Alternative exon skipping biases substrate preference of the deubiquitylase USP15 for mysterin/RNF213, the moyamoya disease susceptibility factor

The deubiquitylating enzyme USP15 plays significant roles in multiple cellular pathways including TGF-β signaling, RNA splicing, and innate immunity. Evolutionarily conserved skipping of exon 7 occurs during transcription of the mRNAs encoding USP15 and its paralogue USP4, yielding two major isoforms for each gene. Exon 7 of USP15 encodes a serine-rich stretch of 29 amino acid residues located in the inter-region linker that connects the N-terminal putative regulatory region and the C-terminal enzymatic region. Previous findings suggested that the variation in the linker region leads to functional differences between the isoforms of the two deubiquitylating enzymes, but to date no direct evidence regarding such functional divergence has been published. We found that the long isoform of USP15 predominantly recognizes and deubiquitylates mysterin, a large ubiquitin ligase associated with the onset of moyamoya disease. This observation represents the first experimental evidence that the conserved exon skipping alters the substrate specificity of this class of deubiquitylating enzymes. In addition, we found that the interactomes of the short and long isoforms of USP15 only partially overlapped. Thus, USP15, a key gene in multiple cellular processes, generates two functionally different isoforms via evolutionarily conserved exon skipping.


Results
The long isoform of USP15 predominantly recognizes mysterin. To elucidate the cellular function of the moyamoya disease-associated protein mysterin, we sought to identify mysterin-binding proteins, reasoning that such proteins would include cofactors and/or substrates of mysterin. To this end, we transiently The long and short isoforms of USP15 consist of 952 and 981 amino acid residues, respectively. The difference between them, the 29-amino acid element encoded by exon 7, is in the inter-region linker between the N-and C-terminal domain clusters. The N-region contains a domain present in USP (DUSP) and a ubiquitin-like motif (UBL), whereas the C-region contains another UBL motif, the enzymatically essential Cys and His boxes, and the enzymatically essential zinc finger (ZF) domain. Four essential ligand cysteines constituting a single ZF are separated by the second UBL domain in the primary structure. The organization of these motifs is conserved among USP4, USP15, and USP11.
Scientific RepoRts | 7:44293 | DOI: 10.1038/srep44293 overexpressed mysterin harboring a FLAG tag at its C-terminus (mysterin-FLAG) in HEK293T human embryonic kidney cells. We precipitated mysterin-FLAG from cell lysates using anti-FLAG antibody and performed LC-MS/MS analysis on total co-precipitates without an isolation process such as SDS-PAGE. Four positive hits, including USP15, LYAR, hnRNPK, and IP3 receptor isoform 3 (IP3R3) in addition to mysterin itself, were reproducibly identified in four trials. In a replication study, we transiently expressed each candidate mysterin-binding protein in HEK293T cells, immunoprecipitated using specific antibodies, and then immunoblotted for mysterin. In this analysis, USP15 and LYAR were detectably physically associated with mysterin, whereas the other two proteins were not ( Fig. 2A and B). However, LYAR is a factor that functions in the nucleolus 23,24 , whereas mysterin is predominantly distributed in the cytosol. Accordingly, although we do not exclude the possibility that LYAR or the other two proteins are functionally associated with mysterin despite their weaker binding and/or inconsistent subcellular distribution under normal condition, we focused on USP15 for the remainder of this study. It is worth noting that mysterin is often detected as multiple bands in immunoblot analysis (see difference in Fig. 2A and B). Independent immunoblot trials even with an identical mysterin protein sample sometimes result in different appearances like those seen in Fig. 2A and B. Thus, we think that it is not caused by any difference in biological conditions but possibly caused by exceptionally large molecular size of mysterin and minor procedural effect in the protein electrophoresis.
We obtained cDNAs encoding both isoforms of USP15 from HEK293 cells by reverse transcription-polymerase chain reaction (RT-PCR). To our surprise, the isoform of USP15 with longer linker region (hereafter USP15L) had significantly stronger affinity for mysterin than the other isoform with the shorter linker (hereafter USP15S; Fig. 2C). Similar alternative skipping in the linker region is observed in USP4, a paralogue of USP15, and the two isoforms of USP4 have different subcellular distributions (see Introduction for details). When we examined the subcellular distributions of the isoforms of USP15, we found that both isoforms were typically localized in the cytosol, along with mysterin, while a minority of cells showed faint signal of USP15S also in the nucleus (Fig. 2D,E, and Supplementary Fig. S1). Thus, USP15L binds more strongly to mysterin than USP15S, and this effect is primarily not a result of isoform-specific differences in subcellular distribution.
The long isoform of USP15 deubiquitylates mysterin. We hypothesized that the alternative exon skipping of USP15 biases its substrate preference. To test this idea, we investigated whether USP15 has an isoform-specific preference for mysterin as a deubiquitylation substrate. For this purpose, the long and short isoforms of USP15 were individually co-expressed in HEK293T cells along with mysterin harboring three tandem FLAG tags at its C-terminus (mysterin-3FLAG). Mysterin-3FLAG was precipitated using FLAG affinity beads, and its ubiquitylation state was examined by immunoblotting using anti-ubiquitin antibody. Overexpression of USP15L significantly decreased the polyubiquitin chain signal associated with mysterin-3FLAG, whereas overexpression of USP15S did not, suggesting that USP15L has the stronger deubiquitylation activity against mysterin (Fig. 3A). The polyubiquitin signal detected here is due to ubiquitylated mysterin protein per se but not to potentially coimmunoprecipitated proteins, since similar polyubiquitylation is detected even after dissociation of potential non-covalent protein interaction by incubation with 0.5% SDS ( Supplementary Fig. 2). Mutations in the enzymatically essential Cys box and zinc finger domain of USP15L abrogated its deubiquitylation activity against mysterin (Fig. 3B), but did not affect the association between USP15L and mysterin (Fig. 3C), supporting the idea that USP15L deubiquitylates its substrate mysterin via its intrinsic enzymatic activity. Moreover, the USP15L mutant increased the level of mysterin-associated polyubiquitin, indicating that, at least in HEK293T cells, this mutant has a dominant-negative effect on the endogenous deubiquitylation activities against mysterin. Together, our data demonstrate that mysterin is a substrate of USP15, and that the conserved skipping of exon 7 significantly decreases its specific affinity for mysterin.
The long isoform of USP15 contributes to stable expression of mysterin. Previous studies demonstrated that USP15 removes K48-linked polyubiquitin chains and thereby protects its substrates against proteasomal protein degradation triggered by K48-linked polyubiquitylation 5,25 . We investigated whether this would also be the case for deubiquitylation of mysterin by USP15L. Immunoprecipitation of mysterin-3FLAG followed by immunoblotting of K48-linked polyubiquitin revealed that mysterin was indeed modified with K48-linked polyubiquitin, which was removed by wild-type USP15L but accumulated in cells expressing the enzymatic mutant (Fig. 4A). Two further observations supported our prediction that the deubiquitylation of mysterin is inversely correlated with polyubiquitylation-induced proteasomal degradation. First, proteasome inhibitor treatment resulted in accumulation of the polyubiquitylated species of mysterin, whereas this species disappeared upon overexpression of USP15L (Fig. 4B). Second, the enzymatic mutant of USP15L destabilized mysterin protein in HEK293T cells, presumably by increasing the polyubiquitylation of mysterin through the mutant's dominant-negative effect on endogenous mysterin-deubiquitylation activity in HEK293T cells ( Fig. 4C and D), while the mutant of USP15S did not affect the stability of mysterin (Supplementary Fig. S3A and B). Thus, USP15L preferentially removes K48-linked polyubiquitin from mysterin and maintains its basal expression level in the cell.
The linker region is necessary but not sufficient for substrate recognition. How does the alternative variation in the linker region bias the affinity of USP15 for mysterin? Although the substrate-binding domain of USP15 has not been definitely identified, one simple model is that the linker region directly mediates mysterin binding. We evaluated the roles of structural regionss and domains in USP15, including the long linker region, in the mysterin-binding process using five truncated mutants of USP15, schematically represented in Fig. 5A. Each variant, harboring three tandem FLAG tags at its C-terminal end, was transiently overexpressed in HEK293T cells. Each USP15 variant was precipitated using FLAG affinity beads, and co-precipitated endogenous mysterin was detected by immunoblotting with anti-mysterin antibody. The results revealed that both the N-and Scientific RepoRts | 7:44293 | DOI: 10.1038/srep44293 C-terminal regions are essential for mysterin binding (Fig. 5B). However, truncated mutants harboring the long linker region, such as Δ C1, Δ C2, Δ N2, and Δ N3 (see Fig. 5A), failed to bind to mysterin, indicating that the long linker is essential but not sufficient for the physical association with mysterin. The DUSP domain and C-region , and LYAR harboring a FLAG epitope tag at their C termini were expressed in HEK293T cells and precipitated using anti-FLAG affinity beads. Endogenous mysterin (591 kDa) co-precipitated with the bait protein was detected with anti-mysterin antibody in the high-molecular weight region of the SDS polyacrylamide gel (top panel). Basal expression levels of mysterin and bait proteins in the cell were detected in the lower panels. Standard molecular sizes are shown on the left side; the same marker was used for all subsequent experiments. Full-length blots are presented in Supplementary Fig. 4A. (B) hnRNPK harboring an EGFP tag at its N terminus was expressed and immunoprecipitated under the same conditions, but no co-immunoprecipitated mysterin was detected. Full-length blots are presented in Supplementary Fig. 4B. (C) Long and short (15S) isoforms of USP15 harboring a FLAG epitope tag at their C termini (USP15L-FLAG and USP15S-FLAG) were transiently expressed and precipitated. Endogenous mysterin preferentially co-precipitated with the long isoform. Fulllength blots are presented in Supplementary Fig. 4C. (D,E) Long and short isoforms of USP15 harboring Myc epitope tag at their C-terminus (USP15L-Myc and USP15S-Myc) were expressed along with mysterin-3FLAG in HEK293T cells. Cells were immunostained using anti-Myc (red) and anti-FLAG antibodies (green). The nucleus was stained using Hoechst 33342 (blue). Both isoforms were predominantly localized to the cytosol and co-localized with mysterin in all cells examined.  were also necessary but not sufficient for binding (Fig. 5B, lanes 3 and 8). These results are consistent with a model in which the DUSP domain and C-terminal region cooperatively clasp mysterin, and this cooperativity is coordinated by the linker region in an isoform-specific manner (see Discussion).
The two USP15 isoforms function in different protein networks. To further characterize the functional divergence of the two major isoforms of USP15, we performed an interactome analysis on each isoform,  Supplementary Fig. 4G. (B) Mysterin-3FLAG was expressed with or without USP15L-Myc in HEK293T cells, which were treated with proteasome inhibitor for the indicated periods before harvest. Cell lysates were analyzed by immunoprecipitation using anti-FLAG affinity beads and immunoblotting using anti-K48 linked polyubiquitin antibody. Lower panels show basal expression of the proteins and the internal marker GAPDH. Proteasome inhibitor treatment resulted in accumulation of mysterin polyubiquitylation, which was antagonized by USL15L. Full-length blots are presented in Supplementary Fig. 4H. (C) HEK293T cells expressing mysterin-3FLAG with or without USP15Lmt-Myc were pulse-labeled with radioactive sulfur-containing cysteine and methionine for 1 hour, and then chased with non-radioactive medium for indicated periods. The radiolabeled mysterin-3FLAG was isolated by immunoprecipitation (anti-FLAG), separated by SDS-PAGE, and analyzed by autoradiography. Full-length autoradiographs are presented in Supplementary Fig. 4I. (D) Quantification of (C) USL15Lmt destabilized mysterin, presumably through a dominant-negative effect.
based on the expectation that each interactome will include substrates and/or cofactors specific for each isoform. We transiently overexpressed the long and short isoforms of USP15, C-terminally tagged with three tandem FLAG epitopes, in HEK293T cells; precipitated them from cell lysate using anti-FLAG antibody; and performed LC-MS/MS analysis on total co-precipitates without isolation. Positive hits reproducibly obtained in all four trials are shown in Fig. 6 in Venn diagram format. Proteins shown in blue and red are substrates and/or cofactors previously identified by interactome analysis 26 or experimentally evaluated in previous studies 9,27,28 , respectively. Although faint physical interaction of mysterin with the short isoform of USP15 was detected in immunoblot analysis (Fig. 2C), the LC-MS/MS analysis identified mysterin only in the long-interactome (Fig. 6), which is possibly due a difference between detection limits or sensitivities of the two experimental procedures. We do not distinguish between substrates and cofactors in these interactomes. However, the partial (but not complete) overlap between the interactome patterns suggests that the functional diversity between the two evolutionarily conserved isoforms of USP15 is not limited to deubiquitylation of mysterin.

Discussion
Alternative skipping of exon 7 in USP15 resulted in deletion of a short element consisting of 29 amino acids in the inter-region linker of USP15 (Fig. 1), which significantly decreased the affinity and deubiquitylation activity Full-length USP15L has a DUSP (domain present in USP), two UBLs (ubiquitin-like domains), a set of Cys and His boxes that constitutes the enzymatic core common throughout the USPs, and a zinc finger domain containing four ligand cysteines (ZF). (B) Wild-type and deletion mutants of USP15L-3FLAG (USP15L harboring three tandem FLAG tags at its C-terminus) were transiently expressed in HEK293T cells and isolated by precipitation using FLAG affinity beads. Endogenous co-precipitated mysterin was analyzed by immunoblotting using anti-mysterin antibody. None of the mutants exhibited significant affinity to endogenous mysterin. Full-length blots are presented in Supplementary Fig. 4J.
Scientific RepoRts | 7:44293 | DOI: 10.1038/srep44293 of USP15 toward mysterin. Previous studies demonstrated that even the short isoforms of USP15 and its paralogue USP4 retain deubiquitylation activities, indicating that the exon skipping does not abrogate the enzymatic activities of these USPs 5,14 . Instead, it is more likely that the exon skipping modulates the substrate specificity of this class of USPs. Our LC-MS/MS analysis revealed that the interactome patterns of the two isoforms of USP15 overlapped partially, but not completely, i.e., some potential substrates are recognized preferably by the short or long isoform of USP15.
The exon 7-encoded short element in the linker region of USP15 was necessary but not sufficient for the physical association with mysterin. Likewise, the DUSP domain and the C-terminal region were also essential but not sufficient. The C-region harbors the enzymatic core responsible for deubiquitylation activity 12,13 , so the C-cluster presumably recognizes the ubiquitin moiety of the ubiquitylated mysterin. Previous structural studies demonstrated that the DUSP domain of USP15 exposes multiple hydrophobic patches on its molecular surface, which are predicted to mediate protein-protein interactions via hydrophobic interactions 29,30 . Indeed, USP15 specifically binds to SART3 (also known as Tip110) via DUSP domain and removes ubiquitin from both SART3 and its binding partner, histone H2B 9,27,28 . The N-terminal DUSP domain of USP15 may recognize the backbone polypeptide of mysterin in a similar manner. Mysterin and SART3, however, may be recognized by USP15 in different manners. Our interactome analysis demonstrated that SART3 is recognized by both isoforms, whereas only the long isoform exhibited significant binding to mysterin (Figs 2 and 6). Furthermore, the DUSP domain of USP15 is sufficient for binding to SART3 9,28 , although it is necessary but not sufficient for mysterin. The DUSP domain of USP15 may bind less strongly to mysterin than to SART3, creating a requirement for supportive binding by the C-terminal enzymatic region to the ubiquitin moiety. In this case, the inter-region linker could be involved in regulating the cooperative recognition by the N-and C-regions. To further explore this idea, determination of the overall structure of USP15 and further biochemical assessment are warranted.
Sixma and colleagues recently reported that the DUSP domain of USP4 promotes the deubiquitylation cycle by discharging hydrolyzed free ubiquitin from the enzymatic core of USP4 through inter-domain interactions 14 . However, the requirement for this activity was less important for USP15, and, as those authors discussed, it is possible that a single DUSP domain consecutively handles substrate recognition and ubiquitin release during the deubiquitylation reaction cycle.
Variation in the linker region is likely to have been maintained under natural selection. Phylogenic study suggested that USP4, 11, and 15 arose from a single ancestral USP, conserved between fungi and metazoans, following the emergence of vertebrates 4 . Although the ancestral USP, USP4, and USP15 retain exon 7, USP11 lacks it and constitutively generates only the short isoform. Experimental assessment identified a conserved nucleotide element in the 5′ -terminal region of exon 7 of USP4 that specifically leads to insufficient splicing, resulting in the generation of the two alternative isoforms 17 . This mechanism is conserved at least among mammalians. The proactively generated short and long isoforms of USP4 exhibit clearly distinct subcellular distributions, supporting the hypothesis that this region is not a mere inter-region connection site. The functional relevance of the linker regions of USP4 and 15 could be related to their serine-rich compositions, which include multiple potential phosphorylation sites 4,17 . Specific regulation by serine kinases may occur throughout this region, although no evidence supporting this idea is currently available. Recent studies, including ours, suggest that variation in the inter-region linker is a key functional determinant for this class of USPs, although the molecular mechanism remains to be elucidated.
We initiated this study based on our interest in the function and/or regulation of the moyamoya disease susceptibility factor mysterin. The resultant data suggest that USP15L stabilizes mysterin via removal of K48-linked polyubiquitin chains. These observations were obtained under conventional cell culture conditions, suggesting that USP15L maintains the basal expression level of this ubiquitin ligase, which was previously shown to have autoubiquitylation (possibly, self-destabilizing) activity 19 , although it remains unclear whether the effect of USP15L on mysterin is also influenced by specific regulation or stimulation. Reduced mysterin expression causes abnormal organogenesis, including embryonic angiogenesis and myogenesis, in zebrafish 19,31 . Therefore, it is important to maintain the physiological expression level of mysterin, and the USP15 long isoform may contribute to this process in vivo. Despite the obvious phenotypes in mysterin knockdown/knockout zebrafish, the cellular and molecular functions of mysterin remain controversial. Recent studies proposed roles in insulin production in pancreatic β cells 32 , WNT signaling in endothelial cells 33 , and non-mitochondrial oxygen consumption in breast cancer cells 34 , although the underlying mechanism remains to be determined. How the suppression of mysterin expression causes abnormal organogenesis in zebrafish also remains completely unknown. Further investigation of these issues, as well as the potential involvement of USP15 in moyamoya disease, is warranted.

Methods
Plasmids. Full-length cDNAs for USP15L (NM_001252078.1) and USP15S (NM_006313.2) were obtained simultaneously by RT-PCR using the OneStep RT-PCR kit (QIAGEN, North Rhine-Westphalia, Germany) with template RNA extracted from HEK293 cells using the RNeasy Mini Kit (QIAGEN). USP15L or S with Myc or three tandem FLAG (3FLAG) epitope tag at the C-terminus was generated by PCR using forward primer USP15FNheI (ATGCGCTAGCACCATGGCGGAAGGCGGAGCGGC) and reverse primers USP15RMycNotI

T GC A A GC T T CT A C T T A T CG T C G T C A T C C T T G T A G T C G AT G T C A T G G T C T T T G T AG T C AC C G T C AT G A T C C T T G TA AT CAAGACAGTAAT T TGAGT T T T TCACAT T), USP15R23FLAGHindIII ( AT GC AA GC T T CT AC T T AT C G T C G T C A T C C T T G TA G T C G AT G T C A T G G T C T T T G T AG T C AC C G T C AT G A T C CT TGTAATCAATAATCACGTGCTCTGTATCT TCTGTCC), and USP15R33FLAGHindIII ( A T G C A A G C T T T T A C T T A T C G T C G T C A T C C T T G T A G T C G A T G T C A T G G T C T TTGTAGTCACCGTCATGATCCTTGTAATCGTTAGTGTGCATACAGTTTTC)
. Δ C1, Δ C2, Δ N1, Δ N2, and Δ N3 deletion mutants were generated by PCR using USP15FNheI and USP15R13FLAGHindIII, USP15FNheI and USP15R23FLAGHindIII, USP15F2NheI and USP15R33FLAGHindIII, USP15FdN2NheI and USP15R33FLAGHindIII, and USP15FdN3NheI and USP15R33FLAGHindIII, respectively.
Mysterin-FLAG and -3FLAG were generated in our previous studies 19,22 . EGFP-hnRNPK was a kind gift from Dr. Kenji Irie (University of Tsukuba, Ibaragi, Japan) 35 . HA-ubiquitin was described previously 36 . Full-length cDNAs of IP3R3 and LYAR were cloned using methods and materials similar to those used for the cloning of USP15L/S. IP3R3 was initially amplified as five fragments, which were ligated using intrinsic restriction sites KpnI, NotI, SalI, and BamHI. Exogenous NheI and XbaI sites and the Myc epitope sequence were introduced into the 5′ and 3′ termini by PCR using forward primer CloNheIIP3R3F1 (ATGCGCTAGCCGCCATGAGTGAAATGTCCAGCTTTCTTCACATCGG) and reverse primer CloXbaIIP3R3R53FLAG (TACGTCTAGATCAGATGTCATGATCTTTATAATCACCGTCATGGTCTTT GTAGTCCTTATCGTCGTCATCCTTGTAATCGCGGCTAATGCAGTTCTGG). The PCR product was subcloned into pcDNA3.1+ using the NheI/XbaI sites. Full-length cDNA of LYAR was obtained by RT-PCR using forward primer KpnILYARCloF (ATGCGGTACCGCCACCATGGTATTTTTTACATGC) and reverse primer N ot I3 FL AG LY ARCloR ( TACGGCGGCCGCTTAGATGTCATGATCTTTATAATCACCG TCATGGTCTTTGTAGTCCTTATCGTCGTCATCCTTGTAATCTTTCACAAGCTTGACTTTG), and then subcloned into pcDNA3.1+ using the KpnI/NotI sites.
Cell culture and transfection. HEK293 (used only for cloning of USP15) and HEK293T cells were maintained in DMEM (Life Technologies, California, USA) containing 10% fetal bovine serum, 100 U/mL penicillin, and 100 μ g/mL streptomycin. Plasmids were transfected into HEK293T cells using Lipofectamine LTX (Life Scientific RepoRts | 7:44293 | DOI: 10.1038/srep44293 Technologies) or PEI MAX (Polysciences, Pennsylvania, USA). The amount of DNA was adjusted to obtain equivalent expression levels of the introduced proteins in each experiment. For detection of ubiquitylation, cells were treated with 1 μ M epoxomicin (Peptide Institute, Osaka, Japan) for 8 h (or other indicated period) before harvest.
Antibodies. Antibodies used for immunoprecipitation (IP), immunoblotting (IB), or immunostaining (IS) in this study include anti-FLAG antibody (M2 used for IP and IB, Sigma Aldrich, St. Louis, MO, USA), anti-Myc antibody (rabbit polyclonal for IS, MBL, Aichi, Japan; 9E10 for IP and IB, Santa Cruz Biotechnology, Dallas, TX, USA), andi-HA antibody (6E2 for IB, Cell Signaling, MA, USA), anti-ubiquitin antibody (FK2 for IB, Enzo Life Sciences, NY, USA), anti-K48 linked ubiquitin antibody (Apu2 for IB, Enzo Life Sciences), and anti-GAPDH antibody (6C5 for IB, Hy Test Ltd, Turku, Finland). Anti-human mysterin mouse monoclonal antibody (1C9) was generated with the method previously reported 37 . Briefly, male BALB/C mice were immunized by inoculations of 200 μ g recombinant partial human mysterin protein (amino acids 1-349), which was obtained by affinity purification using the C-terminal FLAG epitope tag. The mice were boosted with additional recombinant protein 4 days before fusion. Spleen cells were fused with a myeloma cell line. Following fusion, clones were evaluated by ELISA and immunoblotting using mysterin antigen.

LC-MS/MS.
Mysterin-FLAG or USP15L/S-3FLAG was expressed in HEK293T cells, and associated proteins were recovered from cell extracts by immunoprecipitation with anti-FLAG antibody. The associated protein complexes with mysterin-FLAG were immediately digested with lysyl endopeptidase (Lys-C, Wako, Osaka, Japan). Protein complexes associated with USP15L/S-FLAG were reduced with 5 mM TCEP, and then alkylated with 10 mM monoiodoacetoamide. The alkylated protein complexes were digested with Lys-C and trypsin (SCIEX, Tokyo, Japan). The resultant peptides were analyzed using a direct nano-flow LC-MS/MS system, as described previously 38 . Immunoprecipitation. Protein-encoding or empty vector was transiently transfected into HEK293T cells.
The cells were harvested 24 or 48 hours after the transfection and lysed with HEPES buffer (150 mM NaCl, 20 mM HEPES, NaOH, pH 7.4) containing 1% NP40 (for ubiquitylation assay) or digitonin (for all other experiments). The lysate was centrifuged at 13,000 g, and the supernatant was incubated for 1 hour with anti-FLAG M2 agarose beads (Sigma Aldrich) or anti-Myc antibody and protein G beads (GE Healthcare, Buckinghamshire, UK). At this point, aliquots of supernatant were resolved with Laemmli's sample buffer containing 1 mM DTT as a control (i.e., representing sample not subjected to immunoprecipitation). The beads were washed three times with digitonin-containing HEPES buffer and resolved with Laemmli's sample buffer containing 1 mM DTT.
Immunoblotting. Samples were separated with SDS-PAGE using 7.5% or 3-10% acrylamide gel (ATTO, Tokyo, Japan) and transferred to nitrocellulose (GE Healthcare) or PVDF (Merck Millipore, Hessen, Germany) membrane for 1.5 hours at 75 V at 4 °C. Membrane was blocked using Blocking One (Nacalai Tesque, Kyoto, Japan) for 1 hour at room temperature, and then incubated sequentially with primary and secondary antibodies diluted in Can Get Signal (TOYOBO, Tokyo, Japan). Detection was performed via the chemiluminescence or the alkaline phosphatase method.

Immunostaining. HEK293T cells expressing USP15L-Myc or S-Myc together with mysterin-3FLAG
were immobilized with 4% paraformaldehyde (Nacalai Tesque) for 10 min at room temperature 24 hours after transfection. The cells were permeabilized with 100 ng/mL digitonin for 5 minutes, and then washed with phosphate-buffered saline (PBS: 137 mM NaCl, 2.68 mM KCl, 8.1 mM Na 2 HPO 4 and 1.47 mM KH 2 PO 4 , pH 7.4). After incubation with Blocking One (Nacalai Tesque) for 4 hours at room temperature, the cells were incubated sequentially with primary and fluorescence-conjugated second antibodies; each of the antibody incubations were followed by PBS washes. The nucleus was stained using Hoechst 33342. Fluorescence images were obtained by confocal microscopy (LSM 700, Zeiss, Oberkochen, Germany).