Angelman Syndrome Protein Ube3a Regulates Synaptic Growth and Endocytosis by Inhibiting BMP Signaling in Drosophila

Altered expression of the E3 ubiquitin ligase UBE3A, which is involved in protein degradation through the proteasome-mediated pathway, is associated with neurodevelopmental and behavioral defects observed in Angelman syndrome (AS) and autism. However, little is known about the neuronal function of UBE3A and the pathogenesis of UBE3A-associated disorders. To understand the in vivo function of UBE3A in the nervous system, we generated multiple mutations of ube3a, the Drosophila ortholog of UBE3A. We found a significantly increased number of total boutons and satellite boutons in conjunction with compromised endocytosis in the neuromuscular junctions (NMJs) of ube3a mutants compared to the wild type. Genetic and biochemical analysis showed upregulation of bone morphogenetic protein (BMP) signaling in the nervous system of ube3a mutants. An immunochemical study revealed a specific increase in the protein level of Thickveins (Tkv), a type I BMP receptor, but not other BMP receptors Wishful thinking (Wit) and Saxophone (Sax), in ube3a mutants. Ube3a was associated with and specifically ubiquitinated lysine 227 within the cytoplasmic tail of Tkv, and promoted its proteasomal degradation in Schneider 2 cells. Negative regulation of Tkv by Ube3a was conserved in mammalian cells. These results reveal a critical role for Ube3a in regulating NMJ synapse development by repressing BMP signaling. This study sheds new light onto the neuronal functions of UBE3A and provides novel perspectives for understanding the pathogenesis of UBE3A-associated disorders.


Introduction
Angelman syndrome (AS) is a neurodevelopmental disorder characterized by severe mental retardation, developmental delay, ataxia, seizures, speech impairment, and happy disposition [1,2]. It is caused by disruption of the ubiquitin protein ligase E3A (UBE3A), which is expressed in the brain primarily from the maternal allele as a result of tissue-specific genomic imprinting [3,4]. While loss of UBE3A causes AS, a maternal 15q11-13 duplication encompassing UBE3A results in autism [5][6][7]. A point mutation of UBE3A identified in an autism proband disrupts its phosphorylation by protein kinase A and increases its ligase activity [8]. Thus, both loss and gain of UBE3A function are associated with neurodevelopmental and cognitive defects.
UBE3A is involved in ubiquitin-proteasome-mediated protein degradation. Although several substrates of UBE3A have been identified, few have been implicated in neural development and function [9]. Ube3a mutant mice, a valid model of AS, show impairment in hippocampal long-term potentiation (LTP) [10,11] and experience-dependent synaptic plasticity in the visual cortex [12,13], together with imbalanced excitatory/inhibitory neurotransmission [14] and altered intrinsic membrane properties [15]. Moreover, Ube3a has been firmly implicated in the regulation of dendrites and spine development in neurons [16][17][18]. While the underpinnings of UBE3A-associated disorders remain to be elucidated, a recent study has shown that Ube3a targets the small-conductance potassium channel SK2 for degradation, and decreased NMDA receptor activation due to an elevated SK2 protein level underlies the impaired LTP in Ube3a mice [11]. Abnormal development and function of excitatory synapses in conjunction with a reduced number of AMPA receptors in Ube3a mutant mice are attributed to decreased Ube3a-mediated proteasomal degradation of RhoA guanine exchange factor (GEF) Ephexin5 and synaptic protein Arc [19,20]. However, later studies showed that Arc is not a direct substrate of UBE3A [21,22]. Thus, the molecular mechanisms by which UBE3A regulates synapse development and function remain unclear.
Null mutants of ube3a, the Drosophila homolog of human UBE3A [23], show defects in locomotive behavior, circadian rhythms, and long-term memory [24,25], as well as reduced dendritic branching and growth of terminal dendritic processes of sensory neurons [26]. Similar to ube3a loss-of-function mutants, overexpression of wild-type ube3a also leads to motor abnormalities, circadian rhythm defects, and decreased dendritic branching [24][25][26]. So far, it has not been shown if Drosophila ube3a regulates synapse development. No direct targets of Ube3a-involved proteasomal degradation have been definitely identified in Drosophila, despite a series of proteomic studies [23,27,28].
Bone morphogenetic protein (BMP) pathway plays a crucial role in neural development throughout evolution [29][30][31]. In this study, we show that Drosophila Ube3a regulates synapse development and function at neuromuscular junctions (NMJs) by ubiquitinating and promoting proteasome-mediated degradation of the type I BMP receptor Tkv, and that the negative regulation of Tkv by Ube3a is conserved in mammalian cells. These results offer novel insight into the intricate regulation of BMP signaling in the nervous system, and suggest a potential intervention strategy for UBE3A-associated disorders by manipulating BMP signaling via genetic and pharmacological means.

ube3a mutants are viable and fertile
To examine the functions of Ube3a, we generated ube3a mutants through P-element-mediated excisions of EP3214, an insertion in the first exon. Multiple imprecise excisions (e.g., ube3a 8 and ube3a 35 ) were obtained ( Fig 1B). In addition, two nonsense mutations (ube3a 549 and ube3a 689 ) induced by the chemical mutagen ethyl methanesulfonate (EMS) were identified via a TILLING service (see Materials and Methods) ( Fig 1B). Western blotting using a mouse monoclonal antibody against the C-terminal peptide containing the HECT (Fig 1B) confirmed the absence of Ube3a protein in hemizygous ube3a 549 /Df(3L)ED4470 (which removes the ube3a locus completely) and ube3a 689 /Df(3L)ED4470 mutants and in homozygous ube3a 35 mutants, whereas a truncated protein of 110 kDa was detected in ube3a 8 mutants (Fig 1C). Immunostaining of larval and adult tissues using the monoclonal Ube3a antibody 8F7 showed that the endogenous Ube3a was cytoplasmic and widely expressed in various tissues including ventral ganglia, muscles, wing discs, ovaries and testes with an obvious enrichment at presynaptic NMJ terminals (S1 and S2 Figs). Both homozygous and hemizygous ube3a 35 mutants, as well as independent null mutants generated in other laboratories [24,26], were fully viable with no apparent developmental abnormalities, and both male and female adults were fertile.

ube3a regulates NMJ development in presynaptic neurons
Given the synaptic localization of Ube3a in cultured hippocampal neurons [16], the impaired LTP and defective synaptic transmission and plasticity observed in Ube3a mutant mice [10,11,14], and the presynaptic enrichment of Ube3a (S1 Fig), we hypothesized that Drosophila ube3a may play an important role at synapses. To test this possibility, we examined the NMJs of wandering third instar larvae by staining them with anti-cysteine string protein (CSP), a synaptic vesicle marker, and anti-horseradish peroxidase (HRP), which labels the presynaptic neuronal membrane. In ube3a 35 null mutants, the total bouton number for muscle 4 NMJs in abdominal segments A2 and A3 was 33.4 ± 1.2, which was significantly higher than 26.8 ± 1.1 in wild-type (WT) larvae (P < 0.001) (Fig 2A and 2B). Similar to homozygous ube3a 35 mutants, allelic combination of ube3a 35 /ube3a 549 and hemizygous ube3a 35 /Df(3L)ED4470 and ube3a 549 /Df(3L) ED4470 also showed more boutons than the wild type. The number of satellite boutons budding from a parental bouton in ube3a 35 mutants was also significantly higher than in wild-type larvae; these satellites were more frequent at the distal ends of NMJ terminals (Fig 2A and 2C).
To confirm that the observed phenotype was attributable to loss of ube3a, we conducted rescue experiments using a genomic transgene. One copy of wild-type ube3a restored the number of total and satellite boutons in ube3a 35 mutants (Fig 2A, 2B and 2C). However, a ube3a genomic transgene carrying the missense mutation C941A disrupting the thioester bond with ubiquitin did not (Fig 2A, 2B and 2C), although it was expressed at a level comparable to the wild-type genomic transgene.
To determine whether ube3a exerts its effect on the pre-or postsynaptic side, we knocked down Ube3a expression by RNA interference (RNAi) in a tissue-specific manner. Targeted knockdown of ube3a in presynaptic neurons by elav-Gal4-driven Thu3266 RNAi resulted in a modest but significant increase in the number of total boutons (20.15%) and a dramatic increase in satellite boutons (371.7%) compared to wild-type larvae (Fig 2). Motoneuron specific knockdown of ube3a by OK6-Gal4 also showed overgrown NMJs (S3 Fig), whereas knockdown of ube3a in postsynaptic muscles using C57-Gal4 had no effect on bouton number (P > 0.05 compared to wild type; Fig 2). Similar results were observed using an independent RNAi line VDRC45876. These findings together demonstrate that Ube3a regulates NMJ synapse growth presynaptically. Intron-exon organization and various mutations of ube3a. Coding and non-coding exons are represented by black and empty rectangles, respectively. Triangles indicate nonsense mutations; the HECT domain and the peptide used for antibody production are indicated. Two deletions, ube3a 8 and ube3a 35 , produced by imprecise excision of the P-element insertion EP3214 are depicted by interrupted lines. Genomic transgenes carrying wild-type (ube3a + ) and mutant (ube3a C941A ) containing a missense mutation in the codon encoding the catalytic cysteine residue ube3a are shown. (C) Western results of adult head extracts from various genotypes using the monoclonal antibody 8F7 against Ube3a. NOE denotes elav-Gal4/UAS-ube3a. No Ube3a expression was observed in hemizygous ube3a 549 (ube3a 549 /Df(3L)ED4470) and ube3a 689 (ube3a 689 /Df(3L) ED4470) mutants as well as homozygous ube3a 35 mutants. A truncated Ube3a of~110 kDa was observed in ube3a 8 homozygous mutants. Tubulin was used as a loading control.

Synaptic endocytosis is impaired in ube3a mutants
An increased number of satellite boutons at NMJs is closely associated with a defect in endocytosis [32][33][34]. To find out if ube3a mutant NMJs exhibited an endocytic defect, we first examined NMJ synapses by electron microscopy. The critical synaptic structures such as active zones and subsynaptic reticulum appeared normal in ube3a mutant boutons. Specifically, the mean density of synaptic vesicles (SVs) within a 200 nm radius of the active zone of transmitter release was normal in ube3a mutants compared to wild type (S4 Fig), as reported previously [35]. However, significantly more enlarged vesicles > 60 nm in diameter (also known as The precise excision line P95 was used as the wild-type control. ube3a 35 homozygotes showed more total boutons and satellite boutons (indicated by arrows) than wild-type larvae. A copy of wild-type genomic ube3a transgene (g + ) rescued the NMJ phenotype of ube3a 35 mutants, however, a copy of ube3a transgene harboring the missense mutation C941A (g C941A ) did not. elav-Gal4-driven Thu3266 RNAi against ube3a in presynaptic neurons led to increased numbers of total and satellite boutons, while C57-Gal4-driven postsynaptic knockdown of Ube3a had no effect on NMJ growth. Scale bar = 10 μm. (B, C) Quantification of total bouton (B) and satellite bouton (C) number in different genotypes including wild type, ube3a 35 , genomic ube3a + /+; ube3a 35 , genomic ube3a C941A /+; ube3a 35 , elav-Gal4/+; RNAi/+, and RNAi/+; C57-Gal4/+. n ! 18 NMJs; **P < 0.01, ***P < 0.001; ns, not significant. Error bars represent SEM. We then examined both evoked and spontaneous synaptic glutamate release at ube3a 35 mutant NMJs using intracellular recordings. We first stimulated motor neurons at a low frequency of 0.3 Hz in the presence of 0.5 mM Ca 2+ . The amplitudes of excitatory junctional potentials (EJPs) and miniature EJPs (mEJPs) in ube3a 35 mutants did not differ from those in the controls (Fig 3A and 3B). The mEJP frequency in ube3a mutants was similar to that in wild-type larvae ( Fig 3C). Representative traces of EJP and mEJP of wild type (A) and ube3a 35 mutant (B) are shown in S5 Fig. Thus, the basal transmission appeared normal in ube3a mutants. We then examined the ability of ube3a mutants to maintain neurotransmitter release during intense stimulation. The EJP amplitudes in ube3a mutants under high-frequency stimulation of 10 Hz in the presence of 0.5 mM extracellular Ca 2+ for 10 min declined much faster and remained at 64% of the initial response (Fig 3). In contrast, control animals sustained release at about 85% of the initial amplitude of EJP ( Fig 3D). The inability of ube3a mutants to maintain normal levels of transmission during intense activity is consistent with a specific defect in vesicle trafficking or recycling. Finally, to examine whether endocytosis is affected in ube3a mutant NMJ synapses, we performed fluorescent dye FM1-43 uptake experiments. There was intense labeling of FM1-43 at wild-type synaptic boutons, whereas ube3a 35 mutants showed significantly reduced uptake of FM1-43 by 34.1% compared to wild type (p < 0.001; Fig 3E and 3F). Importantly, the endocytosis defect could be fully rescued by one copy of the genomic ube3a transgene (Fig 3E and 3F). The failure to maintain neurotransmission under tetanic stimulation and reduced dye uptake support that ube3a is required for efficient synaptic vesicle endocytosis.
ube3a regulates NMJ growth through the BMP signaling pathway ube3a mutations resulted in more satellite boutons and endocytic defects (Figs 2 and 3). We investigated the underlying mechanism of the synaptic defects in ube3a mutants. Synapse formation requires the coordinated activity of several signaling cascades. At the Drosophila NMJ, the BMP pathway acts in a retrograde manner from muscles to motor neurons and promotes NMJ growth [30,34,[36][37][38][39]. We hypothesized that the overgrowth of synaptic boutons in ube3a mutants may be caused by elevated BMP signaling. To test this possibility, we examined genetic interactions between ube3a and the major genes of the BMP signaling pathway, tkv and mad.
Compared to wild type, muscle 4 NMJ terminals were overgrown in ube3a 35 mutants as described above, but severely underdeveloped in tkv 8 /tkv K16713 mutants (Fig 4A and 4B). In ube3a tkv double mutants, bouton numbers were similar to that of tkv single mutants ( Fig 4A  and 4B), suggesting that the overgrowth of NMJ synapses produced by loss of ube3a may be mediated by alterations in BMP signaling. Furthermore, NMJ overgrowth in ube3a mutants was dependent on the level of Tkv and the BMP effector Mad. Loss of one copy of tkv, which alone had no effect on bouton number, reversed the increase in satellite bouton number caused by loss of Ube3a (Fig 4A, 4B and 4C). Mutating one copy of mad also suppressed the overgrowth of NMJ synapses in ube3a mutants (Fig 4B and 4C). These results indicate that the excessive synaptic growth induced by loss of ube3a is mediated by elevated levels of the BMP signaling components Tkv, Mad, or both.
Increased BMP signaling in ube3a mutants BMP receptor activation leads to phosphorylation of Mad (pMad), which then translocates to the nucleus to activate target gene transcription. Elevated pMad levels are thus indicative of active BMP signaling [34,40,41]. To determine if ube3a loss-of-function does indeed lead to NMJ overgrowth through upregulated BMP signaling, we quantified the level of pMad in NMJ synapses and motoneuron nuclei. pMad staining intensity at NMJ synapses was significantly increased by 56.5% in ube3a 35 mutants compared to wild type (P < 0.001; Fig 5A and 5B). The intensity of pMad immunofluorescence was also significantly higher in motoneuron nuclei of ube3a mutants than in wild types (P < 0.001; Fig 5A and 5B). Furthermore, pMad expression was elevated in larval brain extracts as measured by western analysis (Fig 5C and 5D). One copy of the genomic ube3a transgene fully reversed the increased pMad levels at both NMJs and motoneuron nuclei of ube3a mutants (Fig 5A and 5B). These results show that ube3a negatively regulates pMad expression at both NMJs and motoneuron nuclei. Remarkably, the increased pMad staining, together with reduced dye loading and faster rundown of EJP amplitudes, in ube3a mutants is fully rescued by reducing the dose of Tkv by half (S6 Fig).
Given that Ube3a is an E3 ubiquitin ligase, we examined the protein expression levels of BMP signaling components in ube3a mutants to determine whether any are Ube3a substrates. Western blotting of larval brain extracts showed that pMad level was significantly increased by 203%, while Mad level was increased by 71% in ube3a mutants, consistent with the staining results ( Fig  5A-5D). As a high quality of antibody against Tkv was not readily available, we took advantage of a yellow fluorescent protein (YFP) gene trap line that expresses functional YFP-tagged Tkv (Tkv-YFP) in the ventral nerve cord, including motoneurons under the control of the endogenous promoter (S7 Fig). Western blotting with an anti-GFP antibody that also recognizes YFP, showed that the protein level of Tkv-YFP was increased by 216% in ube3a mutants compared to wild type (P < 0.001), while the level of the endogenous Sax and Wit, the other type I and type II BMP receptors, respectively, was unaltered ( Fig 5C and 5D). This suggests that Tkv is a primary target of Ube3a-mediated degradation. Indeed, the fluorescence intensity of elav-Gal4-driven Tkv-GFP was elevated by 59.2% at NMJ synapses of ube3a mutants compared to wild type ( Fig  5E and 5F). Quantitative real-time PCR, however, detected similar levels of tkv mRNA in the ube3a interacts genetically with BMP signaling pathway components tkv and mad in regulating NMJ growth. (A) Confocal images of muscle 4 NMJ synapses co-stained with anti-CSP (magenta) and anti-HRP (green) to reveal synaptic vesicles and presynaptic membrane, respectively. In ube3a tkv double mutants, bouton numbers were similar to those of tkv 8 /tkv K16713 single mutants. One copy of tkv 8 rescued the synapse overgrowth in ube3a 35 mutants. Scale bar = 10 μm. (B, C) Quantification of total bouton number (B) and satellite bouton number (C) of various genotypes including wild type, ube3a 35 , tkv 8 /tkv K16713 , tkv 8 / tkv K16713 ; ube3a 35 , tkv 8 /+, tkv 8 /+; ube3a 35 , mad 12 /+, mad 12 /+; ube3a 35 . n ! 18 NMJ terminals, one-way ANOVA test, ***P < 0.001, ns, not significant,. Error bars represent SEM. brains of ube3a 35 mutant and wild-type larvae (Fig 5G), indicating that the increase in Tkv protein expression in ube3a mutants is likely due to post-transcriptional regulation.
As Tkv acts upstream of Mad, and Tkv protein level is obviously upregulated in ube3a mutants, we focused on the mechanisms mediating negative regulation of Tkv by Ube3a by examining the stability of Tkv protein in S2 cells at various time points after treatment with the protein synthesis inhibitor cycloheximide. In control cells transfected with double-stranded RNAs (dsRNAs) against GFP, Tkv level decreased markedly from 1.5 h to 9 h upon drug  (Fig 6A). In Ube3a knockdown cells, however, a relatively stable level of Tkv was observed (Fig 6A and 6B). Ube3a knockdown resulted in a significantly elevated level of pMad (1.6 times the control) and Tkv (2.7 times the control) at 48 h after dsRNA transfection (Fig 6C  and 6D). This effect was confirmed by using a distinct dsRNA targeting a different sequence of ube3a mRNA, demonstrating the specificity of Tkv downregulation by Ube3a.
To determine whether Tkv stability is regulated by the proteasomal degradation pathway, Tkv protein level was measured in Ube3a-overexpressing S2 cells treated with the 26S proteasome inhibitor MG132. In contrast to the upregulation of Tkv observed upon Ube3a knockdown, Tkv levels were significantly reduced by 50.17% in Ube3a-overexpressing cells compared with GFP-overexpressing control cells, and this effect was abolished by MG132 treatment (Fig 6E). These results indicate that Ube3a negatively regulates Tkv protein levels by targeting the protein for proteasome-mediated degradation.

Tkv is a direct substrate of Ube3a
Given the increased stability of Tkv in Ube3a knockdown cells and the involvement of the proteasome in Tkv degradation (Fig 6), Tkv may be a direct target for Ube3a-mediated ubiquitination and ensuing proteasome-mediated degradation. This model predicts a physical interaction between the two, and direct ubiquitination of Tkv by Ube3a. We investigated the physical interaction between Ube3a and Tkv by co-immunoprecipitation (co-IP). In third instar larval brain extracts, endogenous Ube3a co-precipitated with Tkv-YFP using an antibody against GFP, but was absent in control immunoprecipitates using anti-IgG (Fig 7A). We defined the regions mediating the interaction in S2 cells and found that the N-terminal 201-400 and 400-640 amino acid fragments of Ube3a each interacted with Tkv, while the STYKc domain of Tkv mediated its interaction with Ube3a (S8 Fig). Co-localization of Tkv-GFP and Ube3a driven by elav-Gal4 was observed in the soma of motoneurons in the ventral ganglion (6.3 puncta positive for both Tkv-GFP and Ube3a per neuron, n = 16; Fig 7B), further supporting a physical interaction between Ube3a and Tkv. To establish whether Tkv is a direct substrate of Ube3a-mediated ubiquitination, we measured the level of unmodified and ubiquitinated Tkv in S2 cells. In Ube3a knockdown cells, Tkv poly-ubiquitination was reduced (Fig 7C). Conversely, overexpression of Ube3a caused an increase in Tkv poly-ubiquitination (Fig 7D). Poly-ubiquitination occurred at lysine 48 of ubiquitin ( Fig 7D); a form of poly-ubiquitination that targets proteins for proteasomal degradation [42]. An in vitro ubiquitination assay further revealed that GST-Ube3a induced polyubiquitination of myc-TkvC, while neither GST alone, the control E3 ligase Parkin, nor a cocktail missing the essential component ubiquitin, led to TkvC poly-ubiquitination (Fig 7E). To verify if the negative of BMP signaling by Ube3a occurred in vivo, we treated dissected larvae with MG132 for 4 h and observed a significantly increased level of pMad at both NMJs and motoneuron nuclei (S9 Fig). Blocking proteasome function by expressing a dominant negative proteasomal subunit DTS5 in presynaptic neurons by elav-Gal4 also resulted in a similar increase of pMad staining (S9 Fig). Western blotting of larval brain lysates showed a decreased level of ubiquitinated Tkv in ube3a mutants (Fig 7F). These results demonstrate that Ube3a promotes ubiquitin-mediated degradation of Tkv.
Given that Ube3a ubiquitinated Tkv and promoted its degradation by proteasome pathway (Figs 6 and 7), we aimed to identify specific amino acids of Tkv targeted by UBE3A. Mass spectrometry revealed two potential ubiquitinated residues, lysine 218 and 227, in the C-terminal STYKc region of Tkv (Fig 8A). To verify which specific sites of Tkv were ubiquitinated by Ube3a, we mutated lysine to arginine at two individual sites, namely, K218R and K227R. We compared the stability of wild-type and mutant Tkv protein in S2 cells by western blotting following treatment with cycloheximide to inhibit protein synthesis for 0, 3 and 6 h. We found that the stability of K218R was comparable to that of wild-type Tkv, the levels of both decreased with time, while the amount of K227R remained stable within 6 h treatment of cycloheximide (Fig 8B and 8C). Ubiquitination analysis showed that K218R was as sensitive as wildtype Tkv for ubiquitination in S2 cells by Ube3a in the presence of HA-Ub and MG132 (Fig 8D  and 8E). There was a concomitant significant reduction in wild-type and mutant Tkv protein level, while K227R was stable and ubiquitination resistant (Fig 8D and 8E). These results demonstrate that K227 is a key target residue for ubiquitination by Ube3a.

Human UBE3A binds and ubiquitinates ALK3, a mammalian homolog of Tkv
Given that Ube3a is evolutionarily conserved from Drosophila to vertebrates, we hypothesized that human UBE3A would interact with and negatively regulate ALK3, a mammalian homolog of Tkv [43]. To test this possibility, we examined the co-immunoprecipitation of ALK3 with UBE3A in lysates of human HEK293 cells. An anti-HA antibody, but not a control anti-IgG, co-precipitated myc-ALK3 with HA-UBE3A (Fig 9A). UBE3A knockdown in HEK293 cells using two different small interfering RNAs (siRNAs) produced a significantly increased level of endogenous ALK3 as well as a near two-fold increase in the level of pSmad; the mammalian counterpart of Drosophila pMad (Fig 9B and 9D). Conversely, overexpression of UBE3A in HEK293 cells led to a significant decrease in ALK3 protein level (38.1% of the control; n = 3, P < 0.001), which was blocked by the proteasome inhibitor MG132 (Fig 9C), implying that human UBE3A promotes the proteasomal degradation of ALK3. Overexpression of wild type but not the ligase dead C833A mutant UBE3A significantly increased ubiquitination of myc-ALK3, while GFP control did not (Fig 9E). Despite the fact that the corresponding K227 in ALK3 is not critical for ubiquitination in HEK293 cells (S10 Fig), consistent with the fact that some proteins have little specificity in ubiquitination sites [44], our results demonstrate that Ube3a-mediated suppression of BMP signaling is conserved in mammalian cells.

Ube3a regulates synapse growth and endocytosis by inhibiting BMP signaling
The synapse is a critical substrate for cognitive function, so disorders associated with intellectual and cognitive deficits, such as AS and autism, likely involve altered pathways that disrupt Tkv was more stable than K218R or wild-type Tkv. S2 cells transfected with Ube3a and wild-type or mutant Tkv were treated with cycloheximide to block protein synthesis, and the level of myc-Tkv was detected to determine protein stability. (C) Statistical analysis of wild-type and mutant Tkv protein levels at different time points after protein synthesis blockade by cycloheximide (n = 3, mean ± SEM). (D) Verification of Tkv ubiquitination sites targeted by Ube3a. S2 cells were transfected with plasmids encoding Ube3a and wild-type or mutant Tkv. After 48 h transfection, the level of Tkv was examined by western blotting (upper blot). For ubiquitination assays, S2 cells over-expressing Ube3a and wildtype or mutant Tkv, together with HA-ubiquitin (+HA-Ub) and treated with MG132 (+MG132) to block ubiquitin-mediated degradation were immunoprecipitated with anti-myc antibody. Western blotting was performed to analyze the levels of ubiquitinated and un-ubiquitinated Tkv. (E) Statistical analysis of wild-type and mutant Tkv protein levels normalized to the loading control tubulin when co-expressed with Ube3a in S2 cells (n = 3, one-way ANOVA test, **P < 0.01, mean ± SEM). synaptic development and plasticity. The present study shows for the first time that ube3a regulates synapse formation and synaptic endocytosis at the Drosophila NMJ. A previous study reported normal NMJ growth in ube3a 15B mutants [35]. However, we showed here that the genetic background in homozygous ube3a 15B mutants may suppress expression of satellite phenotype, as trans-allelic ube3a 15B /ube3a 35 and hemizygous ube3a 15B /Def mutants showed excess satellite boutons (S3 Fig). We noted that the satellite phenotype in ube3a mutants is mild compared to the classical endocytic endophilin and dap160/intersectin mutants [32,45,46]. Also, the fewer and larger SVs at active zones accompanied by an increase in amplitudes of mEJPs reported in endocytic mutants such as tweek and dap160/intersectin mutants [45][46][47], were not observed in ube3a mutants. However, the significantly higher number of satellite boutons, reduced dye uptake, and defective recovery of EJP amplitudes under high-frequency stimulation, implicate a regulatory role for ube3a in endocytosis at NMJs. It is worth noting that a similar endocytic function has been reported for mammalian Ube3a at inhibitory synapses in the visual cortex [14].
There appears to be a reciprocal regulation between BMP signaling and endocytosis at the Drosophila NMJ. Endocytic mutants show upregulated BMP signaling as demonstrated by increased pMad staining at NMJs [40]. Trio, a Rac GTPase-specific GEF, is positively regulated by pMad at the transcriptional level in motoneuron nuclei for proper NMJ development [48]. Conversely, upregulation of BMP signaling by mutations in Dad, an inhibitory Mad that negatively regulates BMP signaling, and in S6KL (S6 kinase like) results in apparent endocytic defects [33,34]. In addition to the mutual regulation between the two processes, the multi-domain actin regulator Nwk functions at the interface of endocytosis and BMP signaling by physically interacting with both Tkv and endocytic proteins dynamin and Dap160/intersectin [40]. We found normal expression of endocytic proteins in ube3a mutant NMJs (S11 Fig), and endocytic endophilin A mutants have been shown previously to have a normal level of Tkv [34]. Thus, we favor a model in which upregulated BMP signaling leads to overgrown NMJs and endocytic defects in ube3a mutants, and the NMJ defects in ube3a mutants could be attributable to altered actin-cytoskeletal formation via Trio-Rac-mediated and/or Nwk-mediated pathways.
While ube3a mutation leads to an elevated level of BMP signaling, neuronal overexpression of Ube3a by elav-Gal4 or OK6-Gal4 does not result in downregulation of BMP signaling. This was demonstrated by overgrown NMJ terminals and a normal level of pMad at both NMJs and motoneuron nuclei (S12 Fig). These results indicate that overexpression of Ube3a alone is not sufficient to downregulate BMP signaling in the neuromusculature. How Ube3a overexpression leads to overgrown NMJs remains to be determined. Similar overgrown NMJ phenotypes caused by Ube3a mutation and overexpression are consistent with previous reports of defective locomotion [24] and circadian rhythms [25] upon bidirectional changes in Ube3a expression.

Ube3a ubiquitinates and promotes proteasomal degradation of the BMP receptor Tkv
Using a series of biochemical assays, we showed that Ube3a physically interacts with Tkv and facilitates its proteasome-mediated degradation (Figs 6 and 7). This function of Ube3a was conserved in mammalian cell cultures (Fig 9). We identified K227 but not K218 in the cytoplasmic region of Tkv as being critical for its ubiquitination (Fig 8). We demonstrated that Ube3a normally functions specifically to downregulate Tkv expression by proteasomal degradation, resulting in reduced presynaptic BMP signaling. In the absence of Ube3a, BMP signaling is upregulated, resulting in overgrown NMJs and endocytic defects.
The impact of Ube3a on Tkv is limited to the NMJ, as altered expression of Ube3a does not lead to corresponding changes in Tkv expression in wing discs, where the role of Tkv is well known (S2 Fig). Thus, the negative regulation of Tkv by Ube3a occurs in a tissue specific manner.
We recently identified that an evolutionarily conserved, previously uncharacterized S6KL similarly regulates NMJ growth and endocytosis by promoting proteasomal degradation of Tkv [34]. As Ube3a biochemically and genetically interacts with S6KL (S13 Fig), it remains possible that Ube3a and S6KL work synergistically in inhibiting NMJ growth by targeting Tkv for degradation In addition to Ube3a, another HECT-domain-containing E3 ligase Smurf also inhibits BMP signaling by targeting two independent signaling components, BMP receptors and downstream effector Smad in distinct developmental processes [43,52,53]. However, Smurf does not regulate NMJ growth [34]. As the protein level of Mad was mildly but significantly increased in ube3a mutant brains (Fig 5C and 5D), it would be of interest to determine if Ube3a also targets Mad for proteasome-mediated degradation in NMJ development.
In addition to its role as an ubiquitin E3 ligase, UBE3A has been reported to affect nuclear hormone receptor-mediated transcription by E3 ligase-independent mechanisms [21,54]. For instance, UBE3A was shown to inhibit the expression of the cyclin-dependent kinase inhibitor p27 at both transcriptional and post-translational levels [55]. Moreover, UBE3A inhibits estradiol-mediated transcription of the target gene Arc [21]. As with mammalian UBE3A, Drosophila Ube3a also acts as a transcriptional regulator; it activates transcription of Punch, which encodes GTP cyclohydrolase 1, a component of the monoamine biosynthesis pathway [27]. In support of a transcriptional role, Ube3a orthologs of different species are localized in the nucleus [16,23,27]. However, the present study reveals that Ube3a functions as an ubiquitin ligase for Tkv but not as a transcriptional regulator since the level of tkv mRNA in ube3a brains was similar to wild type (Fig 5G).

Involvement of BMP signaling upregulation in the pathophysiology of AS
Human UBE3A negatively regulated the protein levels of ALK3, and consequently the downstream effector pSmad, in HEK293 cells (Fig 9). Therefore, it is tempting to speculate that elevated BMP signaling may contribute to a subset of the abnormalities observed in patients with AS. BMP receptors are expressed at specific sites, such as the hippocampus and cerebellum, where Ube3a is also expressed [10,56]. It remains to be determined if BMP signaling is upregulated in these brain regions that correlate with the clinical features of seizures, learning deficit and ataxia in patients with AS.
BMP signaling regulates synapse development specifically at the calyx of Held in the auditory system in a way similar to that at Drosophila NMJ, but shows no effect on synapse development in the lateral superior olive of the auditory nucleus [31]. Given that AS mice show an enhanced seizure-like response to audiogenic challenge [22], it would be improtant to determine if the development and function of the calyx of Held synapses are normal in Ube3a mutant mice. At the microscopic level, AS model mice exhibit normal dendritic elaboration but fewer and shorter dendritic spines in cerebellar Purkinje cells and pyramidal neurons in the hippocampus and cortex [16]. It is not known if BMP signaling plays a role in the formation of dendritic spines, although other targets of Ube3a such as Ephexin5 may regulate spine morphogenesis in hippocampal neurons [20].
In summary, we demonstrated that Drosophila ube3a plays an important role in regulating synapse formation and endocytosis by inhibiting BMP signaling via proteasome-mediated degradation of Tkv. The negative regulation of BMP signaling by Ube3a is conserved in mammals. These findings indicate that increased BMP signaling resulting from loss of UBE3A during development may contribute, at least in part, to the etiology of AS.

Drosophila stocks
All Drosophila stocks were fed standard cornmeal food and maintained at 25°C. w 1118 was used as a wild-type control unless otherwise indicated. We generated intragenic ube3a deletions (ube3a 8 and ube3a 35 ) through P-element-mediated excisions of EP3214. Transgenic lines expressing genomic ube3a + and ube3a C941A (containing a missense mutation in the codon for the critical catalytic cysteine residue) were from J. Fischer. UAS-Tkv-GFP was from T. B. Kornberg. UAS-DTS5 was a gift from J. Belote via K. Broadie [51]. The EMS-induced nonsense mutations ube3a 549 and ube3a 689 were obtained from a TILLING (targeting-induced local lesions in genomes) service (http://tilling.fhcrc.org). For tissue-specific rescue experiments, the pan-neural elav-Gal4 (from Bloomington) and muscle-specific C57-Gal4 (from V. Budnik) lines were used. RNAi lines P{GD450}v45876 and Thu3266 were obtained from the Vienna Drosophila RNAi Center (VDRC) and Tsinghua University (Beijing, China), respectively. A yellow fluorescent protein (YFP) genetrap line labeling the endogenous Tkv with YFP was obtained from the Drosophila Genome Research Center (stock number 115298; Kyoto, Japan). EP(3)3214, Df(3L)ED4470, tkv 8 , tkv K16713 , mad 12 , and wit B11 were obtained from the Bloomington Stock Center.
For overexpression studies, UAS-ube3a was obtained by amplifying the full-length cDNA of ube3a from a home-made cDNA pool derived from whole adults, and cloned into the vector pUAST. An UAS-ube3a insertion in the X chromosome was used in this study.

Generation of monoclonal antibodies against Ube3a
To generate specific antibodies against Ube3a, a fusion protein containing amino acids 568-973 of Drosophila Ube3a with an N-terminal 6×His tag was produced in E. coli. Mouse monoclonal antibodies produced by clone 8F7 were shown to be specific for Ube3a by western blotting and immunostaining.
For statistical analysis of NMJ morphology, the number of total boutons and satellite boutons of muscle 4 NMJ in abdominal segments A2 and A3 was quantified as previously described [57].

Electrophysiological and FM1-43 uptake assays
Electrophysiological recordings were performed at 20°C essentially following a conventional procedure with minor modifications [33,34]. For the basal transmission assay, wandering third-instar larvae were dissected in Ca 2+ -free HL3.1 saline and recorded in HL3.1 saline containing 0.5 mM CaCl 2 . Intracellular microelectrodes with a resistance greater than 5 MO filled with 3 M KCl were used for the assay. Excitatory junction potential (EJPs) and miniature EJPs (mEJPs) were recorded from muscle 6 in the abdominal segments A2 or A3 for 120 seconds. EJPs were elicited by low frequency (0.3 Hz) stimulation. We analyzed recordings from muscles cells with physiological resting potentials -60 mV and input resistances > 5 MO. For examining synaptic transmission under tetanic stimulation, synapses were stimulated at 10 Hz for 10 minutes and recorded in a modified HL3.1 saline with 0.5 mM extracellular Ca 2+ . For the FM1-43 uptake assay, we followed a previously published protocol [34,58].
To examine Tkv protein turnover, S2 cells were transfected with dsRNAs against ube3a or gfp (control) at an average transfection efficiency of about 46%, and treated 48 h later with 40 μg/ml cycloheximide (Sigma) for 0, 1.5, 3, 6, and 9 h. To determine whether the degradation of Tkv was proteasome-dependent, S2 cells were co-transfected with flag-Ube3a or flag-GFP plasmid together with myc-Tkv. After 48 h, cells were treated with 20 μM MG132 (Sigma) or mock treated for 6 h before being harvested for western blotting.

Immunoprecipitation (IP) and western blotting
IP assays of larval brains and ventral ganglia were carried out according to a previous report [61]. Immunoprecipitation from S2 or HEK293 cell lysates was performed following a previously described protocol [43]. Antibodies used for western blotting were: anti-myc (1:1000;

In vivo and in vitro ubiquitination assay
For in vivo ubiquitination assay, S2 cells were transfected with plasmids using Cellfectin II reagent (Invitrogen). The ubiquitination assay was performed according to the protocols described previously [43]. In brief, at 48 hr posttransfection, MG132 was added into the media at the final concentration 50μM. Cells were harvested 4 hr later and lysed with a lysis buffer (50 mM Tris [pH 7.5], 120 mM NaCl, and 0.5% NP40) containing 1% (w/v) SDS to disrupt protein-protein interaction and boiled for 10 min, then diluted 10 times with cell lysis buffer.
To detect the levels of ubiquitinated Tkv in vivo, 200 larval brains of each genotype (elav-Gal4; tkv-GFP and elav-Gal4; tkv-GFP; ube3a 35 ) were dissected and incubated with 50 μM MG132 (Sigma) in Schneider's medium for 4 h to inhibit proteasome-mediated degradation before proceeding with IP. The brains were then homogenized in lysis buffer and the lysates were immunoprecipitated with anti-GFP. The ubiquitinated Tkv was examined by anti-ubiquitin antibody. For in vitro ubiquitination assay, the myc-tagged Tkv C-terminus (myc-TkvC) was synthesized by an in vitro translation kit containing rabbit reticulocyte lysates (Promega). Ube3a was fused to GST, then expressed and purified from E. coli. E1, E2, HA-Ub (all three from Boston Biochem), and the potential substrate myc-TkvC, together with the GST control, the E3 ligase Parkin control, or GST-Ube3a, were incubated at 30°C for 1.5 h in 40 μl ubiquitination reaction buffer (50 mM Tris-HCl (pH 7.5), 1 mM dithiothreitol, 50 mM NaCl, 5 mM MgCl 2 , 2 mM ATP). Reactions were terminated by adding SDS-PAGE sample buffer and analyzed by western blotting.

Mass spectrometry for identification of ubiquitination sites
Mass spectrometry was performed as described previously [62]. myc-Tkv expressed in S2 cells was immunoprecipitated with an anti-myc antibody. The target protein was excised, digested with trypsin, fractionated by HPLC, and analyzed by a LTQ Orbitrap Elite mass spectrometer (ThermoFisher Scientific, Waltham, MA). To further verify the sites of Tkv targeted by Ube3a, myc-Tkv mutants K218R and K227R were generated by site-directed mutagenesis.
Quantitative RT-PCR of tkv mRNA Total RNA was extracted from brains and ventral nerve ganglia of 40 third instar larvae following the standard Trizol reagent protocol (Invitrogen). Power SYBR Green PCR Master Mix (Applied Biosystems) was used for quantitative real-time PCR. Actin mRNA level was amplified as an internal control using published primers [33], whereas tkv cDNA was amplified with primers 5'-GTG ATA GGG CAG GGC GTA GT-3' and 5'-AGT GGG TCT CGT TCT GTG GG-3'.

Statistical analysis
Student's t tests were used for statistical comparisons between two groups. Multiple group means were compared by one-way ANOVA. Asterisks above a column indicate comparisons between a specific genotype and wild type, whereas asterisks above a bracket denote comparisons between two specific genotypes. Data are presented as mean ± standard error of the mean (SEM). P-values < 0.05 were considered statistically significant.