Cell-type-specific dysregulated gene expression in the frontal cortex of an Angelman syndrome pig model

Abstract

Angelman syndrome is a neurodevelopmental disorder caused by the loss of the maternal allele of the ubiquitin-protein ligase E3A (UBE3A) gene. UBE3A is imprinted with maternal-allelic expression in neurons of the central nervous system (CNS) and biallelic expression in other cell types. Consequently, in Angelman syndrome, UBE3A is substantially reduced in CNS neurons and reduced by half in other cells. It is unclear how cell-type-specific gene expression in the brain is dysregulated in Angelman syndrome, as previous studies have lacked cell type resolution. Using single nuclei RNA-sequencing, we show that gene expression is dysregulated in neuronal subtypes in the frontal cortex of neonatal pigs with a UBE3A maternal deletion. A total of 3812 unique genes were dysregulated across ten cell type clusters, with most of the dysregulated genes (3154 genes) in excitatory neurons. Pathway analysis revealed alterations in oxidative phosphorylation, proteasome function, and synaptic function. Additionally, somatostatin (SST)—a secreted neuropeptide involved in GABAergic inhibition and synaptic plasticity— was reduced in inhibitory neurons in the cortex and hypothalamus, which correlated with lower circulating SST protein in neonates but not adolescent pigs. Overall, these findings provide greater clarity on the cell-type-specific dysregulation of gene expression and cellular pathways caused by the loss of UBE3A in CNS neurons, expanding our understanding of the molecular pathology in Angelman syndrome.

Introduction

Angelman syndrome (OMIM # 105830) is a severe neurogenetic disorder characterized by profound developmental delay, cognitive impairment, motor coordination deficits, absent or reduced speech, ataxic gait, and a uniquely cheerful disposition [1]. It is caused by mutations or epimutations leading to the loss of function or expression of the maternally inherited allele of the ubiquitin-protein ligase E3A (UBE3A) gene [2].

The UBE3A gene is imprinted in neurons of the central nervous system (CNS), where the maternal allele is expressed and the paternal allele is repressed. In all other cell types, UBE3A is biallelically expressed. Transcription of the UBE3A paternal allele is repressed by the UBE3A antisense (UBE3A-AS) transcript, which represents the distal end of the small nucleolar host gene 14 (SNHG14) polycistronic transcript [2, 3]. Transcription of UBE3A-AS is thought to inhibit transcriptional elongation of the paternal UBE3A allele [3]. As such, individuals with Angelman syndrome have substantially reduced UBE3A in CNS neurons and approximately fifty percent reduction of UBE3A in all other cell types [2, 4].

The UBE3A protein functions as an E3 ligase that polyubiquitinates proteins for degradation by the 26S proteasome [5]. Additionally, UBE3A functions as a transcriptional coactivator of nuclear steroid hormones and can mono-ubiquitinate at least one protein in a non-degradative context [6, 7]. UBE3A has a broad array of protein and gene targets involved in regulating calcium signaling, cytoskeletal dynamics, proliferation, neuronal homeostasis, mitochondrial function, and retinoic acid signaling, amongst other pathways [8–13]. Changes to these pathways following maternal UBE3A loss result in deficits in long-term potentiation and synaptic structure and function [14, 15], although the dysregulated genes and molecular pathways underlying the symptoms observed in individuals with Angelman syndrome are poorly understood.

Several studies have examined dysregulated gene expression in Angelman syndrome models using bulk RNA-sequencing or proteomics, which lack cell type resolution [8–10, 12, 16–24]. It is thus unclear how much of the dysregulation stems from neuronal vs non-neuronal UBE3A loss in the brain. Our laboratory recently developed a pig model of Angelman syndrome that mirrors the developmental trajectory of the syndrome [25]. Using single nuclei RNA-sequencing, we examined the cell-type-specific dysregulation of genes and pathways in the frontal cortex of neonatal pigs (10-day-old) with a deletion of the maternal UBE3A allele. Our results indicate that loss of maternal UBE3A dysregulates gene expression primarily in CNS neurons.

Materials and methods

Study animals

The study animals included wild-type pigs (UBE3A^+/+), pigs with maternally inherited 97 kilobase deletion of the UBE3A gene (UBE3A^−/+ [NC_010443.5:g.141887260_141984697del]), and pigs with a paternally inherited 97 kilobase deletion of the UBE3A gene (UBE3A^+/−) [25]. The study animals were housed in small groups (approximately 3–4 pigs) in climate-controlled rooms under twelve-hour light cycles. Genotyping was performed through a polymerase chain reaction (PCR), as previously described [25]. Feed was provided ad libitum through day 40 and afterward, was fed at 2–4% daily of body weight. Animals past weaning age (approximately 21 days) were food deprived the night before necropsying. Animals did not receive treatment for Angelman syndrome phenotypes. All animal work was conducted under the guidance, supervision, and approval of the Texas A&M University Institutional Animal Care and Use Committee. A table of all animals used in this study is available in Supplementary File 1.

Postnatal animals were sedated (intramuscular tiletamine) and anesthetized (isoflurane ventilation), then euthanized via cardiac perfusion with phosphate-buffered saline. Animals used in the bulk RNA-sequencing were euthanized with phenytoin/pentobarbital. Fetal pigs were collected by a caesarean-section while the sow was sedated (intramuscular delivery of ketamine, butorphanol, and midazolam) and anesthetized (isoflurane ventilation). Areflexic fetal pigs were euthanized with 5 mL of intravenous potassium chloride prior to tissue collection. Brain region-specific and body tissue samples were taken in 4 mm punches, flash-frozen in liquid nitrogen, and then stored at −80°C until use.

Single nuclei library preparation

Brain region-specific samples were taken in 4 mm punches, flash-frozen in liquid nitrogen, and then stored at −80°C until use. Using the 10× Genomics Nuclei Isolation Kit (#PN-1000493, 10× Genomics, CA, USA), nuclei were isolated from 25 mg frontal lobe cortex tissue from four WT and four UBE3A^−/+ 10-day-old piglets. Each genotype was comprised of two males and two females. The 10× Chromium Connect was used for automated Single Cell Gene Expression 3′ v3 library preparation (#PN-1000075, 10× Genomics), followed by sequencing on Illumina NovaSeq 6000, XP Workflow (Illumina, CA, USA).

Single nuclei RNA-sequencing computational analysis

Cellranger (7.2.0) mkref was used to create a reference genome from NCBI S. scrofa 11.1 (GCF_000003025.6), along with a custom annotation of UBE3A and UBE3A-AS. Cellranger count was used for generating count matrices with default settings applied, except for intron exclusion with –include-introns = false. SoupX (1.6.2) ambient RNA correction was applied to Cellranger count outputs. Scater (1.0.3) was used to calculate median absolute deviations (MADs) for thresholding feature counts per sample. Using Seurat (5.1.0), barcodes were excluded that had feature counts > 5 or < 3 MADs per sample and ≥ 5% mitochondrial genes (annotated with KEF22- prefix in pigs). After initial Seurat CCA integration (dims = 30, resolution = 0.5), clusters with an average of < 900 features per cell were removed. After filtering and computational analysis, a final count of 16 874 cells (8932 WT; 7942 UBE3A^−/+) was retrieved. Final figures and analysis were then generated with Seurat CCA integration (dims = 20, resolution = 0.5). scType was used to generate initial cell type labels for each cluster, and labels were confirmed to agree with published markers [55]. Based on the clustering results, the neuronal clusters were further sub-typed into ERBB2+ (Inhibitory_1) and SST+ (Inhibitory_2) inhibitory neurons, and layer 2/3 (CUX2+; Excitatory_1), layer 4 (RORB+; Excitatory_2), and layer 5/6 (TLE4+; Excitatory_3) excitatory neurons consistent with published markers [26–28]. Differential expression analysis was performed using the FindMarkers function in Seurat using the default settings (Wilcox test with a threshold of log2FC of 0.1) and significance was adjusted for multiple comparisons using a Bonferroni-adjusted P-value of < 0.05. A permissive log2FC threshold was used to capture subtle transcriptional differences, which may include changes of modest biological effect size. Further analyses were performed using Augur (1.0.3) for perturbation analysis and Speckle (1.6.0) for cell type proportion analysis. Genes reported with LOC prefix have not been assigned a gene symbol by NCBI. Average UBE3A and UBE3A-AS expression values were returned using the Seurat AverageExpression function. To illustrate UBE3A expression, values were displayed relative to each wild-type cell type. To illustrate UBE3A-AS expression, values were displayed relative to the Excitatory wild-type to compare expression by cell type. For the list of high-confidence genes, data were sorted in Qiagen Ingenuity Pathway Analysis, where only genes with expression > 0 and with known human orthologues were kept, then sorted by P-value and average log2FC.

Custom annotation of UBE3A and UBE3A-AS

The custom annotation of UBE3A-AS was generated from frontal cortex bulk RNA-sequencing using methods previously described [56]. Briefly, FASTQ sequences generated from strand-specific poly-A enriched RNA were aligned to S. scrofa 11.1 with Hisat2 (2.1.0), then aligned SAM sequences were then converted to binary BAM sequences, indexed, and sorted using samtools. Aligned sequences were filtered remove non-uniquely aligned reads. The 5′ and 3′ reads were split using samtools. Transcript assemblies were generated using Stringtie (1.3.4.d), with the setting: -g 10. Single-exon transcripts and transcripts with non-canonical splice-sites were filtered using gffread (GFF utilities). Transcript units were defined as overlapping transcripts expressed from the same strand of a transcriptional unit and sharing a common terminal exon. Cellranger mkref was used to create a custom reference with the UBE3A/UBE3A-AS overlapping exons removed, which correspond to the following coordinates: chr1:141984591–141 987 465; chr1:141974996–141 975 160; chr1:141925289–141 925 484; chr1:14188718–141 887 772.

Quantitative reverse transcription polymerase chain reaction (qRT-PCR)

RNA from flash-frozen samples was isolated with Qiagen RNeasy Plus kit (#74134, Thermo Fisher Scientific, MA, USA). RNA was quantified with Qubit RNA BR Assay kit (Thermo Fisher Scientific, #Q10210) and a Qubit 4 Fluorometer. First-strand cDNA synthesis was performed with the SuperScript IV kit (Thermo Fisher Scientific, #18091200) with the Oligo(dT)20 template for poly-A transcript enrichment.

For SST in all tissues except the pancreas, TaqMan reactions (TaqMan Gene Expression Master Mix, #4369016, Thermo Fisher Scientific) were performed according to the manufacturer’s instructions and normalized to RPL4 levels. The TaqMan assays used were SST (Ss03391856_m1) and RPL4 (Ss03374067_g1). For UBE3A and UBE3A-AS in the frontal cortex and SST in the pancreas, SYBR green qRT-PCR reactions (PowerUp SYBR Green Master Mix for qPCR, #A25742, Thermo Fisher Scientific) were performed according to the manufacturer’s instructions and used PPIA for measuring relative expression. Primers sequences used are available in Supplementary File 11). Reactions were run in triplicate, and replicate values were excluded if the standard deviation exceeded a quarter of a cycle. Twenty μL reactions with 2 μL of template (equivalent to 8 ng or 16 ng of RNA) were run on a Bio-Rad CFX96 Touch Real-Time thermocycler (Bio-Rad, CA, USA). The parameters for the SYBR assays are as follows: 50°C for 2 minutes, 95°C for 2 minutes, then 40 cycles of 95°C for 15 seconds, 60°C for 1 minute, with readings taken at the end of every 60°C step. The parameters for the TaqMan assays are as follows: 50°C for 2 minutes, 95°C for 10 minutes, then 40 cycles for 95°C for 15 seconds, 60°C for 1 minute, with readings taken at the end of every 60°C step. Results were analyzed with Bio-Rad CFX Maestro software.

Western blot and densitometry

Tissue samples were lysed in NP40 buffer (comprised of 1% Nonidet P40 and 0.01% SDS in 100 mM Tris–HCl [pH 7.2]) with Roche protease inhibitor (#11697498001, Sigma-Aldrich, MA, USA). Lysates were mixed with 4× Laemmli buffer (#1610747, Bio-Rad,) with 10% β-mercaptoethanol added and heated at 95°C for 10 minutes. Proteins were separated on Bio-Rad 10% Mini-PROTEAN TGX precast gels (#4561033, Bio-Rad) at 30 V for 30 minutes, then 100 V for 45 minutes, and transferred to a nitrocellulose membrane using the Bio-Rad Trans-Blot Turbo System. The membrane was blocked in a buffer comprised of 5% nonfat dry milk in 0.1 M Tris-buffered saline with 0.1% Tween-20 (TBST), rotating for 1 hour at room temperature, then incubated overnight rotating at 4°C with mouse anti-UBE3A antibody (#611416, BD Biosciences, CA, USA) diluted 1:1000 in the blocking buffer. After washing in TBST for three washes of 15 minutes, the membrane was incubated with goat anti-mouse HRP-conjugated secondary antibody (#626520, Thermo Fisher Scientific) diluted 1:10000 in the blocking buffer for 1 hour at room temperature, rotating. Protein detection was performed using Bio-Rad Clarity Western ECL Substrate (#1705061, Bio-Rad) and visualized on a Bio-Rad ChemiDoc Imaging System. Densitometry analysis was performed in ImageJ. UBE3A protein levels were normalized to the amount of Ponceau S stain (#K793, VWR, OH, USA) total protein per lane.

Bulk RNA-sequencing

Samples were flash-frozen in liquid nitrogen and then stored at −80°C until use. RNA was extracted from 20 mg of tissue using the Qiagen RNeasy kit (#74104, Thermo Fisher Scientific) following the manufacturer’s protocol. RNA was eluted from the columns using 60 μL of nuclease-free water. RNA quantity was verified using the Qubit RNA BR assay (#Q10210, Thermo Fisher Scientific). RNA quality was checked using an Agilent TapeStation (Agilent, CA, USA). RNA libraries were prepared from 1 μg of total RNA with a RIN value > 6.1 using the TruSeq mRNA stranded kit (20 020 595, Illumina). Paired-end, 150 base pair sequencing on the NovaSeq 6000 at a depth of 10–20 million reads per sample was performed. The Qiagen IPA RNA Portal was used for Ensembl Sscrofa11.1.105 reference alignment and differential expression analysis using the default parameters. Gene names with the ENSSSCG prefix represent unnamed genes or genes in which a human ortholog has not been identified. Ensemble gene IDs were converted to RefSeq gene IDs using conversion resources from the HGNC Comparison of Orthology Predictions (HCOP) tool (genenames.org/tools/hcop).

Gene ontology and pathway analysis of dysregulated genes

DAVID functional annotation was performed with default settings for gene ontology. Dysregulated pathways were identified using QIAGEN Ingenuity Pathway Analysis [57]. The DEGs were entered by cell type. Enriched pathways represent pathways with a greater number of differentially expressed genes than would be expected by random chance. Pathway activation and inhibition scores are predicted based on the observed upregulation or downregulation of the genes. Top pathways with both significant Z-scores and P-values for high confidence DEGs (P-adj < 0.05; log2FC > ±1) are shown. Pathways not relevant to CNS tissue and pathways enriched with genes inherent to variation in single-cell data (i.e. ribosomal genes) were excluded from the analysis but are available in Supplementary File 6. Pathway analysis for all DEGs (P-adj < 0.05) is included in Supplementary File 8. Genes related to synaptic function were determined with SynGO (syngoportal.org).

Enzyme-linked immunosorbent assay (ELISA)

Blood and CSF samples were taken prior to euthanasia. Blood samples from 10-day-old piglets were collected into serum separator tubes (#367814, BD Biosciences), while blood samples from 120-day-old animals were collected into K2 EDTA tubes (#367861, BD Biosciences). Blood samples were centrifuged at 1200× g for 10 minutes, aliquoted, and stored at −80°C. CSF was taken from suboccipital cisterna magna collection. A competitive ELISA from Novus Bio specific for SST (#nbp2–80 269, Novus Bio, CO, USA) was used to quantify SST, according to the manufacturer’s instructions. Absorbance at 450 nm was measured with an Agilent BioTek Cytation5. Absorbance values were interpolated to pg per mL by a standard curve in GraphPad Prism with an asymmetric five-parameter logistic equation fit test. The standard curve was run in duplicate, and samples were run in duplicate or triplicate. Triplicate measurements exceeding 10% coefficient of variation were excluded.

Statistical analyses

Statistical analyses were performed in GraphPad Prism (GraphPad, CA, USA) or JMP (JMP Statistical Discovery, NC, USA). For comparison between two groups, a Student’s t-test was performed. For comparisons between three groups, an ANOVA (Analysis of Variance) test with a Tukey’s multiple comparison’s post-hoc test was used. For multiple comparisons with different tissues from an individual animal, Mixed-Effect Logistic Regression model was used to account for repeated measures.

Results

Experimental design and classification of cell types

Since neurons are predicted to be more severely impacted than other central nervous system (CNS) cell types by the loss of the maternal UBE3A allele, we performed single nuclei RNA-sequencing on tissue punches isolated from the frontal cortex of 10-day-old wild-type (n = 4 [male, n = 2; female n = 2]) and maternal UBE3A deletion (UBE3A^−/+) pigs (n = 4 [male, n = 2; female n = 2]) to identify cell-type-specific differentially expressed genes (DEGs). We then performed multiple pathway analyses to identify and characterize dysregulated pathways in each cell type (Fig. 1A, Supplementary File 1).

Figure 1.

Experimental design and classification of cell types (A) graphical abstract of the experimental design of the study.(B) UMAP of combined single nuclei RNA sequencing data (n = 8), including annotated neuronal and non-neuronal cell types in the frontal cortex. (C) UMAP of single nuclei RNA sequencing data for the WT (n = 4) and UBE3A^−/+ (n = 4) pigs. (D) Percent and average expression of marker genes by cell type and neuronal subtype. Abbreviations: WT, wildtype; UBE3A^−/+, maternal UBE3A deletion.

After single nuclei isolation and sequencing, the sequence reads were mapped to the pig NCBI reference genome assembly (Sus scrofa 11.1, RefSeq GCF_000003025.6) with custom gene annotations of the UBE3A and UBE3A-AS transcripts (see Materials and Methods and Supplementary Fig. 1). Analysis of 8932 WT and 7942 UBE3A^−/+ cells revealed ten cell types (Fig. 1B), which clustered independently of genotype (Fig. 1C), permitting the downstream analysis of differentially expressed genes. The cell types were identified by marker genes expressed in specific cell types, including excitatory neurons (SLC17A7), inhibitory neurons (GAD1), astrocytes (GFAP), oligodendrocytes (MAG), oligodendrocyte precursor cells ([OPC] PDGFRA), microglia (AIF1), and endothelial cells (APOLD1) (Fig. 1D, Supplementary Fig. 2) [26–28]. The excitatory and inhibitory neurons were further clustered into different subtypes based on the expression of additional cell-type-specific marker genes (Excitatory_1: CUX2, cortical layer 2/3; Excitatory_2: RORB, cortical layer 4; Excitatory_3: TLE4, cortical layer 5/6; Inhibitory_1: ERB4, and Inhibitory_2: SST) (Fig. 1D, Supplementary Fig. 2) [29, 30].

The pig UBE3A gene is specifically imprinted in cortical neurons

Analysis of UBE3A and UBE3A-AS expression by cell type revealed that UBE3A was reduced by 86–93% in the neuronal subtypes and by 39–77% in the non-neuronal cell types in the UBE3A^−/+ pigs (Fig. 2A and B, Supplementary Fig. 1C). As expected, UBE3A-AS expression was limited to the neuronal cell types (Fig. 2C and D). Interestingly, UBE3A-AS was slightly increased in the UBE3A^−/+ pigs, which aligns with a previous finding [31]. We confirmed the imprinting of UBE3A in the frontal cortex of 10-day-old WT and UBE3A^−/+ pigs using quantitative RT-PCR and western blot (Fig. 2E and F, Supplementary File 1). Altogether, these findings confirm the neuron-specific imprinting of the UBE3A gene and expression of the UBE3A-AS transcript in pigs.

Figure 2.

The pig UBE3A gene is specifically imprinted in cortical neurons. (A) UMAP heatmap of cell type specific UBE3A expression in the WT and UBE3A^−/+pigs. (B) Relative UBE3A expression by cell type between WT and UBE3A^−/+pigs. Data are presented as the mean UBE3A expression (reads per cell/cell type) normalized to WT in each cell type. (C) UMAP heatmap of cell type specific UBE3A-AS expression in the WT and UBE3A^−/+pigs. (D) UBE3A-AS expression by cell type between WT and UBE3A^−/+pigs. Data are presented as the mean UBE3A-AS expression (reads per cell/cell type) normalized to WT in excitatory neurons. (E) Strand-specific quantitative RT-PCR of UBE3A and UBE3A-AS expression in the frontal cortex of 10-day-old WT (n = 4) and UBE3A^−/+ (n = 4) pigs, normalized to WT. data are presented as mean ± the standard error of the mean (SEM); Student’s t-test, ***P < 0.001; ns, not significant. Western blot quantification of UBE3A protein in the frontal cortex of 10-day-old WT (n = 4) and UBE3A^−/+ (n = 4) pigs, normalized to total protein. Data are presented as mean ± the SEM; Student’s t-test, ***P < 0.001. Abbreviations: UMAP, uniform manifold approximation and projection; WT, wildtype; UBE3A^−/+, maternal UBE3A deletion.

Excitatory neurons have the most dysregulated genes in the UBE3A^−/+ pigs

Next, we analyzed each cell type cluster for differentially expressed genes (DEGs). The relative ratios of each cell type were similar and not significantly different between the WT and UBE3A^−/+ pigs (Fig. 3A), indicating that loss of maternal UBE3A does not affect the differentiation of cells in the frontal cortex. Differential expression analysis identified 3992 DEGs, including 1527 downregulated and 2465 upregulated genes in the UBE3A^−/+ pigs with a Bonferroni-adjusted P-value < 0.05 (Fig. 3B, Supplementary File 2). One hundred and eighty genes showed dual patterns of dysregulation across the cell types (i.e. downregulated and upregulated), reducing the number of unique DEGs to 3812 (Supplementary Fig. 3, Supplementary File 3). Excitatory neuron clusters 1 and 2 had the largest number of DEGs, followed by astrocytes. Overall, most DEGs (3154/3812 = 82.7%) were detected in the excitatory neuronal clusters (Supplementary File 3). The top upregulated and downregulated genes, sorted by fold-change and adjusted P-value, are shown in Fig. 3C and Supplementary File 4. As expected, UBE3A was one of the top downregulated genes in the neuronal clusters, except for the Inhibitory_2 cluster (Fig. 3C). Using quantitative RT-PCR and bulk RNA-sequencing, we confirmed the dysregulation of several genes in the UBE3A^−/+ cortex in two age groups (10 and 140-day-old [Supplementary Fig. 4, Supplementary File 5]), including the downregulation of the integral membrane protein 2B (ITM2B) and somatostatin (SST) genes. Because cluster size influences statistical power for DEG detection, these results should be interpreted in the context of cell-type abundance. Since larger clusters will often yield a greater number of DEGs, we also performed a perturbation analysis, which regresses the cluster size to normalize the number of DEGs per cluster [32]. The results confirmed that both excitatory and inhibitory neurons are among the most transcriptionally impacted cell types after controlling for cluster size (Fig. 3D). These findings suggest that the majority of detected dysregulated genes occur in neuronal cells in the UBE3A^−/+ pigs, likely reflecting both the neuron-specific imprinting of UBE3A and greater statistical power in more abundant clusters.

Figure 3.

Excitatory neurons have the most dysregulated genes in the UBE3A^−/+ pigs. (A) Proportions of cell types in the WT and UBE3A^−/+ pigs. (B) DEGs identified in the UBE3A^−/+ pigs by cell type. (C) Top 15 upregulated and downregulated high confidence DEGs per excitatory neurons, inhibitory neurons, glia, oligodendrocytes and precursor cells, and endothelial cells. Color corresponds to P-value; red = upregulated; blue = downregulated. (D) Perturbation analysis of DEGs by cell type. Abbreviations: DEGs, differentially expressed genes; WT, wildtype; UBE3A^−/+, maternal UBE3A deletion.

Multiple pathways are dysregulated in excitatory neurons in the UBE3A^−/+ pigs

To identify molecular and cellular pathways dysregulated in each cell type of the UBE3A^−/+ pigs, we first performed a molecular function gene ontology (GO) analysis of the DEGs in each cell type using DAVID (Fig. 4A). The term “ubiquitin protein ligase binding” had one of the lowest FDR values in the Astrocytes, Excitatory_1, and Excitatory_2 clusters. Significant GO terms were also observed for “unfolded protein binding,” “actin filament binding,” and “small GTPase binding” as well as the modulation of RNAs (e.g. “mRNA 3’−UTR binding,” “mRNA binding,” and “RNA binding”). We next performed a Qiagen Ingenuity Pathway Analysis of high-confidence DEGs (P-adj < 0.05; log2FC > ±1) (Fig. 4B), which generates a pathway score (activation or inhibition) based on the measured upregulated or downregulated genes in a pathway. Results showed multiple dysregulated pathways for the excitatory neuronal clusters, whereas only a few dysregulated pathways were detected in the Inhibitory_2, Microglia, and Oligodendrocytes clusters, and no dysregulated pathways in the remaining clusters (Fig. 4B, Supplementary File 6). The top dysregulated pathways in the neuronal clusters involved mitochondrial function (Oxidative Phosphorylation, Mitochondrial Dysfunction, and Respiratory Electron Transport), Hedgehog ‘Off’ State, NCAM Signaling for Neurite Out-Growth, Microautophagy Signaling, and TP53 Regulates Metabolic Genes. Perhaps not surprisingly, Excitatory neuron clusters 1 and 2 contained the largest number of dysregulated pathways and the most significant P-values and Z-scores (Supplementary File 6). Further analysis of the high-confidence DEGs revealed an enrichment of dysregulated genes involved in mitochondrial function, protein ubiquitination, and synaptic function (Fig. 4C – E). The top 40 DEGs for Excitatory Synaptic Function are shown in Fig. 4E. Altogether, these findings indicate that excitatory neurons are the most highly affected cell type in the frontal cortex of neonatal UBE3A^−/+ pigs, with several dysregulated genes in pathways involving protein ubiquitination, mitochondrial function, and synaptic function. While these pathways are consistent with known functions of UBE3A, many are commonly enriched in transcriptomic studies of neuronal perturbation and should therefore be interpreted as indicative of general classes of dysregulation rather than specific mechanistic effects.

Figure 4.

Multiple pathways are dysregulated in excitatory neurons in the UBE3A^−/+ pigs. (A) Molecular function gene ontology (GO) enrichment terms by cell type. (B) Pathway analysis results are shown by cell type for pathways with significant -log(P-values) (> 1.3) and significant Z-scores (< −2 or > 2). Blue denotes pathway inhibition; red denotes pathway activation. (C) Mitochondrial function DEGs with log2 fold change > 1 for the combined excitatory neuron analysis, excitatory_1, excitatory_2, and excitatory_3 are shown. (D) Protein ubiquitination DEGs with log2 fold change > 1 for the combined excitatory neuron analysis, Excitatory_1, Excitatory_2, and Excitatory_3 are shown. (E) the top 20 downregulated and top 20 upregulated DEGs (log2 fold change > 1) for synaptic function are shown for the combined excitatory analysis and the three excitatory subclusters, and all synaptic function genes are shown for the combined inhibitory analysis and the two inhibitory neuron subclusters. Abbreviations: DEGs, differentially expressed genes.

Somatostatin (SST) expression is reduced in UBE3A maternal deletion (UBE3A^−/+) pigs

In the single nuclei RNA-sequencing, we found that SST was reduced in the Inhibitory_2 cluster and that the somatostatin receptors SSTR1, SSTR2, and SSTR3 were also differentially expressed in excitatory neurons of the frontal cortex of the UBE3A^−/+ pigs (Supplementary Fig. 4, Supplementary File 2), suggesting that the loss of maternal UBE3A expression disrupts this pathway. Somatostatin is a neuromodulator released from a subpopulation of inhibitory GABAergic neurons, where it functions to increase the strength of the inhibitory signal when co-released with GABA [33–35]. Outside of the CNS, it also regulates digestive hormones in the gastrointestinal tract. To confirm these findings and evaluate SST as a candidate biomarker for Angelman syndrome, we analyzed SST expression in different brain regions and in the blood and cerebrospinal fluid (CSF) of UBE3A^−/+ and WT pigs at different ages.

Quantitative RT-PCR analysis showed that SST expression was significantly reduced in the cortex of 10-day-old UBE3A^−/+ pigs relative to WT pigs (26% reduction, P = < 0.0001, Mixed-effect linear regression model [Fig. 5A]), including the frontal, temporal, and occipital regions (P = 0.002; P = 0.02; P = 0.02). SST expression was slightly reduced in the parietal region but not significantly different (P = 0.2). Similarly, SST expression was slightly reduced in the frontal cortex of embryonic day 103 UBE3A^−/+ pigs (14% reduction, P = 0.3, Student’s t-test [Fig. 5B]) and significantly reduced in the frontal cortex of 60-day-old (27% reduction, P = 0.002, Student’s t-test [Fig. 5B]) and 140-day-old UBE3A^−/+ pigs (26% reduction, P = 0.04, Student’s t-test [Fig. 5B]), indicating that dysregulation of SST persists into adult animals. To determine if reduced SST expression in the brain correlates with lower levels in circulation, we quantified SST protein levels in 10-day-old pigs using an ELISA assay. Results showed that the average serum SST protein levels were significantly lower in the UBE3A^−/+ pigs compared to the WT pigs (WT mean = 190.6 pg/mL; UBE3A^−/+ mean = 85.2 pg/mL, P = 0.04, Student’s t-test [Fig. 5C]); however, plasma SST protein levels in 120-day-old pigs were only slightly lower and not significantly different (WT mean = 1123 pg/mL; UBE3A^−/+ mean = 908.5 pg/mL, P = 0.3, Student’s t-test [Fig. 5C]). SST protein levels in the CSF were similar and not significantly different between the UBE3A^−/+ and WT pigs (WT mean = 38.1 pg/mL; UBE3A^−/+ mean = 37.2 pg/mL, P = 0.7, Student’s t-test [Fig. 5C]). Likewise, bulk RNA-sequencing showed that SST expression was not reduced in the spinal cord of UBE3A^−/+ pigs compared to the WT pigs (Supplementary File 7).

Figure 5.

Somatostatin SST expression is reduced in UBE3A maternal deletion (UBE3A^−/+) pigs. (A) Quantitative RT-PCR of SST in day 10 frontal, temporal, and occipital cortex, but not parietal cortex (WT n = 5–8, UBE3A^−/+ n = 7–10). Data are presented as mean ± SEM; Student’s t-test and mixed-effect linear regression model *P < 0.05; **P < 0.01; ns, not significant. (B) Quantitative RT-PCR of SST in gestation day 103 (WT n = 5, UBE3A^−/+ n = 6) and postnatal day 60 (WT n = 9, UBE3A^−/+ n = 12) and 140 (WT n = 6, UBE3A^−/+ n = 6). Data are presented as mean ± SEM; Student’s t-test, *P < 0.05; **P < 0.01; ns, not significant. (C) SST protein measured by ELISA in the serum of 10-day-old (WT n = 7, UBE3A^−/+ n = 9), plasma of 120-day-old (WT n = 5, UBE3A^−/+ n = 5) and CSF measured in 120-day olds pigs (WT n = 10, UBE3A^−/+ n = 9). Data are presented as mean ± SEM; Student’s t-test, *P < 0.05; ns, not significant. (D) SST qRT-PCR results in frontal cortex, hypothalamus, pancreas, and stomach in adolescent (120-day-old) WT (n = 6–15), UBE3A^+/− (n = 4–6), and UBE3A^−/+ (n = 5–11) animals. Data are presented as mean ± SEM; ANOVA followed by a Tukey post-hoc test, *P < 0.05; ns, not significant. Abbreviations: WT, wildtype; UBE3A^+/−, UBE3A paternal deletion; UBE3A^−/+, UBE3A maternal deletion; SEM, standard error of the mean; ELISA, enzyme-linked immunosorbent assay; ANOVA, analysis of variance.

Serum SST is derived from both the hypothalamus and gastrointestinal (GI) tract [36, 37]. To determine if the reduced serum SST protein expression in the UBE3A^−/+ pigs is due to the loss of UBE3A in the hypothalamus or haploinsufficiency in the GI tract, we examined SST expression in the frontal cortex, hypothalamus, pancreas, and stomach of 120-day-old WT, UBE3A^−/+, and UBE3A^+/− pigs. In the frontal cortex and hypothalamus, SST expression was similar and not significantly different between the WT and UBE3A^+/− pigs, but it was reduced in the UBE3A^−/+ pigs, with several pigs having almost no detectable SST expression (Fig. 5D). In contrast, SST expression was similar and not significantly different across all three genotypes in the pancreas and stomach (Fig. 5D). Thus, loss of UBE3A expression in the hypothalamus of the UBE3A^−/+ but not the UBE3A^+/− pigs is associated with a reduction in SST expression in the hypothalamus, which in turn leads to reduced SST protein in circulation. However, variability across animals and developmental stages suggests that additional factors influence SST expression and secretion.

Discussion

Using single nuclei RNA-sequencing, we identified 3812 differentially expressed genes (DEGs) in the frontal cortex of the neonatal UBE3A^−/+ pigs. Excitatory and inhibitory neurons were the main cell types affected by the loss of maternal UBE3A, consistent with the neuron-specific imprinting of UBE3A. Pathway analyses revealed that these DEGs were enriched in several pathways, including the production of ATP in mitochondria, protein ubiquitination, and synaptic function.

Gene dysregulation is greatest in neuronal cell types

Our findings suggest that the majority of DEGs (83%) and affected pathways were detected in excitatory neurons. Likewise, the perturbation analysis, which corrects for the abundance of cells in the analysis, indicated that both the excitatory and inhibitory neurons were the most affected cell types. This large effect on gene dysregulation in neurons—and not non-neuronal cells—is consistent with the neuron-specific imprinting of UBE3A, further supporting the notion that loss of UBE3A expression in CNS neurons is the primary cause of the symptoms associated with Angelman syndrome. Indeed, DEGs were detected in non-neuronal cells, but our pathway analyses only detected a few perturbed pathways, suggesting that UBE3A is haplosufficient in these cell types. This conclusion is supported by the lack of overt phenotypes in both animal models and people with a paternally inherited deletion of UBE3A [25, 38–40].

Functions of dysregulated genes

Our findings indicate that loss of maternal UBE3A is associated with dysregulation of genes involved in ATP production, protein ubiquitination, and synaptic function, consistent with known roles of UBE3A. Several of these affected pathways overlap with those reported by Pandya et al. [41], who used proteomic approaches to identify dysregulated proteins and pathways in mouse and rat models of Angelman syndrome and iPSC-derived neurons from individuals with Angelman syndrome. The concurrent findings of these transcriptomic and proteomic data support the involvement of mitochondrial, protein homeostasis, and synaptic pathways in Angelman syndrome pathophysiology. However, notable differences were also observed between the datasets. For example, while Pandya et al. reported increased abundance of several proteasome-associated proteins, we observed predominantly reduced expression of genes related to proteasomal function. It remains unclear if the changes in gene expression precede protein abundance alterations or are secondary responses to protein alterations. Unlike Pandya et al., we did not observe dysregulated gene expression of aminoacyl tRNA synthetases (AIMP1, MARS1, YARS, WARS), suggesting that changes in these protein abundances occur post-transcriptionally (Supplementary File 2). These discrepancies may reflect differences between transcriptomic and proteomic measurements, model systems, developmental stage, or compensatory regulatory mechanisms.

Our findings reveal broad transcriptional dysregulation across multiple molecular pathways, which may contribute to the pathology of Angelman syndrome. Several of the top dysregulated genes in Fig. 3C are related to oxidative stress in the brain, including SOD1, PRDX4, PRDX2, PRDX5, GPX1, and HBB. SOD1, superoxide dismutase 1, has previously been identified as a substrate of UBE3A [42]. SOD1 mutations are known for causing familial amyotrophic lateral sclerosis, a fatal and progressive neurodegenerative disease [43]. Previously, mitochondrial dysfunction and elevated reactive oxygen species have been found to be affected in Angelman models [44]. Our top DEGs are consistent with these findings and highlight the transcriptional changes in oxidative reactivity caused by the loss of UBE3A. We also observe dysregulation of genes in the S100 family, a group of calcium-binding proteins that play key roles in calcium modulation and the formation of dendrites and synapses. Previously, signatures of disrupted calcium signaling have been observed in an Angelman mouse model [11]. Both S100A1 and S100B are among the genes showing dual patterns of dysregulation with upregulation in some cell types and downregulation in others (Supplementary Fig. 3, Supplementary File 3). Additionally, we found that the S100 Family Signaling pathway is the top activated pathway in excitatory neurons (Supplementary File 8). The pattern of dysregulation in S100 signaling across the different cell types may indicate degraded cell-to-cell interactions and poor regulation of the intercellular calcium concentrations. Finally, one of the top upregulated genes across the cell types is CELF2 (CUGBP Elav-like family member 2), an RNA binding protein which regulates mRNA translation and alternative polyadenylation. Our pathway and functional annotation analyses of DEGs also returned terms related to RNA binding (Fig. 4A and B). Additional studies should examine if the loss of maternal UBE3A alters splicing through the dysregulation of CELF2 expression.

Reduced somatostatin (SST) suggests impaired inhibitory signaling

Our results showing that SST expression is reduced in the cortex and hypothalamus of the UBE3A^−/+ pigs are also consistent with prior studies [45–47]. SST functions as a neurotransmitter in GABAergic neurons, where it is released concurrently with GABA, thereby strengthening the inhibitory signal of these neurons [45]. Our finding that SST is reduced in SST+ inhibitory neurons is consistent with prior observations that inhibitory signals are altered in a mouse model of Angelman syndrome, although the number of SST+ neurons is unaffected [46]. Kim et al. [47] previously found that Sst is reduced in the hippocampus of Ube3a^−/− mice. SST is crucial for long-term potentiation, plasticity, and motor learning, and a reduction of SST may contribute to the cognitive deficits observed in individuals with Angelman syndrome [48–50]. Furthermore, specific loss of SST+ neurons in the dentate gyrus has been associated with seizure activity, and SST and its analogs been investigated for antiepileptic properties [51, 52]. Cortical SST is regulated independently of sex hormones; however, hypothalamic SST levels are transcriptionally regulated by sex hormones including testosterone and estrogen [53, 54]. In our data, we did not observe an effect due to sex; however, the animals investigated were not sexually mature (Supplementary File 9).

Study limitations

We acknowledge several limitations of the study. First, although we confirmed several DEGs in older animals, the results of this study are mainly limited to the frontal cortex of neonatal animals, and thus, additional studies are needed to evaluate the dysregulation of genes in additional brain regions and at different ages. Second, we had limited success in validating several DEGs using the bulk tissue methods (i.e. quantitative RT-PCRs and bulk RNA-sequencing). We expect this is due to the heterogeneous cell populations present in the tissue samples, but further investigation using techniques with single-cell resolution is required. Third, the single-nuclei isolation protocol and distribution of cells within the tissue inevitably lead to variability in the proportions of cell types, with cells present in higher proportions having greater statistical power to distinguish differential gene expression. For instance, inhibitory neurons, which were the least abundant cell type in this study, also had the fewest number of DEGs. In the perturbation analysis, however, inhibitory neurons had the highest perturbation score after excitatory neurons, indicating that a high proportion of genes are dysregulated in this cell type but likely missed due to their abundance. We attempted to overcome this situation by combining the excitatory and inhibitory clusters (Supplementary File 10). Indeed, additional DEGs were identified by combining the cell types; however, DEGs were also lost (e.g. SST in Inhibitory_2), highlighting the utility of the cell type clustering approach in identifying DEGs.

Conclusions

In conclusion, we demonstrate that the loss of maternal UBE3A expression primarily dysregulates gene expression in neurons, but not non-neuronal cells, in a neonatal pig model of Angelman syndrome. Our study and findings highlight the use of a single-cell approach to identify dysregulated genes and pathways in a large animal model of Angelman syndrome, providing greater clarity on the cell-type-specific dysregulation caused by the loss of maternal UBE3A expression, which is subject to a complex cell-type-specific pattern of regulation.

Author contributions

Ashley Nicole Coffell (Conceptualization, Data curation, Formal analysis, Writing—original draft), Livia Schuller (Data curation), Sarah Christian (Project administration, Writing—review & editing), and Scott Victor Dindot (Funding acquisition, Supervision, Writing—review & editing)

Conflicts of interest

S.V.D. has an equity interest and is an employee at Ultragenyx Pharmaceutical.

Funding

This work was supported by the Chancellor’s EDGES Fellowship Program at Texas A&M University.

Data availability

The bulk and single cell RNA-sequencing data generated during this study will be made publicly available in the Gene Expression Omnibus (GEO) database upon publication.

Use of artificial intelligence

Microsoft Copilot was used to proofread and suggest edits to the final draft of the article.