Direct regulation of TWIST by hypoxia-inducible factor-1
(HIF-1)
promotes metastasis
Muh-Hwa Yang2,3,6,9, Min-Zu Wu1,9, Shih-Hwa Chiou2,4, Po-Min Chen3,6, Shyue-Yih
Chang5,6, Chung-Ji Liu7, Shu-Chun Teng8 and Kou-Juey Wu1,2,6,10
1Institutes of Biochemistry & Molecular Biology, and 2Clinical Medicine, National
Yang-Ming University, Taipei 112; 3Division of Hematology-Oncology, Departments
of Medicine, 4Medical Research & Education, 5Otolaryngology, and 6Genomic
Research Center, Taipei Veterans General Hospital, Taipei 112; 7Department of
Dentistry, Taipei Mackay Memorial Hospital, Taipei 104; 8Graduate Institute of
Microbiology, College of Medicine, National Taiwan University, Taipei 100, Taiwan
9: These authors contributed equally to this work.
10: Correspondence should be addressed to: K.J.W. (Email: [email protected])
Kou-Juey Wu, Institute of Biochemistry and Molecular Biology, National Yang-Ming University, No.155, Li-Nong St., Sec.2, Peitou, Taipei 112, Taiwan; Tel: 886-228267328; Fax: 886-228264843
Stabilization of hypoxia-inducible factor-1 (HIF-1) transcriptional complex caused by intratumoral hypoxia promotes tumor progression/metastasis, which leads to treatment failure and mortality in different types of human cancer. TWIST is a master regulator of gastrulation and mesoderm specification and is recently implicated to be essential to mediate cancer metastasis. The mutant phenotypes of HIF-1 and TWIST null mice show overlapping similarities. Here
we show that hypoxia/HIF-1 overexpression promotes epithelial-mesenchymal
transition (EMT) and metastastic phenotypes. HIF-1 regulates the expression of
TWIST through direct binding to the hypoxia response element (HRE) in the TWIST proximal promoter. siRNA mediated repression of TWIST in
HIF-1overexpressing or hypoxic cells reverts EMT and metastastic phenotypes.
Co-expression of HIF-1/TWIST/Snail in primary tumors of head and neck cancer
patients correlates with metastasis and the worst prognosis. These results provide a key signaling pathway of HIF-1, and implicate the critical role of TWIST in metastasis caused by intratumoral hypoxia.
Introduction
Intratumoral hypoxia followed by stabilization and activation of HIF-1 transcriptional complex is one of the most important mechanisms promoting tumor aggressiveness, leading to metastasis and patient mortality1-3. Epithelial-mesenchymal transition
(EMT), one of the central mechanisms to induce invasion and metastasis of tumors, is a process by which epithelial cells lose their polarity and are converted to a mesenchymal phenotype4-6. Snail is a zinc-finger transcriptional repressor shown to
induce epithelial-mesenchymal transition through the repression of E-cadherin expression, one of the hallmarks of EMT7-9. Overexpression of Snail mediated
invasion and metastasis of different types of human cancer7-10. Other EMT inducers
(or regulators) (e.g. ZEB1, SIP1, Slug, E47, TWIST) were also shown to induce EMT and metastasis through the repression of E-cadherin9. Snail was implicated in the
initial migratory phenotype of tumor cells; whereas maintenance of the migratory phenotype and malignancy could be mediated by Zeb1 and other EMT regulators, suggestive of the differential and hierarchical role for these EMT regulators9. Hypoxia
(or HIF-1 induction) is an important microenvironmental factor which induces the expression of certain EMT regulators such as Snail, Zeb1, SIP1, E47, and LOX (lysyl oxidase) and coordinates the interplay of these EMT regulators9,11-14. However,
whether hypoxia could activate other critical EMT regulators which may play a differential and non-redundant role remains to be explored.
TWIST is a master regulator of gastrulation and mesoderm specification15,16 and
is recently demonstrated to be essential to mediate cancer metastasis17,18. The mutant
phenotypes of HIF-1 and TWIST null mice show overlapping similarities (i.e. defects in somites and abnormal neural fold formation)19-21 and implicate that these
two genes may lie in the same pathway to regulate development as well as tumor progression and metastasis.
In this report, we show that hypoxia or HIF-1 overexpression promotes epithelial-mesenchymal transition and metastastic phenotypes in vitro and in vivo. HIF-1 regulates the expression of TWIST through a hypoxia response element (HRE) located in the TWIST proximal promoter. siRNA mediated repression of TWIST in HIF-1overexpressing or hypoxic cells reverts EMT and metastastic phenotypes. Co-expression of HIF-1, TWIST, and Snail in primary tumors of head and neck squamous cell carcinoma (HNSCC) patients correlates with metastasis and the worst prognosis. These results provide a key signaling pathway elucidating the promotion of tumor progression and metastasis by HIF-1, and implicate the critical role of TWIST in metastasis caused by intratumoral hypoxia.
Results
Hypoxia or HIF-1 overexpression induces EMT. Previous results showed that HIF-1 overexpression induces invasion and is associated with repression of E-cadherin12,22. Here we further demonstrated that hypoxia or HIF-1overexpression
promotes epithelial-mesenchymal transition (EMT), one of the major mechanisms mediating metastasis4-6. Cell lines expressing a low level of endogenous HIF-1
(hypopharyngeal cancer FADU and breast cancer MCF-7) were selected for hypoxia or HIF-1 overexpression experiments. Induction of EMT was demonstrated in both cells after hypoxia as shown by a shift of epithelial markers (E-cadherin and plakoglobin) to mesenchymal markers (vimentin and N-cadherin)5,6,17,18(Fig. 1a).
Immunofluorescence staining of E-cadherin and vimentin confirmed the EMT-associated shift in markers (Fig. 1b). We also demonstrated the increase in migration and invasion of both cell lines, a trait of EMT phenotype, under hypoxia (Fig. 1c). To confirm that HIF-1 plays a significant role in the induction of EMT, FADU cells with constitutive expression of a HIF-1 mutant containing deletion of oxygen degradation domain (∆ODD)23 were generated, in which HIF-1 functions in the
presence of aerobic atmosphere. The results showed that overexpression of HIF-1(∆ODD) in FADU cells induced EMT and increased migration and invasion (Fig. 1d-f). HIF-1(∆ODD) overexpression in FADU cells also increased the number of
pulmonary metastatic lesions under normal oxygen concentration using a tail-vein metastasis assay (Fig. 2a,b). To confirm that HIF-1 activation increases in vivo metastasis, a green-fluorescence protein (GFP)-traced tail vein metastasis assay was performed in NOD-SCID mice by injection of GFP expressing FADU-HIF1(∆ODD) clones. A significant increase in the intensity of green fluorescence was noted in the experimental group compared with the FADU controls (Fig. 2c,d). Immunohistochemistry showed that HIF-1 was overexpressed in the metastatic lung lesions of the experimental mice group, but not in the control mice group (see Supplementary Information, Fig. S2a). To further confirm the induction of metastasis by HIF-1, orthotopic injections of FADU-HIF1(∆ODD) vs. FADU control cells into mice showed that there was a significant increase in local invasion, distant metastasis to lung, and tumor size in the FADU-HIF1(∆ODD) vs. the control FADU cells (Fig. 2e,f and Fig. 1g,h). Short-interference RNA (siRNA) mediated repression of HIF-1 expression in a lung cancer H1299 cell line (H1299-HIF1-si cells) was also performed and the results were consistent with the above observations: a shift of mesenchymal markers to epithelial markers, and a decrease in migration, invasion, and metastasis of H1299 cells when HIF-1 expression was repressed (see Supplementary Information, Fig. S1a-c). Orthotopic implantation assays also showed a significant decrease in local invasion (data not shown), distant metastasis to the
opposite lung and tumor size in H1299-HIF1-si cells vs. H1299 control cells (see Supplementary Information, Fig. S1c,d). Finally, to test whether HIF-1 was mostly responsible for the EMT phenotypes induced by hypoxia, siRNA mediated repression of HIF-1 was performed in FADU or MCF-7 cells undergoing hypoxia. The results showed that repression of HIF-1 caused a complete or partial shift of EMT markers back to the pre-hypoxic status (see Supplementary Information, Fig. S1e,g) and a significant decrease in migration and invasion induced by hypoxia (see Supplementary Information, Fig. S1f,h), demonstrating that HIF-1 is mostly responsible for EMT and metastatic phenotypes induced by hypoxia. EMT was also observed in the invasive front of head and neck squamous cell carcinoma (HNSCC) samples as demonstrated by a shift in EMT markers (downregulation of E-cadherin and upregulation of vimentin expression) and morphologic changes (conversion to a more fibroblastoid morphology, a decrease in keratinization and a loss of cellular cohesiveness)24(see Supplementary Information, Fig. S2b). Our results demonstrated
that hypoxia or HIF-1 overexpression induced the process of EMT and promoted migration, invasion, and metastasis.
Activation of TWIST by hypoxia or HIF-1 Due to the implications of HIF-1 and TWIST in cancer metastasis1-3,17,18 and the overlapping phenotypes of HIF-1 and
upregulates TWIST expression using quantitative real-time PCR and Western blot analysis. The expression levels of a HIF-1 downstream target (VEGF)25 and a TWIST
downstream target (N-cadherin)26 were also examined. Upregulation of mRNA and
protein levels of TWIST, VEGF and N-cadherin was shown in cells under hypoxia or constitutive expression of HIF-1(∆ODD) (Fig. 3a,b). The correlation of HIF-1 and TWIST expression was confirmed in HIF-1 repressed H1299 cells (H1299-HIF1-si cells) (Fig. 3c) and in tail vein-injected FADU-HIF1(∆ODD) cells metasta(H1299-HIF1-sizing to lung (see Supplementary Information, Fig. S2a). Finally, repression of endogenous
HIF-1by siRNA in either FADU or MCF-7 cells under hypoxia showed that TWIST RNA and protein levels were decreased to pre-hypoxic status (Fig. 3d), demonstrating that HIF-1is the major regulator of TWIST expression. The RNA and protein levels of other EMT regulators (such as Snail, LOX, Zeb1, SIP1, and E47)11-14
were examined in cell lines under hypoxia or overexpression of HIF1(∆ODD). The results showed that the RNA levels of LOX, Zeb1, and SIP1 as well as the protein levels of LOX, Snail, Zeb1, and SIP1 increased in FADU or MCF-7 cells under hypoxia and in FADU-HIF1(∆ODD) vs. the control FADU cells, and decreased correspondingly in H1299-HIF1-si vs. H1299 cells (see Supplementary Information, Fig. S3a-c). Finally, repression of HIF-1 in FADU and MCF-7 cells under hypoxia showed the corresponding decrease in the RNA and protein levels of the same EMT
regulators activated by hypoxia (with the exception of moderate decrease of SIP1 in MCF-7 cells, see Supplementary Information, Fig. S3d). These results demonstrated that hypoxia or HIF-1 is responsible for the induction of TWIST and certain EMT regulators.
Regulation of TWIST by HIF-1 through a HRE. In order to demonstrate whether
TWIST is directly regulated by HIF-1, a putative hypoxia response element (HRE)
was identified in the proximal promoter of the TWIST gene (Fig. 4a; for detailed sequence information, see Supplementary Information, Table S1). Transient transfections were performed to investigate whether the TWIST promoter is activated by hypoxia or HIF-1 overexpression. A two to three fold increase in the TWIST promoter activity was demonstrated after hypoxia, transient transfection with a wild type HIF-1 or HIF-1(∆ODD) expression vector. A further increase in the promoter activity (~5 fold) was observed in cells undergoing hypoxia plus HIF-1 overexpression. Transfection with a HIF-1 inactive mutant (HIF-1(LCLL))27 did
not activate the TWIST promoter. Site-directed mutagenesis of the putative HRE in the TWIST promoter eliminated the activation under hypoxia or HIF-1 overexpression (Fig. 4b). Similar results were observed in transfections of the same plasmids into SAS (a HNSCC cell line), MCF-7, and H1299 vs. H1299-HIF1-si cells (see Supplementary Information, Fig. S4a,b). To investigate whether HIF-1
upregulates TWIST expression by direct binding to its promoter, electrophoretic mobility shift assay (EMSA) was performed. The result showed that increased HIF-1 binding was observed after incubation of nuclear extracts of hypoxic cells with the HRE-containing oligonucleotide from the TWIST promoter, and a supershifted band was noted after adding an anti-HIF-1 or anti-HIF-1 specific antibody to the nuclear extracts of hypoxic cells or FADU-HIF1(ODD) cells (Fig. 4c and see Supplementary Information, Fig. 4c,d). Competition of HIF-1 binding by cold competitors abolished the HIF-1 shifted band, and probes containing mutated HRE did not show any HIF-1 binding band (Fig. 4c, data not shown, and see Supplementary Information, Fig. S4e). Chromatin immunoprecipitation (ChIP) assays confirmed the direct binding of HIF-1 to the HRE of the TWIST promoter. PCR amplification of the anti-HIF-1antibody immunoprecipitants showed that the PCR-amplified fragment in the TWIST promoter containing the HRE (198 bp) existed in the FADU-HIF1(∆ODD) sample but not in the FADU-cDNA3 sample; whereas knockdown of HIF-1 in H1299 cells attenuated the intensity of the 198 bp fragment (H1299-control vs. H1299-HIF-1-si)(Fig. 4d, upper panels). Control experiments to PCR-amplify the chromatin immunoprecipitants from the same set of samples showed the positive control band (262 bp) in the VEGF promoter (Fig. 4d, lower panels). All the results demonstrated that HIF-1 directly activated the expression of TWIST by
binding to the HRE in the TWIST proximal promoter.
TWIST is critical for HIF-1hypoxia mediated EMT. To demonstrate whether TWIST is critical in HIF-1 mediated EMT and metastatic phenotypes, siRNA mediated repression of TWIST expression was performed in FADU-HIF1(∆ODD) and H1299 cells. Down-regulation of TWIST in FADU-HIF1(∆ODD) and H1299 clones (i.e. FADU-HIF1(∆ODD)-Twist-si and H1299-Twist-si cells) caused the decrease in N-cadherin expression and the shift of mesenchymal markers (vimentin, N-cadherin) to epithelial markers (E-cadherin, plakoglobin) compared to the control FADU-HIF1(∆ODD) and H1299 cells (Fig. 5a-c). Induction of metastatic phenotypes caused by HIF-1(∆ODD) overexpression was abolished by repression of TWIST as shown by migration, invasion, tail vein injections or orthotopic implantation assays (Fig. 5d,e). Similar results were observed in H1299 cells undergoing TWIST repression (Fig. 5d). To further confirm the role of TWIST in HIF-1 mediated EMT, reconstitution of TWIST expression was performed by transfection of a TWIST expression vector containing siRNA target-site mismatched sequence into FADU-HIF1(∆ODD) cells receiving siRNA against wild type TWIST (FADU-HIF1(∆ODD)-Twist-si clone). Induction of EMT and metastatic phenotypes was observed again after reconstitution of TWIST (Fig. 6a-c). Overexpression of TWIST also induced EMT and metastatic phenotypes in FADU cells (data not
shown), consistent with a previous report17. Finally, repression of endogenous TWIST
caused almost complete reversion of EMT markers and inhibition of migration and invasion in FADU or MCF-7 cells under hypoxia (see Supplementary Information, Fig. S5a,b and data not shown). Takentogether, these results demonstrated the critical role of TWIST in the induction of EMT and metastatic phenotypes caused by constitutive expression of HIF-1(∆ODD) or hypoxic conditions.
TWIST is non-redundant vs. other EMT inducers. To test the functional relations
between TWIST and other EMT inducers, TWIST overexpression was generated in Snail repression status by siRNA in FADU and H1299 cells. The results showed that under Snail repression status TWIST overexpression or hypoxia only partially rescued migration and invasion (see Supplementary information, Fig. S5c,d and data not shown). When Snail was repressed by siRNA (i.e. in FADU-Snail-si cells), combination of TWIST overexpression and hypoxia could not achieve the full range of migration and invasion in FADU cells induced by hypoxia and TWIST overexpression (see Supplementary information, Fig. S5d). The full range of migration and invasion in H1299 cells could not be recovered by TWIST overexpression in H1299 cells under Snail repression (data not shown). In addition, in TWIST repressed FADU (FADU-Twist-si) or H1299 (H1299-Twist-si) cells repression of Snail by siRNA further decreased migration and invasion (under
normoxia or hypoxia in FADU-Twist-si cells) (Fig. 6d-f and data not shown). Similar observations were shown in FADU-Twist-si or H1299-Twist-si cells undergoing further repression of LOX (under normoxia or hypoxia in FADU-Twist-si cells) (data not shown). All the above results suggested that Snail or LOX was responsible for part of the migration and invasion activity not regulated by TWIST. Snail, LOX or TWIST may regulate different pathways to mediate metastatic phenotypes.
Co-expression of HIF-1/TWIST/Snail as a marker. Tumors with abundant HIF-1 stabilization through intratumoral hypoxia are more likely to develop metastasis and correlate with poor survival1-3, especially in HNSCC28 and breast cancer29. TWIST or
Snail overexpression also correlates with a worse prognosis of cancer patients10,17,18,30.
To investigate whether activation of TWIST and Snail by HIF-1 indeed occurs in human cancers and to evaluate the prognostic significance of the expression profile of 1, TWIST, and Snail, tissue-microarray immunohistochemistry analysis of HIF-1 and TWIST expressions was performed in 147 sets of HNSCC samples including 56 metastatic cases (see Supplementary Information, Table S2). Snail expression was investigated in the same population30.The functional indicator of HIF-1, carbonic
anhydrase IX (CA IX)31-33, was stained in all samples to further correlate the hypoxic
zones of tumors. The result showed that tumors with 50% hypoxic areas (CA IX 50% staining) significantly correlated with HIF-1 overexpression (50% HIF-1
nuclear expression in tumor cells) (P=0.001, data not shown). A representative case of IHC staining of all four markers was shown in Fig. 7. This result confirmed the good correlation between HIF-1 and CA IX IHC results, which is consistent with previous reports32,33. This result also identified the patient group with hypoxic tumors
(48.3% of primary and 91.1% of metastatic HNSCCs; Fig. 8a). TWIST and Snail overexpression (50% nuclear expression in tumor cells) was shown in 35.4% and 37.4% of primary tumors and in 85.7% and 82.1% of metastatic ones, respectively (Fig. 8a). Considering the correlation between tumor hypoxia or HIF-1 overexpression and activation of TWIST or Snail, tumors with more than 50% hypoxic areas significantly correlated with TWIST or Snail overexpression (a representative case in Fig. 7; P<0.001, 0.001, respectively) (data not shown). The expression gradient of HIF-1 also correlated significantly with that of TWIST or Snail (P<0.001, <0.001, respectively; see Supplementary Information, Table S3). These results indicated that tumor hypoxia correlated with HIF-1 overexpression and consequentially with TWIST and Snail expression. A higher proportion of HIF-1, TWIST, or Snail expression was observed in metastatic tumor samples than in primary ones (Fig. 8a). There was no statistical difference in the expression pattern of HIF-1, TWIST, and Snail between the metastatic samples obtained from cervical nodes or visceral organ (P=0.681, data not shown). For metastasis and survival
analysis, overexpression of HIF-1, TWIST or Snail in primary HNSCCs was associated with a shorter metastasis-free period (Fig. 8b), and the same situation was observed in the overall survival analysis (P<0.001, <0.001, <0.001, respectively) (data not shown). To investigate the prognostic significance of the expression pattern of 1, TWIST, and Snail in HNSCCs, we divided patients into five groups: 1(-), 1(+)Snail(-)TWIST(-), 1(+)Snail(+)TWIST(-), HIF-1(+)Snail(-)TWIST(+), and HIF-1(+)Snail(+)TWIST(+). Kaplan-Meier metastasis-free curves were generated and the log-rank test was used to test for significant difference among groups. The results showed that co-expression of HIF-1, TWIST, and Snail correlated with the shortest metastasis-free period compared with other groups (Fig. 8c). Similar trend was observed in the overall survival analysis, which correlated with metastasis-free period analysis (data not shown). The prognostic effect of co-expression of HIF-1, TWIST, and Snail was independent of other prognostic markers (advanced T stage, N stage) (see Supplementary Information, Table S4). These results demonstrated that activation of TWIST and Snail by HIF-1 to promote metastasis indeed occurred in HNSCC patients and co-expression of HIF-1, TWIST, and Snail was associated with the most aggressive outcome. These results also supported our observations from cell lines (Fig. 6 and see Supplementary Information, Fig. S5a-d) that TWIST and Snail may contribute differentially to metastasis.
Discussion
Our data show that TWIST is a HIF-1 downstream target, which plays a critical role in hypoxia or HIF-1 overexpression mediated EMT and metastatic phenotypes. Although different EMT regulators were shown to be regulated by HIF-1, our
results show that TWIST may regulate different pathways (versus other EMT regulators such as Snail) to mediate metastatic phenotypes, suggestive of its differential and non-redundant role. The critical role of TWIST in metastasis highlighted the involvement of master developmental regulators (e.g. Snail and other EMT regulators) in the process of tumor aggressiveness and metastasis5,6,9,17,18,30. We
propose that hypoxia or HIF-1 induces EMT through the direct activation of TWIST expression, which may represent an early step and a critical mechanism causing hypoxia-induced tumor progression and metastasis.
TWIST is a bHLH transcriptional factor which mediates axis and pattern formation15,16. It is essential for multiple steps of mesoderm development in
Drosophila15,16 and C. elegans34, especially specification of muscle types. Our results
point to the possible role of hypoxia and HIF-1 to regulate muscle development. Our findings also provide mechanistic insights into the limb bud defects presented in both 1 null mice and mice with conditional knockout of limb bud mesenchyme HIF-121,35.
Co-expression of HIF-1, TWIST, and Snail could be used as a valuable marker to predict metastasis and prognosis in HNSCC patients. Whether the same scenario could apply to other types of human cancer requires further investigation. Finally, our discovery pinpoints the importance of using anti-HIF-1 inhibitors for therapeutic intervention and prevention of metastasis in hypoxic tumors3.
METHODS
Cell culture and oxygen deprivation. Human hypopharyngeal carcinoma FADU and embryonic kidney 293T cell lines were described30. Breast cancer MCF-7, lung cancer
H1299, and tongue cancer SAS cell lines were obtained from ATCC. The FADU and MCF-7 cell lines were used due to their low endogenous HIF-1 levels and low metastatic activity. Oxygen deprivation was carried out in an incubator with 1% O2,
5% CO2, and 94% N2 for 18 hours.
Plasmids and stable transfection. The pHA-HIF-123, pHA-HIF-1(∆ODD)23 and
pHA-HIF-1(LCLL)27 plasmids were gifts from L.E. Huang (Univ. of Utah). The
pSUPER-Snail-si and pSUPER-top3α-si plasmids were described30. The
pSUPER-HIF1-si, pSUPER-Twist-si, and pSUPER-LOX-si plasmids were generated by inserting the oligonucleotide containing the specific siRNA target sequence (see Supplementary Information, Table S5) into the pSUPER vector. The pFlag-Twist was generated by insertion of a 694 bp fragment of the full-length human TWIST cDNA from the pOBT7-Twist (Genomic Center, National Yang-Ming University) into the
ECoRV/XbaI sites of the pFlag-CMV vector. The pFlag-rTwist was generated by
site-directed mutagenesis of pFlag-Twist (Fig. 6a). All the stable clones were established by transfection of plasmids as designated by the name of the clones. siRNA knockdown clones were designated by the suffix “-si”.
Protein extraction, Western blot analysis, immunofluorescence, RNA extraction and quantitative real-time PCR. All the above procedures were performed as described30. The characteristics of the antibodies used and the sequences of primers
used in the real-time PCR experiments are listed (see Supplementary Information, Table S6, S5).
In vitro migration and invasion, tail vein metastasis, green fluorescence intensity detection, and orthotopic implantation assays. The in vitro migration and invasion as well as in vivo tail vein metastatic assays were performed as described30. The results
performed in Boyden chambers were designated as “migration” and “invasion” in the text. The results of in vivo tail vein metastatic assays were designated as “metastasis” in the text. For GFP-traced experiments, the HIF1(∆ODD) and FADU-cDNA3 clones were stably transfected with GFP-expressing plasmid PE-GFPC1 and injected into the tail vein of mice. Metastatic foci of lungs were visualized by an illuminating device (LT-9500, Illumatool, TLS) after 4 weeks. The green fluorescence intensity was captured and then analyzed by Image Pro-plus software (Media Cybernetics, USA). For orthotopic tumor implantation assays, 1x106 cells of head and
neck cancer cells (FADU-cDNA3, FADU-HIF1(ODD), or FADU-HIF1(ODD)-Twist-si) in 0.1 ml of PBS were inoculated into the dorsal side of the left tongue, and 1x106 cells of H1299 control or H1299-HIF-1-si cells were implanted into the left
upper lung of mice. Mice were sacrificed 6 weeks after implantation. The counting of metastatic lesions in the cervical area and internal organ of each mouse wasevaluated by gross and microscopic examination. The invasive characteristics of the orthotopic tumors (e.g., vascular invasion, stromal invasion, tumor emboli) were also examined microscopically. This study was approved by the Ethics committee of the Taipei Veterans General Hospital.
Transient transfection and luciferase assays. The TWIST promoter region (-139 to +48 bp surrounding the transcription start site) was cloned by PCR amplification of genomic DNA from 293T cells and inserted into the HindIII/BglII sites of the pXP2 vector36 to generate the pXP2-Twist parental construct (see Supplementary
Information, Table S1). The pXP2-Twist-mutant construct was made by site-directed mutagenesis of the pXP2-Twist vector (Fig. 4a). The reporter constructs were co-transfected into 293T, SAS, and MCF-7 cells with different expression vectors and internal control plasmid under normoxic or hypoxic culture as described36. To confirm
the effect of endogenous HIF-1 repression and its influence on TWIST promoter activity, the same set of plasmids was tested on H1299-control vs. H1299-HIF1-si cells.
Electrophoretic mobility shift assay (EMSA). Nuclear extracts were harvested from 293T cells under normoxic or hypoxic culture and from cDNA3 or
FADU-HIF1(ODD) cells. Oligonucleotides containing wild type (CCTCCTCACGTCAGGCCA) or mutated (CCTCCTACAGTCAGGCCA) HRE sequence were used in EMSA as described36. In competition assays, excessive
amounts of unlabeled competitors were added 5 min prior to adding the labeled probes. For supershift assay, 1 g of monoclonal antibody against HIF-1 or HIF-1 was added and incubated at 4 C for 60 min.
Chromatin immunoprecipitation (ChIP). ChIP assay was performed as described37.
The lysates were incubated with no antibody or antibodies specific for HIF-1, HA, or c-MYC. The PCR reaction generated a 198 bp product from TWIST proximal promoter (-144 to +54 bp) containing HRE (-83 to -79 bp), or a 169 bp product from the distal region without HRE (-1383 to -1215 bp) (see Supplementary Information, Table S1). A 262 bp product from the VEGF proximal promoter (-1174 to -913 bp) was used as a positive control. The primers and antibodies used in ChIP assay are listed (see Supplementary Information, Table S5, S6).
Study population, sample collection and tissue microarray construction. One hundred and forty-seven HNSCC patients who underwent treatment at Taipei Mackay Memorial Hospital and Taipei Veterans General Hospital between January 2001 and December 2004 were retrospectively analyzed. This study has been approved by the Institutional Review Board of Taipei Veterans General Hospital. The clinical
characteristics of 147 HNSCC patients are illustrated (see Supplementary Information, Table S2). Primary tumor samples and the corresponding non-cancerous matched tissue were obtained during surgery; whereas 56 metastatic tumor samples (34 of visceral metastasis and 22 of cervical nodes) were obtained when metastasis occurred. A high-density tissue microarray (TMA) wasconstructed as described30.
Immunohistochemstry (IHC), validation of antibodies and scoring. The sample processing and IHC procedure were performed as described30. The interpretation of
HIF-1, CA IX, Snail and TWIST IHC results was performed independently by two pathologists according to the criteria described previously30-33,38,39. To validate the
HIF-1, TWIST and Snail antibodies used in IHC experiments, immunocytochemistry (ICC) of HIF-1 was performed in 293T cells under normoxia vs. hypoxia and ICC of TWIST or Snail was performed in 293T cells transfected with control vector vs. pFlag-Twist or pcDNA3-Snail. Peptide blocking reagent without adding antibody was applied as the negative control of ICC experiments. The results showed that HIF-1, TWIST, or Snail antibodies could detect HIF-1, TWIST, or Snail located in the nucleus of 293T cells under hypoxia, TWIST or Snail overexpression (see Supplementary Information, Fig. S5e-g). IHC of CA IX was used to correlate the hypoxic zones of tumor and interpreted as described31-33. The
staining; 1+, 1~25%; 2+, 26~50%; 3+, >50% nuclear staining) according to nuclear expression, and only 3+ (>50% nuclear staining) was considered as a positive IHC result30-33,38,39 (see Supplementary Information, Table S3). E-cadherin and vimentin
antibodies were used in IHC to identify tumor invasive front (see Supplementary Information, Fig. S2b). All the antibodies used in IHC are listed (see Supplementary Information, Table S6).
Statistical analysis. The independent Student’s t-test was used to compare the continuous variables between two groups, and the 2 test was applied for comparison
of dichotomous variables. The Kaplan-Meier estimate was used for metastasis-free and overall survival analysis, and the log-rank test was used to compare the difference. The Cox’s proportional hazards model was applied in multivariate survival analysis to test independent prognostic factors. The control groups of all the statistical analyses were usually the first groups in the panels unless specified otherwise in the figure legends. The level of statistical significance was set at 0.05 for all tests.
Acknowledgements
We greatly appreciate K.W. Chang for TMA construction and C.H. Huang, D.H. Hsu for excellent technical assistance. We are grateful to T.Y. Chou and W.Y. Li for providing expert opinions on pathology reading and IHC analysis. This work was supported in part by National Research Program for Genomic Medicine (DOH-96-TD-G-111-002)(K.J.W.), Taipei Veterans General Hospital VGH 96-C1-126, V-96-ER2-008 (M.H.Y.), V-96-ER2-009 (S.Y.C.), National Science Council (NSC-95-2320-B-010-065)(K.J.W.); (95-2314-B-075-083)(M.H.Y.), a grant from Ministry of Education, Aim for the Top University Plan (95A-C-D01-PPG-05, 96A-D-D139) (K.J.W.), and National Health Research Institutes (NHRI-EX-96-9611BI)(K.J.W.).
Author Contribution
K.J.W., M.H.Y., and M.Z.W. conceived and designed the experiments. M.H.Y. and M.Z.W. performed the experiments with the assistance of S.H.C. for in vivo work. M.H.Y., M.Z.W., S.C.T. and K.J.W. analyzed the data. K.J.W. and M.H.Y. wrote the paper with the assistance of S.C.T. The treatment and sample collection of head and neck cancer patients were performed by S.Y.C., C.J.L., P.M.C. and M.H.Y.
Author Information
competing financial interests. Correspondence and requests for materials should be addressed to K.J.W. ([email protected]).
References
1. Gupta, G. P. & Massague, J. Cancer metastasis: building a framework. Cell 127, 679-695 (2006).
2. Harris, A.L. Hypoxia-a key regulatory factor in tumor growth. Nat. Rev. Cancer 2, 38-47 (2002).
3. Semenza, G. L. HIF-1 and tumor progression: pathophysiology and therapeutics.
Trends Mol. Med. 8 (4 Suppl), S62-67 (2002).
4. Thompson, E. W., Newgreen, D. F. & Tarin, D. Carcinoma invasion and metastasis: a role for epithelial-mesenchymal transition? Cancer Res 65, 5991-5995 (2005).
5. Thiery, J. P. Epithelial-mesenchymal transitions in tumour progression. Nat. Rev.
Cancer 2, 442-454 (2002).
6. Thiery, J. P. & Sleeman, J. P. Complex networks orchestrate epithelial-mesenchymal transitions. Nat. Rev. Mol. Cell. Biol. 7, 131-142 (2006).
7. Batlle, E. et al. The transcription factor snail is a repressor of E-cadherin gene expression in epithelial tumour cells. Nat Cell Biol 2, 84-89 (2000).
8. Cano, A. et al. The transcription factor snail controls epithelial-mesenchymal transitions by repressing E-cadherin expression. Nat Cell Biol 2, 76-83 (2000). 9. Peinado, H., Olmedo, D. & Cano, A. Snail, ZEB, and bHLH factors in tumour
415-428 (2007).
10. Moody, S. E. et al. The transcriptional repressor Snail promotes mammary tumor recurrence. Cancer Cell 8, 197-209 (2005).
11. Imai, T. et al. Hypoxia attenuates the expression of E-cadherin via up-regulation of SNAIL in ovarian carcinoma cells. Am. J. Pathol. 161, 1437-1447 (2003). 12. Krishnamachary, B. et al. Hypoxia-inducible factor-1-dependent repression of
E-cadherin in von Hippel-Lindau tumor suppressor-null renal carcinoma mediated by TCF3, ZFHX1A, and ZFHX1B. Cancer Res. 66, 2725-2731 (2006).
13. Evans, A.J. et al. VHL promotes E2 box-dependent E-cadherin transcription by HIF-mediated regulation of SIP1 and Snail. Mol. Cell. Biol. 27, 157-169 (2007). 14. Erler, J. T., et al. Lysyl oxidase is essential for hypoxia-induced metastasis.
Nature 440, 1222-1226 (2006).
15. Castanon, I. & Baylies, M.K. Twist in fate: evolutionary comparison of Twist structure and function. Gene 287,11-22 (2002).
16. 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).
17. Yang, J., et al. Twist, a master regulator of morphogenesis, plays an essential role in tumor metastasis. Cell 117, 927-939 (2004).
metastasis through induction of epithelial-mesenchymal transition. Clin. Cancer
Res. 12, 5369-5376 (2006).
19. Ryan, H. E. Lo, J., & Johnson, R. S. HIF-1 is required for solid tumor formation and embryonic vascularization. EMBO J. 17, 3005-3115 (1998).
20. Iyer, N. V. et al. Cellular and developmental control of O2 homeostasis by hypoxia-inducible factor 1. Genes Dev. 12, 149-162 (1998).
21. Chen, Z. F. & Behringer, R. R. Twist is required in head mesenchyme for cranial neural tube morphogenesis. Genes Dev. 9, 686-699 (1995).
22. Krishnamachary, B., et al. Regulation of colon carcinoma cell invasion by hypoxia-inducible factor 1. Cancer Res. 63, 1138-43 (2003).
23. 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 SA 95, 7987-7992 (1998). 24. Sawair, F. A., et al. Invasive front grading: reliability and usefulness in the management of oral squamous cell carcinoma. J. Oral Pathol. Med. 32, 1–9 (2003).
25. Hirota, K. & Semenza, G. L. Regulation of angiogenesis by hypoxia-inducible factor 1. Crit. Rev. Oncol. Hematol. 59, 15-26 (2006).
26. Alexander, N. R., et al. N-cadherin gene expression in prostate carcinoma is modulated by integrin-dependent nuclear translocation of Twist1. Cancer Res. 66,
3365-3369 (2006).
27. Koshiji, M., et al. HIF-1alpha induces cell cycle arrest by functionally counteracting Myc. EMBO J. 23, 1949-1956 (2004).
28. 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). 29. Pugh, C. W., Gleadle, J. & Maxwell, P..H. Hypoxia and oxidative stress in breast
cancer. Hypoxia signalling pathways. Breast Cancer Res. 3, 313-317 (2001).
30. 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).
31. Koukourakis, M. I., et al. Endogenous Markers of Two Separate Hypoxia Response Pathways (hypoxia inducible factor 2 alpha and carbonic anhydrase 9) Are Associated With Radiotherapy Failure in Head and Neck Cancer Patients Recruited in the CHART Randomized Trial. J. Clin. Oncol. 24, 727-735 (2006). 32. Hui, E. P., et al. Coexpression of hypoxia-inducible factors 1 and 2, carbonic
anhydrase IX, and vascular endothelial growth factor in nasopharyngeal carcinoma and relationship to survival. Clin. Cancer Res. 8, 2595-2604 (2002). 33. Trastour, C., et al. HIF-1alpha and CA IX staining in invasive breast carcinomas:
prognosis and treatment outcome. Int. J. Cancer 120, 1451-1458 (2007).
plays an essential role in non-striated muscle development. Development 127, 2041-2051 (2000).
35. Provot, S. et al. Hif-1regulates differentiation of limb bud mesenchyme and joint development. J. Cell Biol. 177, 451-464 (2007).
36. Wu, K. J., et al. Direct activation of TERT transcription by c-MYC. Nature Genet.
21, 220-224 (1999).
37. Chiang, Y. C., Teng, S. C., Su, Y. N., Hsieh, F. J. & Wu, K. J. c-Myc directly regulates the transcription of the NBS1 gene involved in DNA double-strand break repair. J. Biol. Chem. 278, 19286-19291 (2003).
38. Tickoo, S. K., et al. Immunohistochemical expression of hypoxia inducible factor-1alpha and its downstream molecules in sarcomatoid renal cell carcinoma. J.
Urol. 177, 1258-1263 (2007).
39. Maestro, R., et al. Twist is a potential oncogene that inhibits apoptosis. Genes
Figure Legends
Figure 1. Hypoxia or HIF-1(∆ODD) overexpression induces
epithelial-mesenchymal transition and causes an increase in migration, invasion and tumor size. a. Western blot analysis of HIF-1, epithelial markers (E-cadherin and plakoglobin) and mesenchymal markers (vimentin and N-cadherin) of FADU and MCF-7 cells under normoxic (N) or hypoxic (H) culture. -actin was used as a loading control. b. Immuofluorescence staining of E-cadherin and vimentin in FADU and MCF-7 cells under normoxic (N) or hypoxic (H) culture. The green signal represented the staining of E-cadherin; whereas the red signal represented the staining of vimentin. The blue signal represented nuclear DNA staining by Hoechst 33342. c. Fold change of migration and invasion in FADU or MCF-7 cells under normoxic (N) or hypoxic (H) culture (n=3). d. Western blot analysis of HIF-1(∆ODD), epithelial markers and mesenchymal markers in HIF-1(∆ODD) overexpressing clones (FADU-HIF1(∆ODD)) vs. control clones (FADU-cDNA3). e. Immunofluorescence staining of E-cadherin and vimentin in FADU-HIF1(∆ODD) vs. FADU-cDNA3 cells. f. Fold change of migration and invasion in HIF1(∆ODD) vs. FADU-cDNA3 clones (n=3). g. Orthotopic implantation of FADU-HIF1(∆ODD) vs. FADU-cDNA3 cells showed the increased tumor mass of FADU-HIF1(∆ODD) cells (right panel). h. A significant increase in tumor size was observed in
FADU-HIF1(∆ODD) clones vs. the FADU-cDNA3 clones (n=6). The asterisk (*) indicated statistical significance (P<0.05) between experimental and control clones. The error bars indicate s.d. The scale bars represent 20 m in b and e. The full length blots/scans of Western blot are presented in the Supplementary Information, Fig. S6.
Figure 2. Constitutive expression of HIF-1(∆ODD) in FADU cells promotes
metastasis and local invasion in vivo. a. Photographic pictures and H&E staining of
HIF-1(∆ODD) of the lungs of NOD-SCID mice 6 weeks after tail-vein injection of FADU-HIF1(∆ODD) vs. FADU-cDNA3 cells. Black arrows indicated metastatic nodules. b. Quantification of the average numbers of metastatic foci in the lungs of mice (n=6). The statistical difference (*) of the number of metastatic nodules in two independent clones of FADU-HIF1(ODD) vs. control clones is P=0.050, 0.044. c. Green fluorescence emitted from tumor tissues residing in lungs of mice 4 weeks after injection of FADU-HIF1(∆ODD)-GFP (lower) vs. FADU-cDNA3-GFP (upper) cells. The right-sided panels indicated the lungs taken out for examination. d. The average intensity of GFP signals from the tumor tissues residing in lungs between the experimental and control mice (n=6). The statistical difference (*) of GFP intensity in two FADU-HIF1(ODD)-GFP clones vs. control clones is P<0.001, <0.001. e. Orthotopic implantation of FADU-HIF1(∆ODD) vs. FADU-cDNA3 cells showed
the extensive and diffuse local invasion of FADU-HIF1(∆ODD) cells (H&E, right lower panel). Arrows indicate tumor invasion. Tumor emboli were visible in the vessels of FADU-HIF1(∆ODD) cells forming tumor (H&E, right upper panel) but not in the group of FADU-cDNA3 cells (H&E, left upper panel). Arrows indicate nearby venous area. f. The increase in pulmonary metastatic nodules of FADU-HIF1(∆ODD) cells after orthotopic implantation was shown (n=6). The asterisk (*) indicated statistical significance (P<0.05) between experimental and control clones. The error bars indicate s.d. The scale bars represent 400 m in a and lower panels of
e, 200 m in upper panels of e.
Figure 3. Hypoxia or constitutive expression of HIF-1(∆ODD) upregulates TWIST expression. a, Upper: fold change of mRNA levels of HIF-1, VEGF,
TWIST and N-cadherin by real-time PCR analysis in FADU or MCF-7 cells under
normoxia or hypoxia (n=3); lower: Western blot analysis of TWIST expression in FADU or MCF-7 cells under normoxia or hypoxia. b, Upper: relative mRNA expression levels of HIF-1, VEGF, TWIST, and N-cadherin in FADU-HIF1(∆ODD) vs. FADU-cDNA3 cells (n=3); lower: TWIST protein levels in FADU-HIF1(∆ODD) vs. FADU-cDNA3 cells. c, Decreased expression of HIF-1,
H1299-HIF1-si clones (n=3 in upper pane;). Knockdown of unrelated protein topoisomerase 3 (H1299-top3-si) and transfection with an empty vector (H1299-cont.) were used as controls. d. siRNA mediated repression of endogenous HIF-1 abolishes the induction of TWIST (mRNA and protein levels) in FADU or MCF-7 cells under hypoxia (n=3 in upper panels). The different bar graph symbols represented different molecules, which applied to all four panels. The asterisk (*) indicated statistical significance (P<0.05) between experimental and control clones. The error bars indicate s.d. The full length blots/scans of Western blot are presented in the Supplementary Information, Fig. S6.
Figure 4. HIF-1 upregulates TWIST expression by direct binding to the hypoxia response element (HRE) located in the proximal promoter of TWIST gene. a,
Schematic representation of the promoter region of TWIST and the reporter constructs used in HIF-1 transfection experiments. The constructs contained wild type (pXP2-Twist) or mutated (pXP2-Twist-mut) HRE located -83 to -79 bp upstream of the transcription start site of TWIST. b, Activation of pXP2-Twist or pXP2-Twist-mut after co-transfection of wild type HIF-1, HIF-1 ODD deletion mutant (∆ODD), or inactive HIF-1 mutant (LCLL) under normoxic or hypoxic culture (n=3). The luciferase activity/-galactosidase of 293T cells co-transfected with pXP2-Twist and
pcDNA3 control vector under normoxic condition was applied as the baseline control of other experiments, and the asterisk (*) indicated statistical significance (P<0.05) between experimental and control transfections. The error bars indicate s.d. c, Electrophoretic mobility shift assay (EMSA) and supershift assay. Oligonucleotides for EMSA were the 18 bp probe from the TWIST promoter which contained a consensus HRE (lane 1-6). Nuclear extracts prepared from 293T cells under normoxia (lane 2) or hypoxia for 18 hours (lane 3-6) were incubated with -32PATP labeled
probe prior to electrophoresis. The supershift assay (lane 4) was performed in the presence of an anti-HIF-1 antibody. The position of supershifted band was indicated. Competition assay was performed in the presence of 50 fold (lane 5) or 100 fold (lane 6) excess of unlabelled oligonucleotides containing the consensus HRE sequence. No protein extracts were put in lane 1. d, Chromatin immunoprecipitation (ChIP) analysis of FADU-HIF1(∆ODD) vs. FADU-cDNA3 and H1299-HIF1-si vs. H1299-control cells. Two different clones were used for each category. Chromatin was incubated without antibody, with an anti-c-MYC antibody, or with an anti-HA (in FADU samples) or an anti-HIF-1 antibody (in H1299 samples). The 198-bp fragment contains the HRE; whereas the 169-bp fragment does not contain any HRE in the
TWIST promoter. PCR amplification of the 262 bp fragment of VEGF promoter
Figure 5. Knockdown of TWIST in FADU-HIF1(∆ODD) or H1299 cells reverts
EMT and metastasis. a, Western blot analysis of HIF-1(∆ODD), TWIST, epithelial and mesenchymal markers in FADU-HIF1(∆ODD)or H1299 cells receiving siRNA against TWIST (FADU-HIF1(∆ODD)-Twist-si and H1299-Twist-si) vs. control clones (FADU-HIF1(∆ODD) or FADU-HIF1(∆ODD)-top3-si as well as H1299-cont. or H1299-top3-si). b, Immunofluorescence of E-cadherin (green) and vimentin (red) in FADU-HIF1(∆ODD) vs. FADU-HIF1(∆ODD)-Twist-si cells. c, Relative mRNA expression levels of TWIST and N-cadherin in cDNA3, FADU-HIF1(∆ODD)-top3-si, FADU-HIF1(∆ODD) and FADU-HIF1(∆ODD)-Twist-si clones as well as in H1299-cont., H1299-top3-si and H1299-Twist-si clones (n=3).
d, Fold change of migration and invasion of FADU-cDNA3, FADU-HIF1 (∆ODD)-top3-si, FADU-HIF1(∆ODD) and FADU-HIF1(∆ODD)-Twist-si clones as well as in H1299-cont., H1299-top3-si and H1299-Twist-si clones (n=3). The FADU-HIF1(∆ODD)-top3-si and H1299-top3-si clones were selected as the control groups in c & d. e, In vivo metastatic ability of cDNA3, FADU-HIF1(∆ODD) and FADU-FADU-HIF1(∆ODD)-Twist-si clones as assayed by tail vein injection (closed bars;n=6) or orthotopic implantation (open bars; n=6) methods. The FADU-cDNA3-1 was selected as the control group. The asterisk (*) indicated
statistical significance (P<0.05) between experimental and control clones. The error bars indicate s.d. The scale bars represent 20 m in b. The full length blots/scans of Western blot are presented in the Supplementary Information, Fig. S6.
Figure 6. Reconstitution of TWIST expression restores EMT and metastatic phenotypes in FADU-HIF1(∆ODD) cells receiving siRNA to knockdown
endogenous TWIST and repression of Snail by siRNA in FADU-Twist-si cells further decreased migration and invasion. a. Schematic representation of the
construction of a TWIST expression vector (pFlag-rTwist) resistant to siRNA repression by site-directed mutagenesis. b. Western blot analysis of TWIST, epithelial and mesenchymal markers in FADU-HIF1(∆ODD)-Twist-si clones with or without TWIST reconstitution. c. Fold change of migration, invasion, and metastasis of FADU-HIF1(∆ODD)-Twist-si clones with or without TWIST reconstitution (n=3 for migration and invasion assays; n=6 for metastasis assays). d. Western blot analysis of LOX, Snail, and TWIST expression in FADU-Twist-si cells receiving siRNA to further repress Snail under normoxic (N) or hypoxic (H) culture. e. Real-time PCR analysis of the mRNA expression levels of TWIST, Snail, E-cadherin, and
N-cadherin in FADU-Twist-si cells undergoing further repression of Snail under
of FADU-Twist-si cells whose Snail levels were further repressed by siRNA under normoxic (N) or hypoxic (H) culture (n=3). The asterisk (*) indicated statistical significance (P<0.05) between experimental and control clones. The error bars indicate s.d. The full length blots/scans of Western blot are presented in the Supplementary Information, Fig. S6.
Figure 7. Co-expression of CA IX, HIF-1, TWIST, and Snail in a HNSCC case. IHC analysis of co-expression of CA IX , HIF-1, TWIST and Snail in corresponding normal tissue (N), primary tumor (T) and metastatic tumor (M) of a representative HNSCC case. The samples prepared for co-expression analysis were cut and examined at the same region. The black arrows indicate the nuclear expression of HIF-1, TWIST and Snail; whereas the red arrows indicate the membranous expression of CA IX. The photographs were taken at the magnification of 50x (left panels) and 400x (right panels). The scale bars represent 800 m in left panels and 100 m in right panels.
Figure 8. Co-expression of HIF-1, TWIST, and Snail correlates with metastasis
of HNSCC cases. a, Percentage of IHC positivity of CA IX, HIF-1, TWIST and Snail in the N (normal tissue), T (tumor tissue), M (metastatic tumor) samples of
HNSCC cases. b, Kaplan-Meier analysis of metastasis-free period of HNSCC cases with HIF-1(-) vs. HIF-1(+), TWIST(-) vs. TWIST(+), and Snail(-) vs. Snail(+) in primary tumors. c. Subgroup analysis of HNSCC cases according to the expression profile of HIF-1, TWIST, and Snail in primary tumors. Co-expression of HIF-1, TWIST, and Snail (Group 5) indicated the shortest metastasis-free period compared with the other groups. The P values of comparison between each group were shown in the inset.
