A comprehensive method for detecting ubiquitinatedsubstrates using TR-TUBEYukiko Yoshidaa,1, Yasushi Saekib, Arisa Murakamia,b, Junko Kawawakia, Hikaru Tsuchiyab, Hidehito Yoshiharab,Mayumi Shindoc, and Keiji Tanakab,1

aProtein Metabolism Project, bLaboratory of Protein Metabolism, and cCenter for Basic Technology Research, Tokyo Metropolitan Institute of MedicalScience, 2-1-6 Kamikitazawa, Setagaya-ku, Tokyo 156-8506, Japan

Edited by Aaron Ciechanover, Technion-Israel Institute of Technology, Bat Galim, Haifa, Israel, and approved March 10, 2015 (received for review November21, 2014)

The identification of substrates for ubiquitin ligases has remainedchallenging, because most substrates are either immediately de-graded by the proteasome or processed by deubiquitinatingenzymes (DUBs) to remove polyubiquitin. Although a methodol-ogy that enables detection of ubiquitinated proteins using ubiq-uitin Lys-e-Gly-Gly (diGly) remnant antibodies and MS has beendeveloped, it is still insufficient for identification and characteriza-tion of the ubiquitin-modified proteome in cells overexpressinga particular ubiquitin ligase. Here, we show that exogenouslyexpressed trypsin-resistant tandem ubiquitin-binding entity(ies)(TR-TUBE) protect polyubiquitin chains on substrates fromDUBs and circumvent proteasome-mediated degradation in cells.TR-TUBE effectively associated with substrates ubiquitinated by anexogenously overexpressed ubiquitin ligase, allowing detectionof the specific activity of the ubiquitin ligase and isolation of itssubstrates. Although the diGly antibody enabled effective identi-fication of ubiquitinated proteins in cells, overexpression of anubiquitin ligase and treatment with a proteasome inhibitor didnot increase the level of diGly peptides specific for the ligase rel-ative to the background level of diGly peptides, probably due todeubiquitination. By contrast, in TR-TUBE–expressing cells, thelevel of substrate-derived diGly peptides produced by the over-expressed ubiquitin ligase was significantly elevated. We developeda method for identifying the substrates of specific ubiquitin li-gases using two enrichment strategies, TR-TUBE and diGly remnantantibodies, coupled with MS. Using this method, we identified tar-get substrates of FBXO21, an uncharacterized F-box protein.

ubiquitin-binding protein | ubiquitin ligase | ubiquitination

Posttranslational modification by ubiquitin regulates diverseprocesses in cells (1, 2). Ubiquitination is catalyzed by three

types of enzymes—E1, E2, and E3, with the selectivity for thetarget protein provided by E3 ubiquitin ligases. Although thehuman genome encodes more than 600 ubiquitin ligases, manyof them remain to be studied (3). The Skp1–Cul1–F-box protein(SCF) complex, one of the best-characterized ubiquitin ligases, iscomposed of three invariable components (Skp1, Cul1, andRbx1) and a variable component F-box protein that serves as thesubstrate recognition module. Among the over 70 F-box proteinsfound in humans, less than half have been characterized (4).The identification of substrates for a specific ubiquitin ligase

has been challenging despite considerable efforts. To date, thephysical interaction between an ubiquitin ligase and its substrateshas been exploited as the major approach for substrate identi-fication (5–7). In these studies, immunoprecipitation followed byMS has been used to isolate ligase–substrate complexes. How-ever, there are several difficulties associated with this approach:Most ligase–substrate interactions are generally too weak andtransient to isolate the substrates by immunoprecipitation, andthe abundances of relevant in vivo substrates are often low dueto proteasomal degradation.Recently, an antibody that recognizes the ubiquitin remnant

motif Lys-e-Gly-Gly (diGly), which is exposed upon tryptic digestion

of ubiquitinated proteins, has been developed for global pro-teomic applications aimed at identifying ubiquitinated substrates(8, 9). Although a few quantitative proteomics studies haveidentified a particular ubiquitin ligase substrate using stableisotope labeling utilizing amino acids in cell culture and the anti-diGly antibody (10), these examples required large quantities ofsamples and advanced techniques.Tandem ubiquitin-binding entity(ies) (TUBE) based on ubiquitin-

associated domains have been developed for isolation of poly-ubiquitinated proteins from cell extracts (11). Notably, TUBEreagents protect polyubiquitin-conjugated proteins in cell lysatesfrom both proteasomal degradation and deubiquitinating en-zymes (DUBs) as efficiently as specific inhibitors of these en-zymes (11). In this paper, we applied the TUBE technology to invivo capture of ubiquitinated proteins. To develop a versatilemethod for identifying substrates of a specific ubiquitin ligase, wedesigned a mammalian expression vector encoding a FLAG-tagged trypsin-resistant (TR) TUBE, which protects ubiquitinchains from trypsin digestion under native conditions. Using twoenrichment methods, TR-TUBE and the anti-diGly antibody, wesucceeded in identifying the target substrates of the uncharac-terized F-box protein FBXO21.

ResultsProtection of Polyubiquitin Chains on Substrates by TR-TUBE. Ourmethod is based on stabilization of ubiquitinated substrates invivo by masking of ubiquitin chains with exogenously expressed

Significance

The identification of specific ubiquitin ligase–substrate pairs iscrucial for understanding the roles of protein ubiquitination inthe regulation of diverse biological processes. Despite thedevelopment of various methodologies for substrate identi-fication, it remains challenging to determine ubiquitin ligasesubstrates. Based on previously described tandem ubiquitin-binding entity(ies) (TUBE), we designed the trypsin-resistant(TR)-TUBE for expression in cells. The coexpression of TR-TUBEwith an ubiquitin ligase stabilizes the ubiquitinated substratesby masking the ubiquitin chains. Using a combination of twostrategies for enriching ubiquitinated substrates, TR-TUBE andanti–Lys-e-Gly-Gly antibody, we successfully identified specificubiquitin ligase–substrate pairs. Our methodology provides aneffective means for the identification of ubiquitin ligase sub-strates and the detection of ubiquitin ligase activity.

Author contributions: Y.Y., Y.S., and K.T. designed research; Y.Y., Y.S., A.M., J.K., H.T., andH.Y. performed research; Y.Y., Y.S., H.T., H.Y., and M.S. analyzed data; and Y.Y., Y.S., andK.T. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence may be addressed. Email: yoshida-yk@igakuken.or.jp ortanaka-kj@igakuken.or.jp.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1422313112/-/DCSupplemental.

4630–4635 | PNAS | April 14, 2015 | vol. 112 | no. 15 www.pnas.org/cgi/doi/10.1073/pnas.1422313112

Dow

nloa

ded

by g

uest

on

May

14,

202

1

TR-TUBE (Fig. 1A). We confirmed that the TR-TUBE can bindto all eight types of ubiquitin chain linkages (Fig. S1A). To ex-amine the effect of overexpressed TR-TUBE on ubiquitin ho-meostasis and cytotoxicity, we first expressed TR-TUBE or anubiquitin-binding–deficient TR-TUBE mutant (12) in 293T cells(Fig. S1 B and C). We then detected the cellular levels of freeand conjugated ubiquitin by immunoblotting with an ubiquitin-specific antibody, and also analyzed cell death by propidiumiodide staining (Fig. 1 B and C). The level of conjugatedubiquitin was increased by 48 h posttransfection, with a concom-itant reduction in free ubiquitin (Fig. 1B). By contrast, we detectedlittle accumulation of ubiquitin conjugates in cells expressing theubiquitin-binding–deficient TR-TUBE mutant or in cells treatedwith the proteasome inhibitor MG132. The accumulation ofubiquitin conjugates upon TR-TUBE expression did result insome cell death at 48 h, comparable to the levels resulting fromlong-term treatment with MG132 (Fig. 1C). Although prolongedexpression of TR-TUBE did not induce significant cell death, theaccumulation of ubiquitin conjugates 72 h posttransfection incells was reduced relative to the accumulation of ubiquitin con-jugates at 48 h (Fig. 1 B and D), suggesting that cells highlyexpressing TR-TUBE gradually undergo cell death.To confirm that ubiquitinated substrates were actually in-

cluded in the high-molecular-weight ubiquitin conjugates, wemonitored the ubiquitination levels of p27/CDKN1B, one ofthe best-characterized ubiquitinated substrates. To this end, weexpressed FLAG-tagged TR-TUBE or an ubiquitin-binding–deficient TR-TUBE mutant in cells with or without exogenouslyexpressed Skp2/FBXL1, the F-box protein that recognizes p27(13). Cell lysates were immunoprecipitated with an anti-FLAG

antibody, and both whole-cell lysates and precipitates were an-alyzed by immunoblotting (Fig. 1D and Fig. S2).Both ubiquitin conjugates and ubiquitinated p27 were effi-

ciently enriched in the TR-TUBE immunoprecipitates. Althoughubiquitin conjugates were detected in TR-TUBE mutant–expressingcells that were treated with proteasome/DUB inhibitors,ubiquitinated p27 was barely detectable, even when cells over-expressing Skp2 were treated with inhibitors (Fig. S2). By contrast,ubiquitinated p27 was detected in lysates of TR-TUBE– and Skp2-overexpressing cells in the absence of inhibitor treatment, sug-gesting that TR-TUBE both protects the polyubiquitin chains onp27 from DUBs and circumvents proteasome degradation. Asshown in Fig. 1D, ubiquitin conjugates and ubiquitinated p27 weredetectable in TR-TUBE immunoprecipitates from cells harvested24 h posttransfection in the absence of exogenous Skp2 expres-sion. Notably, ubiquitinated p27 was present at high levels 48 hposttransfection in cells coexpressing Skp2 and TR-TUBE, sug-gesting that the ubiquitination of p27 detected in these sampleswas carried out by SCFSkp2, primarily exogenous Skp2.

Detection of Ubiquitination Activity Using TR-TUBE. Conventionally,detection of specific ubiquitin ligase activity has been conductedusing in vitro reconstitution or overexpression systems consistingof ubiquitin, a substrate, and an ubiquitin ligase in the presenceof proteasome inhibitors. Indeed, we detected ubiquitination ofp27 by SCFSkp2 in ubiquitin immunoprecipitates only whenthe substrate was overexpressed (Fig. 1E, lanes 5–8). However,overexpression of Skp2 failed to increase the ubiquitinationof p27 to a detectable level. By contrast, in the presence ofTR-TUBE, exogenous expression of Skp2 increased ubiquitination

CBTR-TUBE mutantTR-TUBE +

++

++

++

MG132 (h) -------- 4 8

empposttransfection (h) 0 24 48 4872

WCL

Ubiquitin

conjugates

kDa

Ub

D

E

A

IP: TR-TUBEE3

Proteasome

DUB

TR-TUBE

Lyse

TR-T

UBE

SubstrateUbUbUbUbUb

UbUbUb

TR-T

UBE

SubstrateUbUbUbUb

18898624938281714

6

αUbiquitin

- +HA-Skp2 - + - +posttransfection (h)24 48 72

WCL

- + - + - +24 48 72

FLAG-TR-TUBEIP: αFLAG

- + - + - +24 48 72

WCL

- + - + - +24 48 72

FLAG-TR-TUBEIP: αFLAG

αp27αUbiquitin

(Ub)n

-p27

18898624938281714

6

kDa

25015010075

5037

2520

kDa

- + - + - + - + - + - + - + - +

emp- + - + - + - +

p27 empUb TR-TUBE

IP: αFLAG

HA-Skp2HA-taggedFLAG-tagged

311 11 10 9 8 7 6 5 4 3 412 12 16 15αp27

p27

MG132

25015010075

5037

2520

kDa

(Ub)n

-p27

Substrate

020406080

0 4 8020406080

100 TR-TUBEmutant

Via

bilit

y (%

)

0 24 48 72posttransfection (h)

Via

bilit

y (%

)

MG132 treatment (h)

100

Fig. 1. Protection of polyubiquitin chains on substrates by TR-TUBE. (A) TR-TUBE method for isolation of ubiquitinated substrates. Polyubiquitin chains onsubstrates are masked by exogenously expressed TR-TUBE, and thereby protected from DUBs and the proteasome. Ubiquitinated proteins are enriched byimmunoprecipitation (IP) of TR-TUBE from cells expressing E3 ubiquitin ligase and TR-TUBE. Exogenously expressed proteins are shown in red. Ub, ubiquitin.(B) Accumulation of ubiquitin conjugates in TR-TUBE–expressing cells. 293T cells were transfected with FLAG-TR-TUBE or ubiquitin-binding–deficient FLAG-TR-TUBE mutant plasmid, and the transfected cells were harvested at the indicated times. Cells transfected with HA-empty (emp) vector were treated with10 μM MG132 for the indicated time before harvesting. Whole-cell lysates (WCLs) were analyzed by immunoblotting using antiubiquitin antibody. (C) Effectof TR-TUBE expression or MG132 treatment on cell viability, as determined by propidium iodide staining. Three independent plates of transfected or MG132cells were analyzed. Error bars represent means ± SEM. (D) Detection of ubiquitin conjugates and ubiquitinated endogenous p27. Cells (1.3 × 106) werecotransfected with 3.5 μg of FLAG-TR-TUBE and 3.5 μg of HA-empty or HA-Skp2 expression plasmids, and the transfected cells were harvested at the indicatedtimes. WCLs and anti-FLAG immunoprecipitates were analyzed by immunoblotting. The arrow indicates the position of p27. (E) Detection of ubiquitination ofendogenous and overexpressed p27. Cells expressing FLAG-ubiquitin or FLAG-TR-TUBE with or without HA-Skp2 and/or HA-p27 were treated with or withoutMG132, and the cells were harvested at 48 h posttransfection. Anti-FLAG immunoprecipitates were analyzed by immunoblotting. The arrow indicates theposition of p27.

Yoshida et al. PNAS | April 14, 2015 | vol. 112 | no. 15 | 4631

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

May

14,

202

1

of p27 (Fig. 1E, lanes 9–12): Ubiquitinated p27 was detectable evenin the absence of exogenously overexpressed p27 or MG132. Be-cause the elevated level of ubiquitin ligase was reflected by anincrease in the accumulation of ubiquitinated substrates, theTR-TUBE system appears to be useful for quantitative detectionof ubiquitin ligase activity toward a specific substrate.To suppress ubiquitination by endogenous Skp2, we next

expressed the dominant-negative mutant Skp2ΔF, which lacksthe F-box domain essential for binding to Cul1, thereby inhib-iting formation of the SCF ubiquitin ligase complex. Althoughexpression of the ΔF mutant failed to suppress the accumulationof ubiquitinated p27 in the presence of MG132, the absence ofthe inhibitor altered accumulation of ubiquitinated substrate butdid not necessarily alter activity of the ligand (Fig. 2A). A similarubiquitination pattern was detected for another Skp2 substrate,CDT1 (14). Next, we examined the ubiquitination activity of theother well-characterized F-box proteins, FBXW7 and FBXW1/βTrCP1 (Fig. 2 B–D). The ubiquitination of c-Myc (15), a knownsubstrate of FBXW7, was clearly increased by overexpression ofFBXW7 and decreased by overexpression of its dominant-neg-ative mutant (Fig. 2B). Overexpression of FBXW1 also increasedthe ubiquitination of NFKBIA (16) and PDCD4 (17), knownFBXW1 substrates, even in the absence of stimuli (Fig. 2 C andD). Furthermore, we investigated the ubiquitination activity ofRING-type E3 MDM2 for p53 (18) (Fig. 2E). The level of poly-ubiquitinated p53 was somewhat increased by overexpression ofMDM2, whereas monoubiquitinated p53 was present at highlevels, suggesting that MDM2 preferentially catalyzes mono-ubiquitination of p53 under these conditions (19). Furthermore,we examined the differences in the ubiquitination pattern ofsubstrates due to overexpression of E3 in lysates prepared atdifferent time points (Fig. S3). Although c-Myc was highlyubiquitinated 24 h posttransfection in cells expressing FBXW7,other ubiquitinated substrates accumulated at high levels 48 hposttransfection. Therefore, in subsequent experiments, we usedlysates prepared from cells harvested 48 h posttransfection.

Development of a Method to Identify Substrates for a Ubiquitin Ligase.Recently, a large number of ubiquitination sites and ubiquitinatedproteins were identified using the anti-diGly antibody (8, 9, 20). Inthose studies, this antibody was used for direct immunoprecipi-tation of trypsinized lysates prepared from cells treated withproteasome inhibitor. Therefore, to assess the efficiency of iden-tification of ubiquitinated proteins using TR-TUBE, we com-pared three methods for enriching ubiquitinated peptides forliquid chromatography (LC)-tandem MS (MS/MS) analysis(Fig. 3A). These analyses were performed using starting materialfrom ∼1 × 107 293T cells. The first and second methods useddirect peptide immunoprecipitation, utilizing anti-diGly anti-body, of trypsinized cell lysates from MG132-treated cells [diGly(MG132)] or TR-TUBE–overexpressing cells [diGly (FLAG-TR-TUBE)]. In the direct peptide immunoprecipitation, cellswere lysed in the presence of 9 M urea and the denatured pro-teins were diluted with Hepes buffer before trypsin digestion. Bycontrast, the third method used anti-FLAG antibody enrichmentof TR-TUBE–associated proteins before peptide immunopre-cipitation [FLAG and diGly (FLAG-TR-TUBE)]. Although thenumbers of unique peptides containing the K-e-GG motif (diGlypeptides) and ubiquitinated proteins did not differ significantlybetween these three methods, the ratios of diGly peptides tototal identified peptides and ubiquitinated proteins weredrastically different (Fig. 3B). In the dual-enrichment method,more than 95% of identified peptides included the K-e-GGmotif. In addition, overexpression of TR-TUBE markedly de-creased the abundance of identified peptides derived fromubiquitin, because TR-TUBE protected polyubiquitin chains onsubstrates from trypsin digestion (Fig. S4). These results show

that initial enrichment of ubiquitinated proteins using TR-TUBEreduces the background without loss of diGly peptides.To evaluate the performance of the TR-TUBE system in iden-

tification of ubiquitin ligase substrates, we next compared thenumber of peptide spectrum matches (PSMs) as a semiquantitativeindex of three known Skp2 substrates (p27, p21/CDKN1A, andCDT1), in mock-transfected cells, Skp2-expressing cells, andSkp2 ΔF-expressing cells (Fig. 3C). In the first method, usingMG132-treated cells, p27 and p21 were barely detectable, whereasCDT1 was reproducibly observed. However, levels of the diGlypeptides of these proteins were not always elevated in Skp2-expressing cells. By contrast, in the methods using cells expressingTR-TUBE, especially in the dual-enrichment method, the levels ofdiGly peptides derived from the substrates were elevated in Skp2-expressing cells. Although p27 itself was stabilized by MG132treatment, ubiquitinated p27 was barely detectable in MG132-treated cell lysates (Fig. 3D). By contrast, high-molecular-weightsmears of both p27 and CDT1 were clearly detected in cellscoexpressing TR-TUBE and WT Skp2. Notably, low levels ofubiquitinated CDT1 were detected in MG132-treated cellsexpressing either WT or mutant Skp2, but not in untreated cells,consistent with the results of the MS analyses. Thus, althoughthere was little difference between the three methods with regardto the efficiency of identification of ubiquitinated proteins, en-richment of TR-TUBE–associated proteins before diGly peptideimmunoprecipitation is an effective method for identifying sub-strates of an overexpressed ubiquitin ligase.In addition, we found that the levels of two ubiquitinated

peptides derived from CKS1B were markedly elevated in Skp2-expressing cells. CKS1B is an essential cofactor of SCFSkp2 that isnecessary for the ubiquitination of p27 (21), and CKS1B itself isubiquitinated by the APC/CCdh1 ubiquitin ligase (22). Hence, weinvestigated whether Skp2 ubiquitinates CKS1B by coexpressingmyc-tagged CKS1B and TR-TUBE in cells, followed by immu-noprecipitation of TR-TUBE and immunoblot analysis (Fig. 3E).

250150100755037

25

kDa

- +MG132emp W7

- + - +ΔFHA-tagged

- +MG132empSkp2

- + - +ΔFHA-tagged

250150100755037

kDa

FLAG-TR-TUBEIP: αFLAG

- +empSkp2

- + - +ΔF

250150100755037

25

kDa

FLAG-TR-TUBEIP: αFLAG

FLAG-TR-TUBEIP: αFLAG

αc-Mycαp27 αCDT1

BA

D EC

25015010075503725

kDa

- +MG132empMDM2

- +HA-tagged

IP: αFLAGFLAG-TR-TUBE

- +++ΔF

250150100755037

25

kDa

- +MG132

empW1

- + - +

HA-tagged

- + - +------ ++++

empW1Serum depl

ΔF

IP: αFLAG

MG132TNF

25015010075503725

kDa

- +

empW1

- + - +

HA-tagged

- + - + - +------ ++++++

empW1 ΔFΔF

IP: αFLAGFLAG-TR-TUBE FLAG-TR-TUBE

αPDCD4αNFKBIA αp53

25

Fig. 2. Detection of ubiquitination activity using TR-TUBE. Ubiquitinationassays using the TR-TUBE method are shown. Cells (1.3 × 106) were cotrans-fected with 3.5 μg of plasmid encoding FLAG-TR-TUBE in combination with3.5 μg of plasmid encoding emp, WT F-box protein [Skp2 (A), FBXW7 (W7; B),and FBXW1 (W1; C and D)], its dominant-negative mutant (ΔF), or MDM2(E). Transfected cells were treated with or without MG132 for 4 h beforeharvesting. Cell lysates obtained 48 h posttransfection were immunoprecipi-tated with anti-FLAG antibody, and the immunoprecipitates were analyzed byimmunoblotting. Vertical bars and arrows denote the positions of ubiquiti-nated substrates and unmodified substrates, respectively.

4632 | www.pnas.org/cgi/doi/10.1073/pnas.1422313112 Yoshida et al.

Dow

nloa

ded

by g

uest

on

May

14,

202

1

As predicted, the expression of Skp2 stimulated the ubiquitina-tion of CKS1B.The identified ubiquitination sites of Skp2 substrates are listed

in Table S1, all of which were included in the PhosphoSite Plusand neXtProt databases.

Identification of Substrates for Uncharacterized Ubiquitin Ligases.Our next goal was to develop a method for the systematic iden-tification of ubiquitin ligase–substrate pairs. For this purpose,we examined the F-box proteins constituting the SCF complex in293T cells and identified 12 F-box proteins from a FLAG-Cul1immunoprecipitate (Table S2). We attempted to find substratesfor FBXO21, which is ubiquitously expressed and has not yet beenwell characterized.To screen for substrates of FBXO21, we performed LC-MS/MS

analysis of peptides prepared by the dual-enrichment methodfrom cells coexpressing FLAG-TR-TUBE and HA-empty,HA-FBXO21ΔF, or HA-FBXO21. In three independent analyses,we selected substrate candidates whose PSM numbers and pro-tein scores increased in FBXO21-expressing cells but decreasedin FBXO21ΔF-expressing cells (Table S3). Of these candidates,we picked threonyl-tRNA synthetase (TARS) and EP300 inter-acting inhibitor of differentiation 1 (EID1) because of their re-producibility (Fig. 4A). We also confirmed that the interactionof FBXO21 with TARS or EID1 was stabilized by Skp1 coex-pression (23) and treatment with the Nedd8 E1 enzyme inhibitor

MLN4924, which stabilizes Cullin-RING ligase substrates (20)(Fig. S5A). We cloned several F-box proteins and investigatedwhether they bound EID1 or TARS in the presence of MLN4924(Fig. S5B). Although TARS was detectable at low levels in a fewF-box protein immunoprecipitates, the levels of TARS and EID1were most prominent in FBXO21 immunoprecipitate.EID1, first cloned as an RB1-binding protein, has the ability to

inhibit p300 (24) and can also interact with the orphan nuclearreceptor SHP (25). To determine whether the interaction ofEID1 with FBXO21 is mediated by RB1 or SHP, we sought toidentify the regions on EID1 that are necessary for FBXO21interaction (Fig. S5C). Deletion of the C-terminal RB1-bindingsite of EID1 did not affect the binding to FBXO21, but deletionof an additional nine residues abolished interaction, suggestingthat their interaction is independent on RB1 and SHP. Next, weused immunoblot analysis to evaluate the ability of FBXO21 toubiquitinate EID1. Ubiquitinated EID1 was clearly detected inboth FBXO21-expressing and MG132-treated cells (Fig. 4B),and knockdown of FBXO21 by siRNA suppressed accumulationof ubiquitinated EID1 in MG132-treated cells (Fig. 4C). Further,ubiquitination of EID1 was clearly increased by overexpressionof FBXO21, but not other the F-box proteins tested (Fig. 4D).As shown in Fig. 4E, knockdown of FBXO21 by siRNA stabi-lized EID1 protein levels but did not affect EID1 mRNA levels(Fig. 4F). Consistent with the steady-state levels of EID1, FBXO21knockdown, as well as treatment of cells with MLN4924, led

K -GG

B

C ED

A I. diGly (MG132) II. diGly (FLAG-TR-TUBE) III. FLAG & diGly (FLAG-TR-TUBE)

empty

ΔF

F-box protein

FLAG-TR-TUBE / empty

FLAG-TR-TUBE / ΔF

FLAG-TR-TUBE / F-box proteinMG132 treatment

Lyse (urea lysis buffer)

Trypsin digestion

Peptide IP with anti-diGly antibody

Trypsin digestion

Protein IP with anti-FLAG antibody

LC-MS/MS analysis

Trypsin digestion

LC-MS/MS analysis

LC-MS/MS analysis

Peptide IP with anti-diGly antibody

Peptide IP with anti-diGly antibody

Lyse (TBS-N)Lyse (urea lysis buffer)

48 h post-transfection 48 h post-transfection48 h post-transfection

MG132 - +++ ---

emp

Skp

2ΔFHA-tagged

emp

Skp

2ΔFem

p

TR-TUBE

- +++ ---

emp

Skp

2ΔFem

p

Skp

2ΔFem

p

TR-TUBEWCL WCL

(Ub)n

-CDT1

αp27 αCDT1

25015010075

5037

2520

kDa

- +emp Skp2

- + - +ΔFHA-tagged

6xMyc-tagged CKS1BFLAG-TR-TUBE

IP: αFLAG

(Ub)n

-CKS1B

αMyc

25015010075

5037

2520

kDa

0

2000

4000

6000

8000

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

peptidediGly peptide

0

500

1000

1500

2000

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

proteinub-protein

diGly(MG132)

diGly(TR-TUBE)

FLAG & diGly(TR-TUBE)

# experiment

# pe

ptid

e

# experimentdiGly

(MG132)diGly

(TR-TUBE)FLAG & diGly

(TR-TUBE)

# pr

otei

n

# P

SM

s

010203040

0

4

8

12

1 2 3 1 2 3 1 2 3# experimentdiGly

(MG132)diGly

(TR-TUBE)FLAG & diGly

(TR-TUBE)

# P

SM

s

emptySkp2ΔFSkp2

1 2 3 1 2 3 1 2 3diGly

(MG132)diGly

(TR-TUBE)FLAG & diGly

(TR-TUBE)

12p72p

0

4

8

12

16

1 2 3 1 2 3 1 2 3diGly

(MG132)diGly

(TR-TUBE)FLAG & diGly

(TR-TUBE)

0

20

40

60

1 2 3 1 2 3 1 2 3diGly

(MG132)diGly

(TR-TUBE)FLAG & diGly

(TR-TUBE)

emptySkp2ΔFSkp2

# P

SM

s

# P

SM

s

CDT1 CKS1B

# experiment

TR-T

UBE

SubstrateUbUbUbUb

K -GG

K -GG

Fig. 3. Development of a method for identifying the substrates of an ubiquitin ligase. (A) Schematic indicating the steps in the substrate identificationprocesses. In the diGly (MG132) method, 293T cells were transfected with HA-empty (empty), ΔF, or WT F-box protein, and the transfectants were treatedwith MG132 for 4 h before harvesting. In the diGly (TR-TUBE) and FLAG and diGly (TR-TUBE) methods, cells were transfected with FLAG-TR-TUBE in com-bination with HA-empty, ΔF, or WT F-box protein. In the diGly (MG132) and diGly (TR-TUBE) methods, cells were lysed 48 h posttransfection in urea-basedsolution and diluted WCLs were digested with trypsin. In the FLAG and diGly (TR-TUBE) method, cells were lysed and immunoprecipitated with anti-FLAGantibody and the eluted proteins from immunoprecipitates were digested with trypsin. The tryptic peptides were further enriched in another immunoaffinitystep for peptides containing the K-e-GG motif (peptide IP with anti-diGly antibody), followed by MS analysis. (B) Comparison of the numbers of diGly peptide(Upper) and ubiquitinated protein (Lower) numbers identified by the three methods described in A. Five individual experiments were performed for eachmethod. (C) Comparison of the efficiency of identification of Skp2 substrates by the three methods described in A. Total peptide spectrum match numbers(# PSMs) of the indicated proteins (p27, p21, CDT1, and CKS1B) obtained from LC-MS/MS analysis using cells expressing HA-empty (blue bars), dominant-negative mutant (Skp2ΔF, green bars), or WT Skp2 (red bars) with or without FLAG-TR-TUBE are shown. Three individual experiments were performed.(D) Detection of ubiquitinated p27 and CDT1 in MG132-treated cells and TR-TUBE–expressing cells. Vertical bars indicate ubiquitinated substrates, and arrowsindicate the positions of unmodified substrates. (E) Testing the ubiquitination of CKS1B by Skp2. The arrow indicates the position of 6× Myc-CKS1B.

Yoshida et al. PNAS | April 14, 2015 | vol. 112 | no. 15 | 4633

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

May

14,

202

1

to stabilization of EID1 in a cycloheximide chase experiment,extending the t1/2 from <30 min to >3 h (Fig. 4G). These resultsindicate that EID1 is degraded by the proteasome followingubiquitination by SCFFBXO21.MDM2, rather than Cullin-RING ubiquitin ligase, is the

ubiquitin ligase for EID1 (26, 27). However, siRNA-mediatedknockdown of MDM2 or RB1 did not stabilize EID1 protein in293T cells (Fig. 4H). RB1 was stabilized by knockdown ofMDM2 at 24 h, consistent with the fact that MDM2 promotesdegradation of RB1 (28, 29). In addition, interaction with EID1causes a conformational rearrangement of RB1 (29, 30), andtheir interaction may cause RB1 stabilization because the RB1level was also increased by knockdown of FBXO21 (Fig. 4H).Furthermore, we examined the ubiquitination activity of MDM2for EID1 (Fig. 4I). However, ubiquitinated EID1 was barelydetectable in MDM2-expressing cells, suggesting that the stabilityof EID1 is regulated by SCFFBXO21 under normal conditions.TARS is an aminoacyl-tRNA synthetase that is as abundant in

cells as ribosomal proteins (31). Although TARS is a highly abun-dant and stable protein due to its fundamental roles in proteinsynthesis, 13 residues in TARS are ubiquitinated by Cullin-RINGubiquitin ligases (9), and we found that four sites were ubiquitinatedby SCFFBXO21 (Table S1). Indeed, ubiquitination of TARSwas stimulated by overexpressed FBXO21 (Fig. S6 A and B).However, the levels of endogenous TARS protein remainedconstant regardless of whether FBXO21 was knocked down(Fig. S6 C–E). Furthermore, we showed that FBXO21 interacted

with the TARS editing domain, which includes all of the identifiedubiquitination sites within the TARS protein (Fig. S6F and TableS1). In Escherichia coli, severe oxidative stress reduced overalltranslational fidelity by impairing the editing activity of TARS (24).Actually, ubiquitination of TARS by SCFFBXO21 was slightly ele-vated in H2O2-treated cells (Fig. S6G), but the stress did not affectits stability (Fig. S6H). These results suggest that TARS is regulatedby SCFFBXO21 ubiquitination under stress conditions but that itsdegradation is not detectable because the protein is so abundant.Editing-defective aminoacyl-tRNA synthetase causes protein mis-folding and neurodegeneration (32), and stresses other than oxi-dative stress may damage its editing activity. Further studies areneeded to determine how FBXO21 detects damage in the TARSediting domain.

DiscussionIn this study, we developed the TR-TUBE system, which isuseful for detecting ubiquitin ligase activity and identifying sub-strates of specific E3 ubiquitin ligases. In this system, overex-pressed E3 ubiquitinates its endogenous substrates by usingubiquitination-related factors present in cells, and TR-TUBEprevents degradation and deubiquitination of these substrates,allowing detection of the specific activity of an E3 and isolationof its substrates. Although TR-TUBE immunoprecipitates con-tain excess ubiquitin, TR-TUBE can protect ubiquitin chains onsubstrates from trypsin digestion (Fig. S4) and reduce the proportionof peptides derived from ubiquitin, which hinder identification of

DA

F

αEID1

CHX (hr) 0 .5 1 2 3

si cont si FBXO21 MLN4924

0 .5 1 2 3 0 .5 1 2 3

H I

FBXW

1FB

XW7

FBXW

8FB

XW11

Skp2

FBXL

12em

pty

FBXO

5FB

XO7

FBXO

9FB

XO11

FBXO

21FB

XO22

FBXO

44

FLAG-F box

IP: αHAHA-TR-TUBE

αEID1

- + - +siRNA cont F21

FLAG-TR-TUBEIP: αFLAG

(Ub)n

-EID1

C

GαEID1

αFBXO21

αRb1

αMDM2

si

cont

FBXO

21R

b1M

DM

2co

ntFB

XO21

Rb1

MD

M2

24 h 48 h

kDa75

1501001007537

E

si

αEID1αFBXO21

cont

FBXO

21co

ntFB

XO21

Rel

ativ

e ex

pres

sion EID1 mRNA

00.20.40.60.81.0

cont FBXO210

0.20.40.60.81.0

FBXO21 mRNA

si cont FBXO21

αEID1

- +emp F21

- + - +HA-tagged

FLAG-TR-TUBEIP: αFLAG

(Ub)n

-EID1

MDM2

αEID1

(Ub)n

-EID10

10

20

30

40

1 2 3 1 2 3

vec

FBXO21

FBXO21

ΔF

# P

SM

s

TARS EID1

emp F21 ΔFHA-taggedFLAG-TR-TUBE

IP: αFLAG

- + - + - +

(Ub)n

-EID1

αEID1

B

250

150100755037

25

kDa250

15010075

5037

25

kDaMG132

250

1501007550

37

25

kDa

MG132

250

15010075

5037

25

kDaMG132

37

# experiment

Fig. 4. Identification of substrates for FBXO21. (A) Screening of FBXO21 substrates by the dual-enrichment method using TR-TUBE and anti-diGly antibody.In three sets of independent MS analyses, we selected proteins whose PSM numbers (# PSMs) increased in cells expressing FBXO21 and decreased in cellsexpressing FBXO21 mutant. The total PSM numbers of identified substrates (TARS and EID1) obtained from LC-MS/MS analysis using cells expressingHA-empty (blue bars), ΔF (green bars), or WT FBXO21 (red bars) with FLAG-TR-TUBE are shown. (B and C) Ubiquitination assay of EID1. Forty-four hours afterFLAG-TR-TUBE in combination with emp, WT FBXO21 (F21), its mutant (ΔF), or siRNA transfection, cells were treated with or without MG132 for 4 h. Anti-FLAG immunoprecipitates were analyzed by immunoblotting. Arrows show the positions of EID1. cont, control. (D) Ubiquitination assay of EID1 by usingTR-TUBE and various F-box proteins. Cells were transfected with plasmids encoding HA-TR-TUBE and each FLAG-tagged F-box protein. Anti-HA immuno-precipitates were analyzed by immunoblotting. The arrow shows the position of EID1. (E) RNAi-mediated knockdown of FBXO21. Forty-eight hours aftersiRNA (si) transfection, FBXO21 and EID1 protein levels in WCLs were assessed by immunoblotting. (F) Quantitative RT-PCR analysis. Total RNA was preparedfrom 293T cells 48 h after cells were transfected with control or FBXO21-specific siRNA. The data shown are representative of three independent experiments.(G) Forty-eight hours after siRNA transfection, cells were incubated with 2 μg/mL cycloheximide (CHX). In parallel, cells were treated with 1 μM MLN4924 for1 h before addition of cycloheximide. Cells were harvested at the indicated times after cycloheximide treatment, and WCLs were analyzed by immuno-blotting. (H) Stabilization of EID1 by depletion of FBXO21. Twenty-four hours or 48 h after transfection of the indicated siRNAs, protein levels in WCL wereanalyzed by immunoblotting. (I) Ubiquitination of EID1 by MDM2 and F21. The arrow shows the position of EID1.

4634 | www.pnas.org/cgi/doi/10.1073/pnas.1422313112 Yoshida et al.

Dow

nloa

ded

by g

uest

on

May

14,

202

1

substrates in LC-MS/MS analysis. In trypsinized TR-TUBE immu-noprecipitates, substrates of coexpressed Skp2 could be detected byLC-MS/MS analysis; however, we also detected other proteins thatare abundant in complex with these substrates, as well as Skp2 itself,but no diGly peptides other than those diGly peptides derived fromubiquitin and Skp2. Therefore, the second enrichment with anti-diGly antibody is remarkably effective for identification of substratesof a particular ubiquitin ligase.All of the ubiquitination sites of substrates identified in this

study were included among the known ubiquitination sites con-tained in databases (Table S1). However, some sites identifiedin CDT1 and TARS differed from the sites reported to beubiquitinated by the Cullin-RING ubiquitin ligase (9). In addi-tion to three reported ubiquitination sites (K132, K153, andK165) in CDT1, which our system failed to detect, K24 ofCDT1 was ubiquitinated by SCFSkp2. CDT1 is ubiquitinated byboth SCFSkp2 (14) and Cul4-Ddb1 (33), suggesting that the reportedsites are ubiquitinated by Cul4-Ddb1. Notably, although theubiquitination sites of p27 have been analyzed by site-directedmutagenesis (34), these sites have not previously been assignedby proteomic analyses. The diGly peptides derived from p27were barely detectable in peptide immunoprecipitates of trypsi-nized cell lysates from cells treated with proteasome inhibitor,but they were effectively detected in lysates from TR-TUBE–expressing cells, suggesting that deubiquitination of p27, ratherthan instability of the protein itself, hampers detection of thediGly peptides.Recently, the ubiquitin ligase substrate trapping method was

developed for the isolation of ubiquitinated substrates in yeast.In that method, ligase–substrate affinity is increased by fusing theligase to a tandem ubiquitin-associated domain (35). Althoughthis method should also be useful for identification of the

substrates of ubiquitin ligases, it requires parameters such aslinker length and configuration to be optimized for each F-boxprotein. By contrast, the TR-TUBE strategy is very simple,requiring only coexpression of TR-TUBE and the E3. Fur-thermore, the TR-TUBE method is not restricted to SCF-typeubiquitin ligases, and is also potentially useful for other E3 families.By using a combination of two enrichment strategies, TR-TUBEand the anti–K-e-GG antibody, we succeeded in identifying ubiq-uitinated substrates from small amounts of cell lysate. Thus, the TR-TUBE system represents a practical means for obtaining importantinsights into the functions of ubiquitin ligases.

Materials and MethodsFor immunoaffinity purification for ubiquitinated protein identification,WCLprepared from a 10-cm cell culture dish harvested 48 h posttransfection (∼1 ×107 cells) was incubated for 1 h with anti-FLAG monoclonal antibody (anti-DDDDK)–conjugated agarose beads (MBL International). Bead-bound pro-teins were eluted FLAG peptide (Sigma). Proteins were reduced in 5 mM Tris[2-carboxy-ethyl] phosphine hydrochloride for 30 min at 50 °C, and thenalkylated with 10 mM methylmethanethiosulfonate, and alkylated proteinswere digested overnight at 37 °C with 1 μg of trypsin (Promega). After trypticdigestion, ubiquitinated peptides were enriched by using the PTMScan ubiq-uitin remnant motif (K-e-GG) kit (Cell Signaling). The eluted peptides weredesalted using GL-Tip SDB and GL-Tip GC (GL Sciences) prior to LC-MS analysis.Detailed methods are provided in SI Materials and Methods.

ACKNOWLEDGMENTS. This work was supported by a Grant-in-Aid [JapanSociety for the Promotion of Science (JSPS) KAKENHI Grant 24580152] forScientific Research on a Priority Area (to Y.Y.), a Grant-in-Aid (JSPS KAKENHIGrant 24112008) for Scientific Research on Innovative Areas (to Y.S.), Grants-in-Aid (JSPS KAKENHI Grants 2611377 and 13J07852) for JSPS Fellows (to H.T.and H.Y., respectively), and a Grant-in-Aid (JSPS KAKENHI Grant 21000012)for Specially Promoted Research (to K.T.).

1. Grabbe C, Husnjak K, Dikic I (2011) The spatial and temporal organization of ubiquitinnetworks. Nat Rev Mol Cell Biol 12(5):295–307.

2. Komander D, Rape M (2012) The ubiquitin code. Annu Rev Biochem 81:203–229.3. Deshaies RJ, Joazeiro CA (2009) RING domain E3 ubiquitin ligases. Annu Rev Biochem

78:399–434.4. Jin J, et al. (2004) Systematic analysis and nomenclature of mammalian F-box proteins.

Genes Dev 18(21):2573–2580.5. Busino L, et al. (2007) SCFFbxl3 controls the oscillation of the circadian clock by di-

recting the degradation of cryptochrome proteins. Science 316(5826):900–904.6. Davis MA, et al. (2013) The SCF-Fbw7 ubiquitin ligase degrades MED13 and MED13L

and regulates CDK8 module association with Mediator. Genes Dev 27(2):151–156.7. Tan MK, Lim HJ, Harper JW (2011) SCF(FBXO22) regulates histone H3 lysine 9 and 36

methylation levels by targeting histone demethylase KDM4A for ubiquitin-mediatedproteasomal degradation. Mol Cell Biol 31(18):3687–3699.

8. Xu G, Paige JS, Jaffrey SR (2010) Global analysis of lysine ubiquitination by ubiquitinremnant immunoaffinity profiling. Nat Biotechnol 28(8):868–873.

9. KimW, et al. (2011) Systematic and quantitative assessment of the ubiquitin-modifiedproteome. Mol Cell 44(2):325–340.

10. Sarraf SA, et al. (2013) Landscape of the PARKIN-dependent ubiquitylome in responseto mitochondrial depolarization. Nature 496(7445):372–376.

11. Hjerpe R, et al. (2009) Efficient protection and isolation of ubiquitylated proteinsusing tandem ubiquitin-binding entities. EMBO Rep 10(11):1250–1258.

12. Sasaki T, Funakoshi M, Endicott JA, Kobayashi H (2005) Budding yeast Dsk2 proteinforms a homodimer via its C-terminal UBA domain. Biochem Biophys Res Commun336(2):530–535.

13. Carrano AC, Eytan E, Hershko A, Pagano M (1999) SKP2 is required for ubiquitin-mediated degradation of the CDK inhibitor p27. Nat Cell Biol 1(4):193–199.

14. Li X, Zhao Q, Liao R, Sun P, Wu X (2003) The SCF(Skp2) ubiquitin ligase complex in-teracts with the human replication licensing factor Cdt1 and regulates Cdt1 degra-dation. J Biol Chem 278(33):30854–30858.

15. Yada M, et al. (2004) Phosphorylation-dependent degradation of c-Myc is mediatedby the F-box protein Fbw7. EMBO J 23(10):2116–2125.

16. Tan P, et al. (1999) Recruitment of a ROC1-CUL1 ubiquitin ligase by Skp1 and HOS tocatalyze the ubiquitination of I kappa B alpha. Mol Cell 3(4):527–533.

17. Dorrello NV, et al. (2006) S6K1- and betaTRCP-mediated degradation of PDCD4promotes protein translation and cell growth. Science 314(5798):467–471.

18. Hock AK, Vousden KH (2014) The role of ubiquitin modification in the regulation ofp53. Biochim Biophys Acta 1843(1):137–149.

19. Li M, et al. (2003) Mono- versus polyubiquitination: Differential control of p53 fate byMdm2. Science 302(5652):1972–1975.

20. Emanuele MJ, et al. (2011) Global identification of modular cullin-RING ligase sub-strates. Cell 147(2):459–474.

21. Ganoth D, et al. (2001) The cell-cycle regulatory protein Cks1 is required forSCF(Skp2)-mediated ubiquitinylation of p27. Nat Cell Biol 3(3):321–324.

22. Bashir T, Dorrello NV, Amador V, Guardavaccaro D, Pagano M (2004) Control ofthe SCF(Skp2-Cks1) ubiquitin ligase by the APC/C(Cdh1) ubiquitin ligase. Nature428(6979):190–193.

23. Yoshida Y, Murakami A, Tanaka K (2011) Skp1 stabilizes the conformation of F-boxproteins. Biochem Biophys Res Commun 410(1):24–28.

24. Ling J, Söll D (2010) Severe oxidative stress induces protein mistranslation throughimpairment of an aminoacyl-tRNA synthetase editing site. Proc Natl Acad Sci USA107(9):4028–4033.

25. Båvner A, Johansson L, Toresson G, Gustafsson JA, Treuter E (2002) A transcriptionalinhibitor targeted by the atypical orphan nuclear receptor SHP. EMBO Rep 3(5):478–484.

26. Miyake S, et al. (2000) Cells degrade a novel inhibitor of differentiation with E1A-likeproperties upon exiting the cell cycle. Mol Cell Biol 20(23):8889–8902.

27. Ye B, et al. (2014) Pcid2 inactivates developmental genes in human and mouse em-bryonic stem cells to sustain their pluripotency by modulation of EID1 stability. StemCells 32(3):623–635.

28. Uchida C, et al. (2005) Enhanced Mdm2 activity inhibits pRB function via ubiquitin-dependent degradation. EMBO J 24(1):160–169.

29. Sdek P, et al. (2005) MDM2 promotes proteasome-dependent ubiquitin-independentdegradation of retinoblastoma protein. Mol Cell 20(5):699–708.

30. Hassler M, et al. (2007) Crystal structure of the retinoblastoma protein N domainprovides insight into tumor suppression, ligand interaction, and holoprotein archi-tecture. Mol Cell 28(3):371–385.

31. Kulak NA, Pichler G, Paron I, Nagaraj N, Mann M (2014) Minimal, encapsulated pro-teomic-sample processing applied to copy-number estimation in eukaryotic cells. NatMethods 11(3):319–324.

32. Lee JW, et al. (2006) Editing-defective tRNA synthetase causes protein misfolding andneurodegeneration. Nature 443(7107):50–55.

33. Jin J, Arias EE, Chen J, Harper JW, Walter JC (2006) A family of diverse Cul4-Ddb1-interacting proteins includes Cdt2, which is required for S phase destruction of thereplication factor Cdt1. Mol Cell 23(5):709–721.

34. Shirane M, et al. (1999) Down-regulation of p27(Kip1) by two mechanisms, ubiquitin-mediated degradation and proteolytic processing. J Biol Chem 274(20):13886–13893.

35. Mark KG, Simonetta M, Maiolica A, Seller CA, Toczyski DP (2014) Ubiquitin ligasetrapping identifies an SCF(Saf1) pathway targeting unprocessed vacuolar/lysosomalproteins. Mol Cell 53(1):148–161.

Yoshida et al. PNAS | April 14, 2015 | vol. 112 | no. 15 | 4635

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

May

14,

202

1