Autism-linked UBE3A gain-of-function mutation causes interneuron and behavioral phenotypes when inherited maternally or paternally in mice

Abstract

The E3 ubiquitin ligase Ube3a is biallelically expressed in neural progenitors and glial cells, suggesting that UBE3A gain-of-function mutations might cause neurodevelopmental disorders irrespective of parent of origin. Here, we engineered a mouse line that harbors an autism-linked UBE3A^T485A (T503A in mouse) gain-of-function mutation and evaluated phenotypes in animals that inherited the mutant allele paternally, maternally, or from both parents. We find that paternally and maternally expressed UBE3A^T503A results in elevated UBE3A activity in neural progenitors and glial cells. Expression of UBE3A^T503A from the maternal allele, but not the paternal one, leads to a persistent elevation of UBE3A activity in neurons. Mutant mice display behavioral phenotypes that differ by parent of origin. Expression of UBE3A^T503A, irrespective of its parent of origin, promotes transient embryonic expansion of Zcchc12 lineage interneurons. Phenotypes of Ube3a^T503A mice are distinct from Angelman syndrome model mice. Our study has clinical implications for a growing number of disease-linked UBE3A gain-of-function mutations.

Keywords: Angelman syndrome; CP: Neuroscience; UBE3A; autism; neurodevelopmental disorders; single-cell RNA-seq.

Figure 1.

Haplotype phasing the UBE3AT485A mutation in the autism proband SNPs on chromosome 15 near UBE3A identified from whole-exome sequence data of the parents and haplotype phased whole-genome sequence data from the autism proband (family ID: 13873). The T>C mutation resulting in UBE3AT485A co-segregates with the paternal haplotype in the autism proband. Genomic coordinates are based on hgGRCh37/hg19. Clinical data associated with this proband are available from the Simons Simplex Collection.

Figure 2.. Generation of the Ube3a^T503A mouse model

(A) The ACT codon encoding T503 in mouse UBE3A (mouse NP_001380595.1, synonymous with human UBE3A^T485A) was mutagenized to GCA encoding alanine. A silent A>G point mutation was introduced to create an AluI restriction site to identify the Ube3a^T503A allele. aa, amino acid. (B) Mutagenized nucleotides (asterisks) were detected by Sanger sequencing of PCR-amplified genome fragments from a heterozygous Ube3a^T503A mouse. (C) Electrophoresis of DNA fragments amplified from WT and homozygous and heterozygous Ube3a^T503A mice by PCR before and after AluI digestion. (D–F) UBE3A protein levels from the cerebral cortex of (D) E14.5, (E) P0, and (F) adult mutant and WT mice (n = 4). (G) Western blot analyses of endogenous proteins that associate with UBE3A (n = 6). (H) Western blot analyses of UBE3A, PSMD4, and PSMD1 from adult WT and AS model mice (n = 7). For (D)–(H), protein levels were normalized to b-ACTIN. Levels of protein (mean ± SEM) relative to WT are shown below the representative blots. For (G)–(H), proteins with significant changes are in red. See also Figures S1 and S2.

Figure 3.. UBE3A protein levels in cortical neurons and progenitors of Ube3a^T503A mice

(A–C) Immunostaining of UBE3A in the P0 cerebral cortex (A). Progenitor cells in VZ/SVZ were co-labeled by PAX6 (a–d) and UBE3A (a′–d′). Quantification of UBE3A levels in (B) the cortical plate and (C) VZ/SVZ (WT, n = 6; homoT503A, matT503A and patT503A, n = 7). (D–F) Immunostaining of UBE3A (green) in cortical neurons (NEUN, red) of adult brains (D). Examples of nuclear areas (DAPI, blue) were highlighted by dotted lines. (E) Reduced UBE3A protein levels and (F) altered cytoplasmic/nuclear distribution of UBE3A in cortical neurons of homoT503A and matT503A, but not patT503A, mice compared with controls (n = 3). (G and H) Immunostaining of UBE3A in cortical oligodendrocytes (OLIG2⁺, red) of adult brains (G). (H) Reduced UBE3A (green) levels in OLIG2⁺ cells in homoT503A, matT503A, and patT503A mutant mice compared with WT controls (n = 4). Examples of OLIG2⁺ cells are indicated by arrows. Data are represented as mean ± SEM.

Figure 4.. Cortical development is not grossly affected in Ube3a^T503A mutant mice

(A–C) Brain and body weight of homoT503A (WT, n = 12; homo, n = 15; from 4 litters) (A), matT503A (WT, n = 34; matT503A, n = 35; from 9 litters) (B), and patT503A (WT, n = 39; patT503A, n = 53; from 13 litters) (C) mice compared with WT mice at P0. (D and E) Cortical thickness and lamination of P0 matT503A (D) and patT503A (E) mice (main effect of genotype, F (2, 12) = 2.466, p = 0.1267). Layers 2–4, CUX1⁺ in red; layers 5–6, CTIP2⁺ in green (WT, n = 7; matT503A and patT503A, n = 4). (F and G) Body (main effect of genotype, F (3, 71) = 0.7956, p = 0.5004) (F) and brain weight (main effect of genotype, F (3, 71) = 0.8223, p = 0.4859) (G) of mice at 3 months (WT, n = 35; homoT503A, n = 18; matT503A, n = 12; patT503A, n = 10). (H–J) Normal cortical lamination of adult homoT503A, matT503A, and patT503A mutant mice (H). (I) The thickness of cortical layers (main effect of genotype, F (3, 21) = 0.4553, p = 0.7164). (J) Neuronal numbers in each cortical layer (layer 1, CTIP2⁺; layers 2–4, CUX1⁺; layers 5 and 6, CTIP2⁺) (main effect of genotype, F (3, 21) = 0.4231, p = 0.7384). Data are represented as mean ± SEM.

Figure 5.. Differentially expressed genes in the cerebral cortex of Ube3a^T503A mutant mice across the lifespan

(A) Heatmaps summarizing DEGs in homoT503A, patT503A, and matT503A mice (standardized log2 fold change relative to WT control) at E14.5, 1 month,6 months, and 12 months of age. DEGs were categorized into nine clusters based on their temporal changes in transcript levels across all four ages. Examples of genes in each cluster are listed left of the heatmap graph. See also Figure S4. (B) Averaged (centroid, in yellow) and individual trajectories of all differentially expressed genes in each group. (C) GO analysis of DEGs.

Figure 6.. A cortical interneuron subclass is transiently expanded in embryonic UBE3A^T503A mutant mice

(A) Pseudotime analysis of interneurons (Int1, Int2, Int2_Zic⁺, Int3, Int4) and progenitors in the GEs revealed two distinct interneuron developmental trajectories derived from GE progenitor cells. See also Figure S5. (B) Unique gene expression signatures of Int2 and Int2_Zic⁺ interneurons among cells in lineage 1, ordered by pseudotime. Transcripts of Zcchc12, Tmem130, Hap1, Peg1, Sst, and Calb2 are highly enriched in Int2 and Int2_Zic⁺ interneurons. (C–E) Transient increase of Int2 interneurons validated by HCR single-molecule RNA hybridization. (C) A representative image showing HCR detection of Gad2⁺ (magenta) and Zcchc12⁺ (green) cells in the cortex at E14.5. Gad2⁺ cells were highlighted by circles. Gad2⁺ cells that express high levels of Zcchc12 transcripts (R5 particles) counted as Gad2⁺/Zcchc12⁺ cells are indicated by arrows. Bar: 10 μm. (D) A transient increase of Gad2⁺/Zcchc12⁺ cells was observed in matT503A, patT503A, and homoT503A mice at E14.5 (one-way ANOVA, F (3, 28) = 16.96, p < 0.0001) (n = 8 for all genotypes). ****p < 0.0001. (E) Comparable proportions of Gad2⁺/Zcchc12⁺ cells between WT and mutant mice at P0 (one-way ANOVA, F (3, 19) = 0.1053, p = 0.9560) (WT, n = 5; homo, n = 4; matT503A, n = 6; matT503A, n = 8 animals). Data are represented as mean ± SEM. (F–H) HCR detection of Gad2⁺ (blue), Zcchc12⁺ (green), and mature cortical interneuron markers (red), Calb2 (F), Sst (G), or Pvalb (H). Boxed areas (a), (b), and (c) were enlarged and are shown as (a′), (b′), and (c′) for triple labeling, (a″), (b″) and (c″) for Zcchc12/Gad2 double labeling, and (a‴), (b‴), and (c‴) for interneuron subtype-specific markers. Examples of Gad2⁺/Zcchc12⁺ cells are indicated by arrows. Bar: 50 μm.

Figure 7.. Behavioral phenotypes differ based on parent of origin of Ube3a^T503A allele and age

(A) Summary of mouse behavioral tests. homoT503A vs. WT, n = 12; matT503A vs. WT, n = 14 except open-field tests (n = 14 at 18 and 30–32 weeks, n = 8 at 43–44 and 50–52 weeks); patT503A vs. WT, n = 12. (B–D) Total distance traveled as a function of age for (B) homoT503A (genotype × time interaction, F (3, 66) = 0.4913, p = 0.6895; main effect of genotype, F (1, 22) = 13.69, p = 0.0012); (C) matT503A (genotype × time interaction, F (3, 80) = 1.092, p = 0.3574; main effect of genotype, F (1, 80) = 47.17, p < 0.0001); and (D) patT503A (genotype × time interaction, F (3, 66) = 0.5747, p = 0.6337; main effect of genotype, F (1, 22) = 3.636, p = 0.0697) mice and WT control. (E–G) Time spent in the center as a function of age for (E) homoT503A (genotype × time interaction, F (3, 66) = 1.855, p = 0.1459; main effect of genotype, F (1, 22) = 4.919, p = 0.0372); (F) matT503A (genotype × time interaction, F (3, 54) = 2.033, p = 0.1202; main effect of genotype, F (1, 26) = 13.87, p = 0.001); and (G) patT503A (genotype × time interaction, F (3, 66) = 0.2716, p = 0.8457; main effect of genotype, F (1, 22) = 0.5475, p = 0.4671) mice and WT control. (H–J) Rearing movements as a function of age for (H) homoT503A (genotype × time interaction, F (3, 66) = 1.102, p = 0.3546; main effect of genotype, F (1, 22) = 15.96, p = 0.0006); (I) matT503A (genotype × time interaction, F (3, 54) = 0.1444, p = 0.9328; main effect of genotype, F (1, 26) = 0.6869, p = 0.4148); and (J) patT503A (genotype × time interaction, F (3, 66) = 1.295, p = 0.2836; main effect of genotype, F (1, 22) = 2.659, p = 0.1172) mice and WT control. (K–M) Rotarod tests of (K) homoT503A (genotype × trial interaction, F (4, 88) = 0.7928, p = 0.5330; main effect of genotype, F (1, 22) = 1.334, p = 0.2604; main effect of trial, F (4, 88) = 15.40, p < 0.0001); (L) matT503A (genotype × trial interaction, F (4, 104) = 2.232, p = 0.0.705; main effect of genotype, F (1, 26) = 2.592, p = 0.1195; main effect of trial, F (4, 104) = 11.56, p < 0.0001; ***p < 0.005, uncorrected Fisher’s least significant difference [LSD] test); and (M) patT503A mice (genotype × trial interaction, F (4, 88) = 0.2453, p = 0.9118; main effect of genotype, F (1, 22) = 2.220, p = 0.1505; main effect of trial, F (4, 88) = 10, p < 0.0001). (N–S) Time spent in proximity. (N and O) HomoT503A and WT mice. (N) No difference between homoT503A and WT mice in sociability (genotype × side interaction, F (1, 22) = 0.07381, p = 0.7884) and (O) social novelty (genotype × side interaction, F (1, 22) = 2.231, p = 0.1495) tests. (P and Q) matT503A and WT mice. (P) No difference between matT503A and WT mice in sociability (genotype × side interaction, F (1, 26) = 0.1647, p = 0.6882) and (Q) social novelty (genotype × side interaction, F (1, 26) = 0.434, p = 0.5158) tests. (R and S) PatT503A and WT mice. (R) patT503A mice showed deficits in sociability test (genotype × side interaction, F (1, 22) = 7.462, p = 0.0122) but (S) normal social novelty tests (genotype × side interaction, F (1, 22) = 1.786, p = 0.1950). Unlike WT mice, which spend more time exploring the cage with stranger #1, patT503A mice did not show a preference for the social interactor over an empty cage. (T–Y) Entry times as the measurement of three-chamber social behavior tests. (T) Significant differences were observed in homoT503A mice in both sociability (main effect of genotype, F (1, 44) = 6.826, p = 0.0122) and (U) social novelty tests (main effect of genotype, F (1, 44) = 32.98, p < 0.0001) compared with WT controls. (V) The entry times for matT503A mice were not different from WT controls in sociability tests (F (1, 52) = 1.733, p = 0.1938) but were significantly increased in (W) social novelty tests (F (1, 52) = 8.953, p = 0.0042) compared with WT controls. (X) The entry times of patT503A mice did not differ from WT controls in either the sociability test (main effect of genotype: F (1, 44) = 0.009679, p = 0.9221) or (Y) the social novelty test (main effect of genotype: F (1, 44) = 0.4528, p = 0.5045). Data are represented as mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001, ****p < 0.0001. See also Figure S6.