The ubiquitin-specific protease USP36 SUMOylates EXOSC10 and promotes the nucleolar RNA exosome function in rRNA processing

Abstract The RNA exosome is an essential 3′ to 5′ exoribonuclease complex that mediates degradation, processing and quality control of virtually all eukaryotic RNAs. The nucleolar RNA exosome, consisting of a nine-subunit core and a distributive 3′ to 5′ exonuclease EXOSC10, plays a critical role in processing and degrading nucleolar RNAs, including pre-rRNA. However, how the RNA exosome is regulated in the nucleolus is poorly understood. Here, we report that the nucleolar ubiquitin-specific protease USP36 is a novel regulator of the nucleolar RNA exosome. USP36 binds to the RNA exosome through direct interaction with EXOSC10 in the nucleolus. Interestingly, USP36 does not significantly regulate the levels of EXOSC10 and other tested exosome subunits. Instead, it mediates EXOSC10 SUMOylation at lysine (K) 583. Mutating K583 impaired the binding of EXOSC10 to pre-rRNAs, and the K583R mutant failed to rescue the defects in rRNA processing and cell growth inhibition caused by knockdown of endogenous EXOSC10. Furthermore, EXOSC10 SUMOylation is markedly reduced in cells in response to perturbation of ribosomal biogenesis. Together, these results suggest that USP36 acts as a SUMO ligase to promote EXOSC10 SUMOylation critical for the RNA exosome function in ribosome biogenesis.

As a critical player in ribosome biogenesis, human Rrp6 (also known as exosome component 10, EXOSC10) has been shown to function in the turnover of the 5 -external transcribed spacer (ETS) (15) and the processing of internal transcribed spacer (ITS) 1 (16) and ITS2 (7,8,16,17) of the precursor rRNA (pre-rRNA), and mature snoRNA turnover (17). A recent study showed that the main direct targets of human EXOSC10 include 3 -extended 5.8S rRNA and 3 -extended snoRNAs, indicating that EX-OSC10 mainly functions in 5.8S rRNA maturation and the final steps of snoRNA processing (17). In addition, a number of cofactor proteins are required for proper exosome function (3)(4)(5). For example, the DExH helicase Mtr4 and superkiller 2 (Ski2) are required for RNA degradation in the nucleus and the cytoplasm, respectively (18,19). Mtr4 interacts with a non-canonical poly(A) polymerase (Trf4 or Trf5 in yeast and PAPD5 in human) and a Znknuckle RNA-binding protein arginine methyltransferaseinteracting RING finger 1 (Air1) or Air2 (ZCCHC7 in human) to form the Trf4/5-Air1/2-Mtr4 polyadenylation (TRAMP) complex (20)(21)(22)(23). In human, the TRAMP complex consists of MTR4 (also called Skiv2l2), PAPD5 and ZCCHC7, and is exclusively present in the nucleolus and involved in rRNA processing (4,5,24). Thus, rRNA processing involves complex exosome proteins and their cofactors. However, how these processes are regulated in the nucleolus is largely unknown.
In this study, we found that USP36 associates with the RNA exosome through interaction with its catalytic subunit EXOSC10. USP36 does not significantly affect the levels of EXOSC10 and other tested exosome subunits. Instead, it mainly acts as a SUMO ligase to mediate the SUMOylation of EXOSC10 at lysine (Lys, K) 583. Mutating K583 impaired the binding of EXOSC10 to pre-rRNAs, and the K583R mutant failed to rescue the defects in rRNA processing, protein translation and cell growth caused by knockdown of endogenous EXOSC10. These results suggest that USP36 acts as a SUMO ligase to SUMOylate EXOSC10 and promote the nucleolar RNA exosome function in ribosome biogenesis and cell growth.

Generation of Tet-inducible EXOSC10 expression cell lines
To generate Tet-inducible expression of EXOSC10, HeLa cells were first transfected with pcDNA6-TR (Life Technologies) followed by selection in 5 g/ml blasticidincontaining medium to establish HeLa cells stably expressing TR (HeLa-TR). HeLa-TR cells were then transfected with pcDNA4-TO-EXOSC10 si1-res (WT and the K583R mutant) and selected in medium containing 5 g/ml blasticidin and 100 g/ml zeocin for up to 2 weeks. Single colonies were isolated, expanded and screened by immunoblot analysis for Dox (2 g/ml)-induced expression of EXOSC10 using anti-EXOSC10 antibody. All the cells were cultured in DMEM supplemented with 10% tetracycline system-approved FBS.

Transfection, immunoblot (IB) and co-immunoprecipitation (co-IP) analyses
Cells were transfected with plasmids using Lipofectamine 2000 (Life Technologies) or TransIT®-LT1 reagents (Mirus Bio Corporation) following the manufacturers' protocol. Cells were harvested at 36-48 h post-transfection and lysed in NP-40 lysis buffer consisting of 50 mM Tris-HCl (pH 8.0), 0.5% Nonidet P-40, 1 mM EDTA, 150 mM NaCl, 1 mM phenylmethylsulfonyl fluoride (PMSF), 1 mM dithiothreitol (DTT), 1 g/ml pepstatin A and 1 mM leupeptin. Equal amounts of total protein were used for IB analysis. Co-IP was conducted as described previously (29). Bound proteins were detected by IB using antibodies as indicated in the figure legends.

Affinity purification of human USP36-associated protein complexes and mass spectrometry analysis
293 cells stably expressing control or Flag-USP36 were established by transfecting with control pcDNA3-Flag and Flag-USP36 plasmids, respectively, followed by selection in medium containing 0.5 mg/ml neomycin (G418) for single clones. The cells were lysed in NP-40 lysis buffer supplemented with protease inhibitors at 4 • C for 1 h followed by centrifugation. A 20 mg aliquot of the cleared cell lysates from either control or Flag-USP36-expressing cells was incubated with 0.1 ml of anti-Flag (M2) agarose beads at 4 • C for 4 h. The beads were washed four times in lysis buffer containing protease inhibitors. The beadbound proteins were eluted in 0.2 ml of Tris-buffered saline (TBS; 50 mM Tris-HCl, 150 mM NaCl, pH 7.4) containing 0.1 mg/ml Flag peptides. Eluted proteins were run into a sodium dodecylsulfate-polyacrylamide gel electrophorsis (SDS-PAGE) gel for 6 min, the gel was stained, and protein bands at the top of the gel were excised, cut into 1 mm pieces, reduced/alkylated and digested with trypsin for 1 h at 50 • C in the presence of ProteaseMax™ detergent using the method recommended by the manufacturer (Promega). Recovered peptides were then dried by vacuum centrifugation, dissolved in 5% formic acid and loaded onto an Acclaim PepMap 0.1 × 20 mm NanoViper C18 peptide trap (Thermo Scientific) for 5 min at 10 l/min in a 2% acetonitrile (ACN), 0.1% formic acid mobile phase. Peptides were then separated using a PepMap RSLC C18, 2 m particle, 75 m × 50 cm EasySpray column using a 7.5-30% ACN gradient over 90 min in a mobile phase containing 0.1% formic acid and a 300 nl/min flow rate provided by a Dionex NCS-3500RS UltiMate RSLC nano UPLC (Thermo Scientific). Tandem mass spectrometry (MS/MS) data were collected using an Orbitrap Fusion mass spectrometer (Thermo Scientific) configured for data-dependent analysis. MS1 scan resolution was set to 120 000 (at m/z 200) and the MS1 automatic gain control (AGC) target was 200 000 with a maximum injection time of 50 ms. Mass range was set at 400-1500. MS1 data were acquired in profile mode using positive polarity, a MIPS filter on with relaxed conditions and charge states from +2 to +7 accepted. The linear ion trap AGC target value for fragment spectra was set at 10 000, intensity threshold was 5000 and a rapid scan rate was used. Quadrupole isolation width was set at 1.6 m/z. Normalized higher-energy collisional dissociation (HCD) was set at 35%. Dynamic exclusion was set to ±10 ppm with a duration of 30 s. The program Comet (v. 2016.01, rev. 3) was used to search MS2 Spectra against a June 2019 version of a UniProt FASTA protein database containing 20 960 canonical Homo sapiens sequences with the addition of the human USP36 sequence, and 179 common contaminant sequences. To estimate error rates, sequence-reversed forms of all proteins were concatenated to the FASTA file. The database processing was performed with Python scripts available at https: //github.com/pwilmart/fasta utilities.git and Comet results processing used the PAW pipeline from https://github.com/ pwilmart/PAW pipeline.git. Comet searches for all samples were performed with trypsin enzyme specificity. Monoisotopic parent ion mass tolerance was 1.25 Da. Monoisotopic Nucleic Acids Research, 2023, Vol. 51, No. 8 3937 fragment ion mass tolerance was 1.0005 Da. A static modification of +113.084 Da was added to all cysteine residues and a variable modification of +79.9663 Da to serine, threonine and tyrosine residues. Comet scores were combined into linear discriminant function scores, and discriminant score histograms were created separately for each peptide charge state (2+, 3+ and 4+). Separate histograms were created for matches to forward sequences and for matches to reversed sequences for all peptides of seven amino acids or longer. The score histograms for reversed matches were used to estimate peptide false discovery rates (FDRs) and to set score thresholds for each peptide class. After removal of contaminants, this resulted in the identification of ∼953 proteins with at least two unique peptides per protein and an estimated protein FDR <0.5%. Estimation of protein abundance differences in the immunoprecipitates from control or Flag-USP36 cell lysates was performed using the numbers of assigned MS/MS spectra (spectral counts) to each peptide from protein across the two samples. Putative interacting proteins were defined by having three or more spectral counts in the Flag-USP36 sample and either 0 spectral counts in the control sample or a ratio of ≥2.5 when dividing the numbers of spectral counts in the Flag-USP36 sample by the spectral counts in the control sample.

Glutathione S-transferase fusion protein association assays.
GST fusion protein-protein association assays were conducted as described (29,33,54). Briefly, purified His-USP36 proteins (200 ng) were incubated with glutathione-Sepharose 4B beads (Sigma) containing 200 ng of GST-EXOSC10 and GST alone, respectively, in a final volume of 50 l of binding buffer containing 50 mM Tris-HCl 7.5, 5 mM MgCl 2 , 100 mM NaCl, 10% glycerol, 0.5 mg/ml bovine serum albumin (BSA), 5 mM ␤-mercaptoethanol for 45 min at room temperature with gentle agitation. The beads were then washed five times with 500 l of the binding buffer and bound proteins were analyzed using IB with anti-USP36 antibody.

In vivo ubiquitination and SUMOylation assays
In vivo ubiquitination and SUMOylation assays under denaturing conditions were conducted using an Ni 2+ -NTA pulldown (PD) method as previously described (54,55,59). For ubiquitination assay, cells were transfected with His-Ub and the indicated, plasmids and treated with 40 M MG132 for 6 h before harvesting. The cells were harvested at 48 h after transfection; 20% of the cells were used for direct IB and the rest of cells were subjected to Ni 2+ -NTA PD under denaturing conditions. The bead-bound proteins were analyzed using IB. For SUMOylation assay, cells were transfected with His-SUMO1 and the indicated plasmids followed by Ni 2+ -NTA PD under denaturing conditions similar to that above.

Nucleolar fractionation
Nucleolar fractionation was performed as described previously (29). Briefly, freshly harvested cells were washed with phosphate-buffered saline (PBS), resuspended in hypotonic buffer A (10 mM HEPES pH7.8, 10 mM KCl, 1.5 mM MgCl 2 , 0.5 mM DTT) in the presence of PMSF and incubated for 10 min on ice. The cells were homogenized using a B pestle douncer followed by spinning down at 3000 rpm for 5 min at 4 • C. The supernatant (cytoplasmic fraction) was supplemented with 1/10 volume of buffer B (0.3 M Tris-HCl pH 7.8, 1.4 M KCl, 30 mM MgCl 2 ). The nuclear pellets were washed with buffer A and then resuspended in buffer S1 (0.25 M sucrose, 10 mM MgCl 2 ), layered over buffer S2 (0.35 M sucrose, 0.5 mM MgCl 2 ) and centrifuged at 1430 g for 10 min at 4 • C. The pelleted nuclei were resuspended in buffer S2 with PMSF, and sonicated using a microtip probe with the power setting at 50%. The sonicated nuclei were then layered over buffer S3 containing 0.88 M sucrose and 0.5 mM MgCl 2 , and centrifuged at 3000 g for 10 min at 4 • C. The pellet contained purified nucleoli and the supernatant represented the nucleoplasm.

Size exclusion chromatography (SEC)
Nuclear extracts were prepared from the above 293 cells stably expressing Flag-USP36 as previously described (59). Briefly, cells were resuspended in hypotonic buffer A, homogenized and centrifuged as above. The resulting nuclear pellets were lysed in buffer C consisting of 20 mM HEPES (pH 7.9), 420 mM NaCl, 0.2 mM EDTA, 1.5 mM MgCl 2 , 0.5 DTT, 25% glycerol with protease inhibitors. A total of 250 l of the nuclear extracts were loaded onto a Superose 6 10/300 GL column (24 ml, GE) equilibrated with PBS (pH 7.2). The flow rate was 0.03 ml/min, and 56 fractions (300 l each) were collected automatically. A calibration curve was obtained with standard proteins (Sigma-Aldrich) as run on the same system. Every two fractions were analyzed using IB with antibodies as indicated in the figure legends.

Translation assay
Global protein translation was measured by puromycin labeling as previously described (33). Briefly, cells were pretreated with 10 g/ml puromycin (Invitrogen) for 10 min. The cells were then harvested for IB detection of puromycylation of nascent peptides using anti-puromycin (clone 12D10; EMD Millipore).

Cell proliferation assay
Cell viability was measured by 3-(4,5-dimethylthiazol-2yl)-2,5-diphenyltetrazolium bromide (MTT) assays. Briefly, cells were incubated with 0.5 mg/ml MTT in medium for 3 h. After incubation, MTT medium was removed and dimethylsulfoxide (DMSO; 100 l per well) was added to fully dissolve the purple formazan. The absorbance was measured at OD 560 nm and OD 690 nm . The reduced absorbance (Abs 560 nm -Abs 690 nm ) represents the relative number of viable cells per well. For colony formation assays, an equal number of cells transfected with scrambled or EXOSC10 siRNA were cultured in DMEM containing 10% FBS for up to 2 weeks. The colonies were visualized by staining with 0.5% crystal violet in 50% ethanol.

USP36 associates with the nucleolar RNA exosome by directly binding to EXOSC10
To understand the mechanisms underlying the role of USP36 in ribosome biogenesis, we sought to identify USP36-interacting proteins. We performed an affinity purification of USP36-associated protein complexes from 293 cells stably expressing Flag-USP36 using anti-Flag antibody (M2) agarose gels, followed by elution with Flag peptides (Supplementary Figure S1C) and MS analysis. A total of 482 proteins were found in the eluates from Flag-USP36-expressing cells with higher abundance than in the control 293 cells (Supplementary Dataset S1). Gene Ontology (GO) analysis revealed that USP36-associated proteins are mainly enriched in the nucleolus and play critical roles in rRNA processing, snoRNA processing, ribosome assembly and translation ( Figure 1A), highlighting the role of USP36 in ribosome biogenesis. Interestingly, all 10 components of the nucleolar RNA exosome, EXOSC1 to EX-OSC10, but not Dis3, are present in the complexes ( Figure  1B). Co-IP assays confirmed that Flag-USP36 binds to EX-OSC10, the 'cap' subunit EXOSC3/hRrp40 and the 'ring' subunit EXOSC4/hRrp41 in 293 ( Figure 1C) and HeLa (Supplementary Figure S1D) cells, but not with Dis3 (Figure 1D), the catalytical subunit of the RNA exosome in the nucleoplasm, consistent with the notion that USP36 interacts with the nucleolar RNA exosome. Flag-EXOSC10 also co-immunoprecipitated with USP36 using anti-Flag antibody ( Figure 1E). Further, endogenous USP36 coimmunoprecipitated with endogenous EXOSC10 in both 293 ( Figure 1F) and HeLa cells (Supplementary Figure  S1E). To understand USP36 association with the RNA exosome in cells, we extracted the nuclear fraction of 293 cells transfected with Flag-USP36 and performed SEC. As shown in Supplementary Figure S1F, EXOSC10, EXOSC3 and EXOSC4 have the same elution peak, with the corresponding molecular weight of 400 kDa which is similar to the size of Exo10 hRrp6 . Interestingly, these RNA exosome proteins also co-elute with USP36, the exosome cofactor MTR4, ribosomal proteins L5, L11 and S27a, as well as ribosome biogenesis accessory factors such as NPM and Las1L in high molecular weight fractions (≥669 kDa), further supporting the association between USP36 and the exosome complex in pre-ribosome particles.
To further understand how USP36 interacts with the exosome, we tested whether this interaction is mediated by EXOSC10, as both peptide counts and peptide coverage of EXOSC10 in MS analysis are the highest among all the exosome subunits ( Figure 1B) and it is among the top 20 binding proteins based on MS/MS spectral counts (Supplementary Table S1). As shown in Figure 1G, knockdown of EXOSC10 abrogated the interaction of USP36 with the exosome core components EXOSC3 and EXOSC4, suggesting that USP36 interaction with EXOSC10 is critical for its interaction with the core RNA exosome complex. Given that USP36 may interact with the exosome complex in pre-ribosome particles (Supplementary Figure S1F), we next examined whether the USP36-exosome interaction is dependent on RNA. We found that RNase treat-ment indeed reduced, but did not abolish, the interaction between USP36 and EXOSC10 in both 293 ( Figure 1H) and HeLa cells (Supplementary Figure S1G), whereas the interaction of USP36 with RPL30 and RPS27a was abolished by the RNase treatment, suggesting that USP36 may directly interact with EXOSC10 and that this interaction is facilitated by RNA-containing pre-ribosome particles. We then tested whether USP36 directly interacts with EX-OSC10 using GST PD assays. As shown in Figure 1I, recombinant His-USP36 protein purified from bacteria (Supplementary Figure S1H) was specifically bound by purified GST-EXOSC10, but not GST alone, indicating that USP36 directly interacts with EXOSC10 in vitro. Together, these results suggest that USP36 interacts with the RNA exosome via direct binding to EXOSC10 and this binding is facilitated by association with RNA.

EXOSC10 and USP36 interact via their C-terminal domains
To understand how USP36 interacts with EXOSC10, we mapped the binding domains between USP36 and EX-OSC10 using co-IP-IB assays. We first transfected cells with a panel of Flag-tagged USP36 deletion mutants followed by co-IP using anti-Flag antibody and IB with anti-EXOSC10 antibody. As shown in Figure 2A, the C-terminal nucleolar localization signal (NoLS)- (25,26) containing region (amino acids 801-1121), but not the N-terminal USP domain-containing (amino acids 1-420) and the middle (amino acids 421-800) regions, interacts with EX-OSC10, indicating that EXOSC10 binds to the C-terminus of USP36 ( Figure 2B). EXOSC10 contains multiple functional domains including the N-terminal PMC2NT, the central EXO and HRDC domains required for its exonuclease activity, as well as the C-terminal exosome-associated region (EAR) and recently characterized Lasso domains that bind to RNA substrates and stimulate the degradation and processing activities of exosome substrates (60,61). To examine which domain USP36 binds, we constructed a panel of Flag-tagged EXOSC10 deletion mutants. We co-introduced full-length EXOSC10 or its deletion mutants with V5-USP36 into H1299 cells and performed co-IP using anti-Flag antibody. As shown in Figure 2C, V5-USP36 specifically co-immunoprecipitated with the mutants containing the C-terminal Lasso domain (amino acids 738-885), but not the mutants lacking this region. Endogenous USP36 also co-immunoprecipitated with the C-terminal Lasso domain ( Figure 2D). GST PD assays showed that the recombinant His-tagged Lasso domain of EXOSC10 protein purified from bacteria was specifically bound by the purified GST-fused C-terminus of USP36 (GST-USP36 801-1121 ), but not by GST alone (Supplementary Figure S2A, B). Furthermore, RNase treatment also reduced, but did not abolish, the interaction of USP36 with the C-terminal Lasso domain of EXOSC10 (Supplementary Figure S2C) or the interaction of EXOSC10 with USP36 801-1121 (Supplementary Figure S2D). Together, these results indicate that the C-terminal region of USP36 directly interacts with the C-terminal Lasso domain of EXOSC10 ( Figure 2B, E) that can also be facilitated by the association with RNA. H1299 cells were transfected with Flag-USP36 and assayed by co-IP using anti-Flag antibody followed by IB. (E) EXOSC10 interacts with endogenous USP36. Cell lysates from H1299 cells transfected with Flag-EXOSC10 were immunoprecipitated with anti-Flag antibody followed by IB. (F) Endogenous USP36 interacts with endogenous EXOSC10. 293 cell lysates were immunoprecipitated with anti-USP36 antibody (Proteintech) followed by IB with anti-EXOSC10. (G) Flag-USP36 associates with the RNA exosome through EXOSC10. H1299 cells were transfected with empty vector or Flag-USP36 and infected with scrambled or EXOSC10 shRNA lentiviruses, followed by co-IP with anti-Flag and IB. (H) Interaction between USP36 and EXOSC10 is partially dependent on RNA. 293 cells transfected with Flag-USP36 were subjected to co-IP with anti-Flag antibody in the presence or absence of 100 g/ml RNase A and 100 U/ml RNase T1 treatment followed by IB. (I) USP36 directly interacts with EXOSC10 in vitro. Purified GST or GST-EXOSC10 immobilized on glutathione beads was incubated with purified His-USP36. Bound proteins were assayed by IB with anti-USP36 (top). Coomassie staining of GST and GST-EXOSC10 proteins is shown in the bottom panel.

USP36 interacts with the RNA exosome in the nucleolus
USP36 is predominantly a nucleolar protein (26,29). Its Cterminal NoLS-containing region is required for interacting with EXOSC10 ( Figure 2B). IF staining indicates that USP36 co-localizes with EXOSC10 in the nucleolus (Figure 2F). To further examine whether USP36 interacts with the RNA exosome in the nucleolus, we performed cell fractionation assays. As shown in Figure 2G, EXOSC10 is localized in both the nucleoplasm and the nucleolus, and the exosome core component EXOSC4 is located in both the cytoplasm and the nucleus, whereas USP36 is mainly localized in the nucleolus, consistent with the distribution of different RNA exosomes in different cell compartments (4). We performed co-IP assay using the lysates from the isolated nu-cleolar fraction and further confirmed that USP36 interacts with EXOSC10 in the nucleolus ( Figure 2H). Thus, USP36 mainly associates with the nucleolar RNA exosome.

USP36 SUMOylates EXOSC10 in cells and in vitro
We next asked whether USP36 regulates EXOSC10 ubiquitination as USP36 is a DUB (26,29). However, while we did observe the marginal ubiquitination of the exogenously expressed EXOSC10 that can be deubiquitinated by WT USP36, but not the catalytically inactive C131A mutant (Supplementary Figure S3A), the ubiquitination of endogenous EXOSC10 is undetectable (Supplementary Figure S3B), indicating that the steady-state levels of EXOSC10 ubiquitination under normal cell growth conditions is too low to be regulated by USP36. We recently found that USP36 possesses a novel SUMO ligase activity and mediates nucleolar protein group SUMOylation (33). Therefore, we examined whether USP36 promotes the SUMOylation of EXOSC10. H1299 cells were transfected with Flag-EXOSC10 together with control, His-SUMO1 or His-SUMO2 plasmid, followed by Ni 2+ -NTA PD assays under denaturing conditions and detection of SUMOylated EXOSC10 by IB. As shown in Figure 3A, EXOSC10 is mainly modified by SUMO1. Interestingly, USP36 markedly promoted EXOSC10 SUMOylation by both SUMO1 ( Figure 3B) and SUMO2 (Supplementary Figure S3C). Consistently, knockdown of USP36 significantly reduced the levels of EXOSC10 SUMOylation in cells by both exogenously expressed SUMO1 as determined by Ni 2+ -NTA PD assays ( Figure 3C) and endogenous SUMO shown as the modified EXOSC10 band in IB assays ( Figure 3D). This modified band is the SUMOmodified EXOSC10 as it was also abolished by knockdown of either SUMO E1 subunit SAE2 or Ubc9 (Supplementary Figure S3D) and can be immunoprecipitated by using anti-EXOSC10 antibody and detected by anti-SUMO1 antibody (Supplementary Figure S3E). To test whether USP36 directly SUMOylates EXOSC10, we performed in vitro SUMOylation assays using recombinant proteins. As shown in Figure 3E, in vitro SUMOylation reaction using recombinant E1 (SAE1/SAE2), E2 (Ubc9), SUMO1 and ATP resulted in marginal SUMO conjugation of EXOSC10, consistent with the notion that SUMO E1 and Ubc9 can directly mediate SUMOylation in vitro, although less efficiently (62,63). Notably, USP36 markedly increased EXOSC10 SUMOylation by SUMO1 in vitro, demonstrating that USP36 acts as a SUMO E3 for EX-OSC10. To map the SUMOylation sites in EXOSC10, we mutated the consensus SUMO lysines, K168 and K583, predicted by the GPS-SUMO tool (64), as well as other reported SUMO lysines by high-throughput proteomic analysis including K19, K710, K826, K833, K859 and K873 (49,53,65,66), to Arg (R) and examined their SUMOylation in cells. As shown in Figure 3F and Supplementary Figure S3F and G, mutating K583, but not K168, K19, K710, K826, K833, K859 or K873, abolished EXOSC10 SUMOylation, indicating that K583 is the predominant acceptor Lys for EXOSC10 SUMOylation. Of note, mutating K583 did not affect the ubiquitination of EXOSC10 (Supplementary Figure S3H), suggesting that K583 is not subjected to the ubiquitination modification.

USP36 does not significantly affect the levels of EXOSC10 and its association with the RNA exosome
To understand whether USP36-mediated SUMOylation affects EXOSC10 protein levels, we performed USP36 knockdown experiments. As shown in Supplementary Figure  S4A, knockdown of USP36 did not reduce the levels of EX-OSC10 or other tested exosome core subunits including EX-OSC2, EXOSC3 and EXOSC4. These exosome proteins are highly stable (Supplementary Figure S4B, C). Knockdown of USP36 did not reduce the half-life of EXOSC10 (Supplementary Figure S4D). These results suggest that EXOSC10 is not mainly regulated by ubiquitination-mediated proteasome degradation and that USP36-mediated SUMOylation of EXOSC10 does not significantly regulate its protein levels.
To examine whether USP36-mediated SUMOylation regulates EXOSC10 localization and association with the RNA exosome in the nucleolus, we performed IF staining in cells with or without USP36 knockdown. As shown in Supplementary Figure S4E, knockdown of USP36 did not apparently alter the predominant nucleolar localization of EXOSC10 in cells. Also, mutating the K583 SUMO site did not alter the nucleolar localization of EXOSC10 in HeLa (Supplementary Figure S5A) and U2OS (Supplementary Figure S5B) cells. Knockdown of USP36 also does not interfere with the association of EXOSC10 with the core exosome subunit EXOSC4 as determined by co-IP assays using anti-EXOSC10 antibody (Supplementary Figure S5C). Furthermore, the K583R mutant binds to the exosome as efficiently as WT EXOSC10 as probed by the presence of the core exosome subunits EXOSC3 and EXOSC4 (Supplementary Figure S5D). Also, the K583R mutation does not affect the interaction of EXOSC10 with USP36 (Supplementary Figure S5E). Thus, USP36-mediated SUMOylation does not significantly affect the nucleolar localization of EXOSC10 and its association with core exosome and USP36.

Ablation of EXOSC10 SUMOylation inhibits rRNA processing
Next, we sought to examine whether USP36-mediated SUMOylation of EXOSC10 regulates the nucleolar RNA exosome function in rRNA processing. The 47S pre-rRNA is processed to mature 18S, 5.8S and 28S rRNAs via multiple steps of processing cleavage, as illustrated in Figure 4A. The RNA exosome has been shown to play a key role in the processing of the 3 end of 5.8S (from 12S intermediates to 5.8S, Figure 4B) and 18S (from 21S to 18SE) precur- EXOSC10 is required for 12S rRNA processing. HeLa cells transfected with scr or two different EXOSC10 siRNAs were assayed for rRNA processing by northern blot (top panels) using the 5.8S-ITS2 junction (C) and ITS2 (D) probes as indicated. Ethidium bromide (EB) staining of the RNA gels is shown in the bottom panels. (E and F) WT EXOSC10, but not the K583R mutant, rescued EXOSC10 depletioninduced attenuation of 12S rRNA processing. HeLa cells stably expressing Dox-inducible siRNA-resistant EXOSC10 (WT or the K583R mutant) were transfected with scr or EXOSC10 siRNAs and treated with or without Dox. Total RNAs extracted from the cells were assayed by northern blot using the 5.8S-ITS2 junction probe and IB detection of the expression of EXOSC10 (E). The 7S and 5.8S + 40 nt species were quantified and the rescue efficiency is calculated from three independent experiments (F). **P <0.01, comparison between WT and the K583 mutant as determined by Student's t-test. sors (7,16,67) as well as the degradation of the 5 -ETS (15). Using northern blot with a probe hybridizing to the 5.8S-ITS2 conjunction ( Figure 4C) and the ITS2 ( Figure 4D), respectively, we confirmed that knockdown of EXOSC10 by two different siRNAs markedly impaired rRNA processing, leading to accumulation of 5.8S precursors including 12S, 7S and 5.8S + 40 nt ( Figure 4C, D). To examine the role of EXOSC10 SUMOylation in rRNA processing, we performed EXOSC10 knockdown and rescue experiments. We established HeLa cells stably expressing Tet-inducible siRNA-resistant EXOSC10 (WT or the K583R mutant), whose expression can be induced by Dox. These cells were transfected with control scr RNA or EXOSC10 siRNA in the absence or presence of Dox and then assayed by northern blot using a 5.8S-ITS2 probe to monitor the 12S rRNA processing. As shown in Figure 4E and summarized in Figure 4F, the accumulation of 7S and 5.8S + 40 nt rRNA upon EXOSC10 knockdown was markedly alleviated by the Doxinduced expression of siRNA-resistant WT EXOSC10, but not the K583R mutant. Together, these data suggest that EXOSC10 SUMOylation by USP36 plays a critical role in rRNA processing. USP36 is critical for multiple steps of rRNA processing and its depletion results in the reduction of both 21S and 12S rRNA species (27,33). Consistently, we showed that knockdown of USP36 significantly reduced the levels of 12S rRNA (Supplementary Figure S6A, B). Fur-ther analysis showed that knockdown of USP36 markedly impaired 12S rRNA processing, as indicated by the accumulation of 7S species and the marked increase of the ratios of 7S and 5.8S + 40 nt species to 12S rRNA precursors (Supplementary Figure S6C, D). These results suggest that USP36 plays an important role in 5.8S rRNA maturation, correlating with its role in regulating EXOSC10 and RNA exosome function in processing 5.8S precursors.

Ablation of EXOSC10 SUMOylation inhibits its binding to pre-rRNA
To understand how USP36-mediated SUMOylation may regulate EXOSC10's activity in rRNA processing, we tested whether USP36-mediated SUMOylation promotes EX-OSC10 binding to rRNA. To this end, we performed RNA-IP assays followed by RT-qPCR assays using primers amplifying different regions of pre-rRNA ( Figure 5A). As shown in Figure 5B, mutating K583 significantly attenuated the binding of EXOSC10 to pre-rRNA across the 5.8S-ITS2 junction and 18S-ITS1 junction, consistent with the role of EXOSC10 in processing the 5.8S and 18S precursors. These data suggest that USP36-mediated EXOSC10 SUMOylation promotes the targeting of EX-OSC10 to pre-rRNAs. Of note, although EXOSC10 binds to 28S as reported by CLIP-seq assays (17), mutating K583 only slightly, but not significantly, reduced the binding of EXOSC10 to mature 28S rRNA ( Figure 5B), suggesting that EXOSC10 SUMOylation may specifically affect the exosome-mediated maturation of 5.8S and 18S rRNA. We also examined whether USP36 could regulate the exosome function by binding to rRNA precursors. Indeed, RNA-IP assays showed that USP36 strongly binds to pre-rRNAs ( Figure 5C), suggesting that USP36 itself may act as a pre-rRNA-binding protein or associate with pre-rRNA by binding to the exosome in pre-ribosome particles to regulate exosome function at the pre-rRNA.

Ablation of EXOSC10 SUMOylation inhibits translation and suppresses cell proliferation
Given the critical role of EXOSC10 SUMOylation in rRNA processing, we next sought to examine the role of USP36mediated EXOSC10 SUMOylation in protein translation and cell growth. To do so, we first measured nascent peptide synthesis using puromycin pulse labeling in cells followed by IB with anti-puromycin antibody. As shown in Figure 6A, knockdown of EXOSC10 by two different siR-NAs markedly inhibited protein translation. Consistent with the role of EXOSC10 SUMOylation in rRNA processing (Figure 4), Dox-induced expression of WT EXOSC10, but not the SUMOylation-defective K583R mutant, largely restored the protein translation ( Figure 6B, C). Cell proliferation assays measured by MTT ( Figure 6D) and colony formation ( Figure 6E, F) assays also showed that Doxinduced expression of WT EXOSC10, but not the K583R mutant, markedly rescued cell growth inhibition by knockdown of endogenous EXOSC10, demonstrating that EX-OSC10 SUMOylation at K583 plays a critical role in protein translation and cell proliferation.

Ribosomal stress attenuates EXOSC10 SUMOylation by reducing USP36
To further understand the role of USP36-mediated EX-OSC10 SUMOylation in ribosome biogenesis, we asked whether EXOSC10 SUMOylation could be regulated in cells in response to ribosomal stress caused by the perturbation of ribosome biogenesis. We first treated cells with a low dose (5 nM) of Act D, which specifically inhibits rRNA synthesis and causes ribosome stress (56,59,68), and examined the EXOSC10 SUMOylation. Ni 2+ -NTA PD assays showed that Act D treatment markedly reduced the levels of EXOSC10 SUMOylation by exogenously expressed SUMO-1, but not the total levels of EXOSC10, in a timedependent manner ( Figure 7A). EXOSC10 SUMOylation by endogenous SUMO shown as the modified EXOSC10 band in IB is also significantly inhibited by the treatment with Act D in both HeLa and H1299 cells ( Figure 7B, C). To further validate the reduction of EXOSC10 SUMOylation in response to ribosomal stress, we treated cells with the small molecule RNA Pol I inhibitor CX-5461 (69). Indeed, CX-5461 treatment also significantly impaired EXOSC10 SUMOylation by both exogenously expressed SUMO1 ( Figure 7D) and endogenous SUMO ( Figure 7E Figure S7B), indicating that the attenuated EXOSC10 SUMOylation following ribosomal stress is associated with the reduction of USP36 protein. Together, these results suggest that the levels of USP36 and EXOSC10 SUMOylation are reduced in cells in response to ribosomal stress, thus coordinating with ribosome biogenesis.

DISCUSSION
The RNA exosome plays a critical role in RNA processing, degradation and quality control, thus being essential for normal cell growth and proliferation. The nucleolar RNA exosome is critical for ribosome biogenesis by processing pre-rRNAs and snoRNAs (16,17). Although Rrp6 is the only non-essential component among the 11 exosome subunits for yeast cell growth, human EXOSC10 (hRrp6) is essential for cell growth (17). EXOSC10 knockout mice are embryonic lethal, with embryogenesis arrested at the morula stage (70), suggesting that EXOSC10 plays an indispensable role in early embryogenesis and animal development. As ribosome biogenesis is a complex and dynamic cellular process and is subjected to extensive regulation in response to growth signals and cellular stressors, it is conceivable that the function of the nucleolar RNA exosome is also highly regulated.
In this study, we report that the key RNA exosome component EXOSC10 is SUMOylated at K583 by USP36 in the nucleolus. USP36-mediated EXOSC10 SUMOylation is critical for the nucleolar RNA exosome function in rRNA Flag-EXOSC10 (WT and the K583R mutant) plasmids were subjected to RNA-IP with anti-Flag followed by RT-qPCR. Shown is the fold reduction of relative rRNA binding of EXOSC10 K583R compared with EXOSC10 WT and normalized to empty vector-transfected cells from five independent experiments. The relative RNA enrichment was calculated by dividing RNAs in anti-Flag immunoprecipitates by that in control IgG immunoprecipitates. **P <0.01, compared with Flag-EXOSC10 WT -transfected cells. P-values were calculated by Student's t-test. The expression of Flag-EXOSC10 WT and Flag-EXOSC10 K583R assayed by IB is shown on the right. (C) USP36 binds to pre-rRNA. HeLa cells transfected with control or Flag-USP36 were subjected to RNA-IP using anti-Flag or control mouse IgG, followed by RT-qPCR analysis. Shown are percentage enrichment relative to input from three independent experiments. P-values were calculated by Student's t-test. ***P <0.001.
processing and cell growth, as the SUMO-defective K583R mutant of EXOSC10 has significantly impaired function to rescue the growth inhibition and pre-rRNA processing impairment induced by knockdown of endogenous EXOSC10 (Figures 4 and 6), and knockdown of USP36 results in similar defects in the processing of 5.8S rRNA precursors to that by the knockdown of EXOSC10 ( Supplementary Figure S6). We further show that the K583R mutant of EX-OSC10 has impaired pre-rRNA binding activity ( Figure  5). As USP36 does not significantly alter the protein levels and cellular localization of the exosome proteins, the reduced binding of the K583R mutant to pre-rRNA might be due to the impairment of conformational change of EX-OSC10 upon SUMOylation. The K583 residue is located close to the C-terminal Lasso motif (61), which is critical for binding to RNAs. Thus, K583 SUMOylation could facilitate a conformational change of EXOSC10 that favors its high-affinity binding to pre-rRNA. It has been suggested that the exosome can be recruited to different pre-ribosome particles via the binding of the TRAMP complex component MTR4 to the adaptor proteins such as Nop53 and UTP18 (5,15). Both Nop53 and UTP18 as well as MTR4 were present in our purified Flag-USP36-associated protein complexes as determined by MS analysis (Supplementary Dataset 1; Supplementary Table S1). We indeed observed that USP36 interacts with the TRAMP complex (data not shown). Therefore, it is also possible that EX-OSC10 SUMOylation by USP36 could facilitate the assembly and stabilization of the exosome-TRAMP-adaptor protein complexes in the small subunit processome (SSU) and the large subunit processome (LSU). Accumulating evidence suggests that SUMOylation tends to target a group of functionally and physically connected proteins, called protein group SUMOylation (35,71), which allows multiple SUMO-SIM (SUMO-interacting motif) interactions that contribute to the formation and stabilization of the multiprotein complexes (35,71). Previous proteomics analyses observed the SUMOylation of the TRAMP complex proteins ZCCHC7 and PAPD5 (49,66,(72)(73)(74)(75)(76). The TRAMP complex proteins each contain multiple putative SIMs predicted by the GPS-SUMO tool (64) (not shown). Thus, it is interesting to examine whether USP36 also SUMOy- Figure 6. SUMOylation of EXOSC10 plays a critical role in protein translation and cell growth. (A) Knockdown of EXOSC10 inhibits global protein translation. HeLa cells transfected with scr or one of the two EXOSC10 siRNAs were pulse-labeled with puromycin followed by IB with anti-puromycin to detect new protein synthesis. (B and C) The K583R mutant of EXOSC10 rescues the translation inhibition by endogenous EXOSC10 depletion less efficiently compared with WT EXOSC10. HeLa cells stably expressing Dox-inducible siRNA-resistant WT EXOSC10 or the K583R mutant were transfected with scr or EXOSC10 siRNAs and treated with or without Dox, followed by puromycin labeling and IB analysis (B). The percentage rescue is quantified from three independent experiments (C). Data were presented as mean ± standard deviation (SD). The P-value was determined by Student's t-test. **P <0.01. (D-F) The K583R mutant rescues the growth inhibition by knockdown of endogenous EXOSC10 less efficiently compared with WT EXOSC10. HeLa cells stably expressing Dox-inducible siRNA-resistant WT EXOSC10 or the K583R mutant were transfected with scr or EXOSC10 siRNAs and treated with or without Dox, followed by MTT cell proliferation (D) and colony formation (E, F) assays. Shown are the fold changes of absorbance from four independent experiments (D), one representative colony formation (E) and the average percentage changes of the colony numbers (F) from three independent experiments. P-values shown were calculated by Student's t-test. ***P <0.001, compared with cells transfected with EXOSC10 siRNAs without induction of siRNA-resistant EXOSC10.
lates the Nop53, UTP18 and TRAMP components, given that they all physically interact with USP36, thus mediating exosome-TRAMP-adaptor protein group SUMOylation to promote and/or stabilize the exosome-TRAMPadaptor protein complex in pre-ribosome particles via multiple SUMO-SIM interactions. Together, our data reveal that USP36 is a novel regulator of the nucleolar RNA exosome by acting as a SUMO ligase to mediate EXOSC10 SUMOylation. Knight et al. (77) previously reported that EXOSC10 SUMOylation is increased along with global SUMOylation in cells in response to cooling, resulting in reduced levels of EXOSC10 and defects in ribosome biogenesis. However, it is not clear whether such SUMOylation is the consequence or the cause of changes in cells under cooling conditions and whether abolishing EXOSC10 SUMOylation could alleviate cooling-mediated suppression of rRNA processing and ribosome biogenesis. Compared with mutating three lysine residues, K168, K201 and K583, used in the study (77), we showed that mutating K583 alone can abolish EXOSC10 SUMOylation without affecting EXOSC10 levels. Furthermore, we have shown that EX-OSC10 SUMOylation is markedly reduced in cells in response to the perturbation of ribosome biogenesis mediated by treatment with a low dose of Act D or the RNA Pol Ispecific inhibitor CX-5461 (Figure 7), suggesting that EX-OSC10 SUMOylation is tightly regulated and coordinated with ribosome biogenesis.
EXOSC10 and the RNA exosome core subunit proteins are all stable proteins and their levels are not significantly affected by either overexpression or knockdown of USP36. This is similar to the regulation of the snoRNP complex by USP36 that mainly functions to mediate snoRNP SUMOylation, but not ubiquitination (33). Consistently, the steadystate levels of EXOSC10 ubiquitination are below the level of detection (Supplementary Figure S3B), suggesting that under normal conditions, USP36 mainly acts to SUMOylate EXOSC10 to regulate its function, but not its levels, whereas its DUB activity may play a role in maintaining RNA exosome subunit proteins assembled in the RNA exosome in their deubiquitinated state in the nucleolus.
Interestingly, we also found that USP36 associates with pre-rRNA ( Figure 5C) and snoRNAs (not shown), indi- cating that USP36 itself could be an RNA-binding protein. A previous study using microRNA (miRNA) screening of RNA-binding protein has shown that USP36 is able to bind to several miRNA precursors (78). Thus, our finding reveals an additional mechanism underlying USP36's role in ribosome biogenesis: binding to pre-rRNA and mediating rRNA processing, including its critical role in regulating the RNA exosome function to process 5.8S and 18S rRNA. Future studies would aim to evaluate whether USP36 directly binds to pre-rRNA or indirectly through its association with the exosome and other interacting proteins in pre-ribosome particles. As our proteomic data showed that USP36 forms large protein complexes by associating with many ribosome biogenesisrelated proteins and that the complex may contain additional RNAs, such as pre-rRNA and snoRNAs, USP36 may be critical for the nucleolar protein-RNA complex formation via its DUB-, SUMO E3-and RNA-binding activities. Thus, USP36 is a multi-functional nucleolar protein and may play a central role in ribosome biogenesis and translation by acting as a central regulatory hub for nucleolar protein dynamics and ribosome biogenesis.

DATA AVAILABILITY
The mass spectrometry proteomics data have been deposited in the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD031592.