aurora a and nf-kb survival pathway drive chemoresistance...

Therapeutics, Targets, and Chemical Biology

Aurora A and NF-kB Survival Pathway DriveChemoresistance in Acute Myeloid Leukemia viathe TRAF-Interacting Protein TIFATong-You Wade Wei1,3, Pei-Yu Wu1, Ting-Jung Wu4, Hsin-An Hou5,Wen-Chien Chou6,Chieh-Lin Jerry Teng7, Chih-Ru Lin1,3, Jo-Mei Maureen Chen8, Ting-Yang Lin1,3,Hsiang-Chun Su1,3, Chia-Chi Flora Huang1, Chang-Tze Ricky Yu8, Shih-Lan Hsu9,Hwei-Fang Tien5, and Ming-Daw Tsai1,2,3

Abstract

Aurora A–dependent NF-kB signaling portends poor prognosisin acute myeloid leukemia (AML) and other cancers, but thefunctional basis underlying this association is unclear. Here, wereport that Aurora A is essential for Thr9 phosphorylation of theTRAF-interacting protein TIFA, triggering activation of the NF-kBsurvival pathway in AML. TIFA protein was overexpressed con-currently with Aurora A and NF-kB signaling factors in patientswith de novo AML relative to healthy individuals and also corre-lated with poor prognosis. Silencing TIFA in AML lines andprimary patient cells decreased leukemic cell growth and

chemoresistance via downregulation of prosurvival factorsBcl-2 and Bcl-XL that support NF-kB–dependent antiapoptoticevents. Inhibiting TIFAperturbed leukemic cytokine secretion andreduced the IC50 of chemotherapeutic drug treatments in AMLcells. Furthermore, in vivo delivery of TIFA-inhibitory fragmentspotentiated the clearance of myeloblasts in the bone marrow ofxenograft-recipient mice via enhanced chemotoxicity. Collective-ly, our results showed that TIFA supports AML progression andthat its targeting can enhance the efficacy of AML treatments.Cancer Res; 77(2); 494–508. �2016 AACR.

IntroductionAcute myeloid leukemia (AML) is a clonal hematologic malig-

nancy with great variability in the clinical features, pathogenesisand treatment outcomes. Thismalignant disorder is caused by theabnormal differentiation of hematopoietic precursor cells forfeit-ing their ability to respond to the regulators of proliferation. Thestandard therapeutic approach for patients with AML is initialchemotherapy induction followed by post-remission treatment,

with additional chemotherapy cycles or allogeneic stem celltransplantation for relapse prevention (1). Although significantprogress has been achieved, current treatments for AMLmay onlyoffer limited survival benefits rather thanprovide fully satisfactoryresponses, presumably due to chemoresistance and diseaserelapse (1, 2). To reduce recurrence rate and promote therapeuticefficacy, it is highly urgent to identify new targets (3, 4).

NF-kB controls various aspects of immune responses andsuperiorly regulates cell survival, proliferation, anddifferentiation(5). Recent evidences attribute a growing number ofmalignanciesto aberrant activation of NF-kB that cross talks with other signal-ing molecules and pathways (6), thus it is considered a risk factorfor poor prognosis in several types of cancer including leukemia(7). In agreement, constitutively activated NF-kB was shown toprotect tumor cells from apoptotic stimuli and promote theirresistance to chemotherapies and ionizing radiation (8), presum-ably through transcriptional activation of antiapoptotic/prosur-vival factors Bcl-2 and Bcl-XL (9). In line with this notion, the roleof NF-kB in leukemogenesis has also been addressed for AML(10), and the inhibition of NF-kB reattains chemosensitivity inthis hematopoietic malignancy presumably due to attenuatedprosurvival responses and activated proapoptotic signals (11).These observations collectively point to the NF-kB signaling axisas a promising therapeutic target (12).

The Aurora family of serine/threonine kinases promotes tumorproliferation through regulation of chromosome alignment, seg-regation, and cytokinesis during mitosis (13), and has beensuggested as anticancer target (14). Upregulation of Aurora A hasbeen reported for bone marrow mononuclear cells in patientswith AML (15), and shown to be associated with unfavorable-riskcytogenetics and higher white blood cell (WBC) counts (16). Inaddition, Aurora A has been shown to promote in vitro and in vivo

1Institute of Biological Chemistry, Academia Sinica, Taipei, Taiwan. 2GenomicsResearch Center, Academia Sinica, Taipei, Taiwan. 3Institute of BiochemicalSciences, National Taiwan University, Taipei, Taiwan. 4Division of Liver andTransplantation Surgery, Chang Gung Memorial Hospital, Taoyuan, Taiwan.5Division of Hematology, Department of Internal Medicine, National TaiwanUniversity Hospital, Taipei, Taiwan. 6Departments of Laboratory Medicine,National Taiwan University Hospital, Taipei, Taiwan. 7Division of Hematology/Medical Oncology, Department of Medicine, Taichung Veterans General Hos-pital, Taichung, Taiwan. 8Department of Applied Chemistry, National Chi NanUniversity, Nantou, Taiwan. 9Department of Education and Research, TaichungVeterans General Hospital, Taichung, Taiwan.

Note: Supplementary data for this article are available at Cancer ResearchOnline (http://cancerres.aacrjournals.org/).

T.-Y.W. Wei and P.-Y. Wu authors contributed equally to this article.

Corresponding Authors: Ming-Daw Tsai, Academia Sinica, 128, Academia Rd.,Sec. 2, Nankang, Taipei, 115, Taiwan. Phone: 8862-2785-5696, ext. 3070; E-mail:[email protected]; Ting-Jung Wu, Chang Gung Memorial Hospital, 5,Fuxing St., Guishan, Taoyuan 333, Taiwan. Phone: 88-633281200; E-mail:[email protected]; and Hsin-An Hou, National Taiwan University Hospital,7, Chung Shan S. Rd., Zhongzheng Dist., Taipei, 100, Taiwan. Phone: 88-6223123456; E-mail: [email protected]

doi: 10.1158/0008-5472.CAN-16-1004

�2016 American Association for Cancer Research.

CancerResearch

Cancer Res; 77(2) January 15, 2017494

on August 21, 2019. © 2017 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

Published OnlineFirst November 10, 2016; DOI: 10.1158/0008-5472.CAN-16-1004

http://cancerres.aacrjournals.org/

chemoresistance of cancer cells by reducing chemotherapy-induced apoptosis through activation of NF-kB signaling path-ways (17). In support, an increasing ratio of Bax/Bcl-2, as sug-gestive of promoted apoptosis, has been reported upon treatmentwith Aurora A inhibitor VX680 (18), a drug that showed clinicaleffectiveness to chronic myeloid leukemia (CML; ref. 19).

TRAF-interacting proteinwith FHAdomain (TIFA) interactswithTRAF6 andmodulates oligomerization/ubiquitinationofTRAF6 toactivate IkB kinase (IkK) complex, through which IkB is subse-quently phosphorylated and undergoes ubiquitination-dependentdegradation allowing nuclear translocation of NF-kB to transacti-vate downstream factors (20). Our previous study uncovered thatthis functional event requires the phospho-threonine 9 (pThr9)-directed oligomerization of TIFA dimer upon TNFa stimulation(21). In the current study, we further identified Aurora A as anessential kinase for TIFA phosphorylation at Thr9 and showed thatTIFA mediates the Aurora A–dependent NF-kB signaling pathway.The functional link between TIFA and Aurora A prompted us toexplore the role of TIFA in tumorigenesis and in the regulation ofNF-kB–dependent inflammatory and survival signaling. Ourresults demonstrate that TIFA supports leukemic cell growth andis required for chemoresistance throughmodulation of prosurvivalfactors and the positive feedback signaling. This finding points toTIFA as a possiblemarker for unfavorable prognosis and apotentialnovel therapeutic target in the treatment of AML.

Materials and MethodsCell culture, isolationof humanperipheral bloodmononuclearcells, and TNFa stimulation

HeLa, 293T, andHL-60 cell lineswere obtained fromDr. Li-JungJuan (Academia Sinica, Taipei, Taiwan, 2014). KG-1 cell line wasobtained from Prof. Shih-Lan Hsu (Taichung Veterans GeneralHospital, Taichung, Taiwan, 2015), and the rest of cell lines wereobtained from Dr. Shui-Tein Chen (Academia Sinica, Taipei,Taiwan, 2015). All cell lines were expanded and stored in liquidnitrogen when received. Original vials were thawed for theseexperiments and authenticated simultaneously by STR DNA fin-gerprinting using the Promega GenePrint 10 System and analyzedby ABI PRISM 3730 GENETIC ANALYZER and GeneMappersoftware. All resulting STR profiles were matched with knownATCC fingerprints (www.ATCC.org). Maintenance of respectivecell lines followedour previous protocols (21–23). 293T andHeLacells were maintained in DMEM(Gibco). Primary peripheralblood mononuclear cells (PBMC), HL-60, KG-1, THP-1, U937,Jurkat, and K562 cells weremaintained in RPMI1640 (Invitrogen).All media were supplemented with 10% decomplemented FBS(Gibco), 200 mmol/L L-glutamine (Gibco), 100 U/mL penicillin(Gibco) and 100 mg/mL streptomycin (Gibco). All cells wereincubated in a humidified atmosphere at 37�C with 5% CO2. Toisolate humanPBMCs, freshly acquiredwhole-blood sampleswerelayered over Ficoll (GE Healthcare) as described previously (23).Isolated primary cells were maintained according to a previouslydescribed protocol (24) and subjected to experimental analyseswithin twoweeks.Whennecessary, cells were starved in serum-freecondition for 8 hours prior to addition of 10 ng/mLof TNFa (R&DSystems) for 30 minutes or at indicated time points.

Transient transfection of plasmid and siRNAFor transient transfection of 293T cells, 6 � 105 cells were

transfected with 2 mg of indicated plasmid DNA using Jet-PEI

(Polyplus Transfection) for 48 hours (pCDNA3.1 vector, HA-tagged TIFA, and pNF-kB-Luc) or 24 hours (Flag-tagged AuroraA) according to the manufacturer's instructions. For transienttransfection of U937 cells, 106 suspension cells were transfectedwith 3 mg of the pNF-kB-Luc plasmid DNA using Lipofectamine2000 (Invitrogen) for 48 hours according to the manufacturer'sinstructions. To silence specific genes, 6 � 105 293T cells or 106

suspension cells were transfected with 30 pmol (293T cells) or 50pmol (suspension cells) of indicated siRNAs supplementedwith 9or 15 mL of Lipofectamine RNAiMAX (Invitrogen) in the presenceof Opti-MEM (Invitrogen) according to the manufacturer'sinstruction. Six hours after primary transfection, cells were incu-bated with the regular medium for 24 hours, and the secondarysiRNA transfection was performed with the same protocol. Cellsor conditionedmedia were collected after 72 hours of incubation.For details on plasmid construction and synthesis of siRNA, seeSupplementary Methods.

Preparation of retrovirus-based stable linesTo perform retroviral transduction, pseudotyped viruses were

packaged by the pantropic retroviral expression system in thepresence of VSV-G according to the manufacturer's instructions(Clontech) and subjected to infection of indicated cells asdescribed previously (22). After 400 mg/mL G418 selection forweeks, the expressions of ectopic proteins in stable cells wereconfirmed by Western blot analysis before experimental analysisor siRNA transfection for phenotypic rescue experiments.

Treatments of cellsAll chemotherapy drugs used in this study were obtained as

follows: etoposide (VP-16) was from Sigma, idarubicin andcytarabine (Ara-C) were from Pfizer, cisplatin was from TheEuropean Directorate for the Quality of Medicines & HealthCare(EDQM), sorafenib (Rafi) was from Santa Cruz Biotechnology,etanercept (Enbrel, TNFa inhibitor) was fromWyeth, bortezomib(Velcade, NF-kB inhibitor) was from Janssen Pharmaceuticals,and ABT-263 (Navitoclax, BCL-2 inhibitor) was from AdooQBioscience. For cell treatment with X-ray irradiation, irradiationwas carried out in a Faxitron RX-650 irradiator (Faxitron X-rayCorporation) at a dose rate of 0.46 Gy/minute. When necessary,cells were starved in serum-free condition for 8 hours prior to theaddition of 10 ng/mL of TNFa (R&D Systems) for 30 minutesor at indicated time points.

The kinase inhibitors and conditions used in the treatment of293T cells were as follows: 100 nmol/L casein kinase II inhibitorIII TBCA (CK2i), 500 nmol/L 4-amino-5-(bromomethyl)-2-methylpyrimidine dihydrobromide (GRKi), and 25 nmol/L sor-afenib (Rafi) were from Santa Cruz Biotechnology; 200 nmol/Lnecrostatin-1 (RIPK1i) and 10 nmol/L G €O 6976 (PKCi) werefrom Tocris Bioscience; 50 nmol/L Akt inhibitor VI (Akti), 300nmol/L IRAK-1/4 inhibitor (IRAK1/4i), and 10 nmol/L (5Z)-7-oxozeaenol (TAK1i) were from Merck Millipore; 500 nmol/Lcaffeic acid phenethyl ester (CAPE, NF-kBi) was from Sigma; and10 nmol/L VX680 (Aurora Ai) was from Selleck Chemicals; 50nmol/L MK-5108 (Aurora Ai) was from AdooQ Bioscience.

AntibodiesTIFA-specific mAb was raised as described previously (21).

Anti-Aurora A, antiphospho-Aurora A (Thr288), anti-Bcl-2, andanti-Bcl-XLwere fromCell Signaling Technology. Anti-His-tagwas

TIFA Supports NF-kB Survival Pathway in AML

www.aacrjournals.org Cancer Res; 77(2) January 15, 2017 495




from SignalChem. Anti-Myc, anti-Flag, and anti-HA were fromSigma. Anti-IkK alpha (phospho T23), anti-IkB alpha (phosphoS32þS36), anti-human CD45 PerCP/Cy5.5, anti-human CD33FITC, anti-b-actin, and anti-Bax were from Abcam. Anti-NF-kB–active p65 subunit (clone 12H11) and horseradish peroxidase(HRP)-conjugated anti-mouse or anti-rabbit IgG were fromMillipore.

In vitro kinase assayExperiments followed protocols we previously established (21).

293T cells under indicated treatment were lysed with CHAPS lysisbuffer (20 mmol/L PIPES, 1 mmol/L Na3VO4, 1 mmol/L EGTA,50 mmol/L Tris-HCl, 150 mmol/L NaCl, 50 mmol/L NaF, 1%CHAPS, and 10% glycerol) supplemented with protease inhibitorcocktail (Roche) and phosphatase inhibitor (Sigma). The cellextracts were incubated with the recombinant His-tagged wild-typeor T9A mutant TIFA in the reaction buffer (40 mmol/L HEPES pH7.5, 20 mmol/L MgCl2, and 100 mmol/L ATP) in the presence of1mCi/mL [g-32P]ATP (Perkin-Elmer) at 37�C for 30 to 90minutes.His-taggedproteinswere thenpulleddownwithM-280Dynabeadscoatedwith anti-HismAb. The reactionwas terminatedby additionof SDS sample buffer and heating at 95�C for 10 minutes prior toSDS-PAGE. The level of phosphorylation on recombinant proteinswas revealed through autoradiography.

ImmunoprecipitationImmunoprecipitation experiment was conducted as described

previously (23). Subconfluent 293T cells in a p100 petri dish weretransfected with 5 mg of the vectors or HA-TIFA using Jet-PEI(Polyplus Transfection). After 48 hours, cells were treated with orwithout 10ng/mLTNFa (R&DSystems) for 30minutes,washedbyPBS, and lysed with CHAPS lysis buffer (20 mmol/L PIPES, 1mmol/L Na3VO4, 1 mmol/L EGTA, 50 mmol/L Tris-HCl, 150mmol/L NaCl, 50 mmol/L NaF, 1% CHAPS, and 10% glycerol)supplemented with protease inhibitor cocktail (Roche) and phos-phatase inhibitor cocktail 3 (Sigma).Threemicrogramsof indicatedantibody was precoated with 100 mL M-280 Dynabeads (Invitro-gen) overnight at 4�C. The 500 mg cell extracts were incubated with15 mL antibody-coated Dynabeads at 4�C overnight. The protein–bead complex was thenwashedwith Tris-buffered saline Tween 20(TBST) buffer and subjected to Western blot analysis.

Western blot analysisTo obtain whole-cell extracts, 6 � 105 293T cells or 106

suspension cells were washed with ice-cold PBS, lysed in RIPAlysis buffer (50 mmol/L Tris pH 7.4, 150 mmol/L NaCl, 1% NP-40, 0.5% sodium deoxycholate, and 0.1% SDS) supplementedwith protease inhibitor cocktail (Roche) and phosphatase inhib-itor cocktail (Sigma). After five repeated freeze–thaw cycles, cellextracts were cleared by centrifugation at 4�C. Cell extracts werequantified using Bradford assay, and equal amounts of solubleproteins weremixedwith SDS sample buffer, boiled, separated bySDS-PAGE, and blotted onto polyvinylidene difluoride mem-branes (Millipore). Membranes were then blocked for 1 hourwith 5% dry milk in PBST (PBS supplemented with 0.1% Tween-20) buffer and incubated for overnight at 4�C with primaryantibody diluted in PBST containing 1% BSA (Sigma). After threewashes with PBST, membranes were incubated with HRP-conju-gated anti-mouse or anti-rabbit IgG.Membraneswerewashedfivetimes with PBST and the blotted protein bands were revealedby the ECL system (Millipore). To blot the large number of

signaling factors from the same experiment, some samples wererun on more than one gel in parallel. The loading control wasobtained from each gel, and a representative one is shown in thesame figure panel (Figs. 1 and 4). All Western blot analyses wereperformed independently at least three times except that ofprimary PBMCs.

Immunofluorescence stainingImmunofluorescence staining experiment was performed

as described previously with minor modifications (21). Sub-confluent HeLa cells seeded on coverslips were treated with100 ng/mL nocodazole (Sigma) or DMSO for 16 hours toinduce G2–M arrest, followed by treatment with PBS or 10ng/mL TNFa (R&D Systems) for 30 minutes. Cells were washedwith ice-cold PBS and fixed with 4% para-formaldehyde (Elec-tron Microscopy Sciences) for 20 minutes. Sequentially, cellswere permeabilized with 0.2% Triton X-100, blocked with PBScontaining 10% BSA, and then incubated with primary anti-bodies at 4�C overnight followed by three washes with TBST.Washed cells were then incubated with secondary antibodiesconjugated with tetramethylrhodamine isothiocyanate or fluo-rescein isothiocyanate (Santa Cruz Biotechnology) and 40,6-diamidino-2-phenylindole (DAPI, Sigma) for 1 hour. Afterwashing with TBST, the coverslips were mounted with 90%glycerol containing antifading reagent p-Phenylenediamine(PPD). The fluorescence images on the coverslips were analyzedusing Materials Analysis Microscope Leica DM2500.

Luciferase reporter assayLuciferase-based NF-kB reporter assay was conducted as

described previously (21). A total of 6 � 105 293T or 106

suspension U937 cells in one well of 6-well plates with indicatedconditions were transiently transfected with 2 mg of pNF-kB-Lucreporter plasmid for 48 hours before harvest. Cells were treatedwith 10 ng/mL TNFa (R&D Systems) for 30 minutes and lysedwith 1�passive lysis buffer (Promega).One hundredmicrogramsof cell lysateswere dispensed into awell of 96-well plate, followedby addition of 100 mL of luciferase assay buffer LARII (Promega)and 100 mL of Stop & Glo reagent (Promega). The triplicatedchemiluminescence were measured by SpectraMax ParadigmMulti-Mode Plate Reader (Molecular Devices) and normalizedto corresponding b-actin level from a parallel Western blot anal-ysis of cell lysates. Results were represented as relative lumines-cence units (RLU) from three independent experiments.

WST-1 cell viability assayA total of 105 suspension cells under indicated condition were

seeded per well in 96-well plates and cell viability was analyzed atthe indicated time point posttreatment. The WST-1 reagent(Roche) was mixed with cells in 10-fold dilution according tothe manufacturer's instructions, and absorbances of triplicatedresults were read by SpectraMax Paradigm Plate Reader (Molec-ular Devices) and normalized to day 0 or mock control.

Patient samplesSixteen patients who were diagnosed as de novo AML at the

TaichungVeteransGeneralHospital, Taiwan (VGHTC;AML#1–9)and National Taiwan University Hospital (NTUH; AML#10–16)from December 2007 to March 2014 were enrolled for theanalyses of Western blot, WST-1 cell viability, and apoptosis

Wei et al.

Cancer Res; 77(2) January 15, 2017 Cancer Research496




assay. The average of blasts in these patients weremore than 60%�18 with 4 exceptions (patient #2, 9, 12, and 16) whose bloodsamples were taken after chemotherapy to suppress acute con-

dition. For analysis of immunocytochemical staining, 85 adultAML patients who received standard chemotherapy and hadavailable cryopreserved bone marrow cells at the NTUH from

Figure 1.

TIFA is required for Aurora A–dependent NF-kB signaling. A, In vitro kinase assay for TIFA phosphorylation. Top, Western blot analysis of lysates from293T cells transfected with Flag-tagged Aurora A (Flag-Aurora A) or Aurora A siRNA (siAurora A), or treated with Aurora A inhibitors (VX-680 and MK-5108).Aurora A pT288 represents the activated kinase (18). Left, in vitro kinase assay using recombinant TIFA wild-type (WT) and the unphosphorylatableT9A mutant proteins incubated with lysates from 293T cells treated with Aurora A inhibitors and TNFa in the presence of [g-32P]ATP. WCE, whole-cell extract; Si,Aurora A siRNA; VX, VX680; MK, MK5108; PPTase, alkaline phosphatase. Right, quantification and statistical analysis of three independent experiments.B, Left, immunofluorescence staining of HeLa cells treated with nocodazole and TNFa. Right, intensity of dashed lines indicated on the left. Green, TIFA; red,phospho-Aurora A Thr288. C, Top, immunoprecipitation of Aurora A from 293T cell lysates transfected with HA-tagged TIFA under different time courses of TNFastimulation. Bottom, Western blot analysis of total lysates used in top panel. S, supernatant; P, precipitate. D, Left, Western blot analysis of lysates from 293T cellstransfected with Flag-tagged Aurora A and TIFA siRNA. Right, the same as left panel except that U937 cells transfected with TIFA siRNA and treated with 10 ng/mLTNFa for 30 minutes were examined. E, The same as D, except that the treated cells were additionally transfected with pNF-kB-Luc plasmid and analyzed byluciferase reporter assay for NF-kB activity. Plots represent relative luminance unit (RLU) normalized to nontransfection or non-TNFa-stimulated controls.






Figure 2.

Overexpression of the TIFA protein is associated with prosurvival factors and poor prognosis in patients with AML. A, Representative protein expressions ofTIFA and prosurvival factors in normal and AML PBMCs (n ¼ 16 in each group). Results of Western blot analyses (Fig. 4C; Supplementary Fig. S2A and S2B)were quantified and normalized to corresponding b-actin levels, and relative intensities of each signal factors normalized tomediumvalue in the normal counterpartswere plotted. Horizontal bars, medium value. B, Representative images of immunocytochemical staining of TIFA protein in bone marrow specimensfrom two patients with higher (left) and lower (right) TIFA expressions. Magnification, �1,000. C, Kaplan–Meier curves of OS (left) and DFS (right) of patientswith de novo AML stratified by the level of TIFA expression. D, The same as C, except that patients with non-M3 subtypes (left), intermediate-risk cytogenetics(middle), and normal cytogenetics (right) were compared.

Wei et al.





March 2009 to January 2012 were enrolled (25, 26). Writteninformed consent was obtained from all participants in accor-dance with the Declaration of Helsinki, and all experimentalstudies followed the institutional protocols (CE11166-2,VGHTC, No. 201408027RIND, NTUH, and AS-IRB02-103189and AS-IRB02-104147, Academia Sinica, Taiwan).

Immunocytochemical stainingImmunocytochemical staining was performed as described

previously with minor modifications (27). Cytospin smears ofbone marrow leukemia cells from 85 patients were fixed for 3minutes in formalin acetone (3%). The specimen was thenincubated with peroxidase blocking enzyme (10 minutes), fol-lowed by TIFAmAb (1:200 in blocking solution, overnight at 4�C;ref. 21). Biotinylated donkey anti-mouse IgG (1 mg/mL, diluted inblocking solution, 30 minutes; DAKO) was used as secondaryantibody and the protein was detected using the streptavidin–peroxidase complex (DAKO). The specimenswere counterstainedwith hematoxylin. A score of 0 to 4 was calculated for eachspecimen, according to the addition of the score of stainingintensity (0 ¼ none, 1 ¼ weak, 2 ¼ strong) and the score ofpercentage of myeloid cells positively stained (0¼ 0%–25%, 1¼26%–50%, 2 ¼ 51%–100%).

Analyses of apoptosis and flow cytometryAt the second day after transfection of siRNAs, 106 PBMCswere

seeded per well in 6-well plates and then treated with differentchemotherapy drugs. After 48 hours, cells were washed with ice-cold PBS and subjected to apoptosis detectionusing FITCAnnexinV apoptosis detection kit according to the manufacturer's instruc-tions (BD Biosciences). Cells were then gated for myeloid lineageby light-scattering following a previously described protocol (28),and apoptosis rate, as suggested by double staining of propidiumiodide and FITC Annexin V, was analyzed by BD FACSCaliburFlow Cytometer (BD FACSCalibur) as described previously (22).For analysis of U937 cells in the xenograft model, bone marrowcells ormouse PBMCswere sequentially stainedwith anti-humanCD33 FITC and CD45 PerCP/Cy5.5, respectively, for 1 hour onice. Cells were washed with ice-cold PBS and subjected to flowcytometry as mentioned above.

Assessment of cytokine and chemokine secretion (cytokinearray and ELISA)

Conditionedmedia from equal numbers of culturedU937 cellswere collected at 72 hours after indicated treatments and assessedby theHumanCytokine Array C1 according to themanufacturer'sinstructions (AAH-CYT-1, RayBiotech). Data were acquired using

the ImageJ software and normalized to the positive controls.When indicated, the amounts of secreted cytokines and chemo-kines in the conditionedmedia were determined by the ELISA testkit following themanufacturer's instructions (R&D Systems). Theabsorbances of triplicated samples were read by SpectraMaxParadigm Plate Reader (Molecular Devices). All raw data werenormalized toPBS control, and resultswere represented as relativesecretion of cytokines from three independent experiments.

Xenotransplantation of human leukemic cells and in vivochemo drug treatment

Six- to 8-week-old athymic nude mice from the NationalLaboratory Animal Center (Taiwan) were housed in the patho-gen-free facility of Chang Gung Memorial Hospital (Protocol no.2014092305). For the myeloid sarcoma model, mice were inoc-ulated subcutaneously with 5 � 106 U937-stable cells in 50%Matrigel (Corning) per animal, followed by intraperitonealadministration of PBS or cytarabine (Ara-C, 50 mg/kg) at day6, 8, and 11 postinoculation. Tumor sizes were documented twiceper week by the formula, width2 � length � 0.52 (29). For theorthotopic leukemic model, mice were injected with 1.5 � 105

U937-stable cells per animal via inferior vena cava (IVC) throughsurgical operation, followed by five daily administrations of thesame treatment at the day 5 after transplantation. Mice werehumanely killed at day 60, and bone marrow cells in femurswere subjected to flow cytometry.

Statistical analysisAll data are represented by the means � SD from three

independent experiments. Statistical significance determinedby the independent samples t test were represented as �, P <0.05; ��, P < 0.01; and ��, P < 0.001 except otherwise specified.For immunocytochemical staining, correlations between vari-ables were assessed by the Spearman rank correlation. Mann–Whitney U test was used to analyze the difference in theexpression of TIFA in patients with AML. Overall survival (OS)was measured from the date of first diagnosis to death from anycause, and disease-free survival (DFS) was calculated from thetime of complete remission (CR) until relapse, death from anycause, or end of study. Kaplan–Meier estimation was used toplot survival curves, and log-rank tests were used to test thedifference between groups. HR and 95% confidence interval(CI) was estimated by Cox proportional hazards regressionmodels to determine independent risk factors associated withOS and DFS in multivariate analyses. Two-sided P values < 0.05were considered statistically significant.

Table 1. Multivariate analysis (Cox regression) on the OS and DFS

OS DFSVariables RR (95% CI) P RR (95% CI) P

Ageb 3.475 (1.326–9.107) 0.011a 2.107 (0.914–4.855) 0.080WBCc 2.452 (1.116–5.385) 0.026a 1.404 (0.661–2.981) 0.377Karyotyped 2.350 (1.193–4.627) 0.013a 2.072 (1.110–3.867) 0.022a

NPM1/FLT3-ITDe 0.941 (0.321–2.758) 0.912 1.403 (0.566–3.475) 0.464Higher TIFAf 2.768 (1.200–6.385) 0.017a 1.931 (0.905–4.118) 0.089

Abbreviation: RR, relative risk.aStatistically significant (P < 0.05).bAge > 50 relative to age �50 (the reference).cWBC greater than 50,000/mL vs. 50,000/mL or less.dUnfavorable cytogenetics vs. others.eNPM1mut/FLT3-ITDneg vs. other subtypes.fHigher TIFA vs. lower TIFA expression.






Figure 3.

TIFA silencing enhances chemotoxicities to AML lines. A, Time courses of cell viabilities upon TIFA silencing for four AML lines were examined by theWST-1 assay. Plots represent relative cell viabilities normalized to day 0 from three independent experiments. B, Plots represent relative cellviabilities normalized to mock controls at the fourth day from experiments in A and the corresponding experiments with VX680 treatment. C, AMLlines used in A and B were analyzed by WST-1 assay for cell viabilities under chemo drugs treatment. Plots represent calculated IC50 of chemo drugs.siCon, control siRNA; siTIFA, TIFA siRNA.

Wei et al.





Figure 4.

TIFA silencing enhances chemotoxicities to AML PBMCs. A, Time courses of cell viabilities upon TIFA silencing for 8 AML PBMCswere examined by theWST-1 assay.Plots represent relative cell viabilities normalized to day zero. B, Plots represent relative cell viabilities normalized to mock controls at the fourth dayfrom the experiments in A and the corresponding experiments for normal PBMCs and with VX680 treatment (n ¼ 8 in each group). C, Western blot analysisfor the levels of prosurvival factors in 8 AML PBMCs upon TIFA silencing and 8 normal PBMCs. D, AML PBMCs used in A and B were analyzed by WST-1 assayfor cell viabilities under chemo drug treatments. Plots represent calculated IC50 of chemo drugs. E, Four AML PBMCs (AML #9, 10, 13, and 15) were treatedwith 10 mmol/L etoposide, 10 nmol/L idarubicin, or 10 mmol/L cytarabine upon TIFA silencing or VX680 treatment, and analyzed by flow cytometry. Plots representpercentages of cells with Annexin V and propidium iodide–positive staining. Horizontal bars, medium value. siCon, control siRNA; siTIFA, TIFA siRNA.






Wei et al.





ResultsTIFA is involved inAuroraAkinase–dependentNF-kB signaling

We previously suggested that phosphorylation of TIFA at thre-onine 9 could involve Ser/Thr kinases in the PI3K/AKT signalingpathway (21). To pinpoint the most direct kinase responsible forThr9 phosphorylation of TIFA, we screened kinase inhibitorsthrough a protocol we established previously (21, 30), and foundthat the Aurora A inhibitor was most effective against in vitrophosphorylation by TNFa-treated 293T cell lysates (Supplemen-tary Fig. S1A and S1B). We further validated that Aurora A isrequired for the Thr9 phosphorylation of TIFA based on threeexperiments: First, lysates of cells treated with Aurora A–specificsiRNA and two different Aurora A inhibitors all showed attenu-ated abilities in TIFAThr9phosphorylation, similar to the result ofthe unphosphorylatable T9A mutant (Fig. 1A, lanes 5–7, withcontrols in Supplementary Fig. S1C). Second, immunofluores-cence staining showed colocalization of endogenous TIFA speck-les (21) and activated Aurora A (phospho-Thr288) upon TNFatreatment (Fig. 1B). Third, immunoprecipitation experimentsrevealed direct interaction between Aurora A and TIFA, whichwas reinforced upon TNFa stimulation (Fig. 1C; SupplementaryFig. S1D–S1E).

It has been reported that Aurora A regulates NF-kB signaling(17), and elevated expression of Aurora A coincides with pro-moted levels of TNFa, NF-kB, and chronic inflammation in gastricneoplasia (31). We also observed amarked elevation in the levelsof activated NF-kB, phosphorylated IkK and IkB, and NF-kB–driven luciferase activity in response to overexpressionof AuroraAor TNFa treatment (Fig. 1DandE, lane 2), and in addition showedthat such elevation was negated upon silencing of TIFA (lane 4).The results collectively suggest that the proposed role of Aurora Ain the regulation of NF-kB signaling requires TIFA.

TIFA is correlated with Aurora A, prosurvival factors, and poorprognosis of AML

Aurora A was found to be highly expressed in AML (18). As wehave shown that TIFA is required for the activation of NF-kB inAML lines U937 and THP-1 (Figs. 1D and E and 5D inHuang andcolleagues, 2012), we further askedwhether the TIFA protein levelis associated with Aurora A and NF-kB signal factors duringleukemogenesis. Because of the cytogenetic clonal heterogeneityin AML (32) and limited volumes of patient blood samples, wechose to perform Western blot analysis with a TIFA-specific mAb(21) to compare freshly isolated PBMCs from 16 patients with denovo AML and 16 normal individuals. We found that the TIFAprotein level is significantly elevated in all of the AML-derivedPBMCs but remains unchanged in those of normal donors. Inaddition, the elevation of TIFA is concomitant with upregulationof the levels of Aurora A protein, activated Aurora A, phosphor-ylated IkK/IkB, activatedNF-kB, andprosurvival factors Bcl-2/Bcl-

XL (Fig. 2A). In support, pairwise comparisons among the levels ofall tested proteins in all PBMCs reveal strong correlations, suggest-ing that these factors are functionally related in AML (Supple-mentary Fig. S2C, Pearson correlation coefficient¼0.862–0.991).

The prognostic implication of the TIFA protein level observedfrom nonfractionized PBMCs was further strengthened by theimmunocytochemical staining of the cryopreserved bonemarrowleukemia cells derived from85patientswith de novoAML. Patientswho received conventional intensive-induction chemotherapy(26) were divided into low (score 0–2, n ¼ 40) and high (score3–4, n¼ 45) expression groups, based on immunocytochemistryresults (Fig. 2B). Patients with AML with higher TIFA proteinlevel seemed to have higher WBCs and blast counts than thosewith lower TIFA level (Supplementary Table S1). In addition,higher TIFA protein expression was closely associated with theFAB M4 subtype (P ¼ 0.0315), but negatively associated withCD7 expression (P ¼ 0.0454) in leukemia cells (SupplementaryTables S1 and S2).

Of all AML patients examined, 54 (63.5%) achieved a CR.Higher TIFA expression was associated with an inferior responserate (CR rate, 51.1% vs. 77.5%, P ¼ 0.0139). After following-upfor a medium of 57.8months (range, 0.3–79.1 months), patientswith higher TIFA expression showed significantly shorter OSand DFS than those with lower TIFA expression (Fig. 2C). Thedifferences remained significant in the subgroups of patients withnon-M3 subtypes, intermediate-risk cytogenetics, and normalkaryotype, though to a lesser extent for disease-free survival(DFS; Fig. 2D). In multivariate analysis, higher TIFA expressionwas an independent factor for poor prognosis on OS irrespectiveof age, WBC counts at diagnosis, cytogenetics, and NPM1/FLT3-ITD (Table 1). These results collectively suggest that TIFA expres-sion is associated with not only Aurora A–dependent NF-kBsignaling but also poor prognosis in AML.

TIFA silencing enhances chemotoxicity to AMLTo validate the tumorigenic role of TIFA in relation to Aurora A

signaling, we targeted TIFA-dependent signaling through RNAinterference and examined the proliferation of AML lines throughWST-1 viability assay in a time-dependent manner. The resultshowed that silencing of TIFA attenuates the growth of AML celllines HL-60, KG-1, THP-1, and U937 (Fig. 3A, with controls inSupplementary Fig. S3A), similar to the outcome of VX680treatment (Fig. 3B; ref. 18). In addition, cell proliferations ofboth acute lymphoblastic leukemia (ALL) line Jurkat and CMLline K562 were also inhibited in response to TIFA silencing andVX680 treatment (Supplementary Fig. S3B and S3C).

Inhibition of Aurora A kinase can promote leukemic chemo-sensitivity, which was considered an anticancer strategy (14, 33).To test whether targeting TIFA also relieves the aggressiveness andchemoresistance of hematologic malignancy like inhibition of

Figure 5.Molecular targeting of TIFA inhibits TIFA-mediated NF-kB activation. A, Plots represent calculated IC50 of chemo drugs for AML lines under TIFA silencingand overexpression of siRNA-resistant (siR) TIFA wild-type (WT) or T9A mutant. B, Schematic representation of DN constructs of TIFA. AML lines were,respectively, infected with pseudotyped retroviruses carrying these constructs to generate stable cells. 2Myc, 2 consecutive Myc-tags. C, Lysates from U937 cellsstably expressing TIFA fragments shown in B were immunoprecipitated using anti-Myc immunobeads and analyzed by Western blot analysis. D, Cells usedinCwere additionally transfectedwith pNF-kB-Luc plasmid, treatedwith 10 ng/mLTNFa for 30minutes, and analyzedby luciferase reporter assay forNF-kBactivity.Luciferase activities were normalized to non-TNFa-stimulated control. Plots represent relative luminance unit (RLU) acquired from three independentexperiments. E, Lysates in Dwere analyzed byWestern blot analysis for levels of NF-kB signaling factors. F, AML lines stably expressing TIFA fragments describedin B were treated with chemo drugs in dose-dependent manners and analyzed by WST-1 assay for cell viabilities (Supplementary Fig. S5D). Plots representcalculated IC50 values of chemo drugs acquired from three independent experiments. siCon, control siRNA; siTIFA, TIFA siRNA.






Wei et al.





Aurora A, we investigated leukemic cell viability upon varioustreatments under TIFA silencing. The WST-1–based survival testshowed that all four tested AML lines exhibit approximately 50%decrease in IC50, similar to the effect of VX680, toward chemotreatments with etoposide, idarubicin, and cytarabine, the threeconventional chemotherapy drugs for AML (Fig. 3C; Supplemen-tary Table S3, with raw data in Supplementary Fig. S3D).Similar results were also obtained from silencing of Aurora A(Supplementary Fig. S3E), from treatments with atypical drugs(Supplementary Fig. S3F), and from ALL and CML lines (Supple-mentary Fig. S3G–S3I; Supplementary Table S4).

We next examined the effect of TIFA inhibition using freshlyisolated AML PBMCs. In addition to growth retardation (Fig. 4A,with calculated results in the right panel of Fig. 4B), a significantretrogression of elevated NF-kB signaling factors was seen fromTIFA-depleted AML PBMCs, even under overexpression and acti-vationof Aurora A kinase (Fig. 4C). Interestingly, the effect of TIFAsilencing on the viability of normal PBMCs appeared to be lessthan that of VX680 treatment, in contrast to the comparablechemotoxicity in the AML counterparts (comparison betweenright and left panels in Fig. 4B), implicating that targeting TIFAmay convey less cytotoxicity toward nonleukemic cells. Moreimportantly, silencing of TIFA consistently promoted chemosen-sitivities of patient-derived PBMCs similar to inhibition of AuroraA by VX680 (Fig. 4D; Supplementary Table S5, with raw data inSupplementary Fig. S4A and S4B; the results of atypical treatmentsare shown in Supplementary Fig. S4C), implicating that TIFA isrequired for leukemic chemoresistance similar to Aurora A (34).These promoted chemosensitivities were most likely caused byenhanced apoptosis (Fig. 4E, calculated from SupplementaryFig. S4D) and downregulated NF-kB signaling factors (Fig. 4C,additional data in Supplementary Fig. S4E and S4F) upon silenc-ing of TIFA. Our results together suggest that TIFA is required forthe maintenance of leukemic cell growth and chemoresistance,and that TIFA can serve as a potential therapeutic target.

Targeting TIFA through dominant-negative inhibitionTo address the molecular mechanism of targeting TIFA, we

phenotypically rescued TIFA-depleted AML lines by wild-typeTIFA and T9A mutant following a silencing-complementationstrategy (22). The silencing-dependent chemosensitivity wasefficiently alleviated upon expression of wild-type TIFA, butnot the T9A mutant, in all tested AML lines (Fig. 5A; Supplemen-tary Table S6, with control in Supplementary Fig. S5A and rawdata in Supplementary Fig. S5B), suggesting that pThr9-directedTIFA oligomerization is a key step required for the chemoresis-tance in AML.

We next attempted to disrupt TIFA oligomerization throughectopic expression of dominant-negative (DN) fragments of TIFA

in AML cells. We designed to fine-map the minimal effectiveregion for molecular targeting of TIFA in AML cells throughretroviral transduction (Fig. 5B). The immunoprecipitation oflysates from the transduced U937 cells showed that fragmentsF1F2 and F2F3, both containing the FHA dimerization-core withadditional C- and N-terminal extensions, respectively, were ableto pull down endogenous TIFA (Fig. 5C, with controls in Sup-plementary Fig. S5C), suggesting that the two fragments are ableto interfere with the self-association of endogenous TIFA (21). Asa consequence, TNFa–dependent NF-kB activation and eleva-tions of NF-kB signaling factors were attenuated by F1F2 or F2F3fragment inU937 cells (Fig. 5D andE).More significantly, the twoDN fragments also significantly promoted chemosensitivities ofAML lines with more than 50% decrease in IC50 (Fig. 5F; Sup-plementary Table S7, with raw data in Supplementary Fig. S5D).These observations collectively demonstrate the therapeuticpotential of molecular targeting of TIFA in AML.

Targeting TIFA potentiates the clearance of leukemicmyeloblasts

Aberrant secretion of cytokines contributes to pathogenesis ofleukemia (35) and is considered a prognostic signature for therecurrence ofAML (35, 36).We askedwhethermolecular targetingof TIFA can perturb NF-kB–dependent secretion of leukemiccytokines and attain better therapeutic efficacy. We performedcytokine antibody array to profile cytokines secreted by U937cells. The result showed that cytarabine treatment promotedsecretion of TNFa, CXCL-1, and IL8 but suppressed that of IL6,while expression of TIFA DN fragments clearly led to oppositeeffects (Fig. 6A, with detailed ELISA analysis in Fig. 6B). Withrespect to promoted TNFa secretion under cytarabine treatment,supplement of TNFa was able to enhance the viability ofU937 cells in response to cytarabine treatment, while expressionof TIFADN fragments significantly antagonized the effect of TNFa(Fig. 6C). In support, inhibition of TNFa and downstreamNF-kBsurvival signaling by etanercept (Enbrel, TNFa inhibitor), borte-zomib (Velcade,NF-kB inhibitor), or ABT-263 (Navitoclax, BCL-2inhibitor) enhanced the cytotoxicity of cytarabine-treatedU937 cells, similar to the effect of TIFA DN fragment (Supple-mentary Fig. S6A, with Western blot analysis of treated cells inSupplementary Fig. S6B). These results suggest that TIFA DNfragments could be therapeutically effective to block secretion ofinflammatory cytokines, leading to inactivation of NF-kB survivalfactors in AML.

We therefore assessed the therapeutic potential of TIFA DNfragments in vivo. Although transplanted hematopoietic tissuesproved difficult to propagate in nude mice compared with NOD-SCID orNSG/NOGmice (37), we chose thismore difficultmodelto examine the effect of targeting TIFA. This is because the intact

Figure 6.Targeting TIFA potentiates the clearance of leukemic myeloblasts. A, Conditioned media collected from 10 mmol/L cytarabine-treated U937-stable cells usedin Fig. 5F were analyzed by cytokine-antibody array for cytokine secretion profile. Boxed cytokines were quantified and are shown on the right. B, Secretionof leukemic cytokines from cells used in A was quntitatively analyzed by ELISA. Amounts of cytokines in conditioned media were normalized to PBScontrol, and plots represent relative secretion of cytokines acquired from three independent experiments. siTIFA, TIFA siRNA transfection. C, U937 cells usedin A were analyzed by WST-1 assay for cell viabilities in the presence of 10 ng/mL TNFa and 10 mmol/L cytarabine. Results were normalized to day zero, and plotsrepresent time course of relative cell viabilities. D, Top, experimental procedure for growth of myeloid sarcoma in nude mice. Mice were subcutaneouslyinjected with retrovirally transduced U937 cells used in A and treated with cytarabine. Left, time course of tumor volumes; right, image of tumors measured in leftpanel upon autopsy. E, The same as D, except that U937-stable cells were injected via IVC, and that bone marrow (BM) engraftment was analyzed. Left, bonemarrow cells from treated mice were assessed by flow cytometry using anti-human CD45 and CD33 antibodies. Right, plots of percentages of human CD45þCD33þ

cells analyzed in left (n ¼ 7 in each group). Horizontal bars, medium value. SC, subcutaneous injection; IP, intraperitoneal injection; IVC, injection of cellsvia inferior vena cava.






inflammatory cytokines required for proliferative myeloid pro-genitors could mimic a microenvironment to support chemore-sistance (38). Ectopic xenograft model showed that retroviraltransduction of TIFA DN fragments in addition to cytarabinetreatment significantly prevented the expansion of myeloid sar-coma relative to cytarabine treatment alone (Fig. 6D, with West-ern blot analysis of harvested tumors in Supplementary Fig. S6C).Such prevention of AML expansion in vivowas also observed frominhibition of TNFa, NF-kB, or BCL-2 (Supplementary Fig. S6D),supporting that the enhanced chemotoxicity by targeting TIFAwaslikely due to blockage of the TNFa-dependent NF-kB survivalpathway. In addition, we also established an orthotopic xenograftmodel through direct injection of U937-stable cells via IVC innudemicewithout immunosuppression by chemical or sublethalirradiation. Flow cytometry analysis of bone marrow cells fromrecipient mice revealed that targeting TIFA significantly promotesthe clearance of engrafted human CD45þCD33þ myeloblastsupon cytarabine treatment (Fig. 6E, with PBMC control in Sup-plementary Fig. S6E). The constrained leukemic myeloblasts andfluctuated cytokines collectively suggest that molecular targetingof TIFA is able toperturb thepositive feedback betweenNF-kBandTNFa that is required for the progression of AML (39).

DiscussionTIFA is a relatively new player in the NF-kB signaling pathway.

Our earlier study uncovered the pivotal role of TIFA oligomeri-zation in the promotion of NF-kB signaling pathway uponTNFa stimulation, mechanistically through the inter-molecularbinding between phosphorylated Thr9 and FHA domain of TIFAdimers (21). A recent study demonstrated that such phosphory-lation-dependent oligomerization of TIFA is also triggered by aGram-negative bacteria–derived monosaccharide heptose-1,7-bisphosphate (HBP), leading to activation of innate immunity(30). In a separate article (40), we further showed that TIFAmediates innate immune response through assembly of NLRP3inflammasome. In this study, we identified TIFA as the functionallink between the Aurora A signaling axis and theNF-kB–regulatedprosurvival factors in AML. In addition, we demonstrated thetherapeutic potential of TIFA targeting in the treatment of AML byshowing that TIFA inhibition perturbs leukemic cytokine secre-tion, significantly enhances chemotoxicity to AML cells, andpotentiates the clearance of leukemic myeloblasts in a xenograftmodel.

Instead of conventional real-time PCR-based detection ofprognostic factors at the transcriptional level, we directly assessedthe TIFA protein level through Western blot analysis of freshlycollected patient samples. We observed elevated levels of the TIFAprotein in AML patients comparedwith normal donors, similar toits kinase Aurora A, and demonstrated a tight correlation betweenTIFA and NF-kB driven prosurvival factors. By using immunocy-tochemical staining of cryopreserved patient myeloid cells, wealso reasoned that the higher protein level of TIFA in bonemarrowis an independent factor of poor prognosis for OS in patients withAML, irrespective of age, WBC counts, karyotype, and othergenetic markers. These observations were corroborated by thefinding that patientswithAMLwith higher TIFAprotein levelweremore refractory to chemotherapy and had lower CR rates andpoorer OS. On the basis of these findings, the higher TIFA proteinlevel in bone marrow could also serve as a novel biomarker thatforesees the clinical outcome of patients with AML. In agreement,

silencing of TIFA or expression of dominant-negative TIFAfragments both impaired NF-kB activation, resulting in stalledleukemic cell proliferation and enhanced chemotoxicity.

Ectopic expression of DN protein can antagonize signal trans-duction and elicit therapeutic potential in cancer treatment. Inparticular, transfection of IkB super-repressor, a DN IkB mutantthat ultimately inhibits NF-kB activity, was shown to selectivelyincrease ALL sensitivity to vincristine treatment, through NF-kB–regulated apoptosis (41). We likewise observed an enhancedchemosensitivity in leukemic cells upon ectopic expression oftwo DN fragments of TIFA. We reasoned that the minimal DNfragment must include FHA dimerization-core plus either theN-terminal Thr9 motif for disruption of endogenous TIFAoligomerization or the C-terminal TRAF6-interacting motif forthe interruptionof signal transduction (20, 21). Intriguingly, bothtranscriptional inhibition (through TIFA siRNA; Figs. 3 and 4)and molecular inhibition (through DN fragments; Figs. 5 and 6)display almost identical targeting efficacies against TIFA, impli-cating the irreplaceable and nonredundant role of TIFA in thesecellular processes.

Our results using AML, ALL, and CML lines consistently dem-onstrated the efficacy of targeting TIFA in both the myeloid andlymphoblastic lineages of leukemia, suggesting that the TIFA-driven NF-kB signaling axis may play a fundamental role duringleukemogenesis. In addition, a positive feedback between NF-kBand TNFa was recently described for leukemia-initiating cells(LIC) in the promotion of leukemogenesis (39). LIC-enrichedleukemic bone marrow cells exhibit a constitutively activated NF-kB signaling, which is considered the cause of postremissionrelapse of AML (39). In agreement, our result in Fig. 6C showedthat TNFa treatment did alleviate cytarabine cytotoxicity toundifferentiated U937 monocytes, while molecular inhibitionof TIFA blocked the effect through perturbation of NF-kB–depen-dent prosurvival signaling (Fig. 5D and E). Leukemic cells wereshown to be difficult to propagate in athymic nude mice due toremaining innate immunity (37). We nevertheless observed thehoming of these undifferentiated monocytes in femurs of micethat was suppressed by cytarabine treatment (Fig. 6E), and thatTIFA inhibition significantly promoted the effect. Although thisxenograft model may not represent the true stemness of LICs, itmay imply that the maintenance of undifferentiated leukemiccells under chemodrug treatment requires TIFA, and that the long-term expansion and self-renewal capacity of LICs could beimpaired through targeting TIFA.

Deregulation of innate immunity and inflammatory signalingmay lead to myelodysplastic syndromes (MDS), a group ofheterogeneous clonal hematologic malignancies that overpopu-lates bone marrow and progresses to AML (42). ConstitutiveNF-kB signaling may pathogenically transactivate inflammatorycytokines and prosurvival factors resulting in deregulated expan-sion ofMDSbonemarrow progenitors, while inhibition ofNF-kBactivity induces apoptosis in normal and MDS bone marrowprecursors (43). Consequently, Bcl-2 and Bcl-XL are consideredindicators for chemoresistance in myeloid malignancies includ-ing AML (9, 12, 44, 45). Because TIFA regulates NF-kB–driveninnate immunity (30), the leukemogenic role of TIFAweobservedin AML is more likely through the NF-kB–dependent antiapop-totic/prosurvival pathways. In agreement, our results using leu-kemic lines andAMLpatient–derivedPBMCs consistently showedthat TIFA is required for cell survival upon chemotherapies. Insupport, silencing of TIFA in patient PBMCs significantly blocked


Wei et al.




NF-kB–dependent prosurvival factors in spite of overexpressedAurora A (Fig. 4C) and chemo drug treatment (SupplementaryFig. S4E), suggesting that the loss of chemoresistance abilities weobserved in Figs. 3C, 4D, and 5A and F was due to the perturbedantiapoptotic/prosurvival factors Bcl-2 and Bcl-XL.

Tumor microenvironments are a milieu of proinflammatoryresponses coordinated by a great variety of inflammatory fac-tors including cytokines and chemokines, and such tumor-promoting inflammation essentially enables cancer develop-ment (46). In agreement, a growing number of inhibitors thattarget proinflammatory TNF superfamily members have beenshown to be therapeutically effective (47). Given the essentialrole of TIFA underlying the signaling chain between TNFaand NF-kB, its inhibition should be able to attenuate inflam-matory factors that are transcriptionally regulated by NF-kB.We observed attenuation of CXCL-1 and IL8 secretions uponthe molecular targeting of TIFA coherent with delayed leukemiccell growth and reduced chemoresistance (Fig. 6A and B),which supports previously suggested tumorigenic function ofthe two inflammatory factors (48, 49).

It has been well established that Aurora A is overexpressed inAML (16, 18, 50), and its inhibition was proposed as the targetedtherapy to treat several hematopoietic malignancies (51–53).Although gene amplification of Aurora Awas frequently observedfrom many types of solid tumors leading to overexpression andtumor progression (54), whether Aurora A is upregulated throughthis manner in AML remains elusive. We identified that TIFA islinked to TNFa-dependent NF-kB survival signal through a site-specific phosphorylation that functionally requires the kinaseactivity of Aurora A, through which a positive feedback loop issustained to trigger inflammatory responses and support che-moresistance of leukemic cells (Supplementary Fig. S7; ref. 36). Insupport, we observed enhanced chemotoxicity of AML cells undertreatments of TNFa, NF-kB, BCL-2, and Aurora A inhibitors,similar to transcriptional inhibition and molecular targeting ofTIFA (Figs. 3–5 and 6C; Supplementary Fig. S6A). The transla-tional implication was further strengthened by the combinedtreatment of cytarabine with inhibitor of TNFa, NF-kB, or BCL-2 in vivo (Supplementary Fig. S6D).

Overall, our results demonstrate the functional role of TIFA insupporting the Aurora A–dependent NF-kB survival and inflam-matory pathways as the molecular basis underlying leukemic cellgrowth and chemoresistance in AML (Supplementary Fig. S7). Asthe initial treatment approach using anthracycline or cytarabinehas remained unchanged for decades and CR is still poorlyachieved due to drug resistance, future AML treatment withTIFA-targeted strategy might provide therapeutic advantages tolower the incidences of chemoresistance, enhance the efficacy of

conventional chemotherapies, and improve the long-term clinicaloutcome.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

Authors' ContributionsConception and design: T.-Y.W. Wei, P.-Y. Wu, H. Hou, C.-L. Teng,C.-C.F. Huang, M.-D. TsaiDevelopment of methodology: T.-Y.W. Wei, P.-Y. Wu, T.-J. Wu, H. HouAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): T.-Y.W. Wei, P.-Y. Wu, T.-J. Wu, H. Hou, W.-C. Chou,C.-L. Teng, C.-R. Lin, J.-M.M. Chen, T.-Y. Lin, C.-T.R. Yu, S.-L. Hsu, H.-F. TienAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): T.-Y.W. Wei, P.-Y. Wu, H. Hou, M.-D. TsaiWriting, review, and/or revision of the manuscript: T.-Y.W. Wei, P.-Y. Wu,H. Hou, W.-C. Chou, C.-T.R. Yu, M.-D. TsaiAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): T.-J. Wu, H. Hou, C.-R. Lin, J.-M.M. Chen,H.-C. Su, C.-T.R. Yu, S.-L. HsuStudy supervision: P.-Y. Wu, H. Hou, M.-D. Tsai

AcknowledgmentsThe authors thank the Taipei Blood Center of Taiwan Blood Service Foun-

dation for providing blood samples of the healthy donors, Drs. Meng-Ru Hoand Shu-Chuan Jao of the Biophysics Core Facility and Dr. Chin-Chun Hung ofthe Imaging andCell Biology Facility (all at the Institute of Biological Chemistry,Academia Sinica) for providing instrumental services and experimental sugges-tions, Yan-Ling Peng of National Taiwan University Hospital and Chia-HuiChen of Chang Gung Memorial Hospital for experimental assistances,Drs. Li-Jung Juan and Shui-Tein Chen of Academia Sinica for providingleukemic lines, Dr. Ling-Pai Tin of National Yang-Ming University for providingpNF-kB-Luc plasmid, and Dr. Ruey-Hwa Chen of Academia Sinica for criticalcomments of the manuscript.

Grant SupportThis work was supported by grant NHRI-EX104-10002NI from National

Health Research Institute, the Academia Sinica Investigator Award, and theTaiwan Protein Project (grant no. MOST105-0210-01-12-01 to M.-D. Tsai),grant CMRPG3E0471 from Chang Gung Memorial Hospital (T.-J. Wu), and inpart by grants MOST 103-2628-B-002-008-MY3, 103-2923-B-002-001, 104-2314-B-002-128-MY4, and 105-0210-01-12-01 from the Ministry of Scienceand Technology (Taiwan), MOHW103-TD-B-111-04, from the Ministry ofHealth and Welfare (Taiwan), and NTUH102P06 from the Department ofMedical Research, National Taiwan University Hospital (H.-F. Tien, H.-A. Hou,and W.-C. Chou).

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received April 20, 2016; revised September 28, 2016; accepted October 14,2016; published OnlineFirst November 10, 2016.

References1. Estey E, Dohner H. Acute myeloid leukaemia. Lancet 2006;368:

1894–907.2. Appelbaum FR, Gundacker H, Head DR, Slovak ML, Willman CL, Godwin

JE, et al. Age and acute myeloid leukemia. Blood 2006;107:3481–5.3. Komanduri KV, Levine RL. Diagnosis and therapy of acute myeloid

leukemia in the era of molecular risk s. Ann Rev Med 2016;67:59–72.4. Lai C, Karp JE, Hourigan CS. Precision medicine for acute myeloid leuke-

mia. Expert Rev Hematol 2016;9:1–3.5. Hayden MS, Ghosh S. NF-kappaB, the first quarter-century: remarkable

progress and outstanding questions. Genes Dev 2012;26:203–34.6. Hoesel B, Schmid JA. The complexity of NF-kappaB signaling in inflam-

mation and cancer. Mol Cancer 2013;12:86.

7. ChaturvediMM, Sung B, Yadav VR, Kannappan R, Aggarwal BB.NF-kappaBaddiction and its role in cancer: 'one size does not fit all'. Oncogene2011;30:1615–30.

8. Tergaonkar V, Pando M, Vafa O, Wahl G, Verma I. p53 stabilization isdecreased uponNFkappaB activation: a role forNFkappaB in acquisition ofresistance to chemotherapy. Cancer Cell 2002;1:493–503.

9. Mehta SV, Shukla SN, Vora HH. Overexpression of Bcl2 protein predictschemoresistance in acute myeloid leukemia: its correlation with FLT3.Neoplasma 2013;60:666–75.

10. Zhou J, Ching YQ, Chng WJ. Aberrant nuclear factor-kappa B activity inacute myeloid leukemia: from molecular pathogenesis to therapeutictarget. Oncotarget 2015;6:5490–500.






11. Griessinger E, Frelin C, Cuburu N, Imbert V, Dageville C, HummelsbergerM, et al. Preclinical targeting of NF-kappaB and FLT3 pathways in AMLcells. Leukemia 2008;22:1466–9.

12. Braun T, Carvalho G, Fabre C, Grosjean J, Fenaux P, Kroemer G. TargetingNF-kappaB in hematologic malignancies. Cell Death Differ 2006;13:748–58.

13. Carmena M, Earnshaw WC. The cellular geography of aurora kinases. NatRev Mol Cell Biol 2003;4:842–54.

14. Hilton JF, Shapiro GI. Aurora kinase inhibition as an anticancer strategy.J Clin Oncol 2014;32:57–9.

15. Ye D, Garcia-Manero G, Kantarjian HM, Xiao L, Vadhan-Raj S, FernandezMH, et al. Analysis of Aurora kinase A expression in CD34(þ) blast cellsisolated from patients withmyelodysplastic syndromes and acutemyeloidleukemia. J Hematop 2009;2:2–8.

16. Yang J, Ikezoe T, Nishioka C, Nobumoto A, Udaka K, Yokoyama A. CD34(þ)/CD38(-) acute myelogenous leukemia cells aberrantly express Aurorakinase A. Int J Cancer 2013;133:2706–19.

17. Briassouli P, Chan F, Savage K, Reis-Filho JS, Linardopoulos S. Aurora-Aregulation of nuclear factor-kappaB signaling by phosphorylation ofIkappaBalpha. Cancer Res 2007;67:1689–95.

18. Huang XF, Luo SK, Xu J, Li J, Xu DR, Wang LH, et al. Aurora kinaseinhibitory VX-680 increases Bax/Bcl-2 ratio and induces apoptosis inAurora-A-high acute myeloid leukemia. Blood 2008;111:2854–65.

19. Giles FJ, Cortes J, Jones D, Bergstrom D, Kantarjian H, Freedman SJ. MK-0457, a novel kinase inhibitor, is active in patients with chronic myeloidleukemia or acute lymphocytic leukemia with the T315I BCR-ABL muta-tion. Blood 2007;109:500–2.

20. Ea CK, Sun L, Inoue J, Chen ZJ. TIFA activates IkappaB kinase (IKK) bypromoting oligomerization and ubiquitination of TRAF6. Proc Natl AcadSci U S A 2004;101:15318–23.

21. Huang CC, Weng JH, Wei TY, Wu PY, Hsu PH, Chen YH, et al. Intermo-lecular binding between TIFA-FHA and TIFA-pT mediates tumor necrosisfactor alpha stimulation and NF-kappaB activation. Mol Cell Biol2012;32:2664–73.

22. Pan WA, Tsai HY, Wang SC, Hsiao M, Wu PY, Tsai MD. The RNArecognition motif of NIFK is required for rRNA maturation during cellcycle progression. RNA Biol 2015;12:255–67.

23. LinCR,Wei TY, Tsai HY,WuYT,WuPY, Chen ST.Glycosylation-dependentinteraction between CD69 and S100A8/S100A9 complex is required forregulatory T-cell differentiation. FASEB J 2015;29:5006–17.

24. Oke A, Pearce D, Wilkinson RW, Crafter C, Odedra R, Cavenagh J, et al.AZD1152 rapidly and negatively affects the growth and survival of humanacute myeloid leukemia cells invitro and invivo. Cancer Res 2009;69:4150–8.

25. Hou HA, Kuo YY, Liu CY, Chou WC, Lee MC, Chen CY, et al. DNMT3Amutations in acute myeloid leukemia: stability during disease evolutionand clinical implications. Blood 2012;119:559–68.

26. Hou HA, Lin CC, Chou WC, Liu CY, Chen CY, Tang JL, et al. Integrationof cytogenetic and molecular alterations in risk stratification of 318patients with de novo non-M3 acute myeloid leukemia. Leukemia2014;28:50–8.

27. Hou HA, Chou WC, Lin LI, Tang JL, Tseng MH, Huang CF, et al.Expression of angiopoietins and vascular endothelial growth factorsand their clinical significance in acute myeloid leukemia. Leuk Res2008;32:904–12.

28. Vidriales MB, Orfao A, Lopez-Berges MC, Gonzalez M, Lopez-Macedo A,Garcia MA, et al. Light scatter characteristics of blast cells in acute myeloidleukaemia: association with morphology and immunophenotype. J ClinPathol 1995;48:456–62.

29. O'Reilly MS, Holmgren L, Chen C, Folkman J. Angiostatin induces andsustains dormancy of human primary tumors in mice. Nat Med 1996;2:689–92.

30. Gaudet RG, Sintsova A, Buckwalter CM, Leung N, Cochrane A, Li J, et al.INNATE IMMUNITY. Cytosolic detection of the bacterial metaboliteHBP activates TIFA-dependent innate immunity. Science 2015;348:1251–5.

31. Katsha A, SouttoM, Sehdev V, Peng D,WashingtonMK, PiazueloMB, et al.Aurora kinase A promotes inflammation and tumorigenesis in mice andhuman gastric neoplasia. Gastroenterology 2013;145:1312–22 e1–8.

32. Zeijlemaker W, Gratama JW, Schuurhuis GJ. Tumor heterogeneity makesAML a "moving target" for detection of residual disease. Cytometry B ClinCytom 2014;86:3–14.

33. Goldenson B, Crispino JD. The aurora kinases in cell cycle and leukemia.Oncogene 2015;34:537–45.

34. Yuan H, Wang Z, Zhang H, Roth M, Bhatia R, ChenWY. Overcoming CMLacquired resistance by specific inhibition of Aurora A kinase in the KCL-22cell model. Carcinogenesis 2012;33:285–93.

35. Van Etten RA.Aberrant cytokine signaling in leukemia. Oncogene 2007;26:6738–49.

36. Kornblau SM, McCue D, Singh N, Chen W, Estrov Z, Coombes KR.Recurrent expression signatures of cytokines and chemokines are presentand are independently prognostic in acute myelogenous leukemia andmyelodysplasia. Blood 2010;116:4251–61.

37. McCormack E, Bruserud O, Gjertsen BT. Animal models of acute myelog-enous leukaemia - development, application and future perspectives.Leukemia 2005;19:687–706.

38. Riether C, Schurch CM, Ochsenbein AF. Regulation of hematopoietic andleukemic stem cells by the immune system. Cell Death Differ 2015;22:187–98.

39. Kagoya Y, Yoshimi A, Kataoka K, Nakagawa M, Kumano K, Arai S, et al.Positive feedback betweenNF-kappaB and TNF-alpha promotes leukemia-initiating cell capacity. J Clin Invest 2014;124:528–42.

40. Lin TY, Wei TY, Li S, Wang SC, He M, Matrin M, et al. TIFA as a crucialmediator for NLRP3 inflammasome. Proc Natl Acad Sci U S A 2016;113:15078–83.

41. Zhou M, Gu L, Zhu N, Woods WG, Findley HW. Transfection of adominant-negative mutant NF-kB inhibitor (IkBm) represses p53-depen-dent apoptosis in acute lymphoblastic leukemia cells: interaction of IkBmand p53. Oncogene 2003;22:8137–44.

42. Corey SJ, Minden MD, Barber DL, Kantarjian H, Wang JC, Schimmer AD.Myelodysplastic syndromes: the complexity of stem-cell diseases. Nat RevCancer 2007;7:118–29.

43. Fabre C, Carvalho G, Tasdemir E, Braun T, Ades L, Grosjean J, et al. NF-kappaB inhibition sensitizes to starvation-induced cell death in high-riskmyelodysplastic syndrome and acute myeloid leukemia. Oncogene 2007;26:4071–83.

44. Kornblau SM, Thall PF, Estrov Z, Walterscheid M, Patel S, Theriault A, et al.The prognostic impact of BCL2 protein expression in acute myelogenousleukemia varies with cytogenetics. Clin Cancer Res 1999;5:1758–66.

45. Tu Y, Renner S, Xu F, Fleishman A, Taylor J, Weisz J, et al. BCL-X expressionin multiple myeloma: possible indicator of chemoresistance. Cancer Res1998;58:256–62.

46. Crusz SM, Balkwill FR. Inflammation and cancer: advances andnew agents.Nat Rev Clin Oncol 2015;12:584–96.

47. Aggarwal BB, Gupta SC, Kim JH. Historical perspectives on tumor necrosisfactor and its superfamily: 25 years later, a golden journey. Blood2012;119:651–65.

48. Kuett A, Rieger C, Perathoner D, Herold T, Wagner M, Sironi S, et al. IL-8 asmediator in the microenvironment-leukaemia network in acute myeloidleukaemia. Sci Rep 2015;5:18411.

49. Dhawan P, Richmond A. Role of CXCL1 in tumorigenesis of melanoma. JLeukoc Biol 2002;72:9–18.

50. Lucena-Araujo AR, de Oliveira FM, Leite-Cueva SD, dos Santos GA, FalcaoRP, Rego EM. High expression of AURKA and AURKB is associated withunfavorable cytogenetic abnormalities and high white blood cell count inpatients with acute myeloid leukemia. Leuk Res 2011;35:260–4.

51. Seymour JF, Kim DW, Rubin E, Haregewoin A, Clark J, Watson P, et al. Aphase 2 study of MK-0457 in patients with BCR-ABL T315Imutant chronicmyelogenous leukemia and philadelphia chromosome-positive acutelymphoblastic leukemia. Blood Cancer J 2014;4:e238.

52. Giles FJ, Swords RT,Nagler A,Hochhaus A,OttmannOG, Rizzieri DA, et al.MK-0457, an Aurora kinase and BCR-ABL inhibitor, is active in patientswith BCR-ABL T315I leukemia. Leukemia 2013;27:113–7.

53. Okabe S, Tauchi T, Ohyashiki K. Efficacy of MK-0457 and in combinationwith vorinostat against Philadelphia chromosome positive acute lympho-blastic leukemia cells. Ann Hematol 2010;89:1081–7.

54. Vader G, Lens SM. The Aurora kinase family in cell division and cancer.Biochim Biophys Acta 2008;1786:60–72.


Wei et al.




2017;77:494-508. Published OnlineFirst November 10, 2016.Cancer Res Tong-You Wade Wei, Pei-Yu Wu, Ting-Jung Wu, et al. Acute Myeloid Leukemia via the TRAF-Interacting Protein TIFA

B Survival Pathway Drive Chemoresistance inκAurora A and NF-

Updated version

10.1158/0008-5472.CAN-16-1004doi:

Access the most recent version of this article at:

Material

Supplementary

http://cancerres.aacrjournals.org/content/suppl/2016/11/10/0008-5472.CAN-16-1004.DC1

Access the most recent supplemental material at:

Cited articles

http://cancerres.aacrjournals.org/content/77/2/494.full#ref-list-1

This article cites 54 articles, 17 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/77/2/494.full#related-urls

This article has been cited by 4 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cancerres.aacrjournals.org/content/77/2/494To request permission to re-use all or part of this article, use this link



http://cancerres.aacrjournals.org/lookup/doi/10.1158/0008-5472.CAN-16-1004

http://cancerres.aacrjournals.org/content/suppl/2016/11/10/0008-5472.CAN-16-1004.DC1

http://cancerres.aacrjournals.org/content/77/2/494.full#ref-list-1

http://cancerres.aacrjournals.org/content/77/2/494.full#related-urls

http://cancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://cancerres.aacrjournals.org/content/77/2/494


aurora a and nf-kb survival pathway drive chemoresistance...

Documents