Disruption of FAT10–MAD2 binding inhibitstumor progressionSteven Setiawan Thenga,b,1, Wei Wangc,1, Way-Champ Mahb,d, Cheryl Chanb,d, Jingli Zhuod, Yun Gaod, Haina Qinc,Liangzhong Limc, Samuel S. Chonge, Jianxing Songb,c,2, and Caroline G. Leea,b,d,f,2

aDepartment of Biochemistry, Yong Loo Lin School of Medicine, National University of Singapore, Singapore 119077, Singapore; bNational Universityof Singapore Graduate School of Integrative Sciences and Engineering, National University of Singapore, Singapore 119077, Singapore; cDepartment ofBiological Sciences, Faculty of Science, National University of Singapore, Singapore 119077, Singapore; dDivision of Medical Sciences, Humphrey Oei Instituteof Cancer Research, National Cancer Centre Singapore, Singapore 169610, Singapore; eDepartment of Pediatrics, Yong Loo Lin School of Medicine, NationalUniversity of Singapore, Singapore 119077, Singapore; and fDuke-National University of Singapore Graduate Medical School, Singapore 169547, Singapore

Edited* by Sherman M. Weissman, Yale University School of Medicine, New Haven, CT, and approved November 4, 2014 (received for review February26, 2014)

FAT10 (HLA-F-adjacent transcript 10) is a ubiquitin-like modifier thatis commonly overexpressed in various tumors. It was found to playa role in mitotic regulation through its interaction with mitotic arrest-deficient 2 (MAD2). Overexpression of FAT10 promotes tumor growthand malignancy. Here, we identified the MAD2-binding interface ofFAT10 to be located on its first ubiquitin-like domain whose NMRstructure thus was determined. We further proceeded to demon-strate that disruption of the FAT10–MAD2 interaction throughmutation of specific MAD2-binding residues did not interfere withthe interaction of FAT10 with its other known interacting partners.Significantly, ablation of the FAT10–MAD2 interaction dra-matically limited the promalignant capacity of FAT10, includingpromoting tumor growth in vivo and inducing aneuploidy, prolif-eration, migration, invasion, and resistance to apoptosis in vitro.Our results strongly suggest that the interaction of FAT10 withMAD2 is a key mechanism underlying the promalignant propertyof FAT10 and offer prospects for the development of anticancerstrategies.

chromosomal instability | ubiquitin | aneuploidy | cancer progression |FAT10

FAT10 (HLA-F-adjacent transcript 10) is a ubiquitin-likemodifier protein that functions as a proteasomal degradation

signal (1–3). Recent studies, however, have suggested thatFAT10’s functions extend beyond protein degradation. FAT10 isexpressed mainly in tissues of the immune system, including thespleen and thymus (4, 5). In immune cells, FAT10 is stronglyinduced by proinflammatory stimuli and facilitates T-cell acti-vation by enhancing antigen presentation of mature dendriticcells (6). FAT10 also is induced by proinflammatory cytokines invarious tissues outside the immune system including the liver andcolon (7, 8), although the physiological functions of this responseremain unknown. What is clear, however, is that constitutiveinduction of FAT10 has deleterious consequences in promotingcellular malignancy. Our group recently has reported that ec-topic expression of FAT10 induced malignant transformation innontumorigenic cells and tumor promotion in tumorigenic cells(9), implicating FAT10 in facilitating tumor growth and pro-gression. This finding is consistent with multiple reports thatfound FAT10 to be up-regulated in several tumor types includingtumors of the liver and colon (5, 8, 10, 11).To date, the mechanism underlying FAT10’s promalignant

characteristic remains unclear. One compelling albeit indirectpiece of evidence stems from the finding that FAT10 interactswith the spindle checkpoint protein mitotic arrest-deficient 2(MAD2) during mitosis and reduces MAD2 localization to thekinetochores, resulting in aneuploidy (7, 12), a phenomenon closelyassociated with tumorigenesis and a hallmark of many solid tumors(13). Given the strong association between aneuploidy, chromo-somal instability, and cancer development (reviewed in ref. 14), wehypothesized that FAT10 induces malignant progression through

its interaction with MAD2. Therefore, the aims of this study wereto elucidate the structure of the MAD2-binding interface of FAT10and subsequently to examine the effect of disrupted FAT10–MAD2interaction on FAT10-induced tumor progression.Because of its poor solubility, no information is available thus

far on the structure of FAT10. In this study, using heteronuclearmultidimensional NMR spectroscopy, we identified the MAD2-binding interface of FAT10 located on its first ubiquitin-likedomain and further determined the solution structure of this firstdomain. Moreover, we demonstrated that this interface is spe-cific for the interaction of FAT10 with MAD2 and that mutationof the MAD2-binding residues of FAT10 did not interfere withthe interaction of FAT10 with its other known key interactionpartners including ubiquitin-activating E1 enzyme (UBA6) (15),ubiquitin-like protein NEDD8 μLtimate buster-1 long (NUB1L)(2), histone deacetylase 6 (HDAC6) (16), autophagosome re-ceptor p62 (17), and tumor suppressor p53 (18). Significantly,abrogation of the FAT10–MAD2 interaction by mutating thecore MAD2-binding residues on FAT10 limited FAT10-induced

Significance

FAT10, a ubiquitin-like modifier, is an oncogene that interactswith mitotic arrest-deficient 2 (MAD2) and confers cellular ma-lignancy. Here we identified theMAD2-binding residues of FAT10and determined the first solution structure, to our knowledge, ofthe first FAT10 ubiquitin-like domain. Importantly, we demon-strated the proof-of-mechanism for a novel and specific drug-targeting strategy that entails the specific inhibition of thepathological activity of a therapeutic target but not its repor-ted physiological function, thus minimizing undesirable sideeffects: Abrogation of the FAT10–MAD2 interaction curtailedtumor progression without affecting FAT10’s interaction withits other known physiological binding partners. This studypresents a paradigm for drug targeting and paves the way forthe development of a novel small-molecule anticancer inhibitortargeting the MAD2-binding interface of FAT10.

Author contributions: S.S.T., J.S., and C.G.L. designed research; S.S.T., W.W., W.-C.M., Y.G.,H.Q., and L.L. performed research; J.Z. and S.S.C. contributed new reagents/analytic tools;S.S.T., W.-C.M., J.S., and C.G.L. analyzed data; C.G.L. supervised students and staff on theproject; and C.C., J.S., and C.G.L. wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.

Data deposition: NMR, atomic coordinates, chemical shifts, and restraints reported in thispaper have been deposited in the Protein Data Bank (PDB), www.pdb.org (PDB ID code2MBE), and the Gene Expression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo(accession no. GSE54167).1S.S.T. and W.W. contributed equally to this work.2To whom correspondence may be addressed. Email: dbssjx@nus.edu.sg or caroline_lee@nuhs.edu.sg.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1403383111/-/DCSupplemental.

E5282–E5291 | PNAS | Published online November 24, 2014 www.pnas.org/cgi/doi/10.1073/pnas.1403383111

Dow

nloa

ded

by g

uest

on

June

11,

202

0

tumor growth, as evidenced by the reduced tumor sizes in micexenografts. Additionally, disruption of the FAT10–MAD2 in-teraction mitigated the promalignant capacity of FAT10 in vitro,including its ability to modulate aneuploidy, cellular proliferation,migration, invasion, and resistance to apoptosis. We furtherfound that the interaction of FAT10 with MAD2 was essential forinducing widespread gene dysregulation that may promote cel-lular malignancy.Taken together, our results unequivocally implicate the in-

teraction of FAT10 with MAD2 as a key mechanism underlyingthe promalignant property of FAT10. Importantly, we have elu-cidated the MAD2-binding interface of FAT10 and have dem-onstrated that it is specific for binding to MAD2, opening upprospects for the development of specific anticancer strategiestargeting the promalignant function of FAT10 by inhibiting thepathological FAT10–MAD2 interaction.

ResultsStructural Characterization of FAT10. FAT10 is a 165-aa multi-functional protein comprising two ubiquitin-like domains (Fig. 1A and B). However, the structure of the FAT10 protein remainselusive, largely because of its insoluble nature. Here, by exten-sively optimizing the buffer conditions, we succeeded in obtain-ing the full-length FAT10 sample at a protein concentration of20 μM and collected its circular dichroism (CD) and NMR 1H-15Nheteronuclear single-quantum correlation (HSQC) spectra (rep-resented by black in Fig. 1 C and D). The full-length FAT10protein contains well-formed secondary structures, as is evidentfrom its far-UV CD spectrum comprising two negative signals at208 and 222 nm as well as a positive signal at 195 nm (Fig. 1C).FAT10’s well-folded structure was evidenced further by its HSQCspectrum with large spectral dispersions at both 1H (∼3.1 ppm)and 15N (∼26 ppm) dimensions (Fig. 1D). However, because the

full-length FAT10 protein tended to aggregate, even at 20 μM,we dissected it into its two ubiquitin-like domains: amino acidresidues 1–82 (domain 1) and 83–165 (domain 2). Althoughdomain 2 was highly soluble and stable, domain 1 was poorlysoluble even at 50-μM concentrations. Hence, we further gen-erated several constructs of domain 1 with sequential deletion ofthe N-terminal residues and found that the construct with sevenN-terminal residues deleted exhibited reasonable solubility andstability for further determination of the NMR structure. Thetwo isolated domains contained well-formed secondary struc-tures and well-folded 3D structures as evident from their far-UVCD spectra (represented by red and blue in Fig. 1C) and HSQCspectra (represented by red and blue in Fig. 1D), respectively.Strikingly, the HSQC peaks of the two isolated domains werealmost superimposable on the corresponding peaks of the full-length FAT10 (Fig. 1D). This finding strongly implies that thetwo ubiquitin-like domains are linked by a flexible loop and that,in the context of the full-length FAT10 protein, there might notbe significant packing between the two domains.

NMR Structures and Identification of the MAD2-Binding Surface ofFAT10. Because FAT10 domain 1 started to unfold at pH below7.2 and showed a detectable coexistence of the folded and un-folded states at pH 7.0, we subsequently conducted all bio-physical studies at pH 7.4. By analyzing 15N-edited HSQC-totalcorrelation spectroscopy (TOCSY) and HSQC-NOESY, we ach-ieved both backbone and side-chain assignments of most non-proline residues except for 10 residues within the long loop fromPro59 to Thr73 for which HSQC peaks were not detected at pH7.4. Next, the 3D structure was determined for FAT10 domain 1.Briefly, NOE-derived distance and TALOS-based dihedralangle restraints (19) were used in the CYANA software pack-age (20) to calculate 50 NMR structures. The eight structures

Fig. 1. Structural characterization of FAT10 and its two isolated domains. (A) Sequence alignment of two ubiquitin-like domains of FAT10 with ubiquitin.FAT10 residues that are identical or homologous to ubiquitin are shown in red and blue, respectively. (B) Domain organization of the 165-residue FAT10containing two ubiquitin-like domains designated FAT10 domain 1 and FAT10 domain 2, respectively. The boundaries of the two domains are based on the3D structure we determined. (C) Far-UV CD spectra of the full-length FAT10 (black trace), domain 1 (red trace), and domain 2 (blue trace) at 20-μM proteinconcentration. (D) Superimposition of 2D 1H-15N HSQC spectra of full-length FAT10 (black), domain 1 (red), and domain 2 (blue). (E) Superimposition of theeight selected backbone structures of domain 1 at the secondary structure regions. N refers to the amino-terminus and C refers to the carboxyl-terminus ofFAT10 domain 1. (F) Three-dimensional structure of FAT10 domain 1 in ribbon format comprising a typical ubiquitin-like fold composed of five β-strands andone α-helix. (G and H) The electrostatic potential surface of FAT10 domain 1 oriented as in F (G) and with the orientation rotated 180° along the z axis (H).Positively charged, negatively charged, and nonpolar/neutral residues are represented by blue, red, and gray respectively. The main MAD2-binding regionsin F and G are outlined in cyan.

Theng et al. PNAS | Published online November 24, 2014 | E5283

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

June

11,

202

0

with the lowest target functions were selected for furtherrefinement with AMBER force field (21). The eight super-imposed NMR structures of FAT10 domain 1 are shown in Fig.1E, and the calculation statistics and structure quality are sum-marized in Table S1. The eight NMR structures consist of well-defined secondary structures comprising five β-strands and oneα-helix, with rms deviations over the secondary structure regionsof 0.45 Å for backbone atoms and 1.15 Å for all atoms ascompared with the mean structure. The loop regions were lessdefined, with distinctive conformations in different structures. In

particular, the loop residues Arg60 and Glu61 were highly vari-able because of the lack of NOE connectivities, and their cor-responding HSQC peaks were very broad, implying possibleconformational changes (22). As a consequence, rms deviationsof the whole FAT10 domain 1 became 1.08 Å for backbone atomsand 2.24 Å for all atoms. Furthermore, no slowly exchanged amideprotons could be identified by the NMR H–D exchange experi-ments at pH 7.4, suggesting that FAT10 domain 1 also under-goes a global folding–unfolding exchange that usually occurson the microsecond-to-second time scale. The presence of the

Fig. 2. NMR identification and mutation of MAD2-binding residues of FAT10. (A) Superimposition of HSQC spectra of 15N-labeled full-length FAT10 in theabsence (black) and presence (red) of unlabeled MAD2 at a FAT10:MAD2 molar ratio of 1:1.5 and FAT10 domain 2 (blue). (B) Superimposition of HSQC spectraof 15N-labeled FAT10 domain 1 in the absence (blue) and presence (red) of the unlabeled MAD2 at a molar ratio of 1:1. Corresponding residues of HSQC peaksthat disappeared or displayed significantly reduced intensity are labeled. (C and D) Three-dimensional structure of FAT10 domain 1 in ribbon (C) and surface(D) format. Residues with significantly altered HSQC intensities at the FAT10 domain 1:MAD2 molar ratios of 1:0.5 and 1:1 are shown in red and brown,respectively. (E) Site-directed mutagenesis to introduce various mutations into human FAT10 cDNA. (Top) Schematic showing mutations at MAD2-bindingregion 1 (M1), region 2 (M2), or at both regions (M12). The specific mutations are shown by alignment of primary amino acid sequences (Middle), sche-matically (Bottom Left), and in a table (Bottom Right). (F, Upper) ITC profiles of the binding reaction of FAT10 domain 1 with MAD2 and (Lower) integratedvalues for reaction heats with subtraction of the corresponding blank results normalized against the amount of ligand injected versus the molar ratio ofFAT10 domain 1:MAD2. The thermodynamic binding parameters obtained from data fitting are shown. (G, Left) Wild-type and the various FAT10 mutant celllines were generated following cloning of the respective FAT10 cDNA into the pcDNA3.1 vector and subsequent transfection into HCT116 cells. (Right)Western blot analysis confirmed comparable expression of wild-type and mutant FAT10 in HCT116 cells. FAT10 mutants containing mutated di-glycineresidues (ΔGly: G164A, and G165A) were generated also.

E5284 | www.pnas.org/cgi/doi/10.1073/pnas.1403383111 Theng et al.

Dow

nloa

ded

by g

uest

on

June

11,

202

0

conformational exchanges on the microsecond-to-second timescale thus rationalizes the relatively low number of long-rangeNOEs, which contributes to the relatively high rms deviations forFAT10 domain 1. Although FAT10 domain 1 possesses thecharacteristic ubiquitin-like fold comprising five β-strands and oneα-helix, the electrostatic potential surface of this domain differsfrom that of ubiquitin (Fig. S1) (23). For FAT10 domain 1, thesurface constituted by some of the residues from β-strands 1, 3,and 5 is highly positive (Fig. 1G), whereas the opposite side islargely negative (Fig. 1H).We also conducted extensive investigations on the binding

interaction between FAT10 and MAD2. FAT10 proteins in-cluding FAT10 (aa 1–165), FAT10 (aa 8–165), FAT10 domain 1(aa 1–82), FAT10 domain 1 (aa 8–82), and FAT10 domain 2(aa 82–165) were subjected to NMR HSQC titration by addingMAD2. The HSQC titration allowed the detection of weakbinding events and mapping of the binding interface (22, 24, 25).Titration of the 15N-labeled full-length FAT10 sample by thegradual addition of unlabeled MAD2 triggered a significant re-duction in intensity or even the disappearance of a portion ofFAT10 HSQC peaks (Fig. 2A). Upon closer inspection, we foundthat these peaks were primarily from domain 1 (Fig. 2B), sug-gesting that FAT10 domain 1 is responsible for binding toMAD2. Furthermore, HSQC peaks of residues 1–7 were notperturbed significantly by the addition of MAD2. To confirm ourfinding, 15N-labeled samples of the two isolated domains ofFAT10 were titrated with MAD2. As expected, the gradual ad-dition of MAD2 induced significant reduction in intensity oreven the disappearance of HSQC peaks of domain 1, but nochemical shift or intensity change was detected in HSQC peaksof domain 2 even at a FAT10 domain 2:MAD2 molar ratio of 1:4,despite FAT10 domain 2 assuming the same ubiquitin-like fold(Fig. S2). These data strongly suggest that only domain 1 ofFAT10 interacts with MAD2. Subsequently, we mapped theMAD2-binding interface of FAT10 domain 1 by examining resi-dues with HSQC peaks that either disappeared or significantlybroadened at FAT10 domain 1:MAD2 molar ratios of 1:0.5 and1:1 (represented by red and brown, respectively, in Fig. 2 C andD). Interestingly, the MAD2-binding surface, which is positivelycharged, is composed mainly of β-strands 1, 2, 3, and 5 with thecore residues located on the β1 (Val9–Arg12) and β5 (Ile74–Lys79) strands (Fig. 2 C and D). Of the core residues, only the sidechains of His11 and Arg13 on β1 and His75, Thr77, and Lys79 onβ5 were exposed to the solvent; the other core residues wereburied within the protein. Because buried hydrophobic residuesusually are involved in folding and stabilization, and exposedresidues often play a central role in binding, we focused ourinvestigations on the five exposed residues, namely His11, Arg13,His75, Thr77, and Lys79, and their possible contributions to theFAT10–MAD2 interaction.We therefore generated various FAT10 domain 1 mutants:

M1 with H11D and R13Q mutations in region 1; M2 with H75D,T77D, and K79Q mutations in region 2; and M12 with all theaforementioned mutations (Fig. 2E). The respective amino acidsubstitutions were chosen to disrupt potential amino acidcharge-dependent interactions while retaining their molecularweight. Importantly, the amino acid substitutions did not resultin significant denaturation or misfolding of the respective mu-tant proteins, as indicated by their CD spectra (Fig. S3A). Toquantify the binding event, the thermodynamic parameters ofthe binding between MAD2 and wild-type FAT10 domain 1 aswell as its mutants were measured by isothermal titration cal-orimetry (ITC). The binding affinity obtained for the wild-typeFAT10 domain 1 (dissociation constant, Kd, of 2.97 μM) (Fig. 2F)was consistent with the observed disappearance of HSQC peaks ofthe FAT10 domain 1 upon binding to MAD2 (22, 25). On theother hand, M1 showed significantly reduced binding affinity(Fig. S3B) and thus could not be fitted to obtain the binding

parameters, and M12 showed almost no binding to MAD2 (Fig.S3C). The M2–MAD2 interaction could not be determined be-cause of its low solubility.

Mutation of MAD2-Binding Residues on FAT10 Does Not AffectFAT10’s Binding to Other Interaction Partners. We next examinedthe involvement of the aforementioned FAT10 residues in theinteraction of FAT10 with MAD2. Mutations of the putativeMAD2-binding regions of FAT10 (Fig. 2E) were introduced intofull-length human FAT10 cDNA using site-directed mutagenesisand were cloned into a pcDNA3.1 vector (Invitrogen) for stableexpression of the wild-type and mutant FAT10 proteins in HCT116parental cells (Fig. 2G). Importantly, the expression levels of wild-type and mutant FAT10 proteins used in the subsequent experi-ments of this study were within the range of FAT10 expressiondetected in hepatocellular carcinoma (HCC) tumors (Fig. S4).To identify the FAT10 residues that are critical for its in-

teraction with MAD2, we examined the interaction of wild-typeand mutant FAT10 with endogenous MAD2 using coimmuno-precipitation (co-IP) and in situ proximity ligation assay (PLA),a sensitive technique that detects sites of protein–protein in-teraction through the visualization of red fluorescent signals.Strikingly, both methods of detecting protein–protein interactionshowed abrogation of the FAT10–MAD2 interaction only uponmutation of both MAD2-binding regions (i.e., mutant M12) (Fig.3 A and B, Upper Left and Fig. S5). FAT10 mutated at eitherbinding region (M1 and M2) retained the ability to interact withMAD2, albeit to a lesser extent for M1 (Fig. 3 A and B, UpperLeft). Furthermore, mutation of both MAD2-binding regions didnot affect the binding of FAT10 to its known substrates namelyUBA6, NUB1L, HDAC6, and p62 (Fig. 3 A and B), suggestingthat the MAD2-binding interface of FAT10 is specific to MAD2.Separately, mutation of the C terminus di-glycine motif of FAT10abolished binding to UBA6 and p62 (Fig. 3 A and B, UpperMiddle and Upper Right), consistent with reports that FAT10binds to these proteins through its di-glycine residues (15, 17). Anirrelevant antibody, anti-p16, was used as a negative control in theinteraction experiments. We further ascertained that FAT10 bindsto free MAD2 and not to MAD2 bound to the spindle checkpointcomplex, because members of the MAD2 complex, MAD1 andBub1 (budding uninhibited by benzimidazoles 1), could not bedetected following immunoprecipitation using FAT10-specificantibodies (Fig. 3C and Fig. S5). Because Li et al. (18) previouslyfound that FAT10 overexpression increased the population oftranscriptionally active p53, we next tested if mutation of theMAD2-binding regions of FAT10 interfered with this function. Tothis end, transcriptionally inactive p53 levels were determinedusing PAB240 antibodies following immunoprecipitation of full-length p53 (FL-393 antibody) in wild-type FAT10, M12, andwild-type parental cells. Consistently, a significant reduction intranscriptionally inactive p53 levels was detected in wild-typeFAT10 cells as compared with parental cells, although total p53levels were similar in both cell lines (Fig. 3D). A similar re-duction in transcriptionally inactive p53 levels also was observedin M12 cells (Fig. 3D), suggesting that the abrogation of theFAT10–MAD2 interaction did not interfere with the ability ofFAT10 to increase the population of transcriptionally active p53.Furthermore, mutation of the MAD2-binding residues of FAT10did not interfere with the interaction of FAT10 with p53 (Fig.3D), supporting the specificity of the MAD2-binding interface ofFAT10 for MAD2.

Abrogation of the FAT10–MAD2 Interaction Ameliorates FAT10-InducedChromosomal Instability and Mitotic Checkpoint Dysregulation. FAT10overexpression has been shown to promote chromosomal in-stability (7, 12), tumor initiation, and malignant features in bothnontumorigenic and tumorigenic cells (9). To examine the ef-fect of disrupted FAT10–MAD2 interaction on FAT10-induced

Theng et al. PNAS | Published online November 24, 2014 | E5285

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

June

11,

202

0

chromosomal instability, chromosomes from parental wild-typeHCT116 cells and from wild-type FAT10 (FAT-A and FAT-B),M1, M2, and M12 (M12-A and M12-B) cells were karyotyped.Consistent with our previous findings (7, 12), overexpression ofwild-type FAT10 showed a markedly higher proportion of aneu-ploid cells compared with parental cells (Fig. 4A). Specifically, themajority (>70%) of wild-type FAT10-expressing cells displayedmore than the usually observed number of chromosomes (80–89),whereas the majority (>80%) of the parental cells retained themodal number of chromosomes (40–49) (Fig. 4A) (12). Further,a majority (>68%) of MAD2-deficient cells (Fig. S6A) similarlyshowed more than the usually observed number of chromosomes(80–89) (Fig. S6B). Although similarly high proportions of aneu-ploid cells were observed in M1 and M2 cells and in wild-typeFAT10-expressing cells, the majority (>79%) of M12 cells de-ficient for FAT10–MAD2 interaction retained modal chromo-some numbers (Fig. 4A), suggesting that the FAT10–MAD2interaction may be critical for FAT10-induced aberrations inchromosome number.A closer examination of the morphology of the metaphase

chromosomes in the various cell lines revealed that wild-typeFAT10-expressing cells contained incompletely condensed, long“thread-like” chromosomes as opposed to the condensed, short“ribbon-like” chromosomes in parental cells (Fig. 4B). Strikingly,the morphology of the metaphase chromosomes in M1 andM2 cells resembled that of wild-type FAT10-expressing cells,whereas the morphology of metaphase chromosomes in M12cells resembled that of parental cells (Fig. 4B). Notably, MAD2depletion has been reported to alter chromosome morphologysimilarly (26). Taken together, these data suggest that the in-

teraction of FAT10 with MAD2 contributes to FAT10-inducedabnormalities in chromosome structure and number.Next, because wild-type FAT10 overexpression was reported

to facilitate escape from nocodazole-induced mitotic arrest (12),we examined if this escape was mediated by the interaction ofFAT10 with MAD2. Mitotic cell populations were obtained bysynchronization at G1/S phase followed by nocodazole-inducedarrest for 8 h. Cells then were assessed by FACS analysis fol-lowing staining with the mitosis marker mitotic protein mono-clonal 2 (MPM-2) (27). Consistent with our earlier observations,a significantly reduced proportion of mitotic cells was observedin wild-type FAT10-expressing cells (7.2%) compared with pa-rental cells (23.7%) following mitotic arrest (Fig. 4C). Similarly,depletion of MAD2 resulted in a significantly reduced pro-portion of mitotic cells (Fig. S6C), phenocopying the reductionseen with FAT10 overexpression. Notably, abrogation of theFAT10–MAD2 interaction (i.e., in the M12 mutant) limited thecell’s ability to escape mitotic arrest to levels comparable to thatof parental cells (Fig. 4C). Mutation of the di-glycine motif ofFAT10 (ΔGly) moderately restored the mitotic cell population,albeit to a lesser extent (Fig. 4C). These data support a role forthe FAT10–MAD2 interaction in disrupting the mitotic check-point and in inducing aneuploidy.

Abrogation of the FAT10–MAD2 Interaction Attenuates FAT10-InducedMalignancy in Vitro. We next investigated the role of the FAT10–MAD2 interaction on FAT10-induced malignancy because FAT10overexpression was previously found to enhance the cell’sanchorage-independent growth, resistance to cell death, andits invasion and migration abilities (9).

Fig. 3. Disruption of the FAT10–MAD2 interaction does not affect the binding of FAT10 with its known interaction partners and its function in protein deg-radation. (A and B) Mutation of both MAD2-binding regions of FAT10 (M12) abrogates the FAT10–MAD2 interaction but not FAT10’s interaction with otherknown interaction partners including UBA6, NUB1L, HDAC6, and p62 as shown by co-IP (A, Right) and PLA (B). Antibody against p16 was included as a negativecontrol. Loading controls for co-IP (input lysate) are shown (A, Left). (B) Right columns show red fluorescence signals representing sites of protein–protein in-teraction within cells. Center columns show nuclear staining with DAPI. Right columns show overlay of PLA and DAPI signals. Single antibody (anti-FAT10 or anti-MAD2) and no primary antibody were included as negative controls (B, Lower Right). (C) FAT10 binds to free MAD2. Members of the MAD2 mitotic spindlecheckpoint complex MAD1 and Bub1 were probed on FAT10-immunoprecipitated lysates obtained in A. (D) Mutation of both MAD2-binding regions of FAT10(M12) does not affect the modulation of p53 by FAT10. Inactive p53 was probed using p53 mutant-specific antibody (PAB-240) in immunoblotting (IB) followingimmunoprecipitation (IP) of full-length p53 (FL-393) of HCT116 cell lysates transiently overexpressing wild-type or mutant FAT10 (M12) for 48 h. To probe for totalp53 interacting with FAT10, p53 (FL393) was used for immunoblotting following immunoprecipitation with FAT10-specific antibodies.

E5286 | www.pnas.org/cgi/doi/10.1073/pnas.1403383111 Theng et al.

Dow

nloa

ded

by g

uest

on

June

11,

202

0

Examination of the cellular proliferation profiles of wild-typeparental, wild-type FAT10-expressing (FAT-A and FAT-B),FAT10 knock-down (FATi), M1, M2, and M12 (M12-A andM12-B) cells revealed enhanced proliferation of cells thatretained the FAT10–MAD2 interaction (FAT-A, FAT-B, M1,and M2) compared with parental cells (Fig. 5A). However, dis-ruption of the FAT10–MAD2 interaction (M12) significantlyattenuated cellular proliferation to levels observed in FATi cells,which were notably lower than those of parental cells (Fig. 5A).Next, investigation of the anchorage-independent growth pro-files of the various cell lines showed that wild-type FAT10overexpression enhanced anchorage-independent growth of cellsrelative to parental cells (Fig. 5B). Mutation of either MAD2-binding region of FAT10 (M1 and M2) reduced colony forma-tion on soft agar to that observed in parental cells, whereas ab-rogation of the FAT10–MAD2 interaction (M12) almost obliteratedanchorage-independent growth of the cells (Fig. 5B).Examination of the cell’s ability to resist camptothecin-induced

cell death revealed that ablation of the FAT10–MAD2 interaction(M12) increased susceptibility to cell death, whereas cells harboring

mutation of either MAD2-binding region of FAT10 (M1 and M2)retained an ability to resist cell death similar to that of wild-typeFAT10-expressing cells (Fig. 5C). Additionally, abrogation of theFAT10–MAD2 interaction (M12) attenuated FAT10-induced cel-lular migration and invasion to levels comparable to or below thoseof parental cells (Fig. 5 D and E, respectively). Notably, mutation ofeither MAD2-binding region of FAT10, particularly M1, only par-tially impaired the invasive and migratory ability of the cells. Fur-ther, analysis of a protein that degrades extracellular matrix andis associated with colon cancer progression (28), matrix metal-loproteinase 9 (MMP-9), showed that wild-type FAT10 inducedcells to secrete significantly higher amounts of MMP-9 com-pared with parental cells (Fig. 5F). Abolishment of the FAT10–MAD2 interaction inhibited FAT10-induced MMP-9 secretionto levels observed in parental cells, whereas mutation of eitherMAD2-binding region of FAT10 only moderately attenuated theincreased MMP-9 secretion (Fig. 5F).Collectively, disruption of the interaction of FAT10 with

MAD2 curtailed various FAT10-induced malignant charac-teristics in vitro, and these findings were observed consistently

Fig. 4. Disruption of the FAT10–MAD2 interaction inhibits FAT10-induced chromosomal instability. (A) Tabulation of the karyotype analysis of wild-typeparental cells, FAT10-expressing cells (FAT-A, FAT-B), and the various stable FAT10 mutant cells (M1, M2, M12-A, M12-B). (B) Three representative metaphasespreads of parental wild-type, stable wild-type, and mutant FAT10-expressing cells. (C) Schematic (Upper) and graphical (Lower) representation of mitoticprofiles of parental wild-type, stable wild-type, and mutant FAT10-expressing cells. Cells were synchronized at G1/S phase by single thymidine block followedby treatment with the M-phase inhibitor nocodazole for 8 h. Cells were stained with the mitosis marker MPM-2 (27) before FACS analysis. All data shown aremean ± SE. *P < 0.05 compared with parental wild type; #P < 0.05 compared with FAT10-expressing cells.

Theng et al. PNAS | Published online November 24, 2014 | E5287

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

June

11,

202

0

in transiently-expressing FAT10 SNU449-transformed liver cells(Fig. S7) suggesting a role for the FAT10–MAD2 interaction intumor progression.

Abrogation of the FAT10–MAD2 Interaction Attenuates FAT10-InducedTumor Formation in Vivo. We previously had demonstrated thatFAT10 overexpression promotes tumor formation in nude mice,recapitulating the malignant phenotype observed in vitro (9). Toinvestigate if the tumor-promoting ability of FAT10 is mediatedthrough its interaction with MAD2, parental and various FAT10-expressing HCT116 stable cells were injected s.c. into oppositeflanks of nude mice. Consistent with our previous report, wild-typeFAT10-expressing cells induced significantly larger tumor forma-tion at the injected sites than did parental cells (Fig. 6). Strikingly,MAD2 binding-deficient FAT10-expressing cells (M12 cells)formed dramatically smaller tumors, even smaller than those arisingfrom parental cells, whereas mutation of either MAD2-bindingregion of FAT10 (M1 and M2) moderately retarded FAT10-induced tumor growth (Fig. 6). These data highlight the tumor-promoting ability of FAT10 and further suggest that its proma-lignant characteristic is mediated by its interaction with MAD2.

Mutation of MAD2-Binding Residues of FAT10 Inhibited FAT10-Induced Global Gene-Expression Changes. To assess the effect ofmutating the MAD2-binding regions of FAT10 on global geneexpression, expression profiling was performed on cells stablyexpressing wild-type FAT10 and on cells expressing the variousFAT10 mutants M1, M2, and M12. All cells were synchronizedusing double thymidine treatment before expression profiling toexclude potential cell cycle-dependent confounding effects ongene expression. Although wild-type FAT10 induced significantchanges in global gene-expression profiles, the gene-expressionprofiles of the MAD2 binding-deficient FAT10 mutant (M12)resembled those of the parental cells (Fig. 7A). Specifically, thenumber of dysregulated genes in M12 cells was markedly reducedcompared with the wild-type counterpart (Fig. 7B). In addition,deregulated gene-expression profiles of the top associated networkfunction deregulated by wild-type FAT10, namely, Cellular Growthand Proliferation, Cellular Development, and Cell Death (TablesS2 and S3) (9), could be restored by abrogating the interactionof FAT10 with MAD2 (Fig. 7C). As evident from Fig. 7D, a ma-jority of the significantly deregulated genes in the top associatednetwork of wild-type FAT10-expressing cells were not significantlyaltered in M12 cells. These data suggest that the interaction

Fig. 5. Abrogated FAT10–MAD2 interaction attenuates FAT10-induced malignancy in vitro. In vitro malignant characteristics of parental wild-type,stable wild-type, and mutant FAT10-expressing cells were examined. (A, Upper) Cell-proliferation profiles were assessed by the colorimetric WST assay inwhich the number of viable cells is measured by the reduction of the tetrazolium compound WST in the colored formazan product. (Lower) The cor-responding cell-doubling time in hours. (B) Anchorage-independent growth profiles on soft agar were assessed. (Upper) Representative pictures ofstained colonies. (Lower) Corresponding graphical representation. (C, Left) Cellular apoptosis profiles following camptothecin-induced cell death wereassessed by FACS detection of activated caspase-3. (Right) Corresponding graphical representation. (D, Left) Cell migration profiles were assessed byscratch-wound assay over 15 h. (Right) Subsequent quantification. (E ) Cellular invasion profiles were assessed by Matrigel invasion assay for 24 h. (Upper)Representative pictures of the invaded cells following staining. (Lower) Quantification. (F ) Cellular MMP-9 secretion profiles were detected by ELISA andare represented graphically. All data are shown as mean ± SE. *P < 0.05 and **P < 0.01 compared with parental wild-type; #P < 0.05 and ##P < 0.01compared with FAT10-expressing cells.

E5288 | www.pnas.org/cgi/doi/10.1073/pnas.1403383111 Theng et al.

Dow

nloa

ded

by g

uest

on

June

11,

202

0

between FAT10 and MAD2 is important for FAT10-inducedderegulation of genes of key pathways associated with tumorprogression such as cellular growth and proliferation, cellulardevelopment, and cell death. Taken together, our data demon-strate that FAT10 promotes tumor progression through its in-teraction with MAD2 and that disruption of the FAT10–MAD2interaction may curtail cellular malignancy.

DiscussionIn recent years, increasing evidence has pointed toward a path-ological role for FAT10 in promoting cellular malignancythrough mechanisms beyond its proteasome-targeting function.Liu et al. (4) first reported the interaction of FAT10 with themitotic spindle assembly checkpoint protein MAD2, and wefurther showed that this interaction induced chromosomal in-stability (7, 12), a hallmark of many solid tumors. Recently,a causal role for FAT10 in tumor initiation and progression wasestablished (9, 29). However, the role of the FAT10–MAD2interaction in promoting malignancy remained unclear, largelybecause the insoluble nature of the FAT10 protein in vitrohampered the identification of the FAT10 protein structure andits interaction interface with MAD2.In this study, we achieved a breakthrough to this conundrum,

and with the FAT10 domain 1 structure presented here we re-

port the successful determination of the first (to our knowledge)solution structures of the FAT10 domains through extensivedissection of the protein. We show through NMR studies thatFAT10 contains two domains that adopt the same ubiquitin-likefold and that they may not possess tight packing between eachother. Importantly, we mapped the MAD2-binding interface ofFAT10 to domain 1 of FAT10 with a Kd of 2.97 μM. The verydifferent electrostatic potential surface of domain 2 may belargely responsible for its inability to bind MAD2 (Fig. S2). Wefurther identified and characterized five core residues within thisinterface that are crucial for the interaction of FAT10 with MAD2.Mutation of these five key residues abolished the FAT10–MAD2interaction both in vitro and in vivo and mitigated the promalig-nant features of FAT10, strongly implicating the interaction ofFAT10 with MAD2 in mediating FAT10-induced malignancy.Importantly, disruption of the FAT10–MAD2 interaction signifi-cantly curtailed FAT10-induced tumor growth in vivo, suggestingthat the role of FAT10 in tumor progression is critically dependenton its interaction with MAD2.In agreement with literature reports implicating FAT10 in

mitotic regulation (7, 12), a recent study revealed FATylation ofmitosis-related cell cycle regulators, including MAD2, duringmitosis (30). Here, we provide further mechanistic insight intoFAT10-induced mitotic dysregulation through its interactionwith the mitotic spindle assembly checkpoint protein MAD2. Wedemonstrated that the capacity of FAT10 to induce numericaland structural chromosomal aberrations is critically dependenton its interaction with MAD2. Perhaps this finding is not sur-prising, given that MAD2 serves as an important checkpointduring mitosis, ensuring proper attachment of chromosomes tothe spindle microtubules during prometaphase. During thisphase, unattached kinetochores prevent metaphase-to-anaphasetransition by inducing the conversion of mitotic checkpointproteins such as MAD2 into diffusible inhibitors of anaphase-promoting complex-cdc20 (APCcdc20) (31). Interestingly, theMAD2-dependent FAT10-induced malignant phenotype strik-ingly resembles the high prevalence of aneuploidy and sponta-neous tumors observed in mice heterozygous for MAD2 andoverexpressing MAD2 (32, 33), implicating a weakened mitoticcheckpoint in promoting cellular malignancy. The significance ofcheckpoint aberration and aneuploidy in promoting neoplastictransformation also has been highlighted by several studies thatused mice with genetically altered levels of checkpoint components(reviewed in ref. 13). Notably, we found that other FAT10-inducedmalignant characteristics such as enhanced proliferation, migra-tion, and invasion and resistance to apoptosis also were dependenton its interaction with MAD2, consistent with the reported pro-malignant phenotype of MAD2-deficient cells (34–36). Thesefindings suggest that MAD2 may have other roles that are in-dependent of its role in mitosis, although the underlying mecha-nisms at present remain unclear. Additionally, our data suggestthat the interaction of FAT10 with MAD2 induces widespreadgene deregulation that may promote cellular malignancy. Althoughit is not immediately clear whether the aberrant gene-expressionpatterns observed were a result of FAT10-induced chromosomalinstability or a consequence of the FAT10–MAD2 interaction, ourdata undoubtedly show that abrogating the FAT10–MAD2 in-teraction protects on cells against global gene deregulation inducedby FAT10.In this study, we have demonstrated that the MAD2-binding

interface of FAT10 is specific to MAD2, thus presenting a pre-viously unidentified and specific strategy for targeting theMAD2-related tumor-promoting activities of FAT10. Mutationof the five core residues of FAT10 that are crucial for theFAT10–MAD2 interaction did not affect the covalent binding ofFAT10 with its E1 enzyme UBA6 or with other known bindingpartners through the FAT10 di-glycine residues, consistentwith the notion that the two FAT10 domains are structurally

Fig. 6. Abrogated FAT10–Mad2 interaction attenuates FAT10-induced tu-mor formation in vivo. Representative pictures of tumors in vivo (Top) andexcised tumors (Middle) following s.c. injection of parental wild-type, stablewild-type (FAT10), and mutant (M1, M2, and M12) FAT10-expressing HCT116cells. (Bottom) Mean tumor weights of six nude mice per group, 3 wk posts.c. injection, are represented graphically. All data are shown as mean ± SE.**P < 0.01 and ***P < 0.001 compared with parental wild-type; ###P < 0.001compared with FAT10-expressing cells.

Theng et al. PNAS | Published online November 24, 2014 | E5289

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

June

11,

202

0

Fig. 7. Abrogated FAT10–MAD2 interaction mitigated FAT10-induced global gene-expression changes. (A) Hierarchical clustering of genes dysregulated in wild-type FAT10-expressing cells (FAT10) compared with parental wild-type cells and mutant FAT10-expressing cells (M1, M2, and M12). Down-regulated genes (relativeto parental wild-type cells) are represented in green; up-regulated genes are represented in red. (B) Table showing the number of dysregulated genes in the in-dicated microarray analysis. (C and D) The top associated network “Cellular Growth and Proliferation, Cellular Development, and Cell Death” of genes differentiallyexpressed in wild-type FAT10 and parental HCT116 cells as identified by IPA. The genes derived from this top associated network and their expression profiles inFAT10,M1, M2, andM12 cells relative to parental wild-type cells are presented graphically (C) and in a table that includes their fold change and False Discovery Rate(FDR)-adjusted P value (D). Green and red represent down- and up-regulation of gene expression, respectively, relative to the parental wild-type cells; white ovalsin C represent genes with no significant change of expression between the indicated comparisons. ***P < 0.001, **P < 0.01, *P < 0.05.

E5290 | www.pnas.org/cgi/doi/10.1073/pnas.1403383111 Theng et al.

Dow

nloa

ded

by g

uest

on

June

11,

202

0

independent. Noncovalent interactions between FAT10 andproteins NUB1L and HDAC6 also were unaffected by thedisruption of the MAD2-binding interface of FAT10. Specifictargeting of the MAD2-related pathological activity of FAT10offers the advantage of minimizing potential undesirable sideeffects for future therapy and is of paramount importance,because our current understanding of the physiological roles ofFAT10 is incomplete. Specific targeting of FAT10 also appearsmore favorable than targeting MAD2 because MAD2 is ubiq-uitously expressed and plays a critical physiological role in cellcycle regulation. In contrast, the frequent selective overexpressionof FAT10 in tumors (5) makes extremely attractive the de-velopment of small-molecule inhibitors targeting the MAD2-binding interface of FAT10 for use as a cytostatic agent inanticancer therapies for patients with premalignant neoplasmsor tumors.In summary, this study has shed light on the mechanism

underlying FAT10-induced promotion of cellular malignancythrough its interaction with MAD2. Importantly, our findingspresent a paradigm for drug targeting as well as the foundationfor the development of novel, small-molecule anticancer drugsthat specifically target the MAD2-related promalignant func-tions of FAT10.

Materials and MethodsDetailed materials and methods are included in SI Materials and Methods.Briefly, HCT116 cell lines were purchased from American Type Culture Col-lection (ATCC). The MAD2-binding interface of FAT10 was obtained by NMRstudies. FAT10 mutants were generated using site-directed mutagenesis.Protein–protein interactions in cells were verified using co-IP and PLA.Chromosome numbers and mitotic index analyses were performed as pre-viously described (12). Cell proliferation was measured using the water-solubletetrazolium salt (WST-1) assay. Anchorage-independent growth was in-vestigated using soft agar colony formation assay. Apoptotic cells weremeasured using FACS analysis following FITC-caspase3 staining. Cell migra-tion was assessed using the scratch-wound assay. Invasion assays were per-formed using the Matrigel Invasion Assay. MMP-9 secretion was measuredusing ELISA. Details of the expression microarray analyses to identify genesaffected by the FAT10–MAD2 interaction are described in SI Materialsand Methods.

ACKNOWLEDGMENTS. This work was supported by National Medical ResearchCouncil (NMRC) Grant NMRC/1306/2011 and block funding from NationalCancer Centre Singapore (to C.G.L.) and by NMRC Grant R154-000-454-213andMinistry of Education of Singapore (MOE) Tier 2 Grant MOE 2011-T2-1-096(to J.S.). W.W. is a recipient of an MOE graduate scholarship under MOEGrant R-154-000-388-112 (to J.S.). S.S.T. and W.-C.M. are recipients ofNational University of Singapore Graduate School for Integrative Sciencesand Engineering Scholarships.

1. Schmidtke G, Kalveram B, Groettrup M (2009) Degradation of FAT10 by the 26Sproteasome is independent of ubiquitylation but relies on NUB1L. FEBS Lett 583(3):591–594.

2. Rani N, Aichem A, Schmidtke G, Kreft SG, Groettrup M (2012) FAT10 and NUB1L bindto the VWA domain of Rpn10 and Rpn1 to enable proteasome-mediated proteolysis.Nat Commun 3:749.

3. Hipp MS, Kalveram B, Raasi S, Groettrup M, Schmidtke G (2005) FAT10, a ubiquitin-independent signal for proteasomal degradation. Mol Cell Biol 25(9):3483–3491.

4. Liu YC, et al. (1999) A MHC-encoded ubiquitin-like protein (FAT10) binds non-covalently to the spindle assembly checkpoint protein MAD2. Proc Natl Acad Sci USA96(8):4313–4318.

5. Lee CG, et al. (2003) Expression of the FAT10 gene is highly upregulated in hepato-cellular carcinoma and other gastrointestinal and gynecological cancers. Oncogene22(17):2592–2603.

6. Ebstein F, Lehmann A, Kloetzel PM (2012) The FAT10- and ubiquitin-dependentdegradation machineries exhibit common and distinct requirements for MHC class Iantigen presentation. Cell Mol Life Sci 69(14):2443–2454.

7. Ren J, Wang Y, Gao Y, Mehta SB, Lee CG (2011) FAT10 mediates the effect of TNF-αin inducing chromosomal instability. J Cell Sci 124(Pt 21):3665–3675.

8. Lukasiak S, et al. (2008) Proinflammatory cytokines cause FAT10 upregulationin cancers of liver and colon. Oncogene 27(46):6068–6074.

9. Gao Y, et al. (2013) FAT10, an Ubiquitin-like Protein, Confers Malignant Properties inNon-tumorigenic and Tumorigenic Cells. Carcinogenesis 35(4):923–34.

10. Qing X, French BA, Oliva J, French SW (2011) Increased expression of FAT10 in colonbenign, premalignant and malignant epithelial neoplasms. Exp Mol Pathol 90(1):51–54.

11. Yan DW, et al. (2010) Ubiquitin D is correlated with colon cancer progression andpredicts recurrence for stage II-III disease after curative surgery. Br J Cancer 103(7):961–969.

12. Ren J, et al. (2006) FAT10 plays a role in the regulation of chromosomal stability. J BiolChem 281(16):11413–11421.

13. Holland AJ, Cleveland DW (2009) Boveri revisited: Chromosomal instability, aneu-ploidy and tumorigenesis. Nat Rev Mol Cell Biol 10(7):478–487.

14. Kops GJ, Weaver BA, Cleveland DW (2005) On the road to cancer: Aneuploidy and themitotic checkpoint. Nat Rev Cancer 5(10):773–785.

15. Gavin JM, et al. (2012) Mechanistic studies on activation of ubiquitin and di-ubiquitin-likeprotein, FAT10, by ubiquitin-like modifier activating enzyme 6, Uba6. J Biol Chem287(19):15512–15522.

16. Kalveram B, Schmidtke G, Groettrup M (2008) The ubiquitin-like modifier FAT10 in-teracts with HDAC6 and localizes to aggresomes under proteasome inhibition. J CellSci 121(Pt 24):4079–4088.

17. Aichem A, et al. (2012) The proteomic analysis of endogenous FAT10 substratesidentifies p62/SQSTM1 as a substrate of FAT10ylation. J Cell Sci 125(Pt 19):4576–4585.

18. Li T, et al. (2011) FAT10 modifies p53 and upregulates its transcriptional activity. ArchBiochem Biophys 509(2):164–169.

19. Cornilescu G, Delaglio F, Bax A (1999) Protein backbone angle restraints fromsearching a database for chemical shift and sequence homology. J Biomol NMR 13(3):289–302.

20. Güntert P (2004) Automated NMR structure calculation with CYANA. Methods MolBiol 278:353–378.

21. Christen M, et al. (2005) The GROMOS software for biomolecular simulation:GROMOS05. J Comput Chem 26(16):1719–1751.

22. Qin H, et al. (2010) Structural characterization of the EphA4-Ephrin-B2 complex re-veals new features enabling Eph-ephrin binding promiscuity. J Biol Chem 285(1):644–654.

23. Komander D, et al. (2009) Molecular discrimination of structurally equivalent Lys 63-linked and linear polyubiquitin chains. EMBO Rep 10(5):466–473.

24. Qin H, Shi J, Noberini R, Pasquale EB, Song J (2008) Crystal structure and NMR bindingreveal that two small molecule antagonists target the high affinity ephrin-bindingchannel of the EphA4 receptor. J Biol Chem 283(43):29473–29484.

25. Williamson MP (2013) Using chemical shift perturbation to characterise ligand bind-ing. Prog Nucl Magn Reson Spectrosc 73:1–16.

26. Michel L, et al. (2004) Complete loss of the tumor suppressor MAD2 causes prematurecyclin B degradation and mitotic failure in human somatic cells. Proc Natl Acad SciUSA 101(13):4459–4464.

27. Tapia C, et al. (2006) Two mitosis-specific antibodies, MPM-2 and phospho-histone H3(Ser28), allow rapid and precise determination of mitotic activity. Am J Surg Pathol30(1):83–89.

28. Wagenaar-Miller RA, Gorden L, Matrisian LM (2004) Matrix metalloproteinases incolorectal cancer: Is it worth talking about? Cancer Metastasis Rev 23(1-2):119–135.

29. Liu L, et al. (2014) As an independent prognostic factor, FAT10 promotes hepatitisB virus-related hepatocellular carcinoma progression via Akt/GSK3beta pathway.Oncogene 33(7):909–920.

30. Merbl Y, Refour P, Patel H, Springer M, Kirschner MW (2013) Profiling of ubiquitin-like modifications reveals features of mitotic control. Cell 152(5):1160–1172.

31. Yu H (2006) Structural activation of Mad2 in the mitotic spindle checkpoint: The two-state Mad2 model versus the Mad2 template model. J Cell Biol 173(2):153–157.

32. Michel LS, et al. (2001) MAD2 haplo-insufficiency causes premature anaphase andchromosome instability in mammalian cells. Nature 409(6818):355–359.

33. Sotillo R, et al. (2007) Mad2 overexpression promotes aneuploidy and tumorigenesisin mice. Cancer Cell 11(1):9–23.

34. Wang L, et al. (2010) Depression of MAD2 inhibits apoptosis and increases pro-liferation and multidrug resistance in gastric cancer cells by regulating the activationof phosphorylated survivin. Tumour biol 31(3):225–232.

35. Schvartzman JM, Duijf PH, Sotillo R, Coker C, Benezra R (2011) Mad2 is a criticalmediator of the chromosome instability observed upon Rb and p53 pathway in-hibition. Cancer Cell 19(6):701–714.

36. Prencipe M, et al. (2009) Cellular senescence induced by aberrant MAD2 levels impactson paclitaxel responsiveness in vitro. Br J Cancer 101(11):1900–1908.

Theng et al. PNAS | Published online November 24, 2014 | E5291

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

June

11,

202

0