of a 94°C/15 sec denaturing step at 94°C for 15 sec, annealing at 65°C for 15 sec, and extension at 68°C or 10 sec in a touchdown PCR protocol, with consecutive cycles decreasing annealing temperature by 0.5°C. This was followed by 25 cycles of a denaturing step at 94°C for 15 sec, annealing at 60°C for 15 sec, and extension at 72°C for 10 sec. The final hold was at 72°C for 2 min, and sample temperatures were lowered to 10°C and held indefinitely. Full reaction volumes were mixed with 6X Loading Dye containing no SDS (New England Biolabs; #B7025S) and ran on a 1% agarose gel + ethidium bromide and imaged for relative intensity of amplicon bands using ultraviolet light.

All genotyping reactions contained two sets of primers: Ube3a transgene forward 5’-CTT GTT CAC TGA TTC ACG CG-3’. Ube3a transgene reverse 5’-GAT CAA GAC CCA TTT GCA GC-3’. Internal positive control 5’-CAC GTG GGC TCC AGC ATT-3’. Internal positive control reverse 5’-TCA CCA GTC ATT TCT GCC TTT-3’. Transgene amplicons were at 149 bp, internal positive control amplicons were at 74 bp.

Intracerebroventricular injection

Mice at P1 were infected with the indicated viruses mixed with 1% Fast Green Dye after genotyping was confirmed. A guiding injection opening was made bilaterally using a 31 g syringe superior to the location of the ventricle (40% of the total distance between the λ-juncture and the ipsilateral eye). A 30 g cannula was then guided through. Tubing connected to a 10 μL Hamilton syringe allowed for the injection of the indicated virus at a rate of 500 nL/min for a total volume of 1 μL. Injected mice were replaced with their respective mothers and allowed to develop until collected for behavior and Golgi.

Three-chamber social test

A three chambered box [65 × 28 × 28 cm] was constructed from 1.9 cm thick white plastic board. Center partitions had a single 4 × 4 inch opening to allow for free movement chambers. Left and right chambers contained wire cages used to contain S1 (right) and S2 (left) mice. Test subject mice were allowed to habituate to the entire apparatus for 5 min a day over 3 days. Mice were placed individually into the center chamber on testing days with a stranger mouse (S1) placed under the right wire cage. Test mice were allowed to move freely for 5 min. Test mice were re-confined to the center chamber while a second stranger mouse/novel mouse (S2) was placed in the left chamber’s wire cage. Test mice were again allowed to roam freely for 5 min. The test surface, chamber, partitions, and wire cages were sanitized with 10% bleach and 70% ethanol between each test mouse to eliminate odorants. Videos were recorded using a Logitech c920 webcam. Time spent in each chamber and interacting with stranger mice was analyzed using TrackMo (https://github.com/zudi-lin/tracking_toolbox).

Ultrasonic vocalization (USV) recording

Test subject mice were individually placed into a recording chamber at random. Mice were allowed to habituate to the recording chamber for 5 min before a 5 min total recording within sight of a stranger mouse of WT genotype. Vocalizations were recorded using a CM16/CMPA microphone (Avisoft Bioacoustics) situated at the top of the chamber (15 × 15 × 15 cm) and connected to a preamplifier Ultra-SoundGate 116Hb (Avisoft Bioacoustics). Sampling rate for recordings was set at 300 kHz in 16-bit formatting. Digitized sonograms were converted to spectrograms using fast Fourier transform according to the following settings: 256 FFT length, 100% frame, FlatTop window, 50% window overlap. Background noise under 30 kHz was reduced by applying a high-pass filter. USVs were automatically detected (criteria: −47 ± 10 dB, hold time 7 ms) and legitimized by manual inspection. The recording chamber was sanitized with 10% bleach and 70% ethanol between each test subject mouse.

Repetitive grooming

Test subject mice were placed into fresh cages 24 hours prior to recording. Recordings were performed at random for each test mouse and analyzed over the course of 5 min between the 25–30 min total recording. Recordings were captured using a Logitech c920 webcam and manually analyzed for grooming behaviors [licking and nibbling of the tail, anal region, genital region, paw, and chest, face washing].

Statistical analysis

Data generated from over three independent experiments were averaged to obtain a mean, with multiple means of identical conditions averaged for analysis of parametric data. Two-tail Student’s t-test, one-way ANOVA, or two-way ANOVA were used where appropriate. Statistical analyses with p-values < 0.05 were considered statistically significant. All data presented are represented as means ± standard error of the mean (SEM).

Significance statement.

Increased gene dosage of Ube3a is tied to autism spectrum disorders (ASDs), yet cellular and molecular alterations underlying autistic phenotypes remain unclear. We show that Ube3a overexpression leads to impaired dendritic spine maturation, resulting in reduced spine density and increased filopodia density. We find that dysregulation of spine development is not neuron autonomous, rather, it is mediated by an astrocytic mechanism. Increased gene dosage of Ube3a in astrocytes leads to reduced production of the spinogenic glycoprotein thrombospondin-2 (TSP2), leading to abnormalities in spines. Astrocyte-specific Ube3a overexpression in the brain in vivo confers dysregulated spine maturation concomitant with autistic-like behaviors in mice. These findings indicate the importance of astrocytes in aberrant neurodevelopment and brain function in Ube3a-depdendent ASD.

Highlights.

• Ube3a overexpression leads to a suppression in spine formation

• The Ube3a effect on spinogenesis in neurons requires astrocytes

• Media of Ube3a-overexpressing astrocytes result in spine reduction

• Heightened Ube3a dosage in astrocytes reduces TSP2 production and release