The authors declare no competing financial interests.
References
1. Thiery, J.P., Acloque, H., Huang, R.Y. & Nieto, M.A. Epithelial-mesenchymal
transitions in development and disease. Cell 139,871-890 (2009).
2. Mani, S.A., et al. The epithelial-mesenchymal transition generates cells with properties of stem cells. Cell 133,704-715 (2008).
3. Morel, A.P., et al. Generation of breast cancer stem cells through epithelial-mesenchymal transition. PLoS ONE 3,e2888 (2008).
4. Valk-Lingbeek, M.E., Bruggeman, S.W., & van Lohuizen, M. Stem cells and cancer: the Polycomb connection. Cell 118,409–418 (2004).
5. Widschwendter, M., et al. Epigenetic stem cell signature in cancer. Nat. Genet.
39:157–158 (2007).
6. Molofsky, A.V., et al. Bmi-1 dependence distinguishes neural stem cell self-renewal from progenitor proliferation. Nature 425,962–967 (2003).
7. Iwama, A., et al. Enhanced self-renewal of hematopoietic stem cells mediated by the polycomb gene product Bmi-1. Immunity 21,843-851 (2004).
8. Sangiorgi, E. & Capecchi, M.R. Bmi1 is expressed in vivo in intestinal stem cells.
Nat Genet. 40,915-920 (2008).
9. Jacobs, J.J., Kieboom, K., Marino, S., Depinho, R. & van Lohuizen, M. The oncogene and Polycomb-group gene Bmi1 regulates cell proliferation and senescence through the ink4a locus. Nature 397,164–168 (1999).
10. Bruggeman, S.W., et al. Ink4a and Arf differentially affect cell proliferation and neural stem cell self-renewal in Bmi1-deficient mice. Genes & Dev. 19,1438–
1443 (2005).
11. Czermin, B., et al. Drosophila Enhancer of Zeste/ESC complexes have a histone H3 methyltransferase activity that marks chromosomal Polycomb sites. Cell 111,185–196 (2002).
12. Müller, J., et al. Histone methyltransferase activity of a Drosophila Polycomb group repressor complex. Cell 111,197–208 (2002).
13. Min, J., Zhang, Y. & Xu, R.M. Structural basis for specific binding of Polycomb chromodomain to histone H3 methylated at Lys 27. Genes & Dev. 17,1823–1828 (2003).
14. Bracken, A.P., et al. The Polycomb group proteins bind throughout the INK4A–
ARF locus and are disassociated in senescent cells. Genes & Dev. 21,525–530 (2007).
15. Jacobs, J.J., et al. Bmi-1 collaborates with Myc in tumorigenesis by inhibiting c-Myc-induced apoptosis via INK4a/ARF. Genes Dev. 13,2678-2690 (1999).
16. Leung, C., et al. Bmi1 is essential for cerebellar development and is overexpressed in human medulloblastomas. Nature 428,337-41 (2004).
17. Prince, M.E., et al. Identification of a subpopulation of cells with cancer stem cell properties in head and neck squamous cell carcinoma. Proc. Natl. Acad. Sci.
U.S.A. 104,973-978 (2007).
18. Chiba, T., et al. The polycomb gene product BMI1 contributes to the maintenance of tumor-initiating side population cells in hepatocellular carcinoma. Cancer Res.
68,7742-7749 (2008).
19. Seiwert, T.Y. & Cohen, E.E. State-of-the-art management of locally advanced head and neck cancer. Br. J. Cancer 92,1341-1348 (2005).
20. Janssen, H.L., Haustermans, K.M., Balm, A.J. & Begg, A.C. Hypoxia in head and neck cancer: how much, how important? Head Neck 27,622-638 (2005).
21. Hoogsteen, I.J., Marres, H.A., Bussink, J., van der Kogel, A.J. & Kaanders, J.H.
Tumor microenvironment in head and neck squamous cell carcinomas: predictive
value and clinical relevance of hypoxic markers. A review. Head Neck 29,591-604 (2007).
22. Furlong, E.E., Andersen, E.C., Null, B., White, K.P. & Scott, M.P. Patterns of gene expression during Drosophila mesoderm development. Science 293,1629-1633 (2001).
23. Yang, J., et al. Twist, a master regulator of morphogenesis, plays an essential role in tumor metastasis. Cell 117,927-939 (2004).
24. Yang, M.H., et al. Direct regulation of TWIST by HIF-1alpha promotes metastasis. Nat. Cell Biol. 10,295-305 (2008).
25. Keith, B. & Simon, M.C. Hypoxia-inducible factors, stem cells, and cancer. Cell 129,465-472 (2007).
26. Chen, Y.C., et al. Aldehyde dehydrogenase 1 is a putative marker for cancer stem cells in head and neck squamous cancer. Biochem. Biophys. Res. Commun.
385,307-313 (2009).
27. Huang, L.E., Gu, J., Schau, M. & Bunn, H.F. Regulation of hypoxia-inducible factor 1alpha is mediated by an O2-dependent degradation domain via the
ubiquitin-proteasome pathway. Proc. Natl. Acad. Sci. U.S.A. 95,7987-7992 (1998).
28. Vermeulen, L., et al. Single-cell cloning of colon cancer stem cells reveals a multi-lineage differentiation capacity. Proc. Natl. Acad. Sci. U.S.A. 105,13427-13432 (2008).
29. Cao, Q., et al. Repression of E-cadherin by the polycomb group protein EZH2 in cancer. Oncogene 27,7274-7284 (2008).
30. Tamura, G., et al. E-Cadherin gene promoter hypermethylation in primary human gastric carcinomas. J. Natl. Cancer Inst. 92,569-573 (2000).
31. Peinado, H., Ballestar, E., Esteller, M. & Cano, A. Snail mediates E-cadherin repression by the recruitment of the Sin3A/histone deacetylase 1 (HDAC1)/HDAC2 complex. Mol. Cell Biol. 24,306-319 (2004).
32. Hou, Z., et al. The LIM protein AJUBA recruits protein arginine methyltransferase 5 to mediate SNAIL-dependent transcriptional repression. Mol.
Cell Biol. 28,3198-3207 (2008).
33. Herranz, N., et al. Polycomb complex 2 is required for E-cadherin repression by the Snail1 transcription factor. Mol. Cell Biol. 28,4772-4781 (2008).
34. Wellner, U., et al. The EMT-activator ZEB1 promotes tumorigenicity by repressing stemness-inhibiting microRNAs. Nat. Cell Biol. 11,1487-95 (2009).
35. Yang, M.H. & Wu, K.J. TWIST activated by hypoxia inducible factor-1 (HIF-1):
implication in metastasis and development. Cell Cycle 7,2090-2096 (2008).
36. Li, J., et al. Oncoprotein Bmi-1 renders apoptotic resistance to glioma cells through activation of the IKK-nuclear factor-kappaB Pathway. Am. J. Pathol. 176, 699-709 (2010).
37. Vormittag, L., et al. Co-expression of Bmi-1 and podoplanin predicts overall survival in patients with squamous cell carcinoma of the head and neck treated with radio(chemo)therapy. Int. J. Radiat. Oncol. Biol. Phys. 73,913-918 (2009).
38. Gustafsson, M.V., et al. Hypoxia requires notch signaling to maintain the undifferentiated cell state. Dev. Cell 9,617–628 (2005).
39. Timmerman, L.A., et al. Notch promotes epithelial-mesenchymal transition during cardiac development and oncogenic transformation. Genes Dev. 18,99-115 (2004).
40. Androutsellis-Theotokis, A., et al. Notch signalling regulates stem cell numbers in vitro and in vivo. Nature 442,823-826 (2006).
41. Sparmann, A. & van Lohuizen, M. Polycomb silencers control cell fate, development and cancer. Nat. Rev. Cancer 6,846–856 (2006).
42. Ansieau, S., et al. Induction of EMT by twist proteins as a collateral effect of tumor-promoting inactivation of premature senescence. Cancer Cell 14,79-89 (2008).
43. Huang, C.H., et al. Regulation of membrane-type 4 matrix metalloproteinase (MT4-MMP) by SLUG contributes to hypoxia-mediated metastasis. Neoplasia 11,1371-1382 (2009).
44. Ginestier, C., et al. ALDH1 is a marker of normal and malignant human mammary stem cells and a predictor of poor clinical outcome. Cell Stem Cell 1,555-567 (2007).
45. Ainbinder, E., Amir-Zilberstein, L., Yamaguchi, Y., Handa, H. & Dikstein, R.
Elongation inhibition by DRB sensitivity-inducing factor is regulated by the A20 promoter via a novel negative element and NF-kappaB. Mol. Cell Biol. 24,2444-2454 (2004).
46. Chen, Y.C., et al. Activation of phosphoinositide 3-kinase by the NBS1 DNA repair protein through a novel activation motif. J. Mol. Med. 86,401-412 (2008).
47. Yang, M.H., et al. Overexpression of NBS1 induces epithelial-mesenchymal transition and co-expression of NBS1 and Snail predicts metastasis of head and neck cancer. Oncogene 26,1459-1467 (2007).
48. Kang, M.K., et al. Elevated Bmi-1 expression is associated with dysplastic cell transformation during oral carcinogenesis and is required for cancer cell replication and survival. Br. J. Cancer 96,126-133 (2007).
49. Foschini, M.P., et al. E-cadherin loss and Delta Np73L expression in oral squamous cell carcinomas showing aggressive behavior. Head Neck 30,1475-1482 (2008).
50. Kumar, B., et al. EGFR, p16, HPV Titer, Bcl-xL and p53, sex, and smoking as indicators of response to therapy and survival in oropharyngeal cancer. J. Clin.
Oncol. 26,3128-3137 (2008).
Figure Legends
Figure 1. Overexpression of Twist1 upregulates Bmi1, and co-expression of Twist1
and Bmi1 in HNSCC cells expressing putative tumor-initiating cell markers. (a) Fold-change of mRNA levels of different stemness-related genes in FaDu cells transfected with pcDNA3-Snail, pcDNA3-Slug or pFLAG-Twist1. Transfection of an empty vector was applied as a control. (b),(d),(f) Left, separation of CD44(+) and CD44(-) cells by FACS in FaDu (b), OECM1 (d) and SAS (e) cells. Right: Fold-change of mRNA levels of Twist1 and Bmi1 in CD44(+) vs. CD44(-)cells. (c), (e), (g) Left, separation of ALDH1 (+) and ALDH1(-) cells by FACS in FaDu (c), OECM-1 (e) and primary HNSCC (g) cells. Right: Fold-change of mRNA levels of Twist1 and Bmi1 in ALDH1(+) vs. ALDH(-)cells. The bars show the mean of three independent
experiments SEM. An asterisk (*) indicates P < 0.05 between experimental and control group.
Figure 2. Direct regulation of Bmi1 by Twist1. (a) Upper: Fold-change of mRNA
levels of Bmi1 and p16INK4A in FaDu-CMV vs. FaDu-Twist1 and OECM1-CMV vs.
OECM1-Twist1. Lower: Western blot analysis of Twist1, Bmi1, E-cadherin and N-cadherin in FaDu-CMV vs. FaDu-Twist1 and OECM1-CMV vs. OECM1-Twist1. -actin was used as a loading control. (b) Upper: schematic of the genomic organization
of the regulatory region of Bmi1, and representation of the reporter constructs used in transient transfection assays. The constructs were wild-type (Bmi1-Luc1000), E-Box-deleted (Bmi1-Luc697) or E-box-mutated (Bmi1-Luc1000Mut). E1, exon 1; E2, exon 2. Lower: promoter activity assay. HEK-293T cells were co-transfected with a promoter construct (Bmi1-Luc1000, Bmi1-Luc697 or Bmi1-Luc1000Mut) and an expression vector (pFLAG-Twist1) or an empty vector (pFLAG-CMV) under normoxic or hypoxic conditions. Luciferase activity/-galactosidase of cells transfected with pFLAG-CMV under normoxia was applied as the baseline control for the experiments using the same promoter. The Western blot of Twist1 was shown to indicate the expression of Twist1 under various conditions. (c) EMSA and supershift assay. Nuclear extracts from HEK-293T cells transfected with pFLAG-CMV (lane 2) or pFLAG-Twist1 (lane 3, 4) were incubated with -P32dCTP-labeled probe from the Bmi1 regulatory region containing a Twist1-binding site. Adding of the anti-Twist1 antibody resulted in the supershifted band. The positions of shifted and supershifted band are indicated. No protein extract was added to lane 1. (d) EMSA and competition assay. Nuclear extracts from HEK-293T cells transfected with pFLAG-CMV (lane 1) or pFLAG-Twist1 (lane 2-4) were incubated with -P32dCTP-labeled probe from the Bmi1 regulatory region containing a Twist1-binding site. Ten-fold (lane 3) or 50-fold (lane 4) excess of unlabelled oligonucleotides were
added. (e) ChIP assay of FaDu-Twist1 vs. FaDu-CMV and FaDu-HIF1(ODD) vs.
FaDu-cDNA3. Schematic of the design of ChIP and control primers is shown in the upper panel. The 166-bp fragment contained the Twist1-binding sequence, whereas the 124-bp fragment did not contain any Twist1-binding sequence. The bars in (a) and (b) show the mean of three independent experiments SEM. An asterisk (*) indicates P < 0.05 between experimental and control group. The full length blots/scans of Western blot of Fig. 2a are presented in the Supplementary Information, Fig. S20.
Figure 3. Overexpression of HIF-1, Twist1 or Bmi1 promotes tumor-initiating
capability of HNSCC cells. (a), (b) Spheroid formation assay in FaDu (a) or OECM-1 (b) cells overexpressing HIF-1, Twist1 or Bmi1. Representative pictures of spheroids are shown at the left of each panel. Scale bars = 100 m. (c), (d) (e) Nude mouse xenotransplantation assay. (c) Representative picture of nude mice receiving subcutaneous injections of different clones of tumor cells. Red arrows indicate the xenotransplanted tumor. Cell dose: 1 x 103/mouse. (d) Histological examination of the implanted sites of the mice in (c) by hematoxylin & eosin stains. The red arrows indicate the subcutaneous colonization and muscle infiltration of tumor cells. Scale bars = 500 m. (e) Incidence of tumor growth in nude mice receiving injections of different doses of cDNA3, HIF1(ΔODD)-2, Twist1-1 and
FaDu-Bmi1-1. The P value indicates the statistical comparison of the result between experimental and control (FaDu-cDNA3) groups receiving injection of the same cellular dose. The bars in (a) and (b) show the mean of three independent experiments
SEM. An asterisk (*) indicates P < 0.05 between experimental and control group.
Figure 4. Twist1 and Bmi1 are mutually essential for maintaining EMT and
tumor-initiating capability, and overexpression of Bmi1 induces EMT. (a) Western blot analysis of Twist1, Bmi1, p16INK4A, epithelial markers (E-cadherin and plakoglobin) and mesenchymal markers (N-cadherin and vimentin) expression in FaDu-Twist1 clone receiving siRNA against Bmi1 (FaDu-Twist1-si-Bmi1) or a scrambled sequence (FaDu-Twist1-si-scr). -actin was used as a loading control. (b) Upper: phase-contrast images of FaDu-CMV, FaDu-Twist1, FaDu-Twist1-si-scr and FaDu-Twist1-si-Bmi1. Scale bars = 50 m. Lower: migration and invasion ability of FaDu-CMV vs. FaDu-Twist1 vs. FaDu-Twist1-si-scr vs. FaDu-Twist1-si-Bmi1. (c) Percentages of CD44-positive, ALDH1-positive, side population cells and the spheroid-forming capacity of FaDu-CMV vs. FaDu-Twist1 vs. FaDu-Twist1-si-scr vs.
FaDu-Twist1-si-Bmi1. (d) Western blot analysis of Bmi1, Twist1, epithelial and mesenchymal markers in FaDu-cDNA3 vs. FaDu-Bmi1. -actin was used as a loading control. (e) Western blot analysis of Bmi1, Twist1, p16INK4A, and epithelial and
mesenchymal markers expression in FaDu-Bmi1 clone receiving siRNA against Twist1 (FaDu-Bmi1-si-Twist1) or a scrambled sequence (FaDu-Bmi1-si-scr). -actin was used as a loading control. (f) Upper: phase-contrast images of FaDu-cDNA3, FaDu-Bmi1, FaDu-Bmi1-si-scr and FaDu-Bmi1-si-Twist1. Scale bars = 50 m.
Lower: migration and invasion ability of cDNA3 vs. Bmi1 vs. FaDu-Bmi1-si-scr vs. FaDu-Bmi1-si-Twist1. (g) Percentages of CD44-positive, ALDH1-positive, side population cells and the spheroid-forming capacity of FaDu-cDNA3 vs.
FaDu-Bmi1 vs. FaDu-Bmi1-si-scr vs. FaDu-Bmi1-si-Twist1. The bars in (b), (c), (f) and (g) show the mean of three independent experiments SEM. An asterisk (*) indicates P < 0.05 between the experimental clones vs. vector control clone (the first bar of each panel), and double-asterisks (**) indicate P < 0.05 between the clones receiving scrambled vs. experimental siRNA sequence. The full length blots/scans of Western blots of Fig. 4a, 4d and 4e are presented in the Supplementary Information, Fig. S20.
Figure 5. The Bmi1-containing polycomb repressive complex (PRC) binds to the E-cadherin promoter and exon-1 of p16INK4A. (a) Schematic of the regulatory region of E-cadherin. (b) Quantitative ChIP (qChIP) assay of the E-cadherin promoter region in FaDu-cDNA3, FaDu-Bmi1, FaDu-Bmi1 receiving siRNA against EZH2
(FaDu-Bmi1-si-EZH2) or a scrambled sequence (FaDu-Bmi1-si-scr) with or without suberoylanilide hydroxamic acid (SAHA) treatment. (c) Schematic of the regulatory region of p16INK4A. (d) qChIP assay of the exon-1 of p16INK4A in FaDu-cDNA3 vs. FaDu-Bmi1 vs. FaDu-Bmi1-si-scr vs. FaDu-Bmi1-si-EZH2 with or without SAHA treatment. The antibodies used in qChIP are shown at the top of each panel in (b) and (d). An irrelevant IgG was applied as a control for qChIP experiments. The binding activity of each protein is given as percentage of total input. The bars in (b) and (d) show the mean of three independent experiments SEM.
Figure 6. Co-occupancy of Twist1, Bmi1 and EZH2 on the E-cadherin promoter and
exon-1 of p16INK4A.(a), (b) Schematic of the regulatory region of E-cadherin (a) and p16INK4A (b). E1, E2, E3, and E indicate the location of E-box. (c), (d) qChIP assay for the E-cadherin promoter region (c) and exon-1 of p16INK4A (d) in FaDu-CMV, FaDu-Twist1, FaDu-Twist1 receiving siRNA against Bmi1, EZH2, or a scrambled sequence (scr). (e), (f) qChIP assay for the E-cadherin promoter region (e) and exon-1 of p16INK4A (f) in FaDu-cDNA3, FaDu-Bmi1, FaDu-Bmi1 receiving siRNA against Twist1, EZH2, or a scrambled sequence. The antibodies used in qChIP are shown at the top of each panel in (c), (d), (e) and (f). An irrelevant IgG was applied as a control for qChIP. The binding activity of each protein is given as
percentage of total input. The bars in (c), (d), (e) and (f) show the mean of three independent experiments SEM.
Figure 7. Twist1 and Bmi1 cooperatively repress E-cadherin transcription. (a) Upper:
schematic of the wild-type E-cadherin(wt)) or multiple E-box–mutated (pXP2-E-cadherin(mut-E1E2E3)) E-cadherin promoter construct. Lower: promoter activity assay. HEK-293T cells were co-transfected with the promoter (pXP2-E-cadherin(wt) or pXP2-E-cadherin(mut-E1E2E3)) and expression vector (pFLAG-Twist1 or/and pcDNA3-Bmi1). Luciferase activity/-galactosidase of cells transfected with the empty vectors pFLAG-CMV/pcDNA3.1 was applied as the baseline control for the experiments using the same promoter. The Western blot of Twist1 and Bmi1 was shown to indicate the expression of Twist1 and Bmi1 under different transfections. (b) Plasmid immunoprecipitation assay. HEK-293T cells co-transfected with wild type or mutated E-cadherin promoter construct, empty vector (pcDNA3.1 or/and pFLAG-CMV) or/and expression vector (pcDNA3-Bmi1 or/and pFLAG-Twist1). The precipitated DNA was amplified by quantitative real-time PCR and the antibodies used for plasmid IP are shown at the top of each panel. An irrelevant IgG was applied as a control for qChIP experiments. The binding activity of each protein is given as percentage of total input. (c) EMSA and supershift assay. Oligonucelotides for EMSA
were -P32dCTP-labeled probes from the E-cadherin promoter containing the different E-boxes (E1, E2, E3). Nuclear extracts from HEK-293T cells transfected with pFLAG-CMV (lane 2 of each panel) or pFLAG-Twist1 (lane 3-5 of each panel) were incubated with different probes. Adding an anti-Twist1 or an anti-Bmi1 antibody resulted in the supershifted bands. The positions of shifted and supershifted band are indicated. No protein extract was added to lane 1 in each panel. (d) Co-immunoprecipitation assays using an anti-Twist1 or an anti-Bmi1 antibody to pull down proteins from nuclear extracts of FaDu cells overexpressing Twist1 (FaDu-Twist1-1; upper panel) or Bmi1 (FaDu-Bmi1-1; lower panel). The immune complex was analyzed using anti-Twist1 and anti-Bmi1 antibodies. An irrelevant IgG was used as a control for the co-immunoprecipitation experiments. The bars in (a) and (b) show the mean of three independent experiments SEM. An asterisk (*) indicates P < 0.05 between experimental groups and vector control group. Double-asterisks (**) indicate in the same promoter analysis, P < 0.05 between cells transfected with pFLAG-Twist1 vs. co-transfected with pFLAG-pFLAG-Twist1+pcDNA3-Bmi1. The full length blots/scans of Western blot of Fig. 7d are presented in the Supplementary Information, Fig. S20.
Figure 8. Clinical significance of co-expression of Twist1 and Bmi1 in HNSCC
patients and a proposed model of Twist1- and Bmi1-mediated suppression of E-cadherin and p16INK4A. (a) Percentages of cases with downregulation of E-E-cadherin
(open bars) or p16INK4A (closed bars) in samples with different expression patterns of Twist1 and Bmi1. An asterisk (*) indicates P < 0.05 between the Twist1(+)Bmi1(+) group and other groups of patients. (b) Left: comparisons of the overall survival periods of patients with different patterns of Twist1 and Bmi1 expression. The P-values of the comparison between each group are shown in the inset. Right: prognostic significance of co-expression of Twist1 and Bmi1 in HNSCC cases. (c) A proposed model of Twist1 and Bmi1 mediated suppression of E-cadherin and p16INK4A. Twist1 activates the transcription of Bmi1, and Twist1 cooperates with Bmi1-containing PRC1 and PRC2 to suppress the transcription of E-cadherin and p16INK4A through binding to the E-box(es) located in the proximal promoter of E-cadherin, p16INK4A and exon-1 of p16INK4A. PRC, polycomb repressive complex.