Chromosome Instability Modulated by BMI1–AURKA Signaling Drives Progression in Head and Neck Cancer

Published OnlineFirst November 30, 2012; DOI: 10.1158/0008-5472.CAN-12-2397

Cancer Tumor and Stem Cell Biology Research

Chun-Hung Chou1, Neng-Kai Yang1, Ting-Yun Liu1, Shyh-Kuan Tai7, Dennis Shin-Shian Hsu1, Ya-Wei Chen8, Yann-Jang Chen3, Cheng-Chi Chang9, Cheng-Hwai Tzeng6, and Muh-Hwa Yang1,2,4,5,6

Abstract Chromosomal instability (CIN) is widely considered a hallmark of cancer, but its precise roles in cancer stem cells (CSC) and malignant progression remain uncertain. BMI1 is a member of the Polycomb group of chromatin- modifier proteins that is essential for stem cell self-renewal. In human cancers, BMI1 overexpression drives stem- like properties associated with induction of epithelial–mesenchymal transition (EMT) that promotes invasion, metastasis, and poor prognosis. Here, we report that BMI1 mediates its diverse effects through upregulation of the mitotic kinase Aurora A, which is encoded by the AURKA gene. Two mechanisms were found to be responsible for BMI1-induced AURKA expression. First, BMI1 activated the Akt pathway, thereby upregulating AURKA expression through activation of the b-catenin/TCF4 transcription factor complex. Second, BMI1 repressed miRNA let-7i through a Polycomb complex-dependent mechanism, thereby relieving AURKA expression from let- 7i suppression. AURKA upregulation by BMI1 exerts several effects, including centrosomal amplification and aneuploidy, antiapoptosis, and cell-cycle progression through p53 degradation and EMT through stabilization of Snail. Inhibiting Aurora A kinase activity attenuated BMI1-induced tumor growth in vivo. In clinical specimens of head and neck cancer, we found that coamplification of BMI1 and AURKA correlated with poorer prognosis. Together, our results link CSCs, EMT, and CIN through the BMI1–AURKA axis and suggest therapeutic use from inhibiting Aurora A in head and neck cancers, which overexpress BMI1. Cancer Res; 73(2); 1–14. 2012 AACR.

Introduction an essential role in maintaining self-renewal of stem cells INK4A-ARF – Recently, one of the most important conceptual advances in through suppressing the locus (5 7). In human BMI1 cancer research is the discovery of a small population of cancer cancers, is often overexpressed and acts as an oncogene cells harboring stem cell properties, that is, cancer stem cells to promote tumorigenesis (8), and overexpression of BMI1 is – (CSC). CSCs are critically involved in the initiation, progres- frequently observed in CSCs (9 11). Importantly, the role of – – sion, recurrence, and therapeutic resistance of human cancers BMI1 in epithelial mesenchymal transition (EMT) generated (1, 2). The regulatory mechanisms of CSCs have been exten- CSCs has been highlighted in pancreatic cancer, the EMT sively investigated. Among them, the role of Polycomb group regulator ZEB1 inhibits the expression of the miRNA-200 proteins in promoting stem-like properties of cancer cells has family, leading to an increased expression of BMI1 and been valued significantly (3). The Polycomb group proteins, stem-like properties (12); in squamous cell carcinoma of the including Polycomb repressive complex 1 and 2 (PRC1 and 2), head and neck (HNSCC), BMI1 is essential for Twist1-induced are chromatin modifiers, which regulate the expression of a EMT and stem-like properties (13); in nasopharyngeal carci- PTEN number of genes during stem cell self-renewal, tumor forma- noma, BMI1 represses expression, resulting in Akt tion, and progression (4). BMI1 is a member of PRC1 and plays activation and EMT (14). However, whether BMI1 possesses a more extended function to facilitate cancer progression beyond INK4A-ARF suppression and EMT-generated CSCs, is Authors' Affiliations: Institutes of 1Clinical Medicine, 2Biotechnology in unclear. 3 4 Medicine, and Genome Sciences; Head And Neck Cancer Research Chromosomal instability (CIN) is the inability to maintain a Program, Cancer Research Center; 5Genomic Research Center, National Yang-Ming University; 6Division of Hematology-Oncology, Department of correct chromosome complement after mitosis. The chromo- Medicine; Departments of 7Otolaryngology and 8Stomatology, Taipei some aberrations in tumor cells have been discovered for Veterans General Hospital; and 9Graduate Institute of Oral Biology, School of Dentistry, National Taiwan University, Taipei, Taiwan decades, and CIN has been considered to be one of the hall- marks of tumor formation (15). However, the understanding of Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/). CIN in tumor progression is relatively limited. Emerging evidence suggests the involvement of CIN in cancer progression. Corresponding Author: Muh-Hwa Yang, Institute of Clinical Medicine, National Yang-Ming University, No. 155, Sec. 2, Li-Nong Street, Taipei 112, CIN is associated with a poor prognosis in patients with lung Taiwan. Phone: 886-228267000, ext. 7911; Fax: 886-228235870; E-mail: and colon cancer (16, 17). Furthermore, overactivation of the [email protected] mitotic checkpoint, a major mechanism for CIN acquisition, doi: 10.1158/0008-5472.CAN-12-2397 has been shown to inhibit certain tumor suppressive signal 2012 American Association for Cancer Research. pathways (15). However, the molecular link between CIN and

www.aacrjournals.org OF1

Chou et al.

tumor progression, and whether CIN is related to CSCs, remain mutant construct was generated by mutating the TCF4-bind- largely elusive. ing site on the AURKA promoter. The full-length 30-untrans- Aurora A is a serine/threonine kinase that regulates mitotic lated region (UTR) of AURKA was cloned into the pMIR- processes in mammalian cells, including centrosome matura- REPORTER to generate the pMIR-AURKA-wt construct, and tion, spindle assembly, and chromosome segregation (18, 19). the putative let-7i-binding site was mutated to generate pMIR- Amplification of AURKA has been shown in various types of AURKA-mut. A plasmid expressing the renilla luciferase gene human cancers (20), and overexpression of Aurora A promotes (pRL-TK) was cotransfected into each transfection experiment CIN, cell-cycle progression, and therapeutic resistance (20, 21). as a control for transfection efficiency. Cells were harvested In addition, Aurora A phosphorylates p53, which leads to after 48 hours of transfection, and luciferase activity was ubiquitination and degradation of p53 and suppresses apo- assayed. The relative promoter activities were expressed as ptosis (22). In this study, we discover a novel link between BMI1 the fold-change of luciferase activity after normalization to and Aurora A through 2 PRC-mediated signal pathways. In renilla luciferase activity. HNSCC, BMI1 expression results in multifaceted changes through regulating Aurora A, including centrosomal amplifi- miRNAs microarray analysis cation/aneuploidy, cell-cycle progression, antiapoptosis, and The Agilent human miRNA array (V2) was conducted in EMT. All these events act collaboratively to facilitate tumor FaDu cells transfected with a BMI1-expressing vector (FaDu- progression. BMI1) or a control vector. The microarray data have been deposited in the National Center for Biotechnology Informa- Materials and Methods tion (NCBI) Gene Expression Omnibus (GEO) database with Cell lines and plasmids the accession number of GSE 31910. The miRNA microarray The human hypopharyngeal cancer cell line FaDu and the data for OECM1-sh-BMI1 versus OECM1-sh-scr have been human embryonic kidney cell line HEK-293T were obtained published (24) and deposited in the NCBI GEO database with from the Bioresource Collection and Research Center of Tai- the accession number of GSE 29586. wan. The human oral cancer cell line OECM-1 was originally from Dr. Ching-Liang Meng of National Defense Medical Immunofluorescence staining College (Taipei, Taiwan; ref. 23). The human oral cancer cell Cells were seeded in poly-L-lysine–coated slides and fixed lines SAS and CAL-27 were from Dr. Cheng-Chi Chang's with 4% paraformaldehyde, permeabilized with 0.5% Triton X- laboratory (National Taiwan University, Taipei, Taiwan). The 100, and blocked with 10% FBS. The primary antibody used was pCDH-BMI1 plasmid was generated by inserting a 981-bp against g-tubulin for evaluating the number of centrosomes, or fragment of the full-length human BMI1 cDNA into the E-cadherin for evaluating the adherent junctions. The second- NheI/BamHI sites of the pCDH lentiviral vector. pmCherry- ary antibody used was fluorescein isothiocyanate (FITC)–con- let7i, pmCherry-spg-ctrl, and pmCherry-spg-let7i were previ- jugated against mouse immunoglobulin G (IgG; Cat# AP124F, ously described (24). The pcDNA3–b-catenin was generated by Millipore Corp.) for detecting g-tubulin, and rhodamine-con- cloning the coding sequence of b-catenin into pcDNA3.1 jugated against rabbit IgG (Cat# AP132R, Chemicon Interna- vector. The pcDNA3-AURKA was generated by cloning the tional Inc.) for detecting E-cadherin. 40,6-Diamidino-2-pheny- coding sequence of AURKA into pCDH-GPF-puro vector. The lindole (DAPI) was used for nuclear staining. The slide images dominant-negative TCF4 plasmid pPGS-dnTcf4(DN31) and were captured on an Olympus 1000i (Olympus Corporation). pcDNA3-HA-GSK-3b was purchased from Addgene. The plasmids for short hairpin RNA (shRNA) experiments were gen- Karyotype analysis erated by inserting a specific shRNA target sequence or a Cells were treated with colcemid at 37C for 2 hours. Cells scrambled sequence into the pSUPER.puro vector. The were harvested, swollen in warm 0.075 mol/L KCl, fixed in cold sequences used to generate the plasmid are listed in Supple- 3:1 methanol/glacial acetic acid, dropped onto slides, and dried mentary Table S1. Stable cell lines were generated by trans- at room temperature. The chromosome images were captured fection/infection of expression or shRNA plasmids into parietal on the Olympus BX51 High Class System Microscope (Olympus cell lines and selected using the appropriate antibiotics. Corporation).

þ Sorting of CD44 HNSCC cells Cell-cycle analysis þ For CD44 cell isolation, cells were suspended in 100 mL1 Cells were treated with the Aurora kinase inhibitor III (Cat# PBS and stained with 5 mL CD44-AlexaFluor488 (Clone 156- C1368, Sigma-Aldrich Corp.) at 3 mmol/L for 24 hours. The cells 3C11, Cat# 3516, Cell Signaling Technology, Inc.), or Alexa- were fixed with 75% ethanol at 20C overnight and were Fluor488 mouse IgG2a (Clone MOPC-173, Cat# 558055, BD stained in propidium iodide (PI; 50 mg/mL; Sigma-Aldrich Biosciences) as isotype control. The stained cells were sepa- Corp.), RNase A (0.02 mg/mL), and 1 PBS at 37C for 30 rated by FACSAria (BD Biosciences). minutes in the dark. The stained cells were analyzed by flow cytometry. Luciferase reporter assay The wild-type human AURKA firefly luciferase promoter Apoptosis analysis construct (pGL-1486) was a gift from Dr. Ishigatsubo (Yoko- For evaluating the effect of Aurora kinase inhibition on hama City University, Yokohama, Japan; ref. 25). The pGL-1486 apoptosis, cells were treated with 5 mmol/L Aurora kinase

OF2 Cancer Res; 73(2) January 15, 2013 Cancer Research

BM11-AURKA Axis in Cancer Progression

A B OECM1- FaDu-BMI1 OECM1-sh-BMI1 OECM-1- sh-BMI1 /FaDu-CDH /OECM1-sh-scr FaDu- FaDu- CDH BMI1 sh-scr 1 2 AURKA 3.0 CENPE BMI1 BMI1 AURKB BUB1 Aurora A Aurora A BUB3 β PLK4 β-Actin -Actin FZR1 OECM1-sh-scr MAD1L1 0 5 FaDu-CDH 1.5 FaDu-BMI1 OECM1-sh-BMI1-1 MAD2L1 4 OECM1-sh-BMI1-2 PLK1 3 1 BUB1B 2 NDC80 0.5 protein level protein level 1

PTTG1 Fold-change of Fold-change of CCNB1 0 0 CDC20 -3.0 BMI1 Aurora A BMI1 Aurora A

8 FaDu C D OECM-1 6 CAL-27 SAS 4

mRNA level 2 Fold-change of 0 BMI1 AURKA FaDu OECM-1 CAL-27 SAS BMI1 CDKN2A BMI1 AURKA

-3.0 0 3.0 Aurora A

β-Actin E F FaDu CD44- 10 Isotype CD44-AlexaFluor488 FaDu CD44+ 250 250

(X 1,000) ＊＊ (X 1,000) 200 200 ＊＊

CD44- 6 150 150

＊＊

100 FSC-A 100 4 FSC-A FSC-A

50 50 CD44+

2 0 0 Fold-change of mRNA level 0 102 10 3 10 4 10 5 0 102 10 3 10 4 10 5 0 FITC-A FITC-A CD44 BMI1 AURKA

Figure 1. BMI1 upregulates Aurora A in cancer cells. A, a heatmap summarizing the results of relative mRNA levels of the mitotic-related genes in FaDu cells transfected with a BMI1-expressing vector (FaDu-BMI1) versus control vector (FaDu-CDH), or OECM-1 cells transfected with a shRNA against BMI1 (OECM1-sh-BMI1) versus a scrambled sequence (OECM1-sh-scr). B, top, Western blot analysis of BMI1 and Aurora A in FaDu and OECM-1 clones. Bottom, quantification of the Western blot analysis results. C, correlation between the relative expression levels of BMI1, CDKN2A,andAURKA in cancer cell lines from the NCI-60 panel. D, top, relative mRNA levels of BMI1 and AURKA in HNSCC cell lines. þ Bottom, Western blot analysis of BMI1 and Aurora A in HNSCC cell lines. E, representative result of flow cytometry for sorting the CD44 cells. F, relative mRNA levels of CD44, BMI1,andAURKA in CD44þ versus CD44 FaDu cells. FSC-A, forward scatter. In each experiment, data represent mean SEM (n ¼ 3). , P < 0.01. inhibitor III (AKI III; Cat# C1368, Sigma-Aldrich Corp.) or In vivo drug sensitivity assay the vehicle control for 48 hours. Then the cells were sus- All animal protocols were carried out in accordance with the pended in 100 mL1 binding buffer containing 10 mmol/L institutional animal welfare guidelines of Taipei Veterans 7 HEPES at pH 7.4, 140 mmol/L NaCl, and 2.5 mmol/L CaCl2, General Hospital (Taipei, Taiwan). A total of 1 10 of each and stained with 5 mL allophycocyanin-conjugated Annexin stable cell lines were injected into the subcutaneous area of the V (Cat# 550475, BD Biosciences) and 5 mLof50mg/mL PI 6-week-old BALB/C nude mice. After the tumor volume for 15 minutes. The stained cells were analyzed by flow reached 200 mm3, we started to treat the mice with the Aurora cytometry. kinase inhibitor VX-680 at the dose of 0, 25, or 50 mg/kg, 4 times

www.aacrjournals.org Cancer Res; 73(2) January 15, 2013 OF3

Chou et al.

ACB OECM1- FaDu- FaDu- sh-BMI1 CDH BMI1 OECM1- 3 Phospho-Akt (ser473) DMSO + + - Aurora kinase A A sh-scr 1 2 LY294002 -- + 2 BMI1 BMI1

Phospho-Akt Phospho-Akt (ser 473) 1 (ser 473) Total Akt Total Akt

Fold-change of protein Fold-change of level 0 Phospho-GSK-3β DMSO +123 + - Phospho-GSK-3β (ser 9) LY294002-+ - (ser 9) FaDu-CDH FaDu-BMI1 Total GSK-3β Total GSK-3β FaDu β-Catenin D pcDNA3- β-Catenin EV β-catenin Aurora A Aurora A β-Catenin

β-Actin β-Actin Aurora A

β-Actin

E pGL-1486 pGL-1486mut AURKA promoter Luc AURKA promoter Luc

- 1486 +1 +354 - 1486 +1 +354 -13 -7 -13 -7

TCF4-binding site CTTTGAA TCF4-binding site CTTTGAA

Mutation AGGGTCC FG 3.5 2.5 3 ** ** 2 2.5 2 1.5

1.5 1 fold-change of

fold-change of 1 firefly luc/renilla luc firefly luc/renilla firefly luc/renilla luc 0.5 0.5 0 0 pcDNA3.1 +-- +- pcDNA3.1 +-- +- pcDNA3-β-catenin -++ -+ pcDNA3-BMI1 -++ -+ pPGS-dnTcf4(Δ31) --+ --pPGS-dnTcf4(Δ31) --+ --

pGL-1486 pGL-1486mut pGL-1486 pGL-1486mut

Figure 2. Activation of Akt pathway by BMI1 enhances AURKA promoter activity through b-catenin/TCF-4 complex. A and B, Western blot analysis of BMI1, total Akt, phosphorylated Akt (serine 473), total GSK-3b, phosphorylated GSK-3b (serine 9), b-catenin, and Aurora A in OECM1-sh-BMI1 versus OECM1-sh- scr (A) and FaDu-CDH versus FaDu-BMI1 (B) with or without LY294002 treatment. Dimethyl sulfoxide (DMSO) was a vehicle control. C, quantiﬁcation of the Western blot analysis results in B. D, Western blot analysis of b-catenin and Aurora A in FaDu cells transfected with pcDNA3–b-catenin or an empty vector (EV). E, schematic representation of the wild-type (left) or TCF4-binding site–mutated (right) AURKA promoter construct used in luciferase reporter assay. þ1 indicates the transcription start site. F and G, luciferase reporter assay. Data represent mean SEM (n ¼ 3). , P < 0.05.

OF4 Cancer Res; 73(2) January 15, 2013 Cancer Research

BM11-AURKA Axis in Cancer Progression

A B

2 8 FADU-PCDHFaDu-CDH OECM-1OECM-1-sh-scr / sh-SCR MIRNAs decreased MIRNAs increased FADU-BMI1FaDu-BMI1 OECM-1OECM-1-sh-BMI1 / sh-BMI1 1.5 6 ** in FaDu-BMI1, 2 in OECM1-sh-BMI1, ** fold-change <0.8: fold-change >1.4 1 4 (n = 4) (n = 9) 0.5 2 Fold-change of MIRNA level let-7i, MIR-15b Fold-change of MIRNA level 0 0 let-7ilet-7i MIR-15b MIr-15b let-7ilet-7i MIr-15bMIR-15b C Promoter of MIRLET7I D qChIP qChIP PRE1 PRE2 15 FaDu-PCDHFaDu-CDH 10 FaDu-PCDHFaDu-CDH +1 FaDu-BMI1FaDu-BMI1 FaDu-BMI1FaDu-BMI1 *** 8 PRE1 PRE2 *** 10 *** *** 6 *** *** Homo sapiens 5’-AGTCAGCCTTCTC-3’AGCCTT 5’-GCTATGCCTTGGC-3’GCCTTG Mesocricetus 5’-AAAAAGCCTTTAA-3’AGCCTTT 5’-TGACCGCCTTTCC-3’GCCTTT 4 auratus 5 Gallus gallus 5’-CAAGCGCCTTCAG-3’CGCCTT 5’-CAGCGGCCTTCTG-3’GGCCTT 2 domestica Percentage of input Percentage of input Sus scrofa 5’-AAGTTGCCTTGCC-3’TGCCTT 5’-ACTTGGCCTTTAA-3’GCCTTT domestica 0 0 IgG BMI1 H3K27me3EZH2 IgG BMI1 H3K27me3E Z H2 E F

5 2 let-7i let-7i FaDu- FaDu- 4 OECM1- OECM1- 1.5 ** mCherry let7i ** spg-ctrl spg-let7i 3 1 2 Aurora A Aurora A 0.5 1 β-Actin β-Actin

0 Fold-change of MIRNA level 0 Fold-change of MIRNA level OECM1-mCherry OECM1-let7i FaDu- FaDu- spg-ctrl spg-let7i

G H 1.5 ~ (+1) (+268) (+1480) (+2111 +2117) * TSS ATG TGA let-7i-binding site 1

0.5 let-7i consensus 3’ U–U–GUCGUGU–U–UGAUGAUGGAGU 5’ Fold-change of Fold-change of

AURKA-3’UTR - wt 5’ ATAGGAACACGTGCTC TACCT CCATTT 3’ luc firefly luc / renilla AURKA-3’UTR - mut 5’ ATAGGAAC ACGT GCTC–A–––T––––TT–3’ 0 pmCherry mcherry + mcherry - mcherry + mcherry - pmCherry-let7i - /Let-7i + - /Let-7i +

pMIR-AURKA-wt pMIR-AURKA-mut I FaDu - BMI1 FaDu- CDH EV let-7i

BMI1

Aurora A

β-Actin

Figure 3. Suppression of let-7i by BMI1 upregulates Aurora A expression. A, schema for identifying the BMI1-downregulated miRNAs. B, expression of let-7i and miR-15b in FaDu and OECM-1 clones. C, organization of the MIRLET7I promoter and sequence alignment of the 2 PREs (PRE1 and PRE2). qChIP indicates the ampliﬁed sequences in qChIP. þ1 indicates the transcription start site. D, qChIP assay. E, left, expression of let-7i in OECM-1 cells transfected with a let-7i–expressing vector (OECM1-let7i) or a control vector (OECM1-mCherry). Right, Western blot analysis of Aurora A. F, left, expression of let-7i in FaDu cells transfected with a sponge vector for neutralizing let-7i (FaDu-spg-let-7i) or a control vector (FaDu-spg-ctrl). Right, Western blot analysis of Aurora A. G, schematic representation of the wild-type or let-7i-binding site–mutated 30-UTR reporter constructs of AURKA.H,30-UTR luciferase reporter assay. I, Western blot analysis of BMI1 and Aurora A in FaDu-CDH, FaDu-BMI1 transfected with an empty vector (EV) or a let-7i–expressing vector. In each experiment, data represent mean SEM (n ¼ 3). , P < 0.05; , P < 0.01; , P < 0.001. www.aacrjournals.org Cancer Res; 73(2) January 15, 2013 OF5

Chou et al.

A γ-tubulin DAPI merge B FaDu- FaDu- FaDu-BMI1 FaDu-BMI1 CDH BMI1 -sh-scr - sh-AURKA

FaDu-CDH 20µm 20µm 20µm

- sh scr 20µm 20µm 20µm N = 46 N = 43 N = 42 N = 46 FaDu-BMI1 FaDu-BMI1

120

20µm 20µm 20µm * FaDu-BMI1 FaDu-BMI1 100 ** - sh - AURKA - sh AURKA

100 FaDu-CDH 80 FaDu-BMI1-sh-scr 80 FaDu-BMI1-sh-AURKA 60 60 aberration, % 40 40 ** Cells with chromosome 20 20

Percentage of cells, % Percentage of cells, 0 0 FaDu- FaDu- FaDu-BMI1 FaDu-BMI1 Centrosome -sh-scr - sh-AURKA 1 centr1 2 centr2 > 2>2 centr CDH BMI1 number

C D FaDu-CDH FaDu-BMI1 3 FaDu–BMI1 5.93 2.08 4.41 1.58 ** FaDu- 2.5 CDH sh-scr sh-AURKA 2 BMI1 1.5 89.55 2.44 92.51 1.50 1 Percentage of Aurora A 0.5 early apoptotic cells, % early apoptotic cells, 0 p53 FaDu-CDH FaDu-BMI1

FaDu-BMI1 + DMSO FaDu-BMI1 + AKl lll 30 β-Actin 3.17 1.44 2.95 6.85 *** MG132 + + + 20 phospho-p53 (ser315) 93.81 1.58 67.87 10 Percentage of β-Actin 22.33 early apoptotic cells, % 0 FaDu-BMI1 FaDu-BMI1 + DMSO + AKI III E 160 160 160 FADU-CDH FADU-BMI1- FADU-BMI1+ sh-scr DMSO 120 FADU-BMI1 120 120 FADU-BMI1- FADU-BMI1+ 80 80 sh-AURKA 80 AKI III Events Events Events 40 40 40

0 0 0 PI 1023 PI 1023 PI 1023 FADU-CDH FADU-BMI1-sh-scr FADU-BMI1 + DMSO FADU-BMI1 FADU-BMI1-sh-AURKA FADU-BMI1 + AKI III 80 2 80 3 80 8 ** *** 60 1.5 * 60 *** 60 6 ** 2 ** ** *** 40 ** 1 40 ** 40 ** 4 1 % Cells % Cells % Cells 20 0.5 20 20 2

0 0 0 0 0 0

G0–G1 S G2–M sub-G1 G0–G1 S G2–M sub-G1 G0–G1 S G2–M sub-G1

OF6 Cancer Res; 73(2) January 15, 2013 Cancer Research

BM11-AURKA Axis in Cancer Progression

a week for 3 weeks. The mice were sacrificed after the 3-week cells attenuated both BMI1 and Aurora A (Supplementary treatment, and the xenotransplanted tumors were collected for Fig. S1B). analysis. Terminal deoxynucleotidyl transferase–mediated To further confirm the relationship between AURKA and dUTP nick end labeling (TUNEL) assay by using the In Situ BMI1, we investigated the correlation between AURKA and Cell Death Detection Kit (No. 11-684-817-910, Roche Applied BMI1 in microarray datasets of the NCI-60 panel (27). We Science) was applied for detecting the apoptotic cells in selected the cell lines from the NCI-60 panel in which the tumors. expression of BMI1 was inversely correlated with that of CDKN2A, a known target suppressed by BMI1 (correlation fi r fi Immunohistochemistry coef cient < 0.7), indicating the functional signi cance of Immunohistochemistry (IHC) for detecting BMI1 and Auro- the BMI1 in these cell lines. Then, we observed the correlation ra A was conducted, and the slides were independently scored between BMI1 and AURKA. Among these cell lines, the expres- BMI1 AURKA by 2 individuals according to the immunoreactive score (IRS; sion of was positively correlated with that of fi r ¼ P ref. 26). For BMI1, only nuclear expression was considered (correlation coef cient 0.422; < 0.05; Fig. 1C). Because the positive, whereas both nuclear and cytoplasmic expression NCI-60 panel does not contain HNSCC cell lines, we compared were considered positive for Aurora A. A moderate to strong the mRNA and protein levels of BMI1 and Aurora A in 4 HNSCC level (IRS > 4) was considered positive, and weak or absent cell lines. Among these cell lines, OECM-1 and SAS cells had a fi expression (IRS, 0–4) was considered negative. signi cant higher expression of BMI1 and Aurora A in both Please see Supplementary Methods for the other methods mRNA and protein levels. In contrast, FaDu and CAL-27 had a lower expression of BMI1 and Aurora A (Fig. 1D). Furthermore, used in this study. þ BMI1 and AURKA were coamplified in CD44 population of the HNSCC cell lines and a primary HNSCC culture (Fig. 1E Results and F; Supplementary Fig. S1C and S1D), indicating the acti- BMI1 upregulates the expression of Aurora A and both of vation of BMI1–AURKA signal in HNSCC stem-like cells. Taken them are coamplified in HNSCC stem-like cells together, the earlier findings indicate that in HNSCC, BMI1 We started this study by aiming to understand whether the upregulates the expression of Aurora A, and both of them are expression of the mitosis/CIN–related genes could be regulat- coamplified in the stem-like population. ed by the "stemness factor" BMI1, to provide a molecular connection between CSCs and CIN. For the major role of BMI1 BMI1 increases AURKA through activating in the CSCs of HNSCC (11, 13), we used HNSCC cell lines as the Akt–b-catenin pathway platform to investigate the mechanism. To this end, 2 HNSCC Next, we sought to find out the mechanism about how BMI1 cell lines were applied for manipulating the expression of BMI1 regulates AURKA expression. Because BMI1 suppresses target and examined the change of the expression of mitosis-related genes expression through chromatin silencing (28, 29), we genes. FaDu cells were selected for ectopic expression of BMI1 speculated that BMI1 indirectly enhances Aurora A expression. (FaDu-BMI1) because of the low endogenous BMI1 level, and BMI1 was shown to suppress PTEN and activate Akt pathway OECM-1 cells was selected as the parental cell line to generate (14). Furthermore, b-catenin, an oncoprotein that is phosphor- the BMI1-knockdown clones (OECM1-sh-BMI1) owing to the ylated and destabilized by glycogen synthase kinase-3b (GSK- high endogenous BMI1 expression levels (24). Quantitative 3b; ref. 30), was reported to upregulate AURKA (31). We reverse transcription PCR (qRT-PCR) analysis of a panel of therefore hypothesized that BMI1 induces the expression of mitosis/CIN–related genes (AURKA, AURKB, BUB1, BUB1B, AURKA through activation of Akt, which in turns results in the BUB3, CDC20, FZR1, CENPE, CCNB1, NDC80, MAD1L1, PTTG1, phosphorylation and inactivation of GSK-3b, leading to accu- PLK1, and PLK4; with reference to ref. 15) showed that AURKA mulation of b-catenin and transactivation of AURKA.To was the only candidate both upregulated in FaDu-BMI1 and confirm this notion, we examined the change of phospohry- downregulated in OECM1-sh-BMI1 (Fig. 1A). The protein level lated Akt, phosphorylated GSK-3b, b-catenin, and Aurora A in of Aurora A was also upregulated in FaDu-BMI1 and down- BMI1 knocked down OECM-1 cells versus control cells, and regulated OECM1-sh-BMI1 (Fig. 1B). Because BMI1 is directly BMI1-overexpressed FaDu cells versus control cells. Knock- regulated by Twist1 (13), we examined if Twist1 upregulates down of BMI1 in OECM-1 reduced the phosphorylation of Akt Aurora A. The result showed that ectopic Twist1 augmented at serine 473 and GSK-3b at serine 9, and decreased the the expression of BMI1 as well as Aurora A in FaDu cells expression of b-catenin and Aurora A (Fig. 2A). Consistently, (Supplementary Fig. S1A), and repression of Twist1 in OECM-1 overexpression of BMI1 in FaDu cells enhanced Akt and GSK-

Figure 4. BMI1–AURKA axis promotes CIN, antiapoptosis, and cell-cycle progression. A, top, representative results of immunoflouresent staining in FaDu clones. The arrows indicate centrosomes. Scale bar, 20 mm. Bottom, quantification of centrosome analysis (100 cells were counted for each clone). B, top, representative pictures of karyotype analysis in FaDu clones. The number of chromosomes was shown in each panel. Bottom, quantification of the percentage of cells with aneuploidy (30 cells were counted in each experiment to determine the percentage of aneuploidy). C, top, Western blot analysis of BMI1, Aurora A, and p53 in FaDu clones. Bottom, Western blot analysis for detecting p53 serine 315 phosphorylation in the above cells treated with MG132. D, left, þ representative results of flow cyotmetry for detecting annexin V or/and PI-positive cells. Right, quantification of the early apoptotic cells (annexin V PI ). E, top, flow cytometry for cell-cycle analysis. Bottom, histograms showing the results of cell-cycle analysis. In each experiment, data represent mean SEM (n ¼ 3). , P < 0.05; , P < 0.01.

www.aacrjournals.org Cancer Res; 73(2) January 15, 2013 OF7

Chou et al.

A BC5 4 ** ** FaDu–BMI1 FaDu- E-cadherin DAPI Merge 3 CDH sh-scr sh-AURKA 2 1 MIgrated cells MIgrated BMI1 Fold-change of 0 FaDu- FaDu-BMI1- FaDu-BMI1- CDH123sh-scr sh-AURKA

Aurora A FaDu-CDH

E-cadherin 5 ** ** 4 3 γ-Catenin 2 -sh-scr 1 FaDu-BMI1 MIgrated cells MIgrated N-cadherin Fold-change of 0 FaDu-CDH DMSO AKI III Vimentin FaDu-BMI1 1503 1252.5 ** β-Actin 2 sh-AURKA 100 FaDu-BMI1- 751.5 501 250.5 invaded cells Fold-change of 00 FaDu-CDH FaDu-BMI1- FaDu-BMI1- D F sh-scr sh-AURKA FaDu–BMI1 300 3 FaDu- 2502.5 * ** CDH sh-scr sh-AURKA 200 2 1501.5 100 1 BMI1 FaDu-BMI1- FaDu-BMI1- 50invaded cells 0.5 FaDu-CDH sh-scr sh-AURKA Fold-change of 0 0 CHX Aurora A FaDu-CDH DMSO AKI III (100 μg/mL) 0 2 4 0 2 4 0 2 4 hr FaDu-BMI1 Snail BMI1 G FaDu–BMI1 β-Actin FaDu- Aurora A CDH sh-scr sh-AURKA

BMI1 E Snail 2 SNAI1 Aurora A 1.5 β-Actin

1 Phospho-GSK-3β (ser 9)

mRNA level mRNA 0.5 Fold-change of Total GSK-3β 0 FaDu- sh-scr sh-AURKA CDH FaDu–BMI1 β-Actin

H I Cell line : OECM-1 Cell line : HEK-293T transfected with Cell line : HEK-293T transfected with pCDH-AURKA pcDNA3-HA-GSK3β IP IP IP β Input IgG Aurora A Input IgG Aurora A Input IgG GSK-3 IB: phosph-GSK-3β IB: Aurora A IB: total (ser 9) GSK-3β

IB: total IB: Aurora A IB : Aurora A GSK-3β

Figure 5. Aurora A contributes to BMI1-induced EMT through stabilizing Snail. A, Western blot analysis of BMI1, Aurora A, epithelial markers (E-cadherin and g-catenin), and mesenchymal markers (N-cadherin and vimentin) in FaDu clones. B, immunoﬂuorescent staining of FaDu clones. Scale bar, 20 mm. C, top 2

OF8 Cancer Res; 73(2) January 15, 2013 Cancer Research

BM11-AURKA Axis in Cancer Progression

3b phosphorylation and increased b-catenin and Aurora A Polycomb responsive element (PRE; ref. 33), we analyzed the expression. Inhibition of the phosphoinositide 3-kinase (PI3K)/ sequence of let-7i promoter and 2 highly conserved PREs were Akt pathway by a PI3K inhibitor (LY294002) partially attenu- found (Fig. 3C). Quantitative chromatin immunoprecipitation ated the BMI1-induced Aurora A upregulation (Fig. 2B and C). (qChIP) assays showed the enrichment of the binding of BMI1 Ectopic b-catenin upregulated Aurora A in FaDu cells (Fig. 2D). as well as EZH2, a PRC2 subunit that methylates lysine 27 of These results indicate that BMI1 upregulates Aurora A through histone 3 (H3K27; ref. 34), and trimethylated H3K27 activation of the Akt–b-catenin pathway. (H3K27me3) on PREs of let-7i promoter (Fig. 3D). These results For T-cell factors (TCF) are the major binding partners of indicated that BMI1 represses let-7i expression through the b-catenin to activate target genes transcription (32), we inves- PRC-mediated chromatin silencing. tigated whether BMI1 activates AURKA promoter through the We next aimed to determine whether AURKA is a target of b-catenin/TCF4 complex. We first confirmed the transactiva- let-7i. Ectopic let-7i downregulated Aurora A in OECM-1 cells tion of AURKA by b-catenin/TCF4. The results showed that (Fig. 3E), and neutralization of let-7i enhanced Aurora A b-catenin activated AURKA promoter, and the dominant-neg- expression in FaDu cells (Fig. 3F). In contrast, miR-15 did not ative TCF4 abrogated the b-catenin–induced AURKA activa- have impact on Aurora A expression (Supplementary Fig. S2A). tion. Mutation of the TCF4-binding site also abolished the A reporter assay showed that ectopic let-7i suppressed the AURKA transactivation (Fig. 2E and F). We next examined activity of the AURKA 30-UTR reporter, and mutation of the let- whether BMI1 could activate AURKA promoter through b-cate- 7i–binding site abrogated the repression (Fig. 3G and H). nin/TCF4. Ectopic BMI1 activated AURKA promoter, and Reconstitution of let-7i in FaDu-BMI1 attenuated Aurora A either transfecting the dominant-negative TCF4 or muta- expression (Fig. 3I). We further investigated whether let-7i ting the TCF4-binding sequence on AURKA promoter abro- repression and Akt–b-catenin pathway act independently to gated the BMI1-induced AURKA activation (Fig. 2G). Taken mediate BMI1-induced Aurora A upregulation. The result together, these results suggest that BMI1 upregulates Aurora showed that suppression of Akt activity did not have significant A through the activation of Akt pathway, leading to the impact on restoring let-7i expression in FaDu-BMI1 cells recruitment of b-catenin/TCF4 complex to the promoter of (Supplementary Fig. S2B). The regulation of Aurora A by BMI1 AURKA to activate its transcription. through both Akt pathway and let-7i repression was confirmed in another HNSCC cell line SAS: knockdown of BMI1 in SAS Repression of let-7i by BMI1 contributes to BMI1- cells repressed Aurora A expression, reduced the levels of induced Aurora A expression phosphorylated Akt, phosphorylated GSK-3b, and b-catenin, Because suppression of Akt activity only partially abrogated and upregulated let-7i expression (Supplementary Fig. S2C and BMI1-induced Aurora upregulation, we hypothesized that S2D). Taken together, these data suggest that Aurora A is a there may exist another mechanism to mediate BMI1-induced bona fide target repressed by let-7i. Akt–b-catenin pathway and Aurora A expression. We reasoned that BMI1-repressed let-7i repression are 2 independent mechanisms induced by miRNA(s) may participate in this regulation as miRNAs are BMI1 to regulate Aurora A expression in HNSCC cells. able to suppress target gene expression. In this scenario, BMI1 represses the expression of miRNA(s) that target(s) AURKA, BMI1 promotes chromosomal instability, antiapoptosis, resulting in upregulation of Aurora A. To this end, miRNA and cell-cycle progression through Aurora A microarray analysis was conducted on BMI1-overexpressing Next, we investigated whether BMI1 could acquire the FaDu cells versus a control cells (Supplementary Table S3). The Aurora A–induced phenotype through regulating AURKA.In result was compared with the miRNA profile in BMI1 knocked cancer cells, one of the major functions of Aurora A is to induce down OECM-1 cells versus control cells (24). Both let-7i and CIN (18, 19). We examined the impact of BMI1–AURKA axis on miR-15b were candidates repressed by BMI1 as they were centrosome number and karyotype in FaDu and OECM-1 downregulated in FaDu-BMI1 and upregulated in OECM1- stable cell lines. In FaDu cells, ectopic BMI1 increased the sh-BMI1 (Fig. 3A). let-7i was more probable to be the target percentage of cells with more than 2 centrosomes, and repres- of BMI1 as let-7i ranked first in the miRNAs downregulated in sion of Aurora A in BMI1 transfectants abrogated such effect FaDu-BMI1 (Supplementary Table S3), and ranked third in the (Fig. 4A). In OECM-1 cells, knockdown of either BMI1 or miRNAs upregulated in OECM1-sh-BMI1 (24). qRT-PCR vali- AURKA caused a reduction of cells with more than 2 centro- dation confirmed that let-7i, but not miR-15b, was consistently somes (Supplementary Fig. S3A and S3B). Karyotype analysis repressed by BMI1 in both FaDu and OECM1 systems (Fig. 3B). showed that overexpression of BMI1 increased the proportion For the Polycomb group proteins repress target genes expres- of aneuploidy in FaDu cells, and silencing Aurora A in BMI1 sion through forming PRC on the regulatory area containing transfectants reversed it (Fig. 4B). Suppression of either BMI1

panels, fold-change of migrated cells in wound-healing assay. Bottom 2 panels, fold-change of invaded cells in invasion assay. D, Western blot analysis of BMI1, Aurora A, and Snail in FaDu clones. E, relative SNAI1 mRNA expression in FaDu clones. F, Western blot analysis of BMI1, Aurora A, and Snail in FaDu clones after cyclohexamide (CHX) treatment for 0, 2, and 4 hours. G, Western blot analysis of BMI1, Aurora A, total GSK-3b, and serine 9-phosphorylated GSK-3b in FaDu clones. H, immunoprecipitation by an anti-Aurora A antibody and Western blot analysis for detecting pulled down Aurora A and serine 9-phosphorylated GSK-3b in OECM-1 cells. The arrows indicate the immunoprecipitated bands. I, coimmunoprecipitation assay in HEK-293T cells. The transfection and immunoprecipitation condition was indicated in the panel. The arrows indicate the immunoprecipitated bands. In C and E, data represent mean SEM (n ¼ 3). , P < 0.05; , P < 0.01.

www.aacrjournals.org Cancer Res; 73(2) January 15, 2013 OF9

Chou et al.

or Aurora A expression in OECM-1 cells significantly decreased upregulated Snail expression, and knockdown of AURKA abro- the percentage of cells with aneuploidy (Supplementary Fig. gated the BMI1-induced Snail expression (Fig. 5D); the other S3C and S3D). EMT regulators was not consistently affected (Supplementary For Aurora A phosphorylates p53 at serine 315, leading to Fig. S5A). Silencing AURKA reduced Snail expression in OECM- ubiquitination and degradation of p53 (22), we investigated 1 (Supplementary Fig. S5B), and overexpression of Aurora A whether Aurora A is important for BMI1-induced cell anti- upregulated Snail but not other EMT regulators in FaDu apoptosis and cell-cycle progression through degrading p53. (Supplementary Fig. S5C). However, there was no difference The results showed that in FaDu cells, ectopic BMI1 upregu- between the mRNA level of SNAI1 in FaDu-CDH, FaDu-BMI1- lated Aurora A, enhanced p53 serine 315 phosphorylation and sh-scr, and FaDu-BMI1-sh-AURKA (Fig. 5E), and ectopic Auro- decreased total p53. Repression of AURKA in FaDu-BMI1 ra A also did not change the mRNA level of SNAI1 (Supple- reduced p53 serine 315 phosphorylation and restored total mentary Fig. S5D). Furthermore, BMI1 increased Snail stability, p53 level (Fig. 4C). In OECM-1 cells, silencing AURKA did not and knockdown of AURKA in FaDu-BMI1 reduced Snail sta- alter the level of BMI1, which confirms that Aurora A is located bility (Fig. 5F). These results indicate that BMI1-regulated downstream to BMI1. Knockdown of either AURKA or BMI1 Aurora A induces EMT through stabilization of Snail. reduced p53 serine 315 phosphorylation and increased total Because GSK-3b phosphorylates Snail to promote degrada- p53 level (Supplementary Fig. S3E and S3F). For analyzing the tion (35), and Aurora A enhances GSK-3b phsophorylation to apoptosis, we found that ectopic BMI1 repressed apoptosis as repress its activity (36), we investigated whether Aurora A expected. Treatment of the cells with an Aurora A inhibitor interacts with GSK-3b and increases its phosphorylation under (AKI III) increased the proportion of apoptotic cells (Fig. 4D). BMI1-expressing situation. In FaDu cells, overexpression of For cell-cycle analysis, overexpression of BMI1 enforced the BMI1 enhanced GSK-3b serine 9 phosphorylation, and knock- – AURKA cells into cell cycle by reducing the proportion of cells in G0 G1 down of in FaDu-BMI1 abrogated this phosphorylation and increasing the proportion of cells in S-phase. In FaDu- (Fig. 5G). Repression of Aurora A in OECM-1 cells attenuated BMI1 cells, knockdown of AURKA increased sub-G1 and G2–M GSK-3b serine 9 phosphorylation (Supplementary Fig. S6A). populations. Treatment with the Aurora A inhibitor AKI III Coimmunoprecipitation experiments confirmed the physical had a more prominent effect than AURKA knockdown in association of Aurora A and GSK-3b among different HNSCC enriching sub-G1 population (Fig. 4E). Collectively, these cell lines (Supplementary Fig. S6B–S6E). Serine 9-phsophory- results suggest that BMI1 promotes CIN through regulating lated GSK-3b was present in the immunoprecipitates obtained Aurora A. Furthermore, Aurora A is crucial for BMI1-induced with an anti-Aurora A antibody from OECM-1 cells (Fig. 5H). p53 stabilization, antiapoptosis, and cell-cycle progression. Ectopic Aurora A pulled down endogenous GSK-3 b, and ectopic GSK-3b also pulled down endogenous Aurora A (Fig. Aurora A is critical in BMI1-induced EMT through 5I). Collectively, these data suggest that Aurora A is important stabilization of Snail in BMI1-induced EMT through interaction with GSK-3b, lead- For BMI1 is a major determinant of EMT and stem-like ing to phosphorylation of GSK-3b at serine 9 and inactivation. phenotype of cancer cells (12–14), we investigated whether The stability of Snail is thereby increased. Aurora A plays a role in BMI1-induced EMT. In FaDu cells, overexpression of BMI1 induced EMT as expected, that is, Significance of BMI1–AURKA axis in vivo and in head upregulation of the epithelial markers E-cadherin and g-cate- and neck cancer patients nin, downregulation of mesenchymal markers N-cadherin and To determine the effect of BMI1-regulated Aurora A in vivo, vimentin, dissociation of E-cadherin, and enhanced migration we injected the stable cell lines FaDu-CDH and FaDu-BMI1 and invasion. Interestingly, knockdown of AURKA in FaDu- into the subcutaneous area of nude mice, and treated the mice BMI1 restored the epithelial phenotype (Fig. 5A and B). Over- with different doses of the Aurora kinase inhibitor VX-680 or a expression of BMI1 enhanced both migration and invasion vehicle control to observe the impact of suppressing Aurora ability of FaDu cells. Either knockdown of AURKA or treatment kinase activity in BMI1-expressing HNSCC cells (Fig. 6A). The with AKI III abrogated BMI1-induced migration and invasion result showed that ectopic BMI1 in FaDu cells increased the (Fig. 5C). Because the role of Aurora A in EMT is not clearly volume of xenotransplanted tumors. Treating mice with VX- defined yet, we investigated whether Aurora A is capable of 680 25 mg/kg repressed the in vivo growth of FaDu-BMI1 to a inducing EMT in HNSCC cells. In OECM-1 cells, repression of similar extent of FaDu-CDH, and 50 mg/kg of VX-680 treat- Aurora A reverted the cells to epithelial phenotype and reduced ment further reduced the volume of FaDu-BMI1–formed migration and invasion (Supplementary Fig. S4A–S4D). Ectop- tumors (Fig. 6B and C). IHC of the harvested tumor samples ic expression of Aurora A in FaDu cells induced EMT (Sup- confirmed upregulation of Aurora A by BMI1 in vivo. BMI1 plementary Fig. S4E–S4H). These results suggest that Aurora A promoted nuclear translocation of b-catenin, reduced p53 is critical in BMI1-induced EMT, and Aurora A itself is suffi- expression, and suppressed apoptosis. Inhibition of Aurora cient to induce EMT in HNSCC cells. kinase activity by VX-680 decreased nuclear b-catenin, To determine the mechanism of BMI1-induced Aurora A in restored p53 expression and enhanced apoptosis (Fig. 6D). promoting EMT, we examined the protein levels of different Finally, we confirmed the clinical significance of BMI1–AURKA EMT regulators in established stable cell lines. Among the EMT axis in 46 patients with HNSCC. A positive correlation was regulators, only Snail was consistently regulated by BMI1– found between the mRNA levels of BMI1 and AURKA (Fig. 6E). high AURKA axis in different cell lines. In FaDu cells, ectopic BMI1 The dCT value of AURKA was lower in the BMI1 group as

OF10 Cancer Res; 73(2) January 15, 2013 Cancer Research

BM11-AURKA Axis in Cancer Progression

A B Tumor volume = 200 mm3 Day 0 Day X Day X+7 Day X+14 Day X+21

Subcutaneous injection of 1 x 107 cells of FaDu-CDH or FaDu-BMI1 VX-680 0, 25, 50 mg/kg i.p. once a day, 4 d/wk for 3 wk FaDu-CDH FaDu-BMI1 FaDu-BMI1 + FaDu-BMI1 + Sacrifice and harvest sample control control 25 mg/kg VX-680 50 mg/kg VX-680

C FaDu-CDH control FaDu-BMI1+ VX 680 D 25 mg/kg BMI1 Aurora A β-Catenin p53 TUNEL FaDu-BMI1 control FaDu-BMI1+ VX 680 50 mg/kg 1,600 FaDu-CDH * control

) 1,400 3 1,200 * FaDu- 1,000 BMI1 control 800 * 600 * FaDu-BMI1 400 + VX-680

Tumor volume (mm 200 (50 mg/kg) 0 0123 FaDu-CDH 120 Treatment duration (wks from day X) FaDu-BMI1 control 100 FaDu-BMI1+VX-680 (50 mg/kg)

E F 80 9 12 8 10 60 7 8 6 40 6 5 4 4 20 2 of AURKA of

of AURKA of 3 T

T 0 C 0 Percentage of positive cells (%) C 2 r = 0.6601 d d P < 0.001 BMI1 Aurora A β catenin-Catenin p53 TUNEL tunel 1 P < 0.0001 -2 0 -4 0.0 2.5 5.0 7.5 10.0 BMI1high BMI1low dCT of BMI1 G H BMI1 BMI1(HPF) Aurora A Aurora A (HPF) 1.0 Other patients (n = 32) 0.8 Case 1 BMI1 (-) 0.6 Aurora A (-)

0.4 Case 2 BMI1highAURKAhigh (n = 14) 0.2 BMI1 (+) 0.0 P = 0.023 Aurora A (+)

Proportion of overall survival 5.0 10.0 15.0 20.0 25.0 30.0 35.0 Month (s)

Figure 6. Inhibition of Aurora kinase activity attenuates BMI1-induced tumor progression in vivo, and clinical significance of BMI1–AURKA axis in HNSCC. A, schema of the animal experiment. i.p., intraperitoneal injection. B, representative photos of xenotransplanted tumors. C, tumor volume curves. Data represent mean SEM (n ¼ 6). , P < 0.05 (compared with FaDu-CDH). D, top, IHC of BMI1, Aurora A, b-catenin, p53, and TUNEL assay in xenotransplanted tumors. Scale bar, 50 mm. The insets show the magnification of each image. Bottom, quantification of the IHC results by percentage of positive cells. E, low high qRT-PCR in HNSCC samples (n ¼ 46), and correlation between the dCT values of BMI1 and AURKA.F,thedCT value of AURKA in BMI1 versus BMI1 patients with HNSCC. G, survival analysis in BMI1highAURKAhigh versus the residual cases. H, representative IHC results of HNSCC cases. HPF, high-power field. Scale bar, 200 mminthefirst and third column, and 50 mm in the second and fourth column (HPF images).

www.aacrjournals.org Cancer Res; 73(2) January 15, 2013 OF11

Chou et al.

P Ser 473 EZH2 me me PTEN AKT BMI1 me BMI1 H3K27 OFF

β-Catenin PTEN

EZH2 me me β-Catenin me let-7i BMI1 H3K27 OFF TCF4 ON AURKA mRNA

MIRLET7I AURKA Aurora A P Ser 315 p53 Degradation p53 P Ser 9 GSK-3β Cell-cycle progression Chromosomal Centrosomal antiapoptosis instability amplification

Snail

Cancer Stabilization progression

EMT Invasion

Figure 7. A model depicting the BMI1–AURKA axis-centered signal network during HNSCC progression.

compared with the BMI1low group, indicating a significant mark of cancer, the role of CIN in cancer progression is unclear. higher AURKA level in the BMI1high group (Fig. 6F). Patients For CSCs, extensive studies support their role in both tumor- with both high BMI1 and AURKA levels had a worse survival igenesis and late-stage progression. Emerging evidence impli- (Fig. 6G). The correlation between the protein levels of BMI1 cates a possible link between CIN and CSCs (37–39), however, and Aurora A was confirmed by IHC in 33 HNSCC samples (P ¼ the association between CIN and stem-like properties of 0.015; the representative IHC results shown in Fig. 6H). cancer cells remains largely elusive. In this study, we have We propose a model to summarize our finding (Fig. 7). In shown the mechanism linking CSCs to CIN through the BMI1– HNSCC cells, BMI1 upregulates Aurora A expression through 2 AURKA axis. For the promising antitumor efficacy of Aurora PRC-mediated pathways, that is, enrichment of BMI1, EZH2, kinase inhibitors in different human cancers especially HNSCC and H3K27me3 on the promoter of target genes. First, BMI1 (40, 41), our study provides a solid rationale for applying Aurora suppresses PTEN and activates the Akt signaling pathway, A inhibitors to disrupt the "evil molecular link" in advanced leading to transactivation of AURKA by b-catenin/TCF4 HNSCC. complex. Second, BMI1 inhibits the expression of let-7i, result- In this study, we discovered that repression of let-7i by BMI1 ing in a release of Aurora A from let-7i suppression. Upregula- contributes to the enhancement of Aurora A expression. let-7i is tion of Aurora A by BMI1 leads to several key events in HNSCC: a member of the let-7 miRNA family, which functions as a tumor CIN, cell-cycle progression, and antiapoptosis through degrad- suppressor and represses self-renew of stem cells (42, 43). We ing p53, and EMT through stabilizing Snail. All these events previously discovered that Twist1 and BMI1 act cooperatively to contribute to the progression of HNSCC. repress the expression of target genes (CDH1, CDKN2A, and let- 7i) through the Twist1-binding sites (13, 24). Here, we found that BMI1 itself is capable of repressing let-7i through PREs, suggest- Discussion ing the independent role of BMI1 on regulating let-7i. The Although mitotic aberration and CIN are critically involved independency versus cooperation between Twist1 and BMI1 in tumorigenesis and they have been considered as the hall- during cancer progression deserves further investigation.

OF12 Cancer Res; 73(2) January 15, 2013 Cancer Research

BM11-AURKA Axis in Cancer Progression

We here showed that under the BMI1-overexpressing situ- Development of methodology: C.-H. Chou, N.-K. Yang, T.-Y. Liu, D.S.-S. Hsu, C.- C. Chang ation, the b-catenin/TCF4 complex regulates the transcription Acquisition of data (provided animals, acquired and managed patients, of AURKA, which indicates that b-catenin is located upstream provided facilities, etc.): C.-H. Chou, N.-K. Yang, T.-Y. Liu, S.-K. Tai, D.S.-S. Hsu, to Aurora A. However, results from the in vivo study revealed Y.-W. Chen, C.-H. Tzeng Analysis and interpretation of data (e.g., statistical analysis, biostatistics, that inhibition of the Aurora kinase activity also caused a computational analysis): C.-H. Chou, N.-K. Yang, T.-Y. Liu, S.-K. Tai, D.S.-S. decrease of nuclear b-catenin expression in xenotransplanted Hsu, M.-H. Yang Writing, review, and/or revision of the manuscript: C.-H. Chou, N.-K. Yang, tumors. A possible explanation is that Aurora A phosphorylates T.-Y. Liu, D.S.-S. Hsu, M.-H. Yang GSK-3b and reduces it activity, leading to stabilization and Administrative, technical, or material support (i.e., reporting or orga- nuclear translocation b-catenin. These results indicate that nizing data, constructing databases): C.-H. Chou, N.-K. Yang, S.-K. Tai, D.S.-S. Hsu, Y.-J. Chen, C.-H. Tzeng when BMI1 is overexpressed, Aurora A and b-catenin form a Study supervision: C.-C. Chang, C.-H. Tzeng, M.-H. Yang positive feedback loop to amplify the signal. In summary, our study identifies a novel signaling network Acknowledgments in HNSCC, which links 3 major events during cancer progres- The authors thank Dr. Ishigatsubo (Yokohama City University, Japan) for the generous gift of the AURKA promoter plasmid pGL-1486. sion, that is, CSCs, CIN, and EMT, through the BMI1–AURKA axis. Scientifically, this finding provides a mechanism for Grant Support understanding how cancer cells develop multifaceted changes This work was supported by National Science Council (100-2321-B-010-015 to during progression through a network governed by a pivotal M.-H. Yang; 99-2811-B-010-014 to T.-Y. Liu), National Health Research Institutes oncoprotein, BMI1. Clinically, this study suggests Aurora A as (NHRI-EX101-10037BI to M.-H. Yang), Taipei Veterans General Hospital (100-C1- 088 and 101-C-005 to M.-H. Yang), Veterans General Hospitals University System an ideal target for inhibiting BMI1-induced malignant pro- of Taiwan Joint Research Program (VGHUST101-G7-4-1 to M.-H. Yang), Tai- gression of HNSCC cells, which will be valuable for the future chung Veterans General Hospital–National Yang-Ming University Joint Research development of personalized medicine in advanced HNSCC. Program (TCVGH-YM1000302 to M.-H. Yang), a grant from Ministry of Educa- tion, Aim for the Top University Plan, and a grant from Department of Health, Center of Excellence for Cancer Research (DOH101-TD-C-111-007 to M.-H. Disclosure of Potential Conflicts of Interest Yang). No potential conflicts of interest were disclosed. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Authors' Contributions Conception and design: C.-H. Chou, N.-K. Yang, T.-Y. Liu, D.S.-S. Hsu, C.-C. Received June 22, 2012; revised October 12, 2012; accepted October 30, 2012; Chang, C.-H. Tzeng, M.-H. Yang published OnlineFirst November 30, 2012.

References 1. Clarke MF, Dickm J, Dirks PB, Eaves CJ, Jamieson CH, Jones DL, et al. stem cell properties in head and neck squamous cell carcinoma. Proc Cancer stem cells-perspectives on current status and future direc- Natl Acad Sci USA 2007;104:973–8. tions: AACR Workshop on cancer stem cells. Cancer Res 2006;66: 12. Wellner U, Schubert J, Burk UC, Schmalhofer O, Zhu F, Sonntag A, 9339–44. et al. The EMT-activator ZEB1 promotes tumorigenicity by repressing 2. Pavet V, Portal MM, Moulin JC, Herbrecht R, Gronemeyer H. Toward stemness-inhibiting microRNAs. Nat Cell Biol 2009;11:1487–95. novel paradigms for cancer therapy. Oncogene 2011;30:1–20. 13. Yang MH, Hsu DS, Wang HW, Wang HJ, Lan HY, Yang WH, et al. Bmi1 3. Richly H, Aloia L, Di Croce L. Roles of the Polycomb group proteins in is essential in Twist1-induced epithelial–mesenchymal transition. Nat stem cells and cancer. Cell Death Dis 2011;2:e204. Cell Biol 2010;12:982–92. 4. Sauvageau M, Sauvageau G. Polycomb group proteins: multi-faceted 14. Song LB, Li J, Liao WT, Feng Y, Yu CP, Hu LJ, et al. The polycomb regulators of somatic stem cells and cancer. Cell Stem Cell 2010;7: group protein Bmi-1 represses the tumor suppressor PTEN and 299–313. induces epithelial–mesenchymal transition in human nasopharyngeal 5. Iwama A, Oguro H, Negishi M, Kato Y, Morita Y, Tsukui H, et al. epithelial cells. J Clin Invest 2009;119:3626–36. Enhanced self-renewal of hematopoietic stem cells mediated by the 15. Schvartzman JM, Sotillo R, Benezra R. Mitotic chromosomal instability polycomb gene product Bmi-1. Immunity 2004;1:843–51. and cancer: mouse modelling of the human disease. Nat Rev Cancer 6. Molofsky AV, Pardal R, Iwashita T, Park IK, Clarke MF, Morrison SJ. 2010;10:102–15. Bmi-1 dependence distinguishes neural stem cell self-renewal from 16. Choi CM, Seo KW, Jang SJ, Oh YM, Shim TS, Kim WS, et al. Chro- progenitor proliferation. Nature 2003;425:962–7. mosomal instability is a risk factor for poor prognosis of adenocarci- 7. Sangiorgi E, Capecchi MR. Bmi1 is expressed in vivo in intestinal stem noma of the lung: fluorescence in situ hybridization analysis of paraffin- cells. Nat Genet 2008;40:915–20. embedded tissue from Korean patients. Lung Cancer 2009;64:66–70. 8. Jacobs JJ, Scheijen B, Voncken JW, Kieboom K, Berns A, van 17. Walther A, Houlston R, Tomlinson I. Association between chromo- Lohuizen M. Bmi-1 collaborates with c-Myc in tumorigenesis by somal instability and prognosis in colorectal cancer: a meta-analysis. inhibiting c-Myc-induced apoptosis via INK4a/ARF. Genes Dev Gut 2008;57:941–50. 1999;13:2678–90. 18. Carmena M, Earnshaw WC. The cellular geography of aurora kinases. 9. Chiba T, Miyagi S, Saraya A, Aoki R, Seki A, Morita Y, et al. The Nat Rev Mol Cell Biol 2003;4:842–54. polycomb gene product BMI1 contributes to the maintenance of 19. Meraldi P, Honda R, Nigg EA. Aurora kinases link chromosome tumor-initiating side population cells in hepatocellular carcinoma. segregation and cell division to cancer susceptibility. Curr Opin Genet Cancer Res 2008;68:7742–9. Dev 2004;14:29–36. 10. Leung C, Lingbeek M, Shakhova O, Liu J, Tanger E, Saremaslani P, 20. Gautschi O, Heighway J, Mack PC, Purnell PR, Lara PN Jr, Gandara et al. Bmi1 is essential for cerebellar development and is overex- DR. Aurora kinases as anticancer drug targets. Clin Cancer Res pressed in human medulloblastomas. Nature 2004;428:337–41. 2008;14:1639–48. 11. Prince ME, Sivanandan R, Kaczorowski A, Wolf GT, Kaplan MJ, 21. Cammareri P, Scopelliti A, Todaro M, Eterno V, Francescangeli F, Dalerba P, et al. Identification of a subpopulation of cells with cancer Moyer MP, et al. Aurora-A is essential for the tumorigenic capacity and

www.aacrjournals.org Cancer Res; 73(2) January 15, 2013 OF13

Chou et al.

chemoresistance of colorectal cancer stem cells. Cancer Res 2010;70: 33. Chang CJ, Yang JY, Xia W, Chen CT, Xie X, Chao CH, et al. EZH2 4655–65. promotes expansion of breast tumor initiating cells through activation 22. Katayama H, Sasai K, Kawai H, Yuan ZM, Bondaruk J, Suzuki F, et al. of RAF1-b-catenin signaling. Cancer Cell 2011;19:86–100. Phosphorylation by aurora kinase A induces Mdm2-mediated desta- 34. Czermin B, Melﬁ R, McCabe D, Seitz V, Imhof A, Pirrotta V. Drosophila bilization and inhibition of p53. Nat Genet 2004;36:55–62. enhancer of Zeste/ESC complexes have a histone H3 methyltransfer- 23. Wong YK, Chang KW, Chen CF, Chang CS. Characterization of two ase activity that marks chromosomal Polycomb sites. Cell 2002;111: new cell lines derived from oral cavity human squamous cell carcino- 185–96. mas—OC1 and OC2. J Oral Maxillofac Surg 1990;48:385–90. 35. Zhou BP, Deng J, Xia W, Xu J, Li YM, Gunduz M, et al. Dual regulation of 24. Yang WH, Lan HY, Huang CH, Tai SK, Tzeng CH, Kao SY, et al. RAC1 Snail by GSK-3beta-mediated phosphorylation in control of epithelial– activation mediates Twist1-induced cancer cell migration. Nat Cell Biol mesenchymal transition. Nat Cell Biol 2004;6:931–40. 2012;14:366–74. 36. Dar AA, Belkhiri A, El-Rifai W. The aurora kinase A regulates GSK-3beta 25. Tanaka M, Ueda A, Kanamori H, Ideguchi H, Yang J, Kitajima S, et al. in gastric cancer cells. Oncogene 2009;28:866–75. Cell-cycle–dependent regulation of human aurora A transcription is 37. Miura M, Miura Y, Padilla-Nash HM, Molinolo AA, Fu B, Patel V, et al. mediated by periodic repression of E4TF1. J Biol Chem 2002;277: Accumulated chromosomal instability in murine bone marrow mes- 10719–26. enchymal stem cells leads to malignant transformation. Stem Cells 26. Remmele W, Schicketanz KH. Immunohistochemical determination of 2006;24:1095–03. estrogen and progesterone receptor content in human breast cancer. 38. Shiras A, Chettiar ST, Shepal V, Rajendran G, Prasad GR, Shastry P. Computer-assisted image analysis (QIC score) vs. subjective grading Spontaneous transformation of human adult nontumorigenic stem (IRS). Pathol Res Pract 1993;189:862–6. cells to cancer stem cells is driven by genomic instability in a human 27. Ross DT, Scherf U, Eisen MB, Perou CM, Rees C, Spellman P, et al. model of glioblastoma. Stem Cells 2007;25:1478–89. Systematic variation in gene expression patterns in human cancer cell 39. Liang Y, Zhong Z, Huang Y, Deng W, Cao J, Tsao G, et al. Stem-like lines. Nat Genet 2000;24:227–35. cancer cells are inducible by increasing genomic instability in cancer 28. Valk-Lingbeek ME, Bruggeman SW, van Lohuizen M. Stem cells and cells. J Biol Chem 2010;285:4931–40. cancer; the polycomb connection. Cell 2004;118:409–18. 40. Dar AA, Goff LW, Majid S, Berlin J, El-Rifai W. Aurora kinase inhibitors- 29. Widschwendter M, Fiegl H, Egle D, Mueller-Holzner E, Spizzo G, Marth rising stars in cancer therapeutics? Mol Cancer Ther 2010;9:268–78. C, et al. Epigenetic stem cell signature in cancer. Nat Genet 2007;39: 41. Han H, Von Hoff DD. Aurora sheds light on head and neck squamous 157–8. cell carcinoma. Clin Cancer Res 2006;12:5003–4. 30. Wu D, Pan W. GSK3: a multifaceted kinase in Wnt signaling. Trends 42. Rybak A, Fuchs H, Smirnova L, Brandt C, Pohl EE, Nitsch R, et al. A Biochem Sci 2010;35:161–8. feedback loop comprising lin-28 and let-7 controls pre-let-7 matura- 31. Dutta-Simmons J, Zhang Y, Gorgun G, Gatt M, Mani M, Hideshima T, tion during neural stem-cell commitment. Nat Cell Biol 2008;10: et al. Aurora kinase A is a target of Wnt/beta-catenin involved in 987–93. multiple myeloma disease progression. Blood 2009;114:2699–708. 43. Yu F, Yao H, Zhu P, Zhang X, Pan Q, Gong C, et al. let-7 Regulates self 32. Shitashige M, Hirohashi S, Yamada T. Wnt signaling in the nucleus. renewal and tumorigenicity of breast cancer cells. Cell 2007;131: Cancer Sci 2008;99:631–7. 1109–23.

OF14 Cancer Res; 73(2) January 15, 2013 Cancer Research

Chromosome Instability Modulated by BMI1−AURKA Signaling Drives Progression in Head and Neck Cancer

Chun-Hung Chou, Neng-Kai Yang, Ting-Yun Liu, et al.

Cancer Res Published OnlineFirst November 30, 2012.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-12-2397

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2012/11/30/0008-5472.CAN-12-2397.DC1

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

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/early/2013/01/10/0008-5472.CAN-12-2397. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.