Abstract
Angelman syndrome (AS) is a rare neurodevelopmental disorder caused by loss of expression of the maternal UBE3A allele and is characterized by a constellation of impactful neurologic symptoms. While previous work has uncovered outsized contributions of GABAergic neuron-selective Ube3a deletion to seizure susceptibility and electroencephalography (EEG) phenotypes in a mouse model of AS, the neuronal populations governing a broader range of behaviors have not been studied. Here, we used male and female mice to test the consequences of Ube3a deletion from GABAergic or glutamatergic neurons across a well-characterized battery of AS-relevant behaviors. Surprisingly, we observed deficits in numerous motor and innate behaviors in mice with glutamatergic Ube3a deletion and relatively few consequences of GABAergic Ube3a deletion. Furthermore, genetic Ube3a reinstatement in glutamatergic neurons rescued multiple motor and innate behaviors. When tested for sleep–wake behaviors, the selective loss of Ube3a from glutamatergic neurons disrupted sleep similarly to that of AS model mice (Ube3am–/p+), and glutamatergic Ube3a reinstatement overcame the lack of active cycle “siesta” and decreased REM phenotypes observed in AS model mice. Altogether, this work demonstrates a major role of glutamatergic neuron UBE3A loss in mediating multiple AS behavioral features, suggesting a divergence from the circuitry underlying enhanced seizure susceptibility. Our findings imply that neuronal cell type-agnostic UBE3A reinstatement is likely required for successful AS genetic therapies—with reinstatement of UBE3A in GABAergic neurons necessary for overcoming epileptic and EEG phenotypes, and reinstatement in glutamatergic neurons necessary for overcoming most other behavioral phenotypes.
Significance Statement
Angelman syndrome (AS), a severe neurodevelopmental disorder caused by loss of neuronal UBE3A, is characterized by symptoms such as motor impairment, lack of speech, seizures, and disrupted sleep. While clinical trials aim to restore UBE3A in AS individuals, the neuronal populations responsible for key symptoms remain unclear. Using an AS mouse model, we identify a key role for excitatory neuron loss of UBE3A in motor, innate behavior, and sleep phenotypes, distinct from the previously described impact of inhibitory neuron loss of UBE3A on seizure and electroencephalography phenotypes. These data improve our understanding of the mechanisms by which UBE3A loss leads to symptoms, potentially guiding future therapies.
Introduction
Many core symptoms of autism spectrum disorder (ASD) and other neurodevelopmental disorders are posited to arise from disruptions in the interplay between excitatory and inhibitory neurons, deemed an “excitatory–inhibitory (E–I) imbalance” (Rubenstein and Merzenich, 2003). While this hypothesis can oversimplify complex network deficits, it has provided a useful framework to study how neuronal populations interact in the neurotypical brain and animal models of disease (Nelson and Valakh, 2015; Sohal and Rubenstein, 2019). One strategy to understand how this interplay between excitatory and inhibitory neuronal populations causes symptoms pertinent to neurodevelopmental disorders is to selectively perturb the expression of key risk genes in these broad neuronal populations and then study the resulting effects on behavior. Indeed, selective deletion of neurodevelopmental disorder-linked genes in excitatory versus inhibitory neurons has revealed their divergent contributions to behavioral phenotypes in multiple mouse models of neurodevelopmental disorders, providing insights regarding the mechanisms underlying these symptoms (Chao et al., 2010; Meng et al., 2016; Kim et al., 2024).
Angelman syndrome (AS), a severe neurodevelopmental disorder caused by loss of the maternal allele of E3 ubiquitin ligase UBE3A, is characterized by a constellation of symptoms including recurrent seizures (>90% of individuals), motor impairment, sleep disruption, intellectual disability, and absence of speech (Williams et al., 2010; Thibert et al., 2013; Buiting et al., 2016). Interestingly, mice with Ube3a deletion restricted to GABAergic neurons display seizure susceptibility that is much more severe than that of mice with pan-neuronal loss of Ube3a, suggesting a key role of GABAergic UBE3A loss in seizures (Judson et al., 2016; Gu et al., 2019). GABAergic Ube3a deletion also mediates increased cortical electroencephalography (EEG) power in the delta frequency (2–4 Hz), a phenotype characteristic of AS individuals and model mice (Judson et al., 2016; Sidorov et al., 2017). This finding has particular clinical relevance, as EEG delta power robustly correlates with the severity of several AS symptoms, implicating this phenotype as a potential biomarker for measuring the efficacy of novel treatments (Hipp et al., 2021; Ostrowski et al., 2021; Spencer et al., 2022).
While deletion of Ube3a from GABAergic neurons plays a key role in seizure and EEG phenotypes of AS model mice, the neuronal populations responsible for a broader range of behaviors have not been assessed. Here, we investigated the contributions of GABAergic and glutamatergic neuron UBE3A loss to AS-related behaviors using a well-characterized behavioral battery (Sonzogni et al., 2018; Judson et al., 2021; Syding et al., 2022). Surprisingly, deletion of Ube3a from GABAergic neurons yielded few behavioral deficits, while Ube3a deletion from glutamatergic neurons drove AS-like behaviors in multiple motor and innate behavioral tasks. Accordingly, genetic reinstatement of Ube3a in glutamatergic neurons improved performance in many behavioral tasks, further suggesting UBE3A expression in this neuronal population plays a key role. To test other clinically relevant phenotypes, we used a noninvasive monitoring system to study sleep–wake behavior in mouse models with cell class-specific conditional Ube3a manipulation. These studies yielded similar results, predominantly implicating glutamatergic deletion of Ube3a in the altered sleep–wake behavior of AS mice, but revealing a role of GABAergic Ube3a deletion in a sleep fragmentation phenotype. Taken together, our data support a predominant role of glutamatergic deletion of Ube3a in many of the motor and innate behaviors of AS model mice, contrasting the circuitry mediating seizure and EEG phenotypes. As clinical trials are currently underway to reinstate UBE3A in AS individuals, our findings suggest that gene delivery strategies must effectively target both excitatory and inhibitory neurons, in a balanced manner, for optimal symptom improvement.
Materials and Methods
Animals
All procedures were approved by the Institutional Animal Care and Use Committee of the University of North Carolina at Chapel Hill and were performed in accordance with the guidelines of the US National Institutes of Health. Mice were raised on a 12 h light/dark cycle (lights on at 7 A.M.) and were housed in groups of 2–5 mice per cage with ad libitum access to food (PicoLab 5V5M chow) and water. Mice of both sexes were used in approximately equal genotypic ratios. AS model mice do not display sexually dimorphic phenotypes in the battery of behavioral tests used (Sonzogni et al., 2018)—open field, marble burying, rotarod, and nest building—so data from male and female mice are shown combined for these tests. All experiments and post hoc data collection were performed by experimenters blinded to genotype.
All mice were maintained on a C57BL/6J background. Mice with neuron type-specific deletion of Ube3a were generated by crossing male mice heterozygous for Gad2-IRES-Cre (JAX 028867; Taniguchi et al., 2011) or Vglut2-IRES-Cre (JAX 028863; Vong et al., 2011) to female mice heterozygous (paternal inheritance) or homozygous for the Ube3a-Flox allele (Judson et al., 2016). To generate mice with glutamatergic neuron type-specific reinstatement of Ube3a, male mice heterozygous for Vglut2-IRES-Cre were crossed to females heterozygous (paternal inheritance) for the Ube3a lox-STOP-lox construct (Silva-Santos et al., 2015). Maternal Ube3a-deficient mice (Ube3am−/p+ = AS model mice) were produced by crossing female paternal Ube3a-deficient mice (JAX 016590) to congenic C57BL/6J males. The experimental models used in each figure are summarized in Table 1.
Summary of experimental models used, nomenclature, and controls used for each figure
We genotyped mice using the following polymerase chain reaction (PCR) primers:
Ube3a: Ube3a P1 (5′-ACT TCT CAA GGT AAG CTG AGC TTG C-3′), Ube3a P2 (5′-GCT CAA GGT TGT ATG CCT TGG TGC T-3′), Ube3a P3 (5′-TGC ATC GCA TTG TGT GAG TAG GTG TC-3′)
Ube3aFLOX: Komp 1.2F (5′-AAA ATT GGG TAT GCG AGC TG-3′), Komp 1.5R (5′-GGG GTC TAA GGG CCT ATG AA-3′)
Ube3aSTOP: LSL F (5′-GTA CAT TGC ATT TGC CGT GA-3′), LSL R (5′-GGG GAA CTT CCT GAC TAG GG-3′)
Cre: All Cre F (5′-GAT GGA CAT GTT CAG GGA TCG CC-3′), All Cre R (5′-CTC CCA CCG TCA GTA CGT GAG AT-3′).
Behavioral testing
Mice 2–4 months of age were tested in a well-characterized battery of behavioral tests. Behaviors were performed in the following order with at least 48 h between experiments. Week 1, open field, marble burying; Week 2, rotarod; Week 3, nest building. All behavioral testing was conducted during the light cycle at a similar time of day within each cohort. Mice were allowed to acclimate to the behavioral testing room for at least 30 min prior to starting each experiment.
Unless otherwise stated, mice were run through all tests in the behavioral battery. Of note, however, Gad2-Cre::Ube3amFLOX/p+ mice sometimes show premature mortality in adulthood, presumably from spontaneous seizures (Judson et al., 2016), and spontaneous mortality was observed in Vglut2-Cre::Ube3amSTOP/p+ mice as well. Therefore, there was some drop-off in sample size for these groups over the period of the behavioral battery.
Separate cohorts of mice were evaluated for sleep phenotypes (described below), except for a small subset of Vglut2-Cre::Ube3amSTOP/p+ mice and littermate controls that were tested in sleep chambers after completing the behavioral battery.
Open field
Mice were placed in a novel, sound-attenuated arena (40 cm × 40 cm, ∼650 lux) for 30 min and were video recorded. Arenas were cleaned with ethanol and water and then dried thoroughly between mice. Time in center was defined as the amount of time the animal's center point resided in the central 20 cm × 20 cm of the arena. Total distance traveled and time in center were quantified using EthoVision XT 15.0 software.
Marble burying
Mice were tested in a clean cage placed within the open field arena with the same lighting conditions. Cages were filled with 3 L of 1/8-inch-diameter irradiated corncob bedding (Anderson Lab), on top of which 20 black marbles were placed in an equidistant 5 × 4 grid. Mice were placed in cages with marbles for 30 min with a clear, ventilated plexiglass lid. Before and after images of each cage were taken and used to objectively compute the percentage of marble area obscured by bedding using ImageJ software. Manual counts of the number of marbles at least two-thirds buried were also performed using these images and are reported in the Extended Data.
Accelerating rotarod
Mice were placed on the accelerating rotarod (Ugo Basile Model 47650) at an initial speed of 3 revolutions per minute (rpm), steadily increasing to 30 rpm over 5 min. Mice underwent three consecutive trials on the first day, and two additional trials 48 h later, with an intertrial interval of ∼1 min. Latency to the animal's first passive rotation and latency to 3× rotations or fall were recorded.
Nest building
Mice were single-housed for 3–5 d before beginning the nest building experiment. On Day 1, each animal's facility-provided nesting material was removed and replaced with 11 ± 1 g of extra-thick blot filter paper (Bio-Rad 1703966) cut into eight evenly sized rectangles. For 5 consecutive days, the amount of unused blot paper was weighed, recorded, and placed back in the cage. Across all experimental groups, three mice were excluded from this task because they showed noticeable weight loss and “ruffled” fur upon single-housing, suggestive of stress (Fig. 1G, one Gad2-Cre::Ube3amFLOX/p+ mouse; Fig. 2F, one Vglut2-Cre control mouse; Fig. 3D, one Ube3amSTOP/p+ mouse).
Sleep recordings
Mice were single-housed in 15.5 cm2 chambers with food, water, and nesting material in a dedicated sleep behavior monitoring room. Mice were given two full dark cycles to acclimate to single housing, after which sleep data were recorded for the following 7–9 d. Cage bedding was changed once during the experiment, and data from 7 A.M. of this day to 7 A.M. of the following day were excluded from analysis.
Sleep–wake behavior was quantified using a noninvasive monitoring system, PiezoSleep 2.0 (Signal Solutions), as previously described (Martin et al., 2024). Briefly, the PiezoSleep system uses a piezoelectric mat underneath the floor of each cage to measure pressure changes due to movements of the mouse. Signals were analyzed using custom software (SleepStats, Signal Solutions) to extract movement and respiratory patterns from pressure data, which was used to determine sleep versus wake states. Sleep was characterized by respiratory patterns typical of a sleeping mouse, and wake was characterized by volitional movements and the absence of sleep-typical signals. Average sleep bout duration was also extracted: sleep bout onsets were identified by 30 s periods composed of >50% sleep and terminations by 30 s periods of <50% sleep. This approach has been validated using EEG, EMG, and visual observation (Donohue et al., 2008; Mang et al., 2014).
Estimates of REM and NREM sleep were extracted using SleepStats version 4.0 (Signal Solutions) and were determined by periods of irregular breathing characteristic sleep state transitions (Yaghouby et al., 2016; Vanneau et al., 2021).
Average percent time sleeping and estimates of REM and NREM were calculated in 1 h bins, and average sleep bout durations were calculated using 4 h bins. One cohort, in which control mice did not demonstrate a clear “siesta” during the dark cycle, was excluded from analysis. This was attributed to a recent transfer of the mouse colony to a new vivarium, as consistent siesta behavior in C57BL/6J mice is a good prognosticator of appropriately quiet and relatively undisturbed animal care conditions.
Immunohistochemistry
Free-floating sections were rinsed twice in PBS (5 min each), followed by PBS containing 0.1% Triton X-100 (PBS-T). Sections were blocked for 30 min at room temperature in PBS-T containing 10% fetal bovine serum and then incubated overnight at room temperature with primary antibodies (GAD, 1:1,000; UBE3A, 1:1,000; NeuN, 1:1,000). After PBS-T rinses, sections were incubated with fluorescence-conjugated secondary antibodies and counterstained with DAPI. Mounted sections were air-dried and coverslipped using Vectashield Plus (Vector Laboratories, H-1900). Images were acquired using a Leica STELLARIS 8 FALCON microscope and quantified using QuPath software (RRID: SCR_018257) and InstanSeg (https://github.com/instanseg).
For UBE3A antigen detection, we used the mouse monoclonal antibody 3E5 (Sigma-Aldrich, catalog #SAB1404508, RRID:AB_10740376). For NeuN antigen detection, we used a guinea pig polyclonal antibody (Millipore, catalog #ABN90, RRID:AB_11205592) generated against a GST-tagged recombinant fragment corresponding to the first 97 amino acids of mouse NeuN. For GAD antigen detection, we used a rabbit monoclonal antibody (Abcam, catalog #ab183999, RRID:AB_3662875) that recognizes both GAD65 and GAD67 isoforms. The immunogen used to produce this antibody is proprietary.
For quantification of cell type-specific UBE3A expression levels in conditional knock-out and reinstatement models, nuclear UBE3A intensity was extracted in individual cells in the somatosensory cortex. For each experimental condition, quantification was performed using three animals per condition, three sections per animal, and both hemispheres of each section. Excitatory neurons were defined as NeuN+, GAD− cells, and inhibitory neurons were defined as NeuN+, GAD+ cells.
Experimental design and statistical analysis
Two-group open field distance traveled, rotarod, nest building (Figs. 1, 2): Two-way repeated-measures ANOVA
Three-group open field distance traveled, rotarod, nest building (Fig. 3): Two-way repeated-measures ANOVA with Tukey's post hoc multiple comparisons for main effects of genotype
Two-group marble burying and open field center time (Fig. 1E): Unpaired, two-tailed t test
Three-group marble burying (Figs. 2D, 3B): Brown–Forsythe ANOVA with Dunnett's T3 multiple comparisons
Three-group open field center time (Extended Data Fig. 3-2B): One-way ANOVA
Sleep studies (Figs. 4–9): Two-way repeated-measures ANOVA with Šidák’s multiple-comparisons tests
Data are presented as mean ± SEM unless otherwise noted. A p-value <0.05 was considered statistically significant. GraphPad Prism 10.3.1 software was used for all statistical analyses. Tukey post hoc tests were used when all comparisons were evaluated (overall comparisons between each genotype). Šidák’s post hoc tests were used for sleep analyses in which post hoc comparisons were only used to assess for differences between groups within the same time point (no comparisons were being made across different time points). Brown–Forsythe ANOVA and Dunnett's T3 multiple-comparisons tests were used when genotypic groups displayed unequal variances as measured by the Brown–Forsythe test. The statistical tests used for each figure and the statistical outputs are reported in Extended Data Figure 1-2.
Sample sizes required for appropriate statistical power were based on a meta-analysis of AS mouse model behavioral tasks (Sonzogni et al., 2018) and previous behavioral studies of AS mice (Sonzogni et al., 2019; Judson et al., 2021).
Results
Deletion of Ube3a in GABAergic neurons yields few deficits in a battery of AS-phenotypic behaviors
To assess the role of GABAergic neurons in AS-related behaviors, we generated mice with maternal inheritance of a floxed Ube3a allele and paternal inheritance of Gad2-Cre, thereby deleting Ube3a from the majority of GABAergic neurons (Taniguchi et al., 2011; Judson et al., 2016; Fig. 1A). This genetic cross efficiently lowered UBE3A expression in inhibitory neurons, while leaving excitatory neuron UBE3A expression intact (Fig. 1B). We then tested these Gad2-Cre::Ube3amFLOX/p+ mice, alongside littermate controls, in a well-characterized behavioral battery in which AS mice exhibit reliable, robust deficits (Fig. 1C; Born et al., 2017; Sonzogni et al., 2018; Judson et al., 2021). For behavioral tests, we made the a priori decision to combine all control groups for statistical analysis. However, we unexpectedly detected a significant effect of Cre expression in some instances and thus reported these groups separately in those cases.
Mice with maternal allele Ube3a deletion from GABAergic neurons (Gad2-Cre::Ube3amFLOX/p+) exhibit only modest behavioral deficits. A, UBE3A immunostaining in Ube3amFLOX/p+ and Gad2-Cre::Ube3amFLOX/p+ mice. Dashed lines indicate the largely GABAergic caudate putamen, in which UBE3A is robustly deleted in Gad2-Cre::Ube3amFLOX/p+ mice. Zoom images demonstrate cell type-specific UBE3A expression in the somatosensory cortex. Arrows indicate NeuN+, GAD+ neurons. Hemi-section scale bar, 1 mm. Zoom image scale bar, 10 µm. B, Nuclear UBE3A intensity of individual excitatory (NeuN+, GAD−) and inhibitory (NeuN+, GAD+) cells in the somatosensory cortex displayed as violin plots with mean ± SD. C, Schematic of behavioral battery. D, Distance traveled in the open field across 5 min bins. Two-way RM ANOVA. E, Quantification of marble burying behavior by threshold-based analysis of area obscured. Unpaired t test. F, Latency to fall or first passive rotation on the rotarod across each acquisition (Day 1) and retest (Day 2) trial. Two-way RM ANOVA. G, Quantification of percent nesting material used across 5 d test. Two-way RM ANOVA. Behavioral data presented as mean ± SEM. *p < 0.05, **p < 0.01. Individual data points labeled by genotype and sex are reported in Extended Data Figure 1-1. Statistical tests and output statistics for all figures are reported in Extended Data Figure 1-2. Created with BioRender (https://BioRender.com/a9rsfhy).
Figure 1-1
Gad2-Cre::Ube3amFLOX/p + behavioral battery labeled by genotype and sex. Open circles = males, closed circles = females. Dark gray = WT, light gray = Ube3amFLOX/p+, dark blue = Gad2-Cre, light blue = Gad2-Cre::Ube3amFLOX/p + . (A) Distance traveled in the open field across 5-minute bins. (B) Total time in the center of the open field. (C) Quantification of marble burying behavior by manual count of buried marbles (left panel) and threshold-based analysis (right panel). (D) Latency to fall or first passive rotation on the rotarod across each acquisition (day 1) and retest (day 2) trial. (E) Quantification of percent nesting material used across 5-day test. (F) Nest building behavior from only maternal heterozygous Ube3am+/pFLOX litters with additional cohort added. WT and Ube3amFLOX/p+ controls (n = 7), Gad2-Cre (n = 8), Gad2-Cre::Ube3amFLOX/p+ (n = 10). Two-way RM ANOVA with Tukey’s post hoc comparisons for effect of genotype. Data presented as means ± SEM. *P < 0.05, **P < 0.01. Download Figure 1-1, TIF file.
Figure 1-2
Statistical table. Summary table of statistical tests and outputs by figure. Download Figure 1-2, XLS file.
Mice were first tested in the open field test, a measure of locomotor ability and exploration behavior in which AS model mice are hypoactive. Gad2-Cre::Ube3amFLOX/p+ mice displayed no significant deficits in this behavioral test, suggesting no overt motor impairment (Fig. 1D; main effect of genotype: F(1,41) = 0.7491, p = 0.3918). These mice also showed no difference in time spent in the center of the arena, suggesting no difference in anxiety-like behavior (Extended Data Fig. 1-1B; t(41) = 1.402, p = 0.1684). Mice were next tested in the marble burying assay, a test evaluating motor coordination and the innate digging behavior of mice (Thomas et al., 2009). In this behavior, Gad2-Cre::Ube3amFLOX/p+ mice again performed similarly to controls, using an unbiased thresholding analysis to quantify the marble area unobscured by bedding (Fig. 1E; t(40) = 0.7272, p = 0.4713; Mossner et al., 2020; Judson et al., 2021).
Mice were next tested in a 2 d variation of the rotarod test assessing motor coordination and learning. In accordance with their having typical open field and marble burying performance, Gad2-Cre::Ube3amFLOX/p+ mice performed similarly to controls in the rotarod task (Fig. 1F; main effect of genotype: F(1,39) = 2.569, p = 0.1171), although they showed a tendency toward decreased latency to fall or passively rotate on the retest day.
After displaying only minor deficits in three largely motor tasks, mice were single-housed and assessed for nest building behavior, an innate, presleep behavioral program performed by male and female mice for protection, thermoregulation, and parenting (Sotelo et al., 2022; Tagawa et al., 2024). In contrast to the previous behavioral tests, Gad2-Cre::Ube3amFLOX/p+ mice exhibited a significant decrease in nest material used, indicating a behavioral deficit similar to AS model mice (Fig. 1G; main effect of genotype: F(1,37) = 20.28, p < 0.0001). Since the majority of control mice used for this experiment were Ube3amFLOX/p+ mice, due to breeding with Ube3amFLOX/pFLOX females, we repeated the nest building assay with an additional cohort of mice from Ube3am+/pFLOX females to ensure this deficit was not due to the Gad2-Cre allele alone. When all litters from heterozygous crosses were combined, Gad2-Cre mice performed similarly to Cre-negative controls, and Gad2-Cre::Ube3amFLOX/p+ mice used significantly less nest material than Gad2-Cre controls (Extended Data Fig. 1-1F; main effect of genotype: F(2,22) = 8.711, p = 0.0016, post hoc tests: Gad2-Cre vs Gad2-Cre::Ube3amFLOX/p+: q(22) = 4.889, p = 0.0061). Taken together, these data suggest that, unlike its documented role in EEG activity and seizure susceptibility (Judson et al., 2016; Gu et al., 2019), Ube3a loss from GABAergic neurons is not the driver of many other known AS mouse behavioral phenotypes, excepting nest building behavior.
Glutamatergic deletion of Ube3a produces multiple behavioral deficits
Given that Ube3a loss in GABAergic neurons yielded few behavioral deficits, we reasoned that Ube3a deletion from glutamatergic neurons might instead drive many AS behavioral phenotypes. Accordingly, we generated mice with maternal inheritance of the floxed Ube3a construct and paternal inheritance of Vglut2-Cre (Slc17a6; Vong et al., 2011). While VGLUT2 is primarily expressed subcortically in adulthood, it is expressed broadly in the cortex and hippocampus during development, allowing us, among other groups, to use this line to broadly target glutamatergic neurons (Meng et al., 2016; Gonzalez et al., 2019; Kim et al., 2024). Indeed, Vglut2-Cre-mediated deletion of Ube3a efficiently lowered UBE3A expression in most glutamatergic neurons, while sparing inhibitory neuron UBE3A expression (Fig. 2A,B).
Mice with Ube3a deletion from glutamatergic neurons (Vglut2-Cre::Ube3amFLOX/p+) exhibit motor and innate behavioral deficits. A, UBE3A immunostaining in Ube3amFLOX/p+ and Vglut2-Cre::Ube3amFLOX/p+ mice. Zoom images demonstrate cell type-specific UBE3A expression in the somatosensory cortex. Asterisks indicate excitatory (NeuN+, GAD−) cells. Arrows indicate inhibitory (NeuN+, GAD+) neurons. Arrowhead indicates example excitatory neuron with intact UBE3A expression in Vglut2-Cre::Ube3amFLOX/p+ mouse. Hemi-section scale bar, 1 mm. Zoom image scale bar, 10 µm. B, Nuclear UBE3A intensity of individual excitatory (NeuN+, GAD−) and inhibitory (NeuN+, GAD+) cells in the somatosensory cortex displayed as violin plots with mean ± SD. C, Distance traveled in the open field across 5 min bins. Two-way RM ANOVA. D, Quantification of marble burying behavior by threshold-based analysis of area obscured in Cre-negative controls, Vglut2-Cre controls, and Vglut2-Cre::Ube3amFLOX/p+ mice. Brown–Forsythe test with Dunnett's T3 post hoc multiple comparisons. E, Latency to fall or first passive rotation on the rotarod across each acquisition (Day 1) and retest (Day 2) trial. Two-way RM ANOVA. F, Quantification of percent nesting material used across 5 d test. Two-way RM ANOVA. Behavioral data presented as mean ± SEM. *p < 0.05, **p < 0.01. Individual data points labeled by genotype and sex are reported in Extended Data Figure 2-1.
Figure 2-1
Vglut2-Cre::Ube3amFLOX/p + behavioral battery labeled by genotype and sex. Open circles = males, closed circles = females. Dark gray = WT, light gray = Ube3amFLOX/p+, dark red = Vglut2-Cre, light red = Vglut2-Cre::Ube3amFLOX/p + . (A) Distance traveled in the open field across 5-minute bins. (B) Total time in the center of the open field. (C) Quantification of marble burying behavior by manual count of buried marbles (left panel) and threshold-based analysis (right panel). (D) Latency to fall or first passive rotation on the rotarod across each acquisition (day 1) and retest (day 2) trial. (E) Quantification of percent nesting material used across 5-day test. Download Figure 2-1, TIF file.
In contrast to mice with GABAergic Ube3a deletion, Vglut2-Cre::Ube3amFLOX/p+ mice showed significantly decreased distance traveled in the open field test (Fig. 2C; main effect of genotype: F(1,45) = 6.053, p = 0.0178). Despite their decreased locomotor performance, these mice showed no difference in time spent in the center of the arena (Extended Data Fig. 2-1B; t(45) = 0.2252, p = 0.8228). When tested in the marble burying task, Vglut2-Cre::Ube3amFLOX/p+ mice also showed a striking decrease in burying behavior (Fig. 2D; main effect of genotype: F(2,33.49) = 28.05, p < 0.0001). Notably, Vglut2-Cre mice showed an intermediate marble burying deficit when compared with Cre-negative littermates (Fig. 2D; t(30.99) = 3.335, p = 0.0066). Nonetheless, Vglut2-Cre::Ube3amFLOX/p+ mice displayed significantly impaired burying when compared with Vglut2-Cre controls (Fig. 2D; t(13.28) = 4.219, p = 0.0029), suggesting a key role for Ube3a loss in glutamatergic neurons for this phenotype.
Similarly to the first two behaviors tested, Vglut2-Cre::Ube3amFLOX/p+ mice showed significantly impaired performance on the rotarod task (Fig. 2E; main effect of genotype: F(1,45) = 6.897, p = 0.0118). Lastly, Vglut2-Cre::Ube3amFLOX/p+ mice were tested in the nest building assay, in which they showed comparable performance to littermate controls (Fig. 2F; main effect of genotype: F(1,44) = 1.397, p = 0.2436). Together, these data suggest glutamatergic neuron loss of UBE3A drives several behaviors, particularly motor behaviors in AS model mice, while loss of UBE3A from GABAergic neurons plays a larger role in regulating nest building behavior.
Reinstatement of Ube3a in glutamatergic neurons rescues multiple AS behavioral phenotypes
To further probe and validate the role of glutamatergic deletion of Ube3a in AS-associated behaviors, we utilized the Ube3amSTOP/p+ model of Ube3a reinstatement (Silva-Santos et al., 2015), crossing female Ube3am+/pSTOP mice with male Vglut2-Cre mice. This cross generated (1) mice with pan-neuronal deletion of Ube3a (Ube3amSTOP/p+, equivalent to AS model mice), (2) mice with Ube3a reinstatement in glutamatergic neurons (Vglut2-Cre::Ube3amSTOP/p+), and (3) littermate controls, all of which were tested in the same battery of behavioral tests. Vglut2-Cre-mediated Ube3a reinstatement substantially restored UBE3A expression in glutamatergic neurons, while leaving GABAergic neurons devoid of UBE3A (Extended Data Fig. 3-1A,B).
When assessed in the open field, Vglut2-Cre::Ube3amSTOP/p+ mice unexpectedly showed comparable distance traveled to Ube3amSTOP/p+ mice, indicating no rescue in behavioral performance (Fig. 3A; main effect of genotype: F(2,59) = 15.53, p < 0.0001, post hoc tests: Controls vs Ube3amSTOP/p+: q(59) = 6.335, p = 0.0001, Controls vs Vglut2-Cre::Ube3amSTOP/p+: q(59) = 6.661, p < 0.0001, Ube3amSTOP/p+ vs Vglut2-Cre::Ube3amSTOP/p+: q(59) = 6.661, p = 0.8988). This suggests that while glutamatergic deletion of Ube3a is sufficient to drive a behavioral deficit in this test, this neuronal population likely does not drive this behavior alone. Interestingly, neither Ube3amSTOP/p+ nor Vglut2-Cre::Ube3amSTOP/p+ mice showed a robust deficit in time spent in the center of the area, consistent with other studies demonstrating absent or weak anxiety-like phenotypes of AS model mice (Extended Data Fig. 3-2B; F(2,59) = 1.347, p = 0.2680; Born et al., 2017; Syding et al., 2022; Tanas et al., 2022). In contrast, when tested in the marble burying test, Vglut2-Cre::Ube3amSTOP/p+ mice showed significantly improved performance compared with Ube3amSTOP/p+ mice, and similar performance to that of littermate controls (Fig. 3B; F(2,42.73) = 13.7, p < 0.0001, post hoc tests: Ube3amSTOP/p+ vs Vglut2-Cre::Ube3amSTOP/p+: t(16.55) = 3.630, p = 0.0061, Controls vs Vglut2-Cre::Ube3amSTOP/p+: t(35.44) = 1.355, p = 0.4500). This experiment was not sufficiently powered to detect the previously observed difference between Vglut2-Cre controls and WT mice; however, Cre+ mice tended to bury fewer marbles on average in this experiment (Extended Data Fig. 3-2C).
Glutamatergic neuron-selective Ube3a reinstatement (Vglut2-Cre::Ube3amSTOP/p+) rescues AS motor and innate behavioral deficits. A, Distance traveled in the open field across 5 min bins. Two-way RM ANOVA with Tukey's post hoc comparisons of genotype effect. B, Quantification of marble burying behavior by threshold-based analysis of area obscured. Brown–Forsythe test with Dunnett's T3 post hoc multiple comparisons. C, Latency to fall or first passive rotation on the rotarod across each acquisition (Day 1) and retest (Day 2) trial. Two-way RM ANOVA with Tukey's post hoc comparisons of genotype effect. D, Quantification of percent nesting material used across 5 d test. Two-way RM ANOVA with Tukey's post hoc comparisons of genotype effect. Data presented as means ± SEM. *p < 0.05, **p < 0.01. Histological validation of glutamatergic UBE3A reinstatement in the Vglut2-Cre::Ube3amSTOP/p+ model is reported in Extended Data Figure 3-1. Individual data points labeled by genotype and sex are reported in Extended Data Figure 3-2.
Figure 3-1
Glutamatergic neuron Ube3a reinstatement in Vglut2-Cre::Ube3amSTOP/p+ mice. (A) UBE3A immunostaining in Control, Ube3amSTOP/p+ and Vglut2-Cre::Ube3amSTOP/p+ mice. Asterisks indicate excitatory (NeuN+, GAD-) cells. Zoom images demonstrate cell type-specific UBE3A expression in the somatosensory cortex. Arrows indicate inhibitory (NeuN+, GAD+) neurons. Arrowheads indicate non-neuronal (NeuN-) cells with persistent, weak UBE3A expression in Ube3amSTOP/p+ mice. Hemi-section scale bar = 1 mm. Zoom image scale bar = 10 µm. (B) Nuclear UBE3A intensity of individual excitatory (NeuN+, GAD-) and inhibitory (NeuN+, GAD+) cells in the somatosensory cortex displayed as violin plots with means ± SD. Download Figure 3-1, TIF file.
Figure 3-2
Vglut2-Cre::Ube3amSTOP/p+ behavioral battery labeled by genotype and sex. Open circles = males, closed circles = females. Gray = WT, dark red = Vglut2-Cre, purple = Ube3amSTOP/p+, light red = Vglut2-Cre::Ube3amSTOP/p + . (A) Distance traveled in the open field across 5-minute bins. (B) Total time in the center of the open field. (C) Quantification of marble burying behavior by manual count of buried marbles (left panel) and threshold-based analysis (right panel). (D) Latency to fall or first passive rotation on the rotarod across each acquisition (day 1) and retest (day 2) trial. (E) Quantification of percent nesting material used across 5-day test. Download Figure 3-2, TIF file.
In the rotarod task, Vglut2-Cre::Ube3amSTOP/p+ mice showed comparable motor performance to controls, representing a full phenotypic rescue (Fig. 3C; main effect of genotype: F(2,59) = 17.84, p < 0.0001, post hoc: Ube3amSTOP/p+ vs Vglut2-Cre::Ube3amSTOP/p+: q(59) = 5.730, p = 0.0004). Finally, when evaluated in the nest building task, Vglut2-Cre::Ube3amSTOP/p+ mice demonstrated a partial phenotypic rescue (Fig. 3D; main effect of genotype: F(2,57) = 34.46, p < 0.0001, post hoc: Ube3amSTOP/p+ vs Vglut2-Cre::Ube3amSTOP/p+: q(57) = 4.962, p = 0.0025, Controls vs Vglut2-Cre::Ube3amSTOP/p+ mice: q(57) = 5.430, p = 0.0009), suggesting a partial role for glutamatergic neuron loss of UBE3A in this behavior. Overall, these data suggest that UBE3A loss in glutamatergic neurons plays a major role in the AS-related motor and innate behaviors studied, while fewer behaviors are impacted by UBE3A loss from GABAergic neurons.
AS mice exhibit altered sleep–wake behavior
Sleep disturbances, such as frequent nighttime awakenings and difficulty falling asleep, are common in individuals with AS (Goldman et al., 2012; Pereira et al., 2020; Qu et al., 2024); however, the mechanisms underlying sleep deficits in AS are poorly understood. As GABAergic circuitry is critical for the onset and maintenance of sleep, likely through the inhibition of wake-promoting centers in the brainstem and hypothalamus (Anaclet et al., 2014; Scammell et al., 2017; Takata et al., 2018; Sulaman et al., 2023), we speculated that the loss of UBE3A from GABAergic neurons might be particularly impactful for AS sleep phenotypes. Furthermore, AS individuals and model mice demonstrate increased delta EEG power (Sidorov et al., 2017; den Bakker et al., 2018; Copping and Silverman, 2021), which has been linked to altered sleep (Huber et al., 2000; Tononi and Cirelli, 2006; Hubbard et al., 2020), and this delta phenotype is largely mediated by GABAergic neuron loss of Ube3a (Judson et al., 2016).
Alterations in sleep patterns and composition have been well-studied in the Ube3am−/p+ mouse model of AS using EEG/EMG recordings (Ehlen et al., 2015; Shi et al., 2015; Lee et al., 2023). These studies, however, are exceptionally labor-intensive and therefore not ideally suited for screening sleep behavior in numerous genetic crosses. Thus, we sought to determine whether sleep phenotypes of AS model mice could be replicated using the PiezoSleep monitoring system, a noninvasive piezoelectric system that uses breathing patterns to monitor sleep in single-housed mice. This technology has been validated with simultaneous EEG/EMG measurements (Mang et al., 2014) and has reproduced sleep phenotypes found using EEG/EMG measurements in a mouse model of ASD (Lord et al., 2022). To evaluate the utility of this technology for our study, AS mice and littermate controls were single-housed in piezoelectric chambers and allowed to acclimate for two dark cycles followed by 7–9 d of data collection. Sleep behavior was quantified by averaging data across days for each animal, providing a robust measurement of daily sleep patterns.
Consistent with previous reports of C57BL/6J mouse behavior, wild-type mice demonstrated a robust sleep period in the latter half of the dark cycle, termed the “siesta” (Wisor et al., 2008; Collins et al., 2020; Fig. 4A, Extended Data Fig. 4-1A,E). AS mice, on the other hand, virtually lacked this behavior, showing relatively stable percent sleep across the dark cycle, as previously described (Ehlen et al., 2015; Shi et al., 2022; Fig. 4A; genotype × time interaction: F(23,874) = 11.67, p < 0.0001). Despite this difference in sleep patterning, AS mice demonstrated comparable total sleep in the light and dark phases (Fig. 4B; main effect of genotype: F(1,38) = 0.2925, p = 0.5918).
AS mice exhibit altered sleep patterns. A, Piezoelectric quantification of hourly percent sleep in AS and WT mice. B, Average percent sleep across light and dark cycles. C, Mean sleep bout durations across 4 h time bins. D, Mean sleep bout duration during light and dark cycle, averaged from 4 h bins. All panels analyzed using two-way RM ANOVA with Šídák's post hoc tests. Data presented as mean ± SEM. *p < 0.05, **p < 0.01. Individual data points labeled by genotype and sex are reported in Extended Data Figure 4-1.
Figure 4-1
AS mouse sleep behavior separated by sex. Open circles = males, closed circles = females. (A) Piezoelectric quantification of hourly percent sleep in male and female AS and WT mice. (B) Average percent sleep across light and dark cycles. (C) Mean sleep bout durations across 4-hour time bins. (D) Mean sleep bout duration during light and dark cycle, averaged from 4-hour bins. (E) Piezoelectric quantification of hourly percent sleep presented in minutes. (F) Average percent sleep across light and dark cycles presented in minutes. Data presented as means ± SEM. Download Figure 4-1, TIF file.
We then quantified mean sleep bout duration, used as a measure of sleep fragmentation. Interestingly, AS model mice showed a significant reduction in mean sleep bout duration across the light cycle and at the end of the dark cycle (Fig. 4C; main effect of genotype: F(1,38) = 31.41, p < 0.0001), corresponding to overall decreased mean bout length in both light and dark phases (Fig. 4D; post hoc: light: t(76) = 6.200, p < 0.0001, dark: t(76) = 2.998, p = 0.0073). As AS mice show no notable difference in total sleep in the light or dark cycles, this decreased mean sleep bout length provides evidence for increased fragmentation of sleep, resembling the nighttime awakenings and decreased sleep efficiency seen in AS individuals (Trickett et al., 2019; O’Rourke et al., 2024).
GABAergic deletion of Ube3a increases sleep fragmentation
We next assessed the sleep behavior of Gad2-Cre::Ube3amFLOX/p+ mice to determine whether the sleep phenotypes observed in AS mice are due to GABAergic neuron deletion of Ube3a. On analysis of hourly sleep patterns, Gad2-Cre::Ube3amFLOX/p+ mice, surprisingly, showed intact “siesta” behavior, although they exhibited a dark cycle sleeping period that was shifted slightly earlier than controls; this effect may be partially due to a slightly different sleep pattern of the Gad2-Cre control mice (Fig. 5A; Extended Data Fig. 5-1A,C; genotype × time interaction: F(46,897) = 3.040, p < 0.0001). Corresponding with the slightly earlier and lengthened “siesta” period, Gad2-Cre::Ube3amFLOX/p+ mice showed a trend toward increased total sleep in the dark cycle (Fig. 5B; genotype × time interaction: F(2,39) = 3.164, p = 0.0533). These results suggest that GABAergic neuron loss of Ube3a does not play an important role in the loss of “siesta” in AS mice.
Upon analysis of sleep bout duration, however, Gad2-Cre::Ube3amFLOX/p+ mice showed decreased mean bout durations across much of the light cycle and the end of the dark cycle (Fig. 5C; main effect of genotype: F(1,40) = 8.654, p = 0.0054), corresponding to an overall decrease in bout duration in the light cycle (Fig. 5D; post hoc: light cycle: t(80) = 3.459, p = 0.0017, dark cycle: t(80) = 1.454, p = 0.2771). Therefore, while GABAergic Ube3a deletion did not drive AS-like dysfunction in sleep pattern, GABAergic neurons appear responsible for at least part of the sleep fragmentation phenotype.
Gad2-Cre::Ube3amFLOX/p+ mice exhibit fragmented sleep. A, Piezoelectric quantification of hourly percent sleep in Cre-negative controls, Gad2-Cre controls, and Gad2-Cre::Ube3amFLOX/p+ mice. B, Average percent sleep across light and dark cycles. C, Mean sleep bout durations across 4-hour time bins in Controls and Gad2-Cre::Ube3amFLOX/p+ mice. D, Mean sleep bout duration during light and dark cycle, averaged from 4 h bins. All panels analyzed using two-way RM ANOVA with Šídák's post hoc tests. Data presented as mean ± SEM. *Cre-negative controls versus Gad2-Cre::Ube3amFLOX/p+, #Gad2-Cre controls versus Gad2-Cre::Ube3amFLOX/p+, &Cre-negative controls versus Gad2-Cre controls. *p < 0.05, **p < 0.01. Individual data points labeled by genotype and sex are reported in Extended Data Figure 5-1.
Figure 5-1
Gad2-Cre::Ube3amFLOX/p + sleep behavior separated by genotype and sex. Open circles = males, closed circles = females. Dark gray = WT, light gray = Ube3amFLOX/p+, dark blue = Gad2-Cre, light blue = Gad2-Cre::Ube3amFLOX/p + . (A) Piezoelectric quantification of hourly percent sleep in males. (B) Average percent sleep across light and dark cycles in males. (C) Hourly percent sleep in females. (D) Average percent sleep across light and dark cycles in females. (E) Mean sleep bout durations across 4-hour time bins. (F) Mean sleep bout duration during light and dark cycle, averaged from 4-hour bins. Data presented as means ± SEM. Download Figure 5-1, TIF file.
Glutamatergic deletion of Ube3a disrupts sleep–wake behavior
To assess the role of glutamatergic neurons in AS sleep behavior, we tested Vglut2-Cre::Ube3amFLOX/p+ mice in the PiezoSleep chambers. Deletion of Ube3a from Vglut2-expressing neurons caused a slightly dysregulated hourly sleep pattern (Fig. 6A; Extended Data Fig. 6-1A, C; genotype × time interaction: F(9.871, 533.1) = 2.634, p = 0.0040), though this effect was not as dramatic as the lack of “siesta” seen in AS mice. Despite this altered sleep pattern, Vglut2-Cre::Ube3amFLOX/p+ mice showed normal amounts of total sleep (Fig. 6B; main effect of genotype: F(1,54) = 0.0002, p = 0.9644). Like AS model mice, Vglut2-Cre::Ube3amFLOX/p+ mice also showed decreased mean sleep bout duration (Fig. 6C; main effect of genotype: F(1,54) = 7.309, p = 0.0092), with the most evident difference during the light cycle (Fig. 6D; post hoc: light cycle: t(108) = 3.372, p = 0.0021, dark cycle: t(108) = 0.9239, p = 0.5874). Similarly to the behavioral battery results, these data implicate glutamatergic neuron loss of Ube3a as an important driver of sleep dysfunction in AS mice.
Vglut2-Cre::Ube3amFLOX/p+ mice exhibit altered sleep behavior. A, Piezoelectric quantification of hourly percent sleep in Vglut2-Cre::Ube3amFLOX/p+ mice and controls. B, Average percent sleep across light and dark cycles. C, Mean sleep bout durations across 4 h time bins. D, Mean sleep bout duration during light and dark cycle, averaged from 4 h bins. All panels analyzed using two-way RM ANOVA with Šídák's post hoc tests. Data presented as mean ± SEM. *p < 0.05, **p < 0.01. Individual data points labeled by genotype and sex are reported in Extended Data Figure 6-1.
Figure 6-1
Vglut2-Cre::Ube3amFLOX/p + sleep behavior separated by genotype and sex. Open circles = males, closed circles = females. Dark gray = WT, light gray = Ube3amFLOX/p+, dark red = Vglut2-Cre, light red = Vglut2-Cre::Ube3amFLOX/p + . (A) Piezoelectric quantification of hourly percent sleep in males. (B) Average percent sleep across light and dark cycles in males. (C) Hourly percent sleep in females. (D) Average percent sleep across light and dark cycles in females. (E) Mean sleep bout durations across 4-hour time bins. (F) Mean sleep bout duration during light and dark cycle, averaged from 4-hour bins. Data presented as means ± SEM. Download Figure 6-1, TIF file.
Glutamatergic reinstatement of Ube3a rescues AS sleep patterning
To further assess the role of glutamatergic loss of Ube3a in AS-related sleep phenotypes, we next studied the sleep of mice with Vglut2-Cre-mediated Ube3a reinstatement (Vglut2-Cre::Ube3amSTOP/p+ mice). First, we confirmed that the Ube3amSTOP/p+ model of Ube3a deletion showed the same characteristic lack of “siesta” observed in AS model mice (Fig. 7A, Extended Data Fig. 7-1A, C). Interestingly, Ube3a reinstatement in this neuronal population was sufficient to induce a robust “siesta” behavior similar to that of controls (Fig. 7A; genotype × time interaction: F(20.01, 480.2) = 3.959, p < 0.0001, post hoc: Vglut2-Cre::Ube3amSTOP/p+ vs Ube3amSTOP/p+: hour 19: t(22.28) = 5.420, p < 0.0001, hour 20: t(24) = 4.344, p = 0.0007). Notably, this “siesta” in Vglut2-Cre::Ube3amSTOP/p+ mice terminated slightly earlier than controls, reminiscent of the sleep pattern of mice with GABAergic Ube3a deletion (post hoc: Controls vs Vglut2-Cre::Ube3amSTOP/p+: hour 22: t(31.69) = 3.968, p = 0.0012). This indicates that glutamatergic Ube3a reinstatement was sufficient to partially recover “siesta” behavior. As expected, Vglut2-Cre::Ube3amSTOP/p+ mice showed similar total sleep to controls and Ube3amSTOP/p+ mice (Fig. 7B; main effect of genotype: F(2,48) = 0.9211, p = 0.4050).
Glutamatergic neuron-selective reinstatement of Ube3a rescues sleep patterns in AS model mice. A, Piezoelectric quantification of hourly percent sleep in Vglut2-Cre::Ube3amSTOP/p+, Ube3amSTOP/p+, and control mice. B, Average percent sleep across light and dark cycles. C, Mean sleep bout durations across 4 h time bins. D, Mean sleep bout duration during light and dark cycle, averaged from 4 h bins. All panels analyzed using two-way RM ANOVA with Šídák's post hoc tests. Data presented as mean ± SEM. *Controls versus Ube3amSTOP/p+, #Ube3amSTOP/p+ versus Vglut2-Cre::Ube3amSTOP/p+, &Controls versus Vglut2-Cre::Ube3amSTOP/p+. *p < 0.05, **p < 0.01. Individual data points labeled by genotype and sex are reported in Extended Data Figure 7-1.
Figure 7-1
Vglut2-Cre::Ube3amSTOP/p + sleep behavior separated by genotype and sex. Open circles = males, closed circles = females. Gray = WT, dark red = Vglut2-Cre, purple = Ube3amSTOP/p+, light red = Vglut2-Cre::Ube3amSTOP/p + . (A) Piezoelectric quantification of hourly percent sleep in males. (B) Average percent sleep across light and dark cycles. (C) Hourly percent sleep in females. (D) Average percent sleep across light and dark cycles in females. (E) Mean sleep bout durations across 4-hour time bins. (F) Mean sleep bout duration during light and dark cycle, averaged from 4-hour bins. Data presented as means ± SEM. Download Figure 7-1, TIF file.
On analysis of mean sleep bout duration, Ube3a reinstatement tended to increase bout duration toward the end of the light cycle (Fig. 7C; genotype × time interaction: F(10,240) = 2.202, p = 0.0184); however, overall bout durations across the light and dark cycles did not reach statistical significance (Fig. 7D; main effect of genotype: F(2,48) = 2.852, p = 0.0675, genotype × time interaction: F(2,48) = 2.343, p = 0.1069). Taken together, these data suggest glutamatergic neurons play a major role in the sleep pattern of AS model mice, but only a partial role in sleep fragmentation.
AS mice show decreased estimated REM sleep
In addition to sleep disturbance, AS individuals also show differences in sleep composition, namely, decreased REM sleep (Miano et al., 2004; Levin et al., 2022); mouse models of AS have likewise previously demonstrated a corresponding decrease in REM sleep during the light cycle (Colas et al., 2005; Lee et al., 2023). To determine whether the PiezoSleep chambers could be used to study sleep composition, we leveraged recent advancements in the analysis of piezoelectric signals to extract estimates of REM and NREM sleep based on changes in breathing rates characteristic of sleep state transitions (Mang et al., 2014; Yaghouby et al., 2016; Vanneau et al., 2021; Martin et al., 2024). In agreement with previous EEG/EMG studies (Colas et al., 2005; Lee et al., 2023), piezoelectric estimations of REM sleep revealed decreased REM in AS mice across the light cycle and dark cycle patterns mimicking the “siesta” phenotype seen in total sleep (Fig. 8A,B; Extended Data Fig. 8-1A,B; two-way ANOVA: main effect of genotype: F(1,38) = 4.889, p = 0.0331, post hoc: light cycle: t(76) = 2.675, p = 0.0182). Estimates of NREM sleep showed a similar lack of siesta in AS mice (Fig. 8C, Extended Data Fig. 8-1C; genotype × time interaction: F(23,874) = 11.84, p < 0.0001) and a subtle, statistically nonsignificant increase in NREM sleep (Fig. 8D; main effect of genotype: F(1,38) = 1.960, p = 0.1696).
AS model mice exhibit decreased REM sleep. A, Piezoelectric estimation of hourly percent REM sleep in AS and WT mice. B, Average percent REM across light and dark cycles. C, Estimated hourly percent NREM sleep in AS and WT mice. D, Average percent NREM across light and dark cycles. All panels analyzed using two-way RM ANOVA with Šídák's post hoc tests. Data presented as mean ± SEM. *p < 0.05, **p < 0.01. Individual data points labeled by genotype and sex are reported in Extended Data Figure 8-1.
Figure 8-1
AS mouse estimated REM and NREM sleep separated by sex. Open circles = males, closed circles = females. (A) Piezoelectric estimation of hourly percent REM sleep in male and female AS and WT mice. (B) Average percent REM across light and dark cycles. (C) Estimated hourly percent NREM sleep in AS and WT mice. (D) Average percent NREM across light and dark cycles. Data presented as means ± SEM. Download Figure 8-1, TIF file.
Given the rescue of sleep patterns by glutamatergic reinstatement of Ube3a, we predicted this would also rescue deficits of sleep composition. Indeed, Vglut2-Cre::Ube3amSTOP/p+ mice showed comparable REM sleep to controls across the light cycle (Fig. 9A, Extended Data Fig. 9-1A), demonstrating significantly increased light cycle REM time compared with Ube3amSTOP/p+ mice (Fig. 9B; main effect of genotype: F(2,48) = 9.496, p = 0.0003, post hoc: Ube3amSTOP/p+ vs Vglut2-Cre::Ube3amSTOP/p+: t(96) = 2.499, p = 0.0418). Interestingly, we observed decreased REM in Ube3amSTOP/p+ mice during the dark cycle (post hoc: controls vs Ube3amSTOP/p+: t(96) = 3.180, p = 0.0059, Ube3amSTOP/p+ vs Vglut2-Cre::Ube3amSTOP/p+: t(96) = 3.421, p = 0.0028) that might be due to a trend toward increased REM sleep in the Vglut2-Cre controls (Extended Data Fig. 9-1A–D). Nonetheless, Vglut2-Cre::Ube3amSTOP/p+ mice were indistinguishable from Vglut2-Cre controls across the light cycle. Analysis of NREM estimates revealed a similar “siesta” phenotype seen in the total sleep trace that was largely rescued by glutamatergic reinstatement of Ube3a (Fig. 9C, Extended Data Fig. 9-2A,C), and total NREM sleep did not differ between groups (Fig. 9D; main effect of genotype: F(2,48) = 0.05327, p = 0.9482). Overall, these findings support a critical role of glutamatergic neurons in the altered sleep composition seen in AS model mice.
Glutamatergic neuron-selective reinstatement of Ube3a normalizes sleep composition. A, Piezoelectric estimation of hourly percent REM sleep in Vglut2-Cre::Ube3amSTOP/p+, Ube3amSTOP/p+, and control mice. B, Average percent REM across light and dark cycles. C, Estimated hourly percent NREM sleep in AS and WT mice. D, Average percent NREM across light and dark cycles. All panels analyzed using two-way RM ANOVA with Šídák's post hoc tests. Data presented as mean ± SEM. *Controls versus Ube3amSTOP/p+, #Ube3amSTOP/p+ versus Vglut2-Cre::Ube3amSTOP/p+, &Controls versus Vglut2-Cre::Ube3amSTOP/p+. *p < 0.05, **p < 0.01. Individual data points labeled by genotype and sex are reported in Extended Data Figure 9-1 (REM sleep estimates) and Extended Data Figure 9-2 (NREM sleep estimates).
Figure 9-1
Vglut2-Cre::Ube3amSTOP/p + REM sleep estimates separated by genotype and sex. (A) Piezoelectric estimation of hourly percent REM sleep in males. (B) Average percent REM across light and dark cycles in males. (C) Estimated hourly REM sleep in females. (D) Average percent REM across light and dark cycles in females. Data presented as means ± SEM. Download Figure 9-1, TIF file.
Figure 9-2
Vglut2-Cre::Ube3amSTOP/p + NREM sleep estimates separated by genotype and sex. (A) Piezoelectric estimation of hourly percent NREM sleep in males. (B) Average percent NREM across light and dark cycles in males. (C) Estimated hourly NREM sleep in females. (D) Average percent NREM across light and dark cycles in females. Data presented as means ± SEM. Download Figure 9-2, TIF file.
Discussion
AS is a neurodevelopmental disorder with no disease-modifying treatment. However, clinical trials are currently underway using antisense oligonucleotides to unsilence the dormant paternal UBE3A allele, thereby normalizing UBE3A levels (Ionis: NCT05127226; Ultragenyx: NCT04259281). While this approach holds exciting promise and shows efficacy in mouse models (Meng et al., 2015; Milazzo et al., 2021), there is currently scant information regarding the key cell types or brain regions that require UBE3A reinstatement to mitigate core symptoms of AS. This holds particular importance, as effective biodistribution is a key concern in genetic therapies for CNS disorders (Roberts et al., 2020; Jafar-Nejad et al., 2021; Ling et al., 2023), and suboptimal targeting of necessary cell classes could hamper success. Moreover, mouse models of AS require early postnatal Ube3a reinstatement to achieve optimal phenotypic recovery (Silva-Santos et al., 2015; Sonzogni et al., 2020); early intervention could be difficult to achieve in the patient population without a corresponding early diagnosis, meaning many AS individuals are likely beyond the critical window to maximally benefit from UBE3A reinstatement-based therapies. Therefore, additional work is needed to better understand how loss of UBE3A leads to symptoms, as these insights will aid both in understanding the cell types that must be targeted for optimal genetic interventions and in developing alternative therapeutic options.
Our laboratory's previous work identified an outsized role of GABAergic loss of UBE3A in hyperexcitability phenotypes. GABAergic loss of UBE3A drives increased delta power on cortical EEG (Judson et al., 2016), a phenotype that correlates with the severity of a range of symptoms in AS individuals (Hipp et al., 2021; Ostrowski et al., 2021). Further, mice with Ube3a deleted from GABAergic neurons show decreased threshold to chemically and acoustically driven seizures, and they also exhibit spontaneous behavioral seizures, a phenotype not observed in AS model mice on a C57BL/6J background (Judson et al., 2016; Gu et al., 2019). These data forewarn that UBE3A reinstatement in a manner biased to glutamatergic neurons could potentially worsen epilepsy-related symptoms and highlight the importance of studying the neuronal populations regulating other behaviors.
Based on the exaggerated role of GABAergic neurons in AS seizure phenotypes, we predicted that GABAergic deletion of Ube3a would underlie a broad range of behavioral phenotypes in AS mice. In the present study, we instead found a larger role of Ube3a deletion from glutamatergic neurons in motor coordination, measured by rotarod and open field behavior, and innate species-specific behaviors such as marble burying. Furthermore, glutamatergic loss of UBE3A appears to mediate alterations in sleep patterning and induces some sleep fragmentation, while UBE3A loss from GABAergic neurons only caused fragmented sleep. Interestingly, glutamatergic reinstatement of Ube3a also rescued the decreased REM sleep observed in AS mice, as estimated by the PiezoSleep system. While this study identified some roles of GABAergic neurons in nest building behavior and sleep fragmentation, our data largely suggest a divergence of the neural circuitry underlying the motor, innate behavior, and sleep phenotypes of AS mice from the circuitry responsible for seizure susceptibility and cortical EEG patterns.
While our data suggest a large role of glutamatergic loss of UBE3A in behavioral phenotypes, they also highlight the challenges of studying cell type contributions to complex behaviors. For example, deletion of Ube3a from VGLUT2-expressing neurons caused decreased distance traveled in the open field test, but Ube3a reinstatement in this population provided no rescue of open field behavior. Conversely, Ube3a deletion from glutamatergic neurons caused no significant effect in nest building behavior and only subtle alterations in sleep patterning, but gene replacement in these neurons substantially improved performance in both measures. This discordance between Ube3a deletion and reinstatement in sleep and nest building behavior suggests that UBE3A loss from glutamatergic neurons is necessary for AS behavioral phenotypes but is not in itself sufficient to drive behavioral impairment. This could imply an important role for a population of neurons that was not assessed in this study, such as neuromodulatory populations. Indeed, serotoninergic, cholinergic, and hypocretin/orexin neurons are critical for establishing sleep–wake patterns and could also orchestrate presleep behaviors like nest building (Eban-Rothschild et al., 2018). While these nuanced results are difficult to fully interpret, they largely point toward a prominent role of glutamatergic neuron loss of UBE3A in many AS behavioral phenotypes.
One limitation of behavioral battery studies such as this is the multifaceted nature of the behavioral tests used. For example, marble burying and nest building are innate behavioral programs that require the coordinated activity of multiple circuit nodes, many of which are likely unknown, but also require sufficient motor ability. Therefore, it is difficult to attribute performance on a particular task to a specific behavioral domain. As glutamatergic loss of UBE3A caused deficits in multiple tasks that require intact gross motor function (open field, marble burying, and rotarod), we intuit that this neuronal population exerts a prominent role in motor performance.
This study also revealed key contributions of glutamatergic loss of UBE3A to sleep phenotypes. Here, we demonstrate that the PiezoSleep system can detect similar sleep phenotypes in AS model mice to those previously reported using EEG/EMG recordings, such as their characteristic lack of “siesta” and decreased REM sleep during the light cycle (Colas et al., 2005; Ehlen et al., 2015; Shi et al., 2022; Lee et al., 2023). Importantly, AS model mice also display dysfunctional sleep homeostasis, measured by their response to sleep deprivation (Ehlen et al., 2015) and 24 h light or dark exposure (Shi et al., 2022). These sleep behaviors were not assessed in the present study but should be addressed in future experiments. Additionally, it must be noted that the PiezoSleep system has only been validated for its accuracy of sleep staging in one publication, which was conducted using rats (Vanneau et al., 2021). Furthermore, this system has been validated for studying sleep using simultaneous EEG/EMG in only wild-type mice, raising the possibility that it may less accurately measure sleep in mouse models with motor impairments. While the piezoelectric system is not the gold standard measure of sleep–wake behavior, our ability to reproduce key AS sleep phenotypes with this system further supports its use as a high-throughput method to study sleep–wake behavior in genetic mouse models.
While much work has been done to dissect the neural circuitry mediating sleep–wake transitions and circadian rhythms, the mechanisms governing sleep–wake behavior during the active period are less understood. One recent study revealed a role for GABAergic vasoactive intestinal peptide (VIP)-expressing neurons in the hypothalamic suprachiasmatic nucleus (SCN) in shaping the “siesta” period sleep of mice during the dark cycle (Collins et al., 2020). However, in the present study, deletion of maternal Ube3a from GABAergic neurons, presumably including those of the SCN, tended to lengthen the “siesta” period rather than eliminate it. Additionally, neuronal silencing of the paternal Ube3a allele is partially relaxed in the SCN relative to other brain regions (Ehlen et al., 2015; Jones et al., 2016). These data suggest the lack of “siesta” in AS mice may be mediated by circuitry outside the SCN, such as those involved in the accumulation of sleep pressure.
This study focuses on the contributions of glutamatergic and GABAergic neurons to AS-related behaviors, but it is possible that certain phenotypes are influenced by altered activity of other neuronal populations, such as neuromodulatory cells. Indeed, AS model mice display altered dopamine release in the mesolimbic and nigrostriatal pathways (Riday et al., 2012), as well as altered firing properties and synaptic transmission in medium spiny neurons of the striatum, which receives substantial dopaminergic input (Hayrapetyan et al., 2014; Rotaru et al., 2023). Interestingly, the majority of midbrain dopaminergic neurons transiently express VGLUT2 during development and are targeted by the Vglut2-Cre mouse line (Steinkellner et al., 2018), indicating behavioral effects of UBE3A loss from midbrain dopaminergic neurons would be reflected by the Vglut2-Cre::Ube3amFLOX/p+ mouse model.
UBE3A is expressed in virtually all brain regions and neuron types and is paternally silenced in nearly all mature neurons (Judson et al., 2014; Gonzalez Ramirez et al., 2024), making the AS mouse model well suited to study the behavioral consequences of neuron type-specific manipulations. Our previous studies demonstrated a major role for GABAergic loss of UBE3A in seizure susceptibility and cortical oscillation patterns of AS model mice, suggesting these phenotypes are mediated by altered E–I balance broadly favoring excitation. In the present study, our data largely implicate glutamatergic loss of UBE3A in mediating many AS behavioral phenotypes, suggesting a different modality of E–I imbalance in the manifestation of these symptoms. Excitatory and inhibitory populations, however, are entangled and complex, and their contributions to behavior can be studied on various scales from inter-region connectivity to regional microcircuitry. These data on the broad cell types driving behaviors can serve as an entry point to more focused studies examining the key neuronal subtypes and brain regions involved. Taken alongside our previous findings, these data underscore the importance of UBE3A reinstatement strategies effectively targeting both excitatory and inhibitory neuron populations for optimal symptom improvement in AS individuals.
Footnotes
The authors declare no competing financial interests.
This work was supported by the Simons Foundation Autism Research Initiative (SFARI 702556), National Institutes of Health (NIH) R01NS131615, R01NS129914, National Research Service Award (NRSA) training grant F30HD111296, and Behavior Phenotyping Core funding P50HD103573. We thank Kathryn Harper, Viktoriya Nikolova, Samuel Harp, and Sheryl Moy of the UNC Mouse Behavioral Phenotyping Core for coordinating sleep monitoring studies. We thank Wendy Salmon and the UNC Hooker Imaging Core Facility, supported in part by NIH P30CA016086 Cancer Center Core Support Grant to the UNC Lineberger Comprehensive Cancer Center, and NIH 1S10OD030300.
This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International license, which permits unrestricted use, distribution and reproduction in any medium provided that the original work is properly attributed.
References
In this issue
