Ub-ProT reveals global length and composition of protein ubiquitylation in cells

Protein ubiquitylation regulates diverse cellular processes via distinct ubiquitin chains that differ by linkage type and length. However, a comprehensive method for measuring these properties has not been developed. Here we describe a method for assessing the length of substrate-attached polyubiquitin chains, “ubiquitin chain protection from trypsinization (Ub-ProT).” Using Ub-ProT, we found that most ubiquitylated substrates in yeast-soluble lysate are attached to chains of up to seven ubiquitin molecules. Inactivation of the ubiquitin-selective chaperone Cdc48 caused a dramatic increase in chain lengths on substrate proteins, suggesting that Cdc48 complex terminates chain elongation by substrate extraction. In mammalian cells, we found that ligand-activated epidermal growth factor receptor (EGFR) is rapidly modified with K63-linked tetra- to hexa-ubiquitin chains following EGF treatment in human cells. Thus, the Ub-ProT method can contribute to our understanding of mechanisms regulating physiological ubiquitin chain lengths and composition.

P rotein ubiquitylation is a dynamic multifaceted posttranslational modification (PTM) responsible for regulating a diverse array of cellular processes, including protein degradation, protein trafficking, signal transduction, and the DNA damage response 1,2 . Ubiquitylation is catalyzed by the concerted action of ubiquitin (Ub)-activating (E1), Ubconjugating (E2), and Ub-ligating (E3) enzymes. Deubiquitylating enzymes (DUB) antagonize ubiquitylation by removing Ub modifications from their substrates. Ub can be covalently conjugated to substrates in several ways: as a single Ub conjugated to a single (monoubiquitylation) or multiple sites (multiple monoubiquitylation), or as polymeric chains (polyubiquitylation). Different Ub chains are formed through isopeptide linkages using seven internal lysine (K) residues, as well as its N-terminal methionine (M1). Effector proteins harboring Ub-binding domains (UBDs) function as readers/decoders by discriminating specific Ub linkages 3 . In addition to homotypic chains, cells contain heterotypic Ub chains in which multiple linkages form mixed or branched chains. Furthermore, Ub undergoes phosphorylation and acetylation at specific S/T and K residues 4 .
Accumulating evidence indicates that linkage type, length, and chemical modification work in concert to affect the topology and dynamics of Ub chains and direct substrates to distinct biological pathways [4][5][6] . In particular, linkage type is a critical determinant of chain function. For example, it is widely accepted that K48-linked chains function as targeting signals for proteasomal destruction, whereas K63-linked chains are generally involved in signal transduction, DNA repair, and trafficking of membrane proteins; other linkage types also have distinct cellular functions. In contrast, despite its fundamental importance, our knowledge regarding the functional relevance of Ub chain length remains limited. Earlier in vitro studies suggested that the proteasome recognizes K48-linked tetraubiquitin as the minimal targeting signal, and that binding strength increases markedly as chain length increases up to octaubiquitin 7 . However, more recent studies showed that monoubiquitylation and multiple short Ub chains also constitute efficient proteasomal targeting signals [8][9][10] . In endocytosis and endosomal targeting, the relative importance of monoubiquitylation and K63-linked polyubiquitylation of receptor proteins remains unclear 11 .
To understand the biological significance of various Ub chain structures, it is essential to determine the linkage types, modifications, and lengths of endogenous, substrate-linked chains. Recent advances in mass spectrometry (MS) and antibodyengineering technologies allow us to determine and quantitate Ub linkages and PTMs in complex biological samples 12 . In contrast, the lengths of substrate-attached Ub chains have only been determined by analyzing their gel mobility. However, since most endogenous substrates have multiple ubiquitylation sites, and attached chains might have heterogeneous lengths 13,14 , more comprehensive and accurate techniques are required. Here we describe a novel biochemical method for determining Ub chain length, "Ub chain protection from trypsinization (Ub-ProT)." By combining this method with quantitative MS analysis, we identified the length and composition of Ub chains in yeast, and of ligand-activated epidermal growth factor receptor (EGFR) in mammalian cells.

Results
Establishing a method for determining Ub chains. Because Ub can form polymeric chains, a given composition of ubiquitylation can form numerous structures, e.g., a substrate protein bearing four Ubs can form five distinct topologies even if linkage types and branching are not considered (Fig. 1a). Thus, the gel mobilities of ubiquitylated proteins do not accurately reflect individual chain Ub lengths. Analysis of Ub chains cleaved from substrate proteins at the proximal Ub moiety would be the optimal way to determine chain length. DUBs could be used for this purpose, but unfortunately, the known DUBs do not discriminate linkage positions. Although the proteasomal DUB Rpn11 can remove Ub chains by cleaving the proximal Ub, the reaction is coupled to substrate unfolding by ATPase subunits 15 . Therefore, we designed an alternative approach using trypsin and a Ub chain protector. A previous study showed that Ub is specifically cleaved at Arg74 by trypsin digestion under native conditions. This cleavage occurs for all different Ub linkage types 16 (Fig. 1b, i). However, if substrate-attached chains are masked by a Ub chain protector, intact polyubiquitin chains should remain after trypsinization, allowing the substrate-bound poly-Ub chains to be easily analyzed using a gel-based assay. We named this approach Ub-ProT (Fig. 1b, ii). As a Ub chain protector, we used the tandem Ub-binding entity (TUBE), a high-affinity probe for Ub chains 17 . TUBE is an artificial protein, consisting of four repeats of the Ub-associated (UBA) domain of human RAD23A or UBQLN1. The UBQLN1 TUBE binds K48-and K63-linked tetraubiquitins with a K d of 9 and 0.7 nM, respectively. To prevent trypsin digestion, we constructed trypsin-resistant (TR)-TUBE consisting of a biotin tag, a hexahistidine tag, and six tandem repeats of the UBQLN1 UBA domain in which the Arg residues were replaced by Ala 18,19 (Supplementary Fig. 1). We confirmed that TR-TUBE could efficiently pull down all eight types of di-Ubs ( Supplementary Fig. 2a). Then, we tested whether free K48-, K63-, M1-, and K11-linked Ub chains, K48/K63 branched chains, and di-Ubs consisting of all eight different linkage types could be protected from trypsin digestion by TR-TUBE (Fig. 1c, Fig. 2c). Greater than 90% of input K11-, K48-, K63-, and M1-linked di-Ubs were protected from trypsinization, and 40-60% of K6-, K27-, K29-, and K33-linked chains were protected (Supplementary Figs. 2c and d; Supplementary Data 1). TR-TUBE could also protect K48/K63 branched chains from digestions, although the protection efficiency was lower than for homogeneous chains ( Fig. 1d; Supplementary Fig. 5).
Since our TR-TUBE construct consists of six Ub-binding domains, one concern about our method was that it might only protect Ub chains up to hexamers. However, as seen in Fig. 1c, TR-TUBE could efficiently protect much longer chains, suggesting that multiple molecules of TR-TUBE can bind to a single chain to restrict trypsin accessibility. To further examine whether the number of Ub-binding domains in TR-TUBE may artificially influence the length of protected Ub chains, we constructed TR-TUBEs with four and eight Ub-binding domains, and compared the lengths of protected Ub chains ( Supplementary Fig. 4). All three different TR-TUBE constructs protected M1-linked chains equally, and protected chains were indistinguishable from untrypsinized control chains.
We next applied the Ub-ProT method to tetraubiquitin-fused Sic1 (4×Ub-T7-Sic1) as a model substrate with a defined Ub chain length (Fig. 2a). The tetraubiquitin of 4×Ub-T7-Sic1 was digested to Ub monomers by trypsinization, but almost completely protected in the presence of TR-TUBE. In contrast, the same construct containing the I44A mutation in each Ub moiety (4×UbI44A-T7-Sic1) was completely digested, validating our method. Several Ub enzymes are efficiently self-ubiquitylated in vitro: Cdc34 is self-ubiquitylated with K48-linked chains 20 , Rsp5 is self-ubiquitylated with K63-linked chains 21 , and MBPtagged Parkin is multiply monoubiquitylated 22 . In each case, ubiquitylated proteins were detected as a smear at high molecular weight, which disappeared following trypsinization (Fig. 2b). In the presence of TR-TUBE and trypsin, the protein substrate was digested leaving typical Ub ladders. Comparison with free Ub chains (used as a length marker) revealed that Cdc34 and Rsp5 were modified with K48-and K63-linked chains, respectively, of up to~10-mers (Fig. 2b, left and middle). In contrast, selfubiquitylated Parkin was modified with monoubiquitin and, to a lesser extent, short Ub chains (Fig. 2b, right). Because TR-TUBE captured almost all ubiquitylated Parkin, we concluded that TR-TUBE can bind not only Ub chains but also multiply monoubiquitylated substrates.

Measurement of chain length of yeast ubiquitylated proteins.
To characterize global Ub chain lengths in an entire organism, we next investigated the mean length of substrate-attached Ub chains in yeast lysates. In this experiment, we used a drug-sensitive pdr5 mutant, which increases sensitivity to the proteasome inhibitor MG132. This allowed us to compare Ub chain lengths in control lysates and in lysates from yeast in which proteasomal activity was inhibited by MG132 23 . Soluble lysates were prepared from exponentially growing cells cultured with or without MG132, and ubiquitylated proteins were captured and pulled down by TR-TUBE. Immunoblotting with anti-Ub antibody revealed that TR-TUBE was unable to pull down Ub monomers, but otherwise     Here the amount of material analyzed was 5-or 15-fold higher than that in a, for the anti-Ub blot and protein staining, respectively. Gel regions subjected to MS quantitation are indicated by red letters (a-l). The position of the ubiquitin monomer (Ub 1 ) was defined as the gel fraction "a." Asterisks denote TR-TUBE. d Length distributions of total ubiquitin and five major linkages at steady state or following MG132 treatment. Gel fractions in c were analyzed by quantitative mass spectrometry (mean ± s.e.m.; n = 3 biological replicates; Supplementary Data 3). Relative positions of K48-and K63-linked chains are labeled in blue and red, respectively Ub chain ladder (Fig. 3a, lanes 7 and 8). In addition, different Ub chains with three or more Ubs exhibited different gel mobilities ( Fig. 3c; Supplementary Fig. 6). Comparison with free Ub chains revealed that the lengths of the substrate-attached Ub chains were mainly in the monomer to heptamer range (Fig. 3c, left and middle). Although MG132 treatment increased the amounts of chains of each length, maximal lengths were unchanged.
To investigate the relationship between linkage types and chain lengths, we cut gel lanes into 12 pieces to separate monomers and different-length Ub chains (Fig. 3c, right). The resultant gel slices were subjected to Ub-AQUA/PRM ( Fig. 3d; Supplementary Data 3). Because the abundances of the M1, K27, and K33 linkages were quite low (<50 attomoles: lower limit of Ub-AQUA/PRM; Supplementary Fig. 7; Supplementary Data 4), we focused on the five major linkage types detected in this experiment. In the absence of MG132, K48, the most abundant linkage type, was detected in di-Ub and longer chains. K11 and K63 linkages were mainly found in di-Ubs, whereas K6 and K29 linkages were mainly detected in longer chains. Proteasome inhibition by MG132 dramatically increased the abundance of each linkage type in each fraction. In particular, the abundance of the K29 and K48 linkages was elevated for dimer to heptameric chains, and the abundance of the K6, K11, and K63 linkages was elevated for longer chains (Fig. 3d). The K29 linkages in long chains may be involved in the Ub-fusion degradation (UFD) pathway, in which heterotypic chains with K29 and K48 linkages are directed to proteasomal degradation 25 . It will be of great interest to determine whether the K6, K11, and K63 linkages mainly detected in the longer chains are homogeneous or heterogeneous. Moreover, it is noteworthy that nearly 50% of Ub was detected as monomers in untreated cells, and the amount of monomeric Ub increased by 1.8-fold in MG132-treated cells (Fig. 3d, total). TR-TUBE does not bind Ub monomers, suggesting that a significant portion of ubiquitylated cellular proteins are modified with multiple monoubiqutins 13 or with terminal monoubiquitin branches off of polyubiquitin chains. Together with a recent finding that~20% of proteasome substrates are degraded upon monoubiquitylation 10 , our results indicate that a wide range of Ub signals, including monoubiquitylation and chains of mainly dimer to heptamer length, function in proteasome targeting.
Cdc48/p97 regulates Ub chain length. We next sought to investigate factors that regulate the global chain length of ubiquitylated substrates. Ubp6, the proteasome-associated DUB, maintains cellular Ub levels by removing Ub chains from substrates en bloc, whereas Rad23 and Dsk2 shuttle substrates to the proteasome 26,27 . The Ub-selective chaperone Cdc48 (p97/VCP in mammalian cells) is involved in endoplasmic reticulum (ER)associated degradation (ERAD), autophagy, and other Ubdependent processes 28 . In lysates, the levels of ubiquitylated  substrates were reduced in ubp6Δ and rad23Δ dsk2Δ cells (Fig. 4a  left; ubp6Δ and r23Δ d2Δ). This result is consistent with those of a previous study demonstrating that Rad23 and Dsk2 protect Ub chains from DUB 29 . In contrast, in cdc48-3 mutant cells, we observed significant accumulation of ubiquitylated substrates, with elevated levels of five Ub linkage types (K6, K11, K29, K48, and K63) (Fig. 4a left and 4b; Supplementary Data 5). Strikingly, a subsequent Ub-ProT assay revealed that Ub chains were significantly elongated in the cdc48-3 mutant, but unaffected in ubp6Δ and rad23Δ dsk2Δ cells (Fig. 4a, middle and right). Linkage quantification suggested that the elongated chains in cdc48 cells contained all five major linkages ( Fig. 4c; Supplementary Data 6). Because chain elongation was not observed in proteasomeinhibited cells, this effect was not simply due to the accumulation of ubiquitylated proteins. Instead, since Cdc48/p97 is involved in segregation/remodeling of protein complexes, and major ubiquitylating enzymes build Ub chains on their substrates through processive elongation 30, 31 , Cdc48-dependent segregation/remodeling likely inhibits processive Ub chain elongation in substrate complexes. Substrate-attached Ub chains could be remodeled on Cdc48 complexes by multiple cofactors, such as Ub-binding cofactors (Ufd1, Npl4, Ufd3, and Shp1), the DUB Otu1, and E4 Ub elongation factor Ufd2 32, 33 . We investigated the role of these cofactors in regulation of substrate-attached Ub chain lengths (Fig. 5), and found that Ub chains were significantly elongated in a npl4-1 mutant, which is defective for extraction of ubiquitylated membrane proteins 34 (Fig. 5a). Ub chains were also slightly elongated in shp1Δ cells, but unaffected in otu1Δ, ufd2Δ, and ufd3Δ cells (Fig. 5b). Previously, we reported that Ub-binding of Npl4 was completely abolished by the npl4-1 mutation (G323S) in vitro 19 . Thus, we conclude that the recognition and segregation of ubiquitylated substrate by Cdc48 and Npl4 is important in regulating Ub chain lengths in cells.
Chain topology in ubiquitylated EGFR. Ubiquitylation of ligand-activated EGFR promotes clathrin-independent endocytosis and endosomal targeting of the receptor for lysosomal degradation 11 . EGFR is modified by multiple mono-and polyubiquitin chains linked through K11, K48, and K63 linkages 35 , but the topology and functional relevance of these modifications are still under debate [36][37][38] . To explore this issue, we applied Ub-ProT to EGFR ubiquitylation. For this purpose, we generated HeLa cells stably expressing EGFR-EGFP-3×FLAG from the AAVS1 locus (Fig. 6a). Upon EGF treatment (100 ng ml −1 ), EGFR-EGFP-3×FLAG was rapidly ubiquitylated, and was subsequently degraded after 60 min (Fig. 6b). Microscopic analysis confirmed that the tagged EGFR initially localized to the plasma membrane, but moved to endosomes after 15 or 30 min of EGF stimulation, and then translocated to lysosomes after 1 h (Supplementary Fig. 8). Ub-AQUA/PRM analysis revealed that ubiquitylated EGFR at the 5-min time point contained the following Ub chain composition: K63 linkage (~50% of total Ub), K48 linkage (~4%), K11 linkage (~2%), and mono/endocap Ub (~40%) ( Fig. 6c; Supplementary Data 7). These results are consistent with those from a previously published study 35 . We next subjected the ubiquitylated EGFR to Ub-ProT and observed a Ub ladder corresponding to K63-linked chains of four to six Ub molecules specifically in EGF-treated cells (Fig. 6d, lane 5). To determine whether these were true K63 chains, we used linkageselective DUBs. Recently, Komander et al. developed a versatile method for analyzing the higher-order architecture of heterotypic chains by linkage-selective DUBs, termed "Ub chain restriction (UbiCRest)" 39 . Among the DUBs, we used K48-chain-selective OTUB1, K63-chain-selective AMSH, and non-selective USP2. As expected, the Ub ladder was completely digested by treatment with AMSH or USP2, but not with OTUB1 (Fig. 6d, lanes 6-8).
Thus, ligand-activated EGFR is primarily modified with K63linked tetra-to hexa-ubiquitins. When present, K48 and K11 linkages are formed at sites distal to the K63 chains.
To further investigate the relationship between EGFR trafficking and Ub chain length, we performed Ub-ProT at various times after EGF stimulation (Fig. 6e). Signals for monoubiquitylation, which may consist of multiple monoubiquitylations on the substrate, or of multiple monoubiquitylations that branch off of chains, and Ub chains were elevated after 5 min of EGF treatment, and reached a maximum after 30 min, suggesting that EGFR is continuously ubiquitylated from the plasma membrane to the endosome. After 60 min of stimulation, the intensities of Ub bands decreased, consistent with deubiquitylation at multivesicular bodies, and lysosomal degradation. Thus, although certain additional bands were also detected, the major Ub modifications to EGFR were multiple monoubiquitylations and K63-linked polyubiquitylation during the trafficking process (Fig. 6f). Consistent with this result, recent studies suggested that the K63-linked chains function in endosomal sorting rather than internalization of EGFR 36,40 . It remains unclear why the K63 chains had intermediate length, i.e., why they were tetra-to hexa-ubiquitins rather than di-or tri-ubiquitins. One possible explanation is that long chains may be advantageous for collaborative recognition by the highly organized ESCRT complexes (0, I, II, and III), which contain multiple UBDs 41 .

Discussion
In this study, we developed a novel method for determining the chain length of ubiquitylated substrates. By combining this method (Ub-ProT) with MS-based Ub quantitation (Ub-AQUA/ a b   (Fig. 3). At steady state, K48, K6, and K29 linkages were predominantly incorporated in dimeric to heptameric chains, whereas K11-and K63-linked chains existed mainly in dimeric chains. Surprisingly, maximum chain lengths were only slightly altered by proteasome inhibition, implying the existence of a proteasome-independent mechanism for regulating chain length. This mechanism may require specific recognition by UBDcontaining proteins. For example, UBD-containing proteins such as Rad23/Dsk2 bind Ub chains with appropriate lengths (in the 2-7-mer range), thereby protecting them from DUBs. Alternatively, the chain-elongating ability of ubiquitylating enzymes might be intrinsically limited in vivo, as suggested by in vitro studies 30,31 . In this context, because attachment of a single Ub costs one molecule of ATP, restricting chain length may benefit the cell by limiting total energy consumption. On the other hand, we found that inactivation of Cdc48 and Npl4 caused a significant accumulation of elongated Ub chains (Figs. 4 and 5), suggesting that chain length is regulated by sequestration by the Cdc48 complex rather than limitations of ubiquitylating enzymes. Cdc48 regulates ubiquitylation of over a thousand proteins 42 , explaining its global effects on yeast Ub chain lengths. Ub-ProT analysis also revealed the Ub chain length and composition of ubiquitylated EGFR in HeLa cells (Fig. 6). Although we also detected K48 and K11 linkages, combined analysis with UbiCRest suggested that a K63-linked chain of four to six Ubs is the core unit of the attached chains. Further combinations of Ub-ProT, Ub-AQUA/ PRM, UbiCRest, and middle-down MS might allow us to investigate the Ub chain architectures of more complex chains, including the positions of linkage branching and PTMs.
In all of our Ub-ProT assays, we used TR-TUBE as the chain protector. The advantages of TR-TUBE are its extremely tight binding to Ub chains, and its ability to bind in tandem to bind and protect long Ub chains. However, it is important to note the limitations to our method. While we demonstrate that Ub-ProT using TR-TUBE can bind and protect all Ub linkage types, there is up to a twofold variability in the amount of protection. While 90% of K11-, K48-, K63-, and M1-linked dimeric Ub is protected by Ub-ProT using TR-TUBE, K6-, K27-, K29-, and K33-linked Ubs are protected at 40-60% efficiency ( Supplementary Fig. 2). Therefore, our method may slightly underestimate the occurrence of certain linkage types. In addition, while TR-TUBE does not bind free monomeric Ub, we see some monomeric Ub when we analyze chain length and composition type after Ub-ProT treatment of lysates (Figs. 3d and 4c). We propose that most of this monomer is generated from multiply monoubiquitylated substrates, such as multiply monoubiquitylated Parkin, which is bound by TR-TUBE (Fig. 2b). However, there are trace amounts (<5%) of K11-and K63-linked Ub in the monomer fraction (Figs. 3d and 4c), indicate that at least some of the monomeric Ub is generated from a slight degradation of K11 and K63 chains that escaped protection by TR-TUBE. Furthermore, it is possible that some of the unlinked monomeric Ub observed may be generated from the trimmed terminal ends of chains instead of from multiple monoubiquitylations of substrates. Finally, while we demonstrate that Ub-ProT can also protect branched Ub chains (Fig. 1d), again this protection is somewhat reduced compared to homogeneous chains (Fig. 1c), raising the possibility that our method may underestimate the occurrence of branched chains.
Despite these caveats, our characterization of global yeast Ub chain lengths, regulation of chain lengths by Cdc48 and Npl4, and modification of EGFR upon EGF treatment indicate that Ub-ProT using TR-TUBE, or a novel chain protector, will be useful for the study of Ub chain lengths.

Methods
Plasmids. TUBE expression plasmids: We modified a previously reported highaffinity probe for Ub chains, TUBE, which consists of four tandem repeats of the UBA domain of UBQLN1 17 . To capture ubiquitylated proteins more efficiently and   prevent trypsinization to allow MS analysis, we mutated all Arg residues in the UBA domain to Ala and fused four to eight repeats together 18 . Next, the gene encoding TR-TUBE was inserted into a modified pRSET-A vector (Life Technologies), which contains a hexahistidine tag and Cys residue for purification and biotinylation, respectively. The protein-coding sequence of TR-TUBE is shown in Supplementary Fig. 1b. Our TR-TUBE constructs will add to Addgene. 4×Ub-Sic1 expression plasmids. To construct tetraubiquitin-fused Sic1, we first generated pET15b-4×Ub and pET15b-4×UbI44A, in which four Ubs were tandemly fused without a linker, followed by T7-Sic1 PY inserted using the In-Fusion cloning kit (Takara). The template plasmid, pET15b-1×Ub, was kindly provided from Dr. K. Sakamoto (RIKEN institute, Japan). The resultant plasmids, pET15b-4×Ub-Sic1 and pET15b-4×UbI44A-Sic1, express 4×Ub-T7-Sic1 PY -His 6 and 4×UbI44A-T7-Sic1 PY -His 6 , respectively. DUB expression plasmids. To construct bacterial expression vectors for linkagespecific DUBs, cDNAs codon-optimized for bacterial expression of human AMSH and OTUB1 (Eurofins Genomics) were cloned into the pGEX6P1 expression vector (GE Healthcare). The bacterial expression vector for USP2, pET15b-USP2cc, was a gift from Dr. R. Baker 43 .
Yeast strains and media. Saccharomyces cerevisiae strains used in this study are listed in Supplementary Table 1. All strains are isogenic to W303 or BY4741. Standard media and genetic techniques were used to manipulate yeast strains 45 . A deletion mutant of PDR5 was used to increase sensitivity to the proteasome inhibitor MG132 23 . Yeast cells were grown at 28°C in SC medium (0.67% yeast nitrogen base without amino acids, 0.5% casamino acids, 2% glucose, 10 mM potassium phosphate (pH 7.5), 400 mg l −1 adenine sulfate, 10 mg l −1 uracil, and 20 mg l −1 tryptophan).
Ub-ProT assay for yeast lysates. For Ub-ProT assay of yeast extracts, 30 OD 600 units of log-phase cells were harvested and lysed with glass beads in 300 μl of lysis buffer (50 mM Tris-HCl (pH 7.5), 100 mM NaCl, 10% glycerol, 10 μM MG132, 10 mM iodoacetamide, and 1× complete protease inhibitor cocktail (EDTA-free, Roche)). After centrifugation, the supernatant (100 μg) was incubated with TR-TUBE (10 μg) for 1 h at 4°C, followed by incubation for 45 min at 4°C with Dynabeads MyOne Streptavidin C1 (1 mg). Bead-bound TR-TUBE-polyubiquitylated proteins were pulled down with a magnet, washed three times with PBS-T, and then incubated overnight at 37°C in 100 μl of trypsin solution (50 mM AMBC, 0.01% Rapigest SF, and 15 ng μl −1 trypsin). We found that streptavidin was not digested by trypsin under this condition; therefore, the polyubiquitin chains were still retained on the beads via the TR-TUBE/streptavidin complex after trypsinization. After the beads were washed with PBS-T, the polyubiquitin chains were selectively eluted by 30-min incubation with 1× NuPAGE LDS sample buffer. The samples were directly subjected to electrophoresis on NuPAGE gels without boiling in order to avoid aggregation of polyubiquitin chains 39 .