Biochemical analysis of Angelman syndrome-associated mutations in the E3 ubiquitin ligase E6-associated protein.

Angelman syndrome is a severe neurological disorder characterized by mental retardation, absent speech, ataxia, seizures, and hyperactivity. The gene affected in this disorder is UBE3A, the gene encoding the E6-associated protein (E6AP) ubiquitin-protein ligase. Most patients have chromosomal deletions that remove the entire maternal allele of UBE3A. However, a small subset of patients have E6AP point mutations that result in single amino acid changes or short in-frame deletions that still allow translation of a full-length protein. By studying these point mutations in E6AP, we found a strong correlation between Angelman-associated mutations and a loss of E3 ubiquitin ligase activity. Interestingly the point mutations affect E6AP activity in different ways. Some mutant proteins cannot form thiol ester intermediates with ubiquitin, others retain the thiol ester formation activity but cannot efficiently transfer ubiquitin to a substrate, and still others are unstable in cells. Our results suggest that the loss of E6AP catalytic activity and likely the improper regulation of E6AP substrate(s) are important in the development of Angelman syndrome.

Angelman syndrome is a severe neurological disorder characterized by mental retardation, absent speech, ataxia, seizures, and hyperactivity. The gene affected in this disorder is UBE3A, the gene encoding the E6-associated protein (E6AP) ubiquitin-protein ligase. Most patients have chromosomal deletions that remove the entire maternal allele of UBE3A. However, a small subset of patients have E6AP point mutations that result in single amino acid changes or short in-frame deletions that still allow translation of a full-length protein. By studying these point mutations in E6AP, we found a strong correlation between Angelman-associated mutations and a loss of E3 ubiquitin ligase activity. Interestingly the point mutations affect E6AP activity in different ways. Some mutant proteins cannot form thiol ester intermediates with ubiquitin, others retain the thiol ester formation activity but cannot efficiently transfer ubiquitin to a substrate, and still others are unstable in cells. Our results suggest that the loss of E6AP catalytic activity and likely the improper regulation of E6AP substrate(s) are important in the development of Angelman syndrome.
Angelman syndrome is a severe neurological disorder characterized by mental retardation, absent speech, ataxia, seizures, and hyperactivity (1). The candidate gene affected is UBE3A, which encodes the E6AP 1 ubiquitin-protein ligase (2,3). The UBE3A locus is imprinted in certain regions of the brain such that only the maternal copy of the gene is expressed (4 -6), and most genetic abnormalities identified in Angelman patients result in the loss of maternal E6AP expression. Approximately 70% of patients possess maternal specific deletions of chromosome 15q11-13, which encompasses the UBE3A locus, 3-4% of patients have two paternal copies of the chromosome, and another 3-4% have defects in which the maternal chromosome possesses the paternal imprint such that no functional maternal copy of UBE3A is present in brain (1).
Although the deletion of other genes in 15q11-13, such as the ␥-aminobutyric acid neurotransmitter receptor ␤ genes, may contribute to the severity of Angelman symptoms (7), the identification of UBE3A point mutations in numerous patients provides the most compelling evidence for E6AP being the Angelman gene (2,3). The majority of these mutations result in frameshifts and/or premature truncations, but some are missense mutations or short in-frame deletions that still allow the translation of a full-length (or nearly full-length) protein (8,9).
E6AP was initially identified as the cellular factor that cooperates with the human papillomavirus E6 oncoprotein to stimulate the ubiquitin-mediated degradation of the tumor suppressor p53 (10). In this process, the E6⅐E6AP complex comprises the E3 ubiquitin ligase activity in a ubiquitin thiol ester cascade (11). E6AP was the first identified member of a family of ubiquitin ligases called hect proteins, all of which contain a domain homologous to the E6AP carboxyl terminus (12). These are modular proteins in which the conserved hect domains catalyze ubiquitin transfer, whereas other divergent domains confer substrate specificity.
Because E6AP is a ubiquitin ligase, it is likely that the improper regulation of E6AP substrate(s) causes Angelman syndrome, although no disease-relevant E6AP targets have yet been identified. Recent structural data show that several of the Angelman-associated point mutations actually map to the E6AP catalytic cleft (13) and may affect E6AP ubiquitin ligase activity. Other Angelman-associated point mutations map to the E6AP amino terminus and are outside the catalytic domain. These mutations might affect E6AP in a manner distinct from ubiquitin transfer perhaps by compromising substrate binding, altering E6AP subcellular localization, or destabilizing the protein. To determine whether the loss of E6AP catalytic activity correlates with Angelman syndrome, we examined the activities of the non-truncating point mutations identified in Angelman patients.

MATERIALS AND METHODS
Protein Purification-His 6 -E6AP (wild type and mutant) baculoviruses were generated using the BaculoGold system (Pharmingen) according to the manufacturer's instructions. E6AP proteins were purified from two T175 flasks (Corning) of baculovirus-infected Hi5 cells. Forty-eight hours after infection, cells were harvested and lysed in Nonidet P-40 lysis buffer (1% Nonidet P-40, 50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 0.5 mM DTT, 50 mM NaF, 0.2 mM Na 3 VO 4 , and protease inhibitor mixture (Pharmingen)). Lysates were rotated for 30 min at 4°C and cleared by centrifugation (25,000 ϫ g for 30 min at 4°C). Imidazole was added to a final concentration of 10 mM, and the lysates were incubated with a 200-l bed volume of Ni 2ϩ -NTA (Qiagen) for 2 h at 4°C. Beads were washed with 20 column volumes of 10 mM imidazole wash buffer (50 mM Tris (pH 8.0), 300 mM NaCl, 10 mM imidazole, 0.5 mM DTT) and 20 column volumes of the same buffer containing 20 mM imidazole and eluted in 0.1-ml fractions of the same buffer containing 250 mM imidazole. Fractions were pooled; dialyzed against a buffer containing 25 mM Tris-HCl (pH 7.6), 50 mM NaCl, 10% glycerol, and 1 mM DTT; concentrated in Centricon YM-30 devices (Amicon); and stored at Ϫ80°C.
UbcH7 was expressed in Escherichia coli strain BL21(DE3) (Novagen). Midlog phase cultures (2 liters) were induced with 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside for 4 h at 26°C. Cells were pelleted and frozen at Ϫ80°C. Pellets were thawed on ice, resuspended in 25 ml of TDE (50 mM Tris (pH 7.6), 1 mM DTT, 0.1 mM EDTA) supplemented with 0.4 mg/ml lysozyme and a protease inhibitor mixture containing 1 g/ml aprotinin, 1 g/ml leupeptin, 1 g/ml pepstatin, and 100 g/ml phenylmethylsulfonyl fluoride), and incubated until bacteria were lysed. MgCl 2 and DNase were added to final concentrations of 10 mM and 20 g/ml, respectively, and incubated until DNA was digested. The lysate was cleared by centrifugation (25,000 ϫ g for 30 min) and passed over Q Sepharose FF (Amersham Biosciences) equilibrated with TDE. The flow-through was then loaded onto S Sepharose FF (Amersham Biosciences) equilibrated with TDE. UbcH7 was eluted stepwise with TDE containing increasing concentrations of NaCl, and UbcH7-containing fractions were concentrated in a Centricon YM-10 device (Amicon). UbcH7 was further purified on a Superdex 200 fast protein liquid chromatography column equilibrated with 50 mM Tris (pH 7.6), 1 mM DTT, and 125 mM NaCl. UbcH7-containing fractions were pooled, concentrated as above, and frozen in 10% glycerol. E1 was purchased from Boston Biochem Inc. Ubiquitin was purchased from Sigma.
GST-hect proteins were expressed in E. coli strain BL21. Midlog phase cultures (500 ml) were induced with 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside for 16 h at 15°C. Lysates were prepared as above. Soluble proteins were incubated for 2 h at 4°C with a 500-l bed volume of glutathione-Sepharose (Amersham Biosciences) equilibrated with phosphate-buffered saline. Beads were washed with 200 column volumes of phosphate-buffered saline containing 1 mM DTT. GST-HA hect wild type was eluted with 10 mM reduced glutathione in 25 mM Tris, pH 8.0. The GST tags were removed from the MYC-hect proteins by cleaving with Precission protease (Amersham Biosciences) overnight at 4°C. The cleaved proteins were dialyzed in TDE and further purified by gel filtration by fast protein liquid chromatography (Amersham Biosciences) with a Superdex 75 column equilibrated with TDE buffer. Pooled fractions were concentrated and frozen as above.
In Vitro Ubiquitylation Assay-HHR23A was in vitro translated in the presence of [ 35 S]methionine and [ 35 S]cysteine (PerkinElmer Life Sciences) using the TNT-coupled wheat germ extract system (Promega). Assays were done in 50-l volumes containing 25 mM Tris, pH 7.6, 50 mM NaCl, 8 mM ATP, 10 mM MgCl 2 , 0.2 mM DTT, 2 l of in vitro translated HHR23A, ϳ20 ng of E1, 125 ng of UbcH7, 300 ng of insectcell expressed E6AP, and 4 g of bovine ubiquitin. Reactions were incubated at 30°C for 90 min and stopped in Laemmli buffer.
For assays using in vitro translated E6AP, His 6 -E6AP was translated from pcDNA4 (Invitrogen) with or without [ 35 S]methionine and [ 35 S]cysteine (PerkinElmer Life Sciences) using the TNT-coupled rabbit reticulocyte lysate system (Promega) in 50-l volumes. The lysate was incubated for 10 min at room temperature with a 30-l bed volume of Q Sepharose FF (Amersham Biosciences) equilibrated in TDE. Beads were washed five times with 500 l of TDE containing 20 mM NaCl, and bound proteins were eluted with TDE containing 500 mM NaCl. The lysate was adjusted to 10 mM imidazole, and the proteins were incubated with a 10-l bed volume of Ni 2ϩ -NTA (Qiagen) equilibrated in the same buffer. Beads were washed five times with 500 l of TDE containing 1 M NaCl and then three times with 500 l of TDE containing 50 mM NaCl. Ubiquitylation reactions were performed as above except that they contained either one-tenth or one-fourth of the bead-bound E6AP as the E3 source, and the reactions were rotated for 2 h at 30°C.
Pulse-Chase Experiments-E6APϪ/Ϫ mouse embryo fibroblasts (15,16) stably expressing wild type or mutant E6AP were plated in 6-well dishes. The cells were starved in medium lacking methionine and cysteine for 1 h before being pulsed with the same medium supplemented with [ 35 S]methionine and [ 35 S]cysteine (PerkinElmer Life Sciences) for an additional hour. The cells were washed twice with phosphate-buffered saline and chased in medium containing excess cold methionine and cysteine for 0, 2, 4, and 8 h. Cells were harvested in 500 l of Nonidet P-40 lysis buffer (prepared as above), and protein concentrations were determined using the Bradford assay reagent (Bio-Rad). 500 g of each lysate were immunoprecipitated with anti-E6AP antiserum, resolved by 8% SDS-PAGE, and detected by autoradiography.
Plasmid Preparations-A plasmid for E6AP isoform III in the pGEM-1 vector has been described (17). All mutants were generated by the QuikChange method (Stratagene) using E6AP/pGEM-1 as a template. Primers used for mutagenesis are available by request. E6AP and mutants were excised from pGEM-1 by digesting with BamHI and HindIII. The fragments were ligated into BamHI/HindIII-digested pIND (Invitrogen). The NotI/PmeI fragment was then ligated into pAcHLT-A (Pharmingen) for baculovirus production. For mammalian expression and in vitro translation, E6AP constructs were excised from pIND (Invitrogen) using BamHI and PmeI and ligated into the BamHI/ EcoRV sites of pDNA4/HisMax-C (Invitrogen).
Wild type E6AP hect domain with an amino-terminal MYC or HA tag was generated from the E6AP/pGEM-1 template by PCR and encoded from residue 491 to the end (according to E6AP isoform I) of the hect domain. The amplified products were digested with BamHI/SmaI and ligated into pET23a (Novagen). The MYC-tagged C820A and Angelman mutant fragments were amplified from the E6AP/pIND templates by PCR, cut with BamHI/SmaI, and ligated into the BamHI/HincII sites of pET23a (Novagen) and the BamHI/SmaI sites of pGEX 6P-1 (Amersham Biosciences) for in vitro translations and bacterial expression, respectively. Mutations were confirmed by sequencing.
For stable cell lines, all E6AP constructs were cloned as follows. The BamHI/NotI fragments were excised from the pcDNA4 constructs described above and ligated into the BamHI/NotI sites of pBluescript (Stratagene). The SalI/NotI fragments were then ligated to the XhoI/ NotI sites of pOZ-C, which also contains the interleukin-2 receptor gene under control of an internal ribosomal entry site. Cells were transduced with retroviruses, and stably expressing cells were sorted using magnetic beads (Dynal) coated with anti-interleukin-2 antibody.
Yeast Two-hybrid Assays-Yeast two-hybrid assays were performed essentially as described previously (18). Briefly E6AP constructs were cloned into the SalI/NotI sites of pPC97 such that they were fused in-frame to the GAL4 DNA binding domain. HHR23A and UbcH7 cloned into pPC86 and fused in-frame to the GAL4 activation domain have been described previously (18). Yeast were co-transformed with two of the plasmids by the lithium acetate procedure and selected by plating on SD-Leu Ϫ Trp Ϫ . Single colonies were then grown in SD-Leu Ϫ Trp Ϫ broth to midlog phase, and aliquots in 10-fold dilutions were spotted on both SD-Leu Ϫ Trp Ϫ and SD-Leu Ϫ Trp Ϫ His Ϫ .
E2 Binding/Competition Assay-Bovine ubiquitin (Sigma) was biotinylated using the BiotinTag microbiotinylation kit (Sigma) according to the manufacturer's instructions. Approximately 0.5 g of GST-HAhect wild type (WT) was used in ubiquitylation assays performed as described above with the following exceptions: the ubiquitin was biotinylated, the reaction time was 30 min, and the assays were performed in the presence of increasing amounts of competitor protein, 0.3, 1, or 3 g of bovine serum albumin or MYC-hect (WT or mutant) in which the GST was removed by cleavage with Precission protease (Amersham Pharmacia). Reactions were stopped in Laemmli buffer and resolved by SDS-PAGE, and proteins were visualized with Extravidin-peroxidase (Sigma) and anti-MYC antibody (Oncogene, Cambridge, MA).

There Is a Strong Correlation between Angelman-associated
Mutations and the Loss of E6AP Ubiquitin Ligase Activity-E6AP ubiquitylation activity consists of several discrete steps. E6AP must bind to an E2 enzyme, accept ubiquitin from the E2 in the form of a thiol ester linkage, and transfer the ubiquitin to a substrate (14). We first tested whether the Angelmanassociated E6AP point mutants could ubiquitylate a model substrate, reasoning that E6AP mutants that retained this activity must be able to bind to E2 enzymes and form ubiquitin thiol esters. Mutants impaired in substrate ubiquitylation were then studied further to determine the biochemical activity affected by the mutation.
To examine E6AP ubiquitin ligase activity, we performed in vitro ubiquitylation assays using full-length wild type and mutant E6AP proteins purified from insect cells (Fig. 1). The in vitro substrate was HHR23A, which had been identified previously as an E6AP-interacting protein in a yeast two-hybrid screen (19). Radiolabeled HHR23A was generated by in vitro translation in wheat germ extract, which lacks endogenous E6AP activity (20). The addition of wild type E6AP to the reaction induced the formation of a high molecular weight ladder of HHR23A (Fig. 1A, lane 2), whereas addition of the catalytically inactive C820A mutant E6AP (14) had no effect (Fig. 1A, lane 3).
To compare the activities of the Angelman-associated E6AP Ubiquitylation reactions were performed as above, using the indicated insect cellexpressed wild type or mutant E6AP proteins. Amount of E6AP used in each reaction is indicated in the immunoblot. The schematic shows the positions Angelmanassociated mutations in the context of the full-length E6AP. The E2-binding region and the catalytic cysteine residue (Cys-820) are indicated. C, Non-hect domain E6AP mutation S349P is inactive in HHR23A ubiquitylation. In vitro translated His 6 -tagged 35 S-labeled or cold E6AP proteins were purified by anion exchange followed by affinity chromatography on Ni 2ϩ -NTA. Purified E6AP was incubated with 35 S-labeled HHR23A under standard ubiquitylation conditions. Lane 1, unprogrammed rabbit reticulocyte lysate subjected to identical purification scheme lacks an activity that ubiquitylates HHR23A. The asterisk denotes a background band. In vitro translated wild type E6AP ubiquitylates HHR23A (lanes 3-5), whereas catalytically inactive (C820A) (lanes 7-9) and Angelman-associated S349P (lanes 11-13) mutants do not. The amount of cold E6AP used was equivalent to the higher concentration of 35 S-labeled E6AP. All reactions were stopped in Laemmli buffer, resolved by 10% SDS-PAGE, and visualized by autoradiography. Ub'd, ubiquitylated; Ub, ubiquitin; Unprog RRL, unprogrammed rabbit reticulocyte lysate. point mutants to the wild type protein (Fig. 1B), we performed ubiquitylation assays using the amount of E6AP equivalent to the lowest concentration of wild type E6AP needed to catalyze the conversion of 99% of the HHR23A starting material into ubiquitylated forms. The positions of the mutants are shown schematically in Fig. 1B. Most of the Angelman-associated mutations lie within the carboxyl-terminal hect domain of the protein, whereas two lie in the non-catalytic amino-terminal portion. We predicted that the former mutations might affect E6AP enzymatic activity, whereas the amino-terminal mutations might affect substrate binding, subcellular localization, or protein stability.
The crystal structure of the hect domain reveals it to be an L-shaped molecule consisting of two lobes, the interface of which forms the catalytic cleft (13). Two Angelman-associated mutations, L502P and E550L, lie in the amino-terminal lobe of the hect domain and were completely inactive in HHR23A ubiquitylation (Fig. 1B, lanes 6 and 7). Glu-550 forms a salt bridge with a nearby arginine residue (13); therefore, the E550L mutation may affect a critical structural element of the hect domain.
The other mutations map to the carboxyl-terminal lobe of the hect domain, two of which were also severely impaired in HHR23A ubiquitylation. They are the F782⌬ mutation (Fig.  1B, lane 9), which removes a residue from the hydrophobic core of the hect domain (13), and an insertion at residue Lys-836 (Fig. 1B, lane 10). The latter mutation introduces a frameshift downstream of the catalytic cysteine residue (Cys-820) that inserts several amino acids and prematurely truncates much of the carboxyl-terminal ␣-helix of the protein.
Two other hect domain mutations consistently had partial defects in HHR23A ubiquitylation. One is the in-frame deletion of Lys-801 (Fig. 1B, lane 4), a mutation that has been identified in two unrelated patients, and the other is a missense mutation that changes Ile-804 for a lysine (I804K) (Fig. 1B, lane 8).
Another patient possesses a mutation causing the insertion of an isoleucine at position 803 (I803ins). This mutation had no discernible effect on E6AP-mediated HHR23A ubiquitylation (Fig. 1B, lane 5).
The remaining two mutations reside in the non-catalytic amino terminus of E6AP. One mutant, C21Y, retained wild type activity in the HHR23A ubiquitylation assay (Fig. 1B, lane  3) whereas the other, S349P, was largely insoluble when expressed in insect cells. As an alternative source of E6AP S349P protein, we translated a His 6 -tagged version in rabbit reticulocyte lysate and purified it by affinity chromatography to separate the in vitro translated E6AP from the endogenous E6AP in the reticulocyte lysate (Fig. 1C). Unprogrammed rabbit reticulocyte lysate subjected to the same purification scheme lacked activity (Fig. 1C, lane 1), but His 6 -wild type E6AP efficiently ubiquitylated both itself and the HHR23A substrate (Fig. 1C,  lanes 3-5). Ubiquitylated wild type E6AP appeared as a slightly slower migrating form, due to the addition of one ubiquitin, and a high molecular weight smear, indicative of polyubiquitylation. The appearance of both the monoubiquitylated form and the smear required that ubiquitin be supplemented in the reaction (data not shown), indicating that these are indeed ubiquitylated forms. Furthermore the high molecular smear could be collapsed to approximately five distinct species by the addition of methylated ubiquitin (data not shown). This indicates that the smear is due to polyubiquitylation (21) and that E6AP places ubiquitin on approximately five of its own lysine residues. The HHR23A substrate was modified by several ubiquitins, most clearly indicated by including a reaction in which His 6 -E6AP WT was translated in the presence of cold methionine so that HHR23A was the only radiolabeled protein (Fig. 1C, lane 5) in the reaction.
In vitro translated His 6 -E6AP S349P, however, was severely impaired in its ability to ubiquitylate HHR23A (Fig. 1C, lanes  11-13), identifying it as the first non-hect domain mutation to affect E6AP catalytic activity. The defect was not due to a loss of binding between E6AP S349P and HHR23A because they still interacted in a yeast two-hybrid experiment (see Fig. 3B).
Overall there was a strong correlation between the loss of substrate ubiquitylation activity and the occurrence of Angelman syndrome-associated E6AP mutations (Table I).
Of the Ubiquitylation-defective Mutants, Two Are Unable to Form Thiol Esters, Whereas Others Retain the Ability to Do So-E6AP forms a covalent thiol ester intermediate between its catalytic cysteine residue and the carboxyl-terminal glycine of ubiquitin prior to transferring the ubiquitin to a substrate (14). To qualitatively determine whether the defect in the ubiquitylation-defective Angelman-associated mutants was in this activity, we tested them in an in vitro thiol ester formation assay. We used in vitro translated, radiolabeled hect domains, which are sufficient for thiol ester formation (22). The wild type and mutant hect domains were incubated under standard ubiquitylation conditions in the presence of GST-ubiquitin and analyzed for the presence of the hect-GST-ubiquitin thiol ester adduct. The thiol ester bond is sensitive to reducing agents, so it is only detectable when the reaction is stopped in the absence of DTT. Under non-reducing conditions, the hect domain appeared as multiple species ( Fig. 2A), which likely correspond to different oxidation states of the protein. After 1 min of reaction, a slowly migrating, DTT-sensitive polypeptide appeared that was the hect-GST-ubiquitin thiol ester ( Fig. 2A, lanes 2-4). As a negative control, a hect domain possessing the C820A mutation, which cannot form a ubiquitin thiol ester (14), was included (Fig. 2B). An additional polypeptide occasionally appeared in this assay as can be seen in the E6AP C820A 4Ј ϩDTT lane. This indicates the presence of an activity in the rabbit reticulocyte lysate that can modify the E6AP hect domain during the reaction. Since E6AP C820A absolutely cannot form thiol esters with ubiquitin, this indicates the background in the reaction.
As expected, mutations that caused either no or partial effects on substrate ubiquitylation (i.e. I803ins, K801⌬, and I804K) retained the ability to form thiol ester intermediates with ubiquitin (data not shown). Two of the Angelman-associated point mutants that were completely inactive in HHR23A ubiquitylation, the in-frame Phe-782 deletion (Fig. 1B, lane 9) and the carboxyl-terminally truncating K836ins mutation (Fig.  1B, lane 10) did not form detectable thiol ester intermediates (Fig. 2, C and D), indicating that these mutations either impaired the ability of the hect domain to bind to the E2 enzyme or to efficiently receive ubiquitin from it.
Surprisingly two other ubiquitylation-defective mutants, L502P and E550L, formed hect-GST-ubiquitin thiol esters as well as wild type (Fig. 2, E and F), indicating that these mutations might impair the transfer of the thiol ester-linked ubiquitin to the substrate. We next considered the fate of the ubiquitin during the ubiquitylation assay. If these mutants formed thiol esters with ubiquitin but did not transfer them to the substrate, where did the ubiquitin go? One possibility was that the ubiquitin remained "trapped" on the E6AP catalytic cysteine, causing an apparent accumulation of thiol esterlinked adducts. However, the ubiquitin-thiol ester adducts on the L502P hect domain did not accumulate any more than on the wild type hect domain (Fig. 2) even at later time points (data not shown), and those on hect E550L actually diminished slightly after 2 and 4 min (Fig. 2E). Because there was no obvious "trapping" of the thiol ester-linked ubiquitin on these mutants, the thiol ester bond must have been either hydrolyzed or the ubiquitin passed to an acceptor other than the HHR23A substrate, perhaps to E6AP itself. To address the latter possibility, we performed HHR23A ubiquitylation assays using in vitro translated, 35 S-labeled E6AP so we could visualize both the enzyme and substrate simultaneously. For each E6AP construct tested, one reaction was performed in which HHR23A was the only radiolabeled protein to highlight the bands which correspond to ubiquitylated HHR23A (Fig. 3A, lanes 4, 8, and 12). As with the insect cell-derived proteins, the in vitro translated E6AP E550L and L502P mutants did not ubiquitylate HHR23A (Fig. 3A, lanes 8  and 12), although they did ubiquitylate themselves. The wild type E6AP and E550L E6AP looked identical, both accumulating a monoubiquitylated form and a polyubiquitylated smear (Fig. 3A, lanes 2 and 3 and lanes 6 and 7, respectively). The L502P mutant, however, ubiquitylated itself in a hyperactive and highly processive manner (Fig. 3A, lanes 10 and 11). The hyperactivity was intrinsic to this mutant because it still occurred even when no substrate was added (data not shown). Together the data indicate that the defect caused by the E550L and L502P mutations was not in catalyzing isopeptide linkages per se because the proteins still attached ubiquitins to themselves; instead the defect was specifically in the transfer of ubiquitin to the substrate.
To rule out the possibility that E6AP E550L and L502P had simply lost the ability to bind to HHR23A, these mutants were tested in a yeast two-hybrid assay (Fig. 3B). This was the original experiment that identified this interaction (18) and remains the most reliable method to detect binding between these proteins. An interaction is denoted by growth on plates lacking leucine, tryptophan, and histidine. Both E6AP L502P and E550L scored in this assay, stimulating growth of the yeast on these plates equivalent to that of the wild type E6AP. This indicates that the mutations do not affect the interaction between E6AP and HHR23A. Moreover Fig. 3B, bottom, shows that the isolated hect domain (in which the L502P and E550L mutations reside) is not sufficient for binding to HHR23A in this assay. This suggests that the interaction with HHR23A is mediated by a non-hect domain portion of E6AP, providing additional evidence that the L502P and E550L mutations do not affect HHR23A binding. The relative levels of GAL4 DBD-E6AP fusion proteins expressed in the yeast are shown in Fig.  3B. The levels of E6AP S349P are lower than that of the other E6AP proteins likely because the protein has a shorter half-life than wild type (see Fig. 5D) when expressed in cells; however, the mutant still scored for binding to HHR23A, indicating the sensitivity of the assay.

Mutants That Cannot Form Thiol Ester Intermediates with Ubiquitin Do Not Functionally Interact with the E2 Enzymes-
The defect in the Angelman-associated mutants (the F782⌬ and K836ins mutants) that could not form thiol ester intermediates with ubiquitin must either be in binding to the ubiquitin-conjugating enzyme or in receiving ubiquitin from it. To discriminate between these possibilities, we first tested for direct binding between bacterially purified hect domains and the ubiquitin-conjugating enzyme UbcH7. We could not reliably detect an interaction by pull-down experiments or gel filtration under the conditions we used likely because of the lower affinity of E3 ligases for uncharged ubiquitin-conjugating enzymes (23,24).
As an alternative to the pull-down assay, we examined the abilities of the mutant hect domains to compete with the wild type for binding to E2 enzymes in a self-ubiquitylation experiment. When wild type GST-hect was incubated with E1, UbcH7, ubiquitin, and ATP, it efficiently ubiquitylated itself (Fig. 4A, lane 1). However, when excess wild type hect protein (from which the GST moiety had been cleaved so that it would migrate distinctly from the GST-hect by SDS-PAGE) was in-FIG. 3. Substrate ubiquitylation-defective Angelman mutants retain self-ubiquitylation activity and substrate binding. A, in vitro translated His 6 -tagged 35 S-labeled or cold E6AP proteins were purified and incubated with 35 S-labeled HHR23A under standard ubiquitylation conditions. Lanes 1, 5, and 10, input E6AP proteins after purification. The amount of HHR23A used in each reaction is equivalent. The amount of cold E6AP used was equivalent to the higher concentration of 35 S-labeled E6AP (e.g. the amounts of E6AP in lanes 3 and 4 are equivalent). Open arrows indicate ubiquitylated HHR23A. All reactions were stopped in Laemmli buffer, resolved by 10% SDS-PAGE, and visualized by autoradiography. B, plasmids encoding E6AP point mutants, hect domain only, or empty vector (fused in-frame to the GAL4 DNA binding domain) were co-transformed with HHR23A (fused in-frame to the GAL4 activation domain) into yeast (see "Materials and Methods"). Transformants were grown to midlog phase in medium lacking Leu and Trp. Equal numbers of cells were spotted in serial 10-fold dilutions on plates either lacking Leu, Trp, and His or lacking only Leu and Trp. Interactions are scored as growth on the former. Relative protein levels were determined by resolving equal amounts of whole cell yeast extracts by SDS-PAGE and immunoblotting with anti-E6AP antiserum. The asterisk denotes a cross-reactive band that serves as a loading control. Ub'd, ubiquitylated; Ub, ubiquitin. cluded in the reaction, self-ubiquitylation of the GST-hect protein greatly diminished, whereas self-ubiquitylation of the smaller, untagged hect protein concomitantly increased (Fig.  4A, lanes 2 and 3). Unlike the addition of wild type hect protein, excess hect F782⌬ or K836ins had no effect on GST-hect selfubiquitylation (Fig. 4A, lanes 4 -7), indicating that these mutants could not compete effectively with the GST-hect protein for binding to the E2. Adding the nonspecific competitor bovine serum albumin had no inhibitory effect on the degree of selfubiquitylation of the GST-hect protein (Fig. 4A, lanes 8 and 9). Because the F782⌬ and K836ins mutations do not map directly to the region shown to be involved in binding to UbcH7 (13), it is possible these mutations affect the overall folding or stability of the hect domain.
Next we performed a yeast two-hybrid assay as an additional test of the binding between these two mutants and UbcH7 (Fig.  4B). This assay has been used previously to examine the E6AP/ ubiquitin-conjugating enzyme interaction (19). Neither of the mutants scored in the two-hybrid assay, suggesting that they do not interact with UbcH7. The levels of E6AP F782⌬ and K836ins protein were lower than that of the C820A mutant in the yeast likely because they have very short half-lives when expressed in cells (around 30 min compared with Ͼ8 h, Fig. 5,  G and H). However, the K836ins mutant was expressed at least as well as the E6AP S349P shown in Fig. 3B that scored for binding in the experiment. Although the levels of E6AP F782⌬ were also low, the complete lack of growth of the yeast suggests that this mutation also affects binding to UbcH7. Together the two-hybrid results and the competition assay are consistent with the finding that these mutations abrogate the interaction of E6AP with UbcH7.
Certain Angelman-associated Mutations Affect the Stability of E6AP in Cells-It remained possible that the Angelmanassociated mutations could affect the stability of E6AP in cells.
To address this, we performed pulse-chase experiments to measure the half-lives of the Angelman-associated mutant E6AP proteins. To assure that we could discriminate the mutant from the endogenous E6AP protein, we reconstituted E6APϪ/Ϫ mouse embryonic fibroblasts with wild type or mutant E6AP. Wild type (Fig. 5A) E6AP was stable over the course of the 8-h chase period in agreement with previously published data (25). Note that in vitro translated E6AP was run on the same gel to indicate the position of the unmodified E6AP.
Several of the Angelman-associated mutants were degraded within 30 min after synthesis. These include the F782⌬ and K836ins mutants (Fig. 5, G and H, respectively), which did not interact with UbcH7, the I804K mutant (Fig. 5J), which had partial activity in the HHR23A ubiquitylation assay (Fig. 1B,  lane 8), and the non-hect domain C21Y mutant (Fig. 5C), which displayed wild type activity in the in vitro ubiquitylation assay (Fig. 1B, lane 3). These results suggest that all these mutations may affect E6AP protein folding.
The other non-hect domain Angelman-associated mutant, S349P, had a half-life of approximately 3 h (Fig. 5D), which is shorter than that of the wild type. This mutant does not ubiquitylate HHR23A in vitro, so both its loss of catalytic activity and decreased half-life may contribute to the E6AP enzyme defects in the Angelman syndrome patient with this mutation. FIG. 4. Two Angelman-associated mutations impair E2 binding. A, bacterially expressed GST-HA hect WT was incubated with E1, UbcH7, and biotinylated ubiquitin in the absence (lane 1) or presence (lanes 2-9) of increasing amounts of the following competitor proteins: MYC-E6AP hect WT (lanes 2 and 3), MYC-E6AP hect ⌬F782 (lanes 4 and 5), MYC-E6AP hect K836ins (lanes 6 and  7), or bovine serum albumin (BSA) (lanes  8 and 9). The GST moieties were cleaved from the competitor hect proteins so they would migrate distinctly from the GSThect protein. Reactions were performed for 30 min at 30°C, stopped in Laemmli buffer, resolved by 10% SDS-PAGE, then probed with Extravidin-peroxidase to detect the ubiquitylated conjugates, and immunoblotted with anti-MYC antibody to detect the competitor hect proteins. Monoubiquitylated MYC-hect conjugates are detectable on the MYC-tagged wild type hect domains in lanes 2 and 3. B, yeast two-hybrid experiments performed essentially as described in Fig. 3B. The exceptions are that yeast were co-transformed with plasmids encoding UbcH7 fused to the GAL4 activation domain and one of the E6AP mutants shown fused to the GAL4 DNA binding domain. Ub, ubiquitin; CA, C820A. The asterisk denotes a cross-reactive protein that serves as a loading control.
Because the L502P mutant self-ubiquitylated so efficiently in vitro, we expected it to be less stable than wild type E6AP when expressed in cells. Consistent with the in vitro data, E6AP L502P was self-ubiquitylated much more than wild type (note the multiple polypeptides in Fig. 5F as opposed to in Fig.  5A); however, E6AP L502P was stable over the course of the chase (Fig. 5F). Since the E6AP L502P protein was highly self-ubiquitylated, why wasn't its half-life decreased? One possibility was that the E6AP L502P protein was monoubiquitylated on multiple sites rather than polyubiquitylated. This would not constitute a degradation signal since a chain of at least four ubiquitins is required for targeting proteins to the proteasome (26). Alternatively the self-ubiquitylation may have occurred after the cells were lysed. Indeed we could prevent self-ubiquitylation of E6AP L502P (and E6AP WT) by lysing the cells under denaturing conditions (data not shown). Therefore, the L502P protein was stable in cells but became readily self-ubiquitylated after cell lysis, likely during the immunoprecipitation. The other mutant that formed thiol ester adducts with ubiquitin was E6AP E550L, but this mutation had no effect on the half-life of the protein (Fig. 5E).
The half-life of the other partially active hect domain mutant, K801⌬ (Fig. 5I), was also shorter than wild type, around 4 h. It is unclear whether or not this difference in half-life would be of physiologic significance. Additional Angelman-relevant substrates will need to be identified to test this further. Finally the I803ins (Fig. 5K) mutation had no effect on the stability of the protein. This mutant had no defects in any of the assays tested in our analysis. DISCUSSION There is a strong correlation between the occurrence of Angelman syndrome and the loss of E6AP catalytic activity, suggesting that the dysregulation of E6AP substrates is an important factor in this disease. Of the E6AP point mutations examined here, seven map to the catalytic hect domain, and they affected all steps in the ubiquitylation process. Four of the hect domain mutations severely impaired the ability of E6AP to ubiquitylate the model substrate HHR23A. Three of these, the F782⌬, K836ins, and I804K mutations, were likely misfolded. The F782⌬ and K836ins mutants did not form detectable thiol ester intermediates with ubiquitin (Fig. 2, C and D), were unable to functionally interact with the ubiquitin-conjugating enzyme UbcH7 (Fig. 4A, lanes 4 -7, and b), and were unstable when expressed in cells (Fig. 5, G and H). The hect domain crystal structure predicts that these two residues would make important contributions to the structural integrity of the protein. Phe-782 resides in the hydrophobic core, and its side chain projects directly into the center of one of the two lobes that comprise the hect domain. The K836ins mutation, on the other hand, deletes the carboxyl-terminal ␣-helix of the hect domain, which makes numerous contacts with residues in the other lobe of the domain (13). The extreme carboxyl terminus is also necessary for hect domain function (in addition to structure) because deleting the six carboxyl-terminal amino acids slows (or stops) the transfer of the thiol ester-linked ubiquitin from the catalytic cysteine residue of the hect domain, causing the accumulation of E6AP-ubiquitin thiol ester adducts (12).
I804K retained partial activity in HHR23A ubiquitylation but was also very unstable when expressed in cells. Qualitatively this mutant still formed thiol ester intermediates with ubiquitin (data not shown) but did not compete as effectively as wild type E6AP for E2 binding. The rapid turnover of this mutant in cells suggests that it may be misfolded, although it still possessed partial activity in vitro.
Two other hect domain mutants, E550L and L502P, retained the ability to form thiol ester adducts with ubiquitin (Fig. 2, E  and F), although they were impaired in transferring the ubiquitin to the substrate (Fig. 1B, lanes 6 and 7). Their defect was not in catalyzing isopeptide linkages because the proteins still Cell extracts were immunoprecipitated with anti-E6AP antiserum, resolved by 8% SDS-PAGE, and visualized by autoradiography. The open arrow indicates a slightly slower migrating form of the E6AP likely due to modification by one ubiquitin. The asterisk denotes an even more slowly migrating form of E6AP, possessing an unknown modification, which persists on the catalytically inactive forms of E6AP (compare A (wild type E6AP) with B (catalytically inactive E6AP)). IVT, in vitro translated. affixed ubiquitin to themselves (Fig. 3, lanes 6 and 7 and lanes  10 and 11, respectively) with the L502P mutant doing so hyperactively. Moreover the mutations did not abrogate E6AP⅐ substrate binding (Fig. 3B). These phenotypes may indicate that the E550L and L502P mutations either impair or sterically limit access of a substrate lysine to the hect domain catalytic cleft. The hyperactivity of the L502P mutant may reflect an increase in the efficiency with which it catalyzes the transfer of thiol ester-linked ubiquitin to itself.
Alternatively the E550L and L502P mutants may not properly undergo conformational changes required for normal E6AP activity. Such changes are predicted from the two known hect domain crystal structures. The structure of the E6AP hect domain bound to the ubiquitin-conjugating enzyme UbcH7 depicted a 40-Å distance between the catalytic cysteine residues of the two enzymes (13). This distance is too great to enable the ubiquitin to be transferred from one enzyme to the other, making a conformational change necessary. In addition, the WWP1 hect domain structure depicts a dynamic protein that swivels about a flexible linker, thereby facilitating ubiquitin transfer from the ubiquitin-conjugating enzyme to itself. Mutations predicted to constrain the flexibility of the linker reduced hect domain activity (27). This study used self-ubiquitylation to assay enzymatic activity, but such a change is likely also required for substrate ubiquitylation. The E550L and L502P mutations might alter the ability of the hect domain to undergo such conformational switches, but the thiol esterlinked ubiquitin would still be passed to the nearest eligible lysine residue, which would be on E6AP itself.
Although the L502P mutant self-ubiquitylated efficiently in vitro (Fig. 3A, lanes 10 and 11) and appeared as a series of slower migrating forms when immunoprecipitated from cells, it was stable over the course of the 8-h chase period (Fig. 5F). We believe that the L502P mutant self-ubiquitylated after the cells were lysed because L502P migrated as a single polypeptide when we lysed the cells under denaturing conditions. We hypothesize that immunoprecipitating the protein stimulated self-ubiquitylation likely by forcing the dimerization of two E6AP molecules similar to the way in which dimerization of kinases induces them to phosphorylate each other (28).
The I803ins mutation had no defects in any of the assays performed here. This mutation may cause only a subtle defect that cannot be detected under the in vitro conditions tested here. Alternatively this mutation could affect ubiquitylation of a specific E6AP substrate relevant to Angelman syndrome or affect an E6AP activity other than ubiquitylation. E6AP has been reported to act as a coactivator in progesterone-dependent transcription and does not require its catalytic activity to do so (29). Interestingly the F782⌬ and I804K mutants retained the coactivator function in that report, although these mutations destabilized the proteins in the pulse-chase experiments performed here (Fig. 5, G and J). This suggests that coactivator activity does not require an intact hect domain and that the E6AP protein must be utilized for coactivator function rapidly after protein synthesis.
A small percentage of the wild type E6AP migrated slightly slower than the unmodified protein on the SDS-polyacrylamide gel (Fig. 5A, open arrowhead). This modification required E6AP catalytic activity because it was not present on the inactive E6AP C820A mutant (Fig. 5, compare A and B) and likely corresponds to E6AP modified by a single ubiquitin. Strikingly a more robust, even slower migrating form of E6AP appeared in the stable cell lines likely due to the addition of an unknown modification to E6AP. This was not a cross-reactive polypeptide for two reasons: it was not present in non-transduced cells, and it reacted with both anti-E6AP antiserum and an epitope tag-specific probe when a His 6 tag was appended to E6AP (data not shown). This modification must have been attached to E6AP during or shortly after E6AP synthesis since it was present at the end of the "pulse" period. Also this modified form had a much shorter half-life than the unmodified E6AP (Fig. 5A), and its disappearance during the chase required its own catalytic activity since it persisted on all the ubiquitylation-defective mutants (Fig. 5, compare A with B and D-F). The nature of this modification is currently under investigation.
The non-hect domain C21Y mutant displayed wild type activity in the in vitro ubiquitylation assay (Fig. 1B, lane 3) but had a half-life of less than 30 min (Fig. 5C) in cells, which is significantly shorter than that of wild type E6AP. Also its rapid degradation was proteasome-dependent (data not shown). Cys-21 is one of four conserved cysteine residues in a 60-amino acid region that is highly conserved among E6AP orthologs in humans, fruit flies, and the soft-shell clam (30). Although the spacing of the cysteines does not conform to any known consensus, the instability of the C21Y mutant in cells suggests that it may be misfolded and that Cys-21 makes important structural contributions to the protein.
S349P, the other non-hect domain mutant, did not ubiquitylate HHR23A although it possesses a wild type hect domain and retained the ability to bind to HHR23A (Fig. 3B). The loss of the activity of this mutant may be due to several possibilities. First Ser-349 may stimulate hect domain activity perhaps via an intramolecular interaction. Alternatively the S349P mutation may disrupt the structure of the protein, which would explain its insolubility and its slightly decreased half-life in cells (Fig. 5D).
In conclusion, we undertook a thorough biochemical analysis of E6AP mutations identified in patients with Angelman syndrome. None of the examined mutations have been found in unrelated control individuals, 2 although normal individuals occasionally have benign polymorphisms in UBE3A. We show here that there is a strong correlation between the occurrence of Angelman syndrome-associated E6AP mutations and the loss of E6AP ubiquitylation activity, indicating that the improper regulation of E6AP substrate(s) is likely to be an important determinant in this disease. It will be important to identify additional E6AP substrates and examine their relevance to the pathology of Angelman syndrome. Finally ϳ15-20% of people with Angelman syndrome have no large deletion, no uniparental disomy, no imprinting defect, and no UBE3A mutation in the exons or intron-exon boundaries. This suggests either that UBE3A is not expressed in or active in the brains of these patients or that additional gene(s) may be involved in causing Angelman syndrome.