C-terminally GFP-tagged DDI1 (DDI1-GFP) vector was generated using the commercial pEGFP-N1 vector (Clontech), while N-terminally GFP-tagged DDI1 (GFP-DDI1) was cloned into a PM-cherry-C1 vector kindly provided by Dr. Daniel Abankwa (see section “Cloning Procedures”). The plasmids used to express the N-terminally FLAG-tagged versions of the wild type (FLAG-UBE3A-pCMV, UBE3A^WT) and catalytically inactive (FLAG-UBE3A^LD-pCMV, UBE3A^LD) human UBE3A (Tomaić et al., 2009) were a gift from Dr. Vjekoslav Tomaic. FLAG-tagged ubiquitin (FLAG-Ub) cloned in pCDNA3.1 vector (Invitrogen) was used (Ramirez et al., 2018). A previously described untagged human Parkin plasmid (Martinez et al., 2017) was used and the empty pCDNA3.1 vector (Invitrogen) was used as control (Ramirez et al., 2018).

Gene for human DDI1 protein (Uniprot Q8WTU0) was synthesized by GenScript and further amplified using 5′-TATAGGTACCATGCTGATCACCGTG-3′ forward primer and 5′-TATAACCGGTATGCTCTTTTCGTCC-3′ reverse primer and inserted between the Acc65I and AgeI sites of the pEGFPN1 vector. For the generation of GFP-DDI1, human DDI1 sequence was amplified using 5′-CGCGTGTACAGTATGAATATAGCGATA-3′ forward primer and 5′-CGCAAGCTTTCAATGCTCTTTTCG-3′ reverse primer. Amplified DDI1 was then inserted between BsrGI and HindIII sites of the PM-cherry-C1 where mCherry had been previously replaced by GFP. All PCR reactions were carried out with Phusion High-Fidelity DNA polymerase (Thermo Scientific). PCR product gel extractions and plasmid purifications were performed with the E.Z.N.A. Gel Extraction Kit and E.Z.N.A. plasmid mini and midi kits (Omega Bio-tek), respectively. Correct sequence for all plasmids was confirmed by sequencing either by the Eurofins GATC Biotech Company (Köln, Germany) or the SGIKER Unit of Sequencing and Genotyping at the University of the Basque Country (Leioa, Spain).

Human Embryonic Kidney cells (HEK293T) were cultured under standard conditions (37°C, 5% CO2) using Dulbecco’s modified Eagle medium/nutrient mixture F-12 (DMEM/F-12) with GlutaMAX (Thermo Scientific) and supplemented with 10% fetal bovine serum (Thermo Scientific), 100 U/ml of penicillin (Invitrogen) and 100 mg of streptomycin (Invitrogen). For ubiquitination sites and chain-types experiments by mass spectrometry, HEK293T cells were seeded in 100 mm plates (2 × 10⁶ cells). After 24 h from seeding, cells were co-transfected with 12,5 μg of DDI1-GFP and 12,5 μg of UBE3A^WT or UBE3A^LD for 72 h using Lipofectamine 3000 transfection reagent (Invitrogen) and according to manufacturer’s instructions. For the DUB silencing experiment, HEK293T cells were seeded the previous day in six-well plates (6 × 10⁵ cells). Each well was then transfected with 10 nM siRNA targeting specific DUBs (Ambion; life technologies). For that purpose Lipofectamine RNAiMAX transfection reagent (Thermo Fischer) was used according to manufacturer’s instructions. After 24 h incubation, 1,5 μg of DDI1-GFP and 1,5 μg of FLAG-Ub were co-transfected for another 48 h using again Lipofectamine 3000 transfection reagent (Invitrogen). For all experiments, cells were washed twice in PBS after the transfection periods and stored at -20°C until required.

Transfected HEK293T cells were lysed with 500 μl of lysis buffer in six-well plates and with 2 mL in 100 mm dishes (50 mM Tris–HCl pH 7.5, 150 mM NaCl, 1 mM EDTA, 0.5% Triton, 1× protease inhibitor cocktail from Roche Applied Science and 50 mM N-ethylmaleimide from Sigma) and centrifuged at 14 000 g for 10 min. Supernatants were mixed with 25 μl/well (6-well plate) and 50 μL/well (100 mm plates) of GFPTrap-A agarose beads suspension (Chromotek GmbH), which had been previously washed twice with a Dilution buffer (10 mM Tris–HCl pH 7.5, 150 mM NaCl, 0.5 mM EDTA, 1× protease inhibitor cocktail, 50 mM N-ethylmaleimide). After incubation for 3 h at room temperature with gentle rolling, samples were centrifuged at 2700 g for 2 min to separate the beads from the unbound material. GFP beads were then subjected to three washing steps; once with the dilution buffer, thrice with washing buffer (8 M urea, 1% SDS in PBS) and once with 1% SDS in PBS. GFP-tagged proteins bound to the beads were eluted by incubating at 95°C for 10 min with 25 μl elution buffer (250 mM Tris–HCl pH 7.5, 40% glycerol, 4% SDS, 0.2% bromophenol blue, 100 mM DTT).

For SDS–PAGE, 4–12% Bolt Bis–Tris Plus pre-cast gels (Invitrogen) were used. Proteins were then transferred to PVDF membranes using the iBlot system (Invitrogen). Membranes were blocked using 5% powdered milk in PBS-Tween (PBS-T). Following primary and secondary antibody incubation, membranes were developed with an ECL kit (Biorad Clarity). To create dual-color westerns independent color channels were assigned to two independent westerns that were developed in the same membrane. The amount of material loaded for western blot analysis was between 10–30% of neat inputs and 30–50% of neat elution samples. The following primary antibodies were used at 1/1000 concentration: mouse monoclonal anti-GFP antibody (Roche Applied Science; catalog number 11814460001); mouse monoclonal anti-FLAG M2-HRP conjugated antibody (Sigma; catalog number A8592); mouse monoclonal anti-UBE3A (clone E6AP-300) antibody (Sigma; catalog number E8655), rabbit monoclonal anti-UCHL-5 (Abcam, ab124931), rabbit polyclonal anti-USP9X (Proteintech, 55054-1-AP), anti-USP7 (kindly provided by Dr. Emilio Lecona), rabbit polyclonal anti-USP42 (Sigma, HPA006752). The following secondary antibodies were used at 1/4000 concentration: goat anti-mouse-HRP-labeled antibody (Thermo Scientific; catalog number 62-6520) and goat anti-rabbit-HRP labeled antibody (Cell Signaling; catalog number 7074).

Prior to MS analysis, the efficiency of GFP pull-downs was evaluated by silver staining. Briefly, the 10% of each neat elution was resolved by SDS–PAGE (4–12% Bolt Bis–Tris Plus precast gels, Invitrogen), and after fixing the proteins within the gel with 40% methanol/10% acetic acid, gel bands were visualized using the SilverQuest kit following manufacturer’s instructions (Invitrogen).

Ninety percent of the neat elution from each GFP pull-down was resolved by SDS–PAGE (4–12% Bolt Bis–Tris Plus precast gels, Invitrogen) and visualized with Colloidal Blue (Invitrogen) following manufacturer’s instructions. Aiming to isolate mono- and poly-ubiquitinated DDI1, each gel lane was excised into two slices above the clear band of approximately 75 KDa that corresponds to non-modified DDI1 (Figure 1). Subsequently, each gel lane was chopped, and proteins were subjected to in-gel digestion protocol using trypsin as described before (Osinalde et al., 2015). Briefly, gel pieces, previously dehydrated with acetonitrile (ACN), were reduced with 10 mM dithiothreitol (DTT), alkylated with 55 mM chloroacetamide, and finally rehydratated in 12.5 ng/ml trypsin and incubated overnight at 37°C. The following day, resulting tryptic peptides were extracted from the gel by serial incubation with 100% ACN and 1% trifluoroacetic acid in 30% ACN. Finally, prior to MS analysis, peptide solutions were dried down in a SpeedVac centrifuge (Thermo Scientific).

Mass spectrometric analyses were performed on an EASY-nLC 1200 liquid chromatography system interfaced with a Q Exactive HF-X mass spectrometer (Thermo Scientific) via a nanospray flex ion source. Dried peptides were dissolved in 0.1% formic acid and loaded onto an Acclaim PepMap100 pre-column (75 μm × 2 cm, Thermo Scientific) connected to an Acclaim PepMap RSLC C18 (75 μm × 25 cm, Thermo Scientific) analytical column. Peptides were eluted from the column using the following gradient: 18 min from 2.4 to 24%, 2 min from 24 to 32% and 12 min at 80% of acetonitrile in 0.1% formic acid at a flow rate of 300 nl min^-1. The mass spectrometer was operated in positive ion mode. Full MS scans were acquired from m/z 375 to 1800 with a resolution of 120,000 at m/z 200. The 10 most intense ions were fragmented by higher energy C-trap dissociation with normalized collision energy of 28 and MS/MS spectra were recorded with a resolution of 15,000 at m/z 200. The maximum ion injection time were 100 ms and 120 ms, whereas AGC target values were 3 × 10⁶ and 5 × 10⁵ for survey and MS/MS scans, respectively. In order to avoid repeat sequencing of peptides, dynamic exclusion was applied for 12 s. Singly charged ions or ions with unassigned charge state were also excluded from MS/MS. Data were acquired using Xcalibur software (Thermo Scientific).

Acquired raw data files were processed with the MaxQuant (Cox and Mann, 2008) software (version 1.5.3.17) using the internal search engine Andromeda (Cox et al., 2011) and searched against the UniProt database restricted to Homo sapiens entries (release 2017_03; 42165 entries) where the sequences of FLAG_UBIQUITIN and DDI1_GFP proteins were manually added. To assess ubiquitin chains types present on DDI1, only raw files corresponding to poly-ubiquitinated DDI1 were processed. By contrast, to infer the DDI1 ubiquitination sites, raw files corresponding to both mono- and poly-ubiquitinated DDI1 were processed together. Carbamidomethylation (C) was set as fixed modification whereas Met oxidation, protein N-terminal acetylation and Lys GlyGly (not C-term) were defined as variable modifications. Mass tolerance was set to 8 and 20 ppm at the MS and MS/MS level, respectively. Enzyme specificity was set to trypsin, allowing for a maximum of three missed cleavages. Match between runs option was enabled with 1.5 min match time window and 20 min alignment window to match identification across samples. The minimum peptide length was set to seven amino acids. The false discovery rate for peptides and proteins was set to 1%. For the analysis of ubiquitination sites, a minimum localization probability of 0.75 was used. Normalized spectral protein label-free quantification (LFQ) intensities were calculated using the MaxLFQ algorithm.

We recently discovered that Ube3a ubiquitinates endogenous Rngo in flies, and also its human homolog DDI1 when transfected in S5-SYHY neuroblastoma cells (Ramirez et al., 2018). Using a GFP-based pulldown protocol (Ramirez et al., 2016), we now confirm that UBE3A ubiquitinates also DDI1 in HEK293T kidney cells. As compared to co-transfection with an empty vector (pCDNA3.1), or another E3 ligase such as PARKIN, a ligase dead version of UBE3A (UBE3A^LD) does not increase the ubiquitination of GFP-tagged DDI1. By contrast, wild type UBE3A (UBE3A^WT) enhances the ubiquitination of DDI1 while the level of non-modified DDI1 remains unchanged (Figure 1 and Supplementary Figure 1). This effect is independent of whether DDI1 is N- or C-terminally fused to GFP (Supplementary Figure 2). Interestingly, UBE3A-mediated ubiquitination of DDI1 does not seem to target it for degradation, as total GFP-tagged DDI1 levels are equivalent in all samples.

According to PhosphositePlus database – the most comprehensive repository of PTM sites – (Hornbeck et al., 2015), human DDI1 can be ubiquitinated on K345 and K346. Nevertheless, DDI1 contains twelve additional lysines that may potentially be modified by ubiquitin. With the aim to identify the ubiquitination sites on DDI1 that are modified by UBE3A, we followed an approach that optimized the GFP pull-down protocol (Ramirez et al., 2016) for mass spectrometry (MS)-based detection (Figure 2).

First, GFP-tagged DDI1 was transfected on HEK293T cells together with wild type or, as control, ligase dead UBE3A (UBE3A^WT and UBE3A^LD, respectively). Then, DDI1 was isolated from both experimental conditions by GFP pull-down procedure, and subsequently resolved by SDS–PAGE. Coomassie staining revealed a band just below 75 kDa corresponding to GFP-tagged DDI1, as earlier identified by western blotting (Figure 1). As we were interested in the ubiquitinated fraction of DDI1, we excised a slice of the gel directly above that band (Figure 2). Aiming to separate mono- and poly-ubiquitinated DDI1, this gel piece was further divided into two subsequent slices based on the observations by silver staining (Supplementary Figure 3) as well as by previous immunoblotting results (Figure 1): a tight band just above the non-modified DDI1 should correspond to mono-ubiquitinated DDI1-GFP and a bigger slice above this one, corresponding to the smear of poly-ubiquitinated DDI1 observed by immunobloting. Then, proteins within each gel slice were subjected to trypsin digestion, resulting peptides were extracted from the gel, and analyzed by MS; this protocol from the transfection of cells to the MS analysis was performed in triplicate for each experimental sample.

To infer DDI1 ubiquitination sites, MS-derived raw files corresponding to the slices of both mono- and poly-ubiquitinated DDI1 were jointly analyzed using MaxQuant software. Based both on MS/MS counts and LFQ intensity values, GFP-tagged DDI1 was, as expected, the most abundant protein detected by mass spectrometry (Figure 3A and Supplementary Table 1). The second most abundant protein was ubiquitin (Figure 3A) despite the lack of ectopic ubiquitin expression in this experiment; given the stringent washing conditions used in the pull-down protocol, it could be postulated that those ubiquitin moieties were covalently attached to DDI1. While the levels of DDI1 were practically the same independently of UBE3A activity, Ub levels were significantly increased in the sample transfected with UBE3A^WT (Figure 3B).

We detected numerous unmodified peptides that covered 76% of the DDI1-GFP sequence (Supplementary Figure 4A), but more interestingly we identified six diGly-modified peptides across the whole sequence of DDI1 (Figure 3C and Supplementary Table 2), and one on the C-terminally fused GFP protein (Supplementary Figure 4B). More precisely, we found that ubiquitin was conjugated to the following residues on DDI1: K77, which localizes in the N-terminal ubiquitin-like (UBL) domain; K133, K161 and K213 within the helical domain (HDD); K291, that resides in the aspartyl protease-like RVP domain; and K382, on the C-terminal end of the protein (Figure 3C). It is remarkable that besides K291 ubiquitination, whose homolog lysine had been previously reported to be ubiquitinated on experiments performed in rat brain (Na et al., 2012), the remaining five lysines detected in the present study (K77, K133, K161, K213, and K382) are novel ubiquitination sites of DDI1. Representative annotated spectra for these diGly modified peptides are shown on Supplementary Figure 4C.

Aiming to evaluate the effect of UBE3A on DDI1 ubiquitination, we compared the intensity of the diGly peptides between UBE3A^WT and UBE3A^LD overexpressing cells. The intensities of DDI1 ubiquitination sites K213-, K291-, and K382- containing diGly peptides were similar in both experimental conditions, indicating that UBE3A is not involved in modifying such resides (Figure 3C). On the contrary, the intensity of K133- and K161-bearing diGly peptides was induced 5- and 2- fold, respectively, in the UBE3A^WT overexpressing condition (Figure 3C), while total DDI1 levels remained constant (Figure 3B). Similarly, ubiquitination on lysine 77 also appeared to be enhanced upon UBE3A^WT overexpression, although this increase did not appear to be statistically significant, most probably due to the low MS/MS counts recorded (Supplementary Table 3). Overall, our data suggest that UBE3A may be involved in the ubiquitination of three (K77, K133, and K161) out of the six ubiquitination sites detected on DDI1.

To confirm that indeed UBE3A is responsible for ubiquitinating DDI1 on K77, K131, and K161, we generated three DDI1 point mutations, each with one of the above-mentioned lysines replaced by arginine. Following the same GFP pull-down procedure mentioned above, we isolated the DDI1-GFP mutants and monitored their ubiquitination by immunoblotting. As shown in Figure 3D, DDI1 ubiquitination upon UBE3A^WT overexpression was still observed for the K77R and K161R mutants to similar level as for wild type DDI1. Having detected by MS in wild type DDI1 a moderate increase of ubiquitination at those two lysines, their removal did, however, not prevent UBE3A to still ubiquitinate DDI1, probably due to the ubiquitination being transferred to an alternative lysine. By contrast, UBE3A-dependent DDI1 ubiquitination was significantly reduced in the case of the K133R mutant (Figure 3D and Supplementary Figure 4D). It must be noted that position K133 was the one at which the highest ubiquitinated peptide intensity increase had been observed by MS upon UBE3A^WT co-expression. In fact, the ubiquitination pattern of K133R co-expressed with UBE3A^WT was very similar to the pattern exhibited by wild type DDI1 co-expressed with UBE3A^LD. These results indicate that UBE3A is responsible for ubiquitinating DDI1 on K133, and that the presence of this lysine 133 is necessary for UBE3A-mediated ubiquitination of DDI1.

Next, we studied the types of ubiquitin linkage formed on poly-ubiquitinated DDI1. For that purpose, we focused on the diGly peptides corresponding to ubiquitin within the gel slice containing poly-ubiquitinated DDI1-GFP (Figure 2). In line with previous results (Figure 1, 3B), the ubiquitin levels in this slice were also more abundant upon overexpression of UBE3A^WT, and the ubiquitin intensity ratio between UBE3A^WT and UBE3A^LD samples displayed an approximately two-fold increase (Figure 4). MaxQuant analysis of this slice confidently detected four ubiquitination sites within ubiquitin: K11, K29, K48, and K63. The levels of K63 chains remained unaffected upon overexpression of UBE3A^WT. K29 ubiquitin linkages appeared to be increased in the presence of UBE3A^WT, in line with some earlier reports for UBE3A and other HECT E3 ligases (Chastagner et al., 2006; Xu et al., 2018), but we should note that those linkages were detected with very few counts and overlapped with a possible ubiquitination event also in K27 (Supplementary Table 3). More convincingly, the average from the three biological replicates analyzed displayed a clear increase of approximately two fold for both K48 and K11 chains in the UBE3A^WT over-expressing cells with respect to the control sample (Figure 4 and Supplementary Table 3). It is worth noting that the intensity fold change measured for the K48 and K11 diGly peptides were similar to the global fold-change observed for ubiquitin itself. Based on this data, and in partial agreement with earlier reports (Kim and Huibregtse, 2009), we conclude that UBE3A^WT overexpression indeed induces both K48 and K11 ubiquitin chains on DDI1.

The ubiquitination state of any given protein is determined by the balance between the conjugating action of E3 ubiquitin ligases and their counteracting deubiquitinating (DUB) enzymes. Having demonstrated that UBE3A ubiquitinates DDI1, we next aimed to discover the DUB capable of deubiquitinating it. We screened an RNAi library in order to identify the DUB whose silencing would enhance DDI1 ubiquitination. For that purpose, we used HEK293T cells expressing DDI1-GFP and FLAG-Ub, in which 41 different DUBs were silenced individually. More precisely, we silenced 36 members of the USP family, as well as 3 and 2 members, respectively, of the UCH and OTU families (Figure 5A). To measure DDI1 ubiquitination, GFP-tagged DDI1 was first isolated by GFP pull-down, and then ubiquitin was detected by immunoblot with an anti-FLAG antibody. Changes in DDI1 ubiquitination were measured by normalizing the FLAG-Ub signal for each sample against its own GFP signal. Additionally, aiming to compare the results obtained in the several 6-well plate based experiments, we normalized this ratio to the average of the 3 lowest ratios on each dataset of 6 silencing experiments. As shown in Figure 5A, silencing of most DUBs did not substantially modify the ubiquitination state of DDI1 (normalized GFP/FLAG intensity between 0.5–2). Interestingly, upon silencing of UCHL-5, OTUD3 and 8 USP family members (USP2, 3, 7, 8, 9X, 30, 42, and 46) ubiquitination of DDI1 was quantitatively enhanced (normalized GFP/FLAG intensity > 2). From those 10 candidates, we focused our attention on the four DUBs having the greatest influence on DDI1 ubiquination: UCHL-5, USP7, USP9X, and USP42.

We then silenced the above mentioned four DUBs separately, and used a scramble siRNA as a control for this validation experiment that was performed in triplicate (Supplementary Figure 5A). The silencing of DUBs – UCHL-5, USP7, USP9X and USP42 – was confirmed by immunoblotting (Supplementary Figure 5B). After DDI1-GFP and FLAG-Ub overexpression in HEK293T cells, DDI1-GFP was isolated by GFP-pulldown, and the ubiquitination of DDI1 monitored by immunoblotting. As shown in Figure 5B, silencing of neither USP7 nor UCHL-5 had any influence on the ubiquitination state of DDI1. Downregulation of USP42, on the other hand, showed an increase on DDI1 ubiquitination, but this increase resulted to be statistically no significant. Importantly, this second round of screening clearly demonstrated that USP9X silencing results in enhanced DDI1 ubiquitination, and hence confirms that USP9X is required for deubiquitinating DDI1 (Figure 5B).

DUB inhibition has been considered as a potential therapeutic strategy for those diseases on which the function of a given E3 ligase is lost, and certain essential substrates are ubiquitinated at levels below their physiological requirement. A compound termed WP1130 has been previously used to experimentally inhibit USP9X (Kapuria et al., 2010; Sun et al., 2011), even if it is known that it also targets other DUBs (Kapuria et al., 2010). Since highly selective USP9X inhibitors are not yet available, we tested whether WP1130 would –presumably through USP9X– enhance the ubiquitination of DDI1. For that purpose, HEK293T cells were incubated for 1 h with 5 μM WP1130 or with DMSO as control. We then checked the ubiquitination pattern of GFP-DDI1 by immunoblotting, using the same GFP-pulldown protocol described above. Ubiquitination of DDI1 was significantly enhanced in cells treated with USP9X inhibitor WP1130 (Figure 6), supporting the earlier observations that DDI1 is an USP9X-regulated substrate.

Overall, our data indicate that UBE3A and USP9X are the E3 ligase and DUB, respectively, involved in modulating the ubiquitination of DDI1.