Oral squamous cell carcinoma (OSCC) is a cancer subtype that lacks validated prognostic and therapeutic biomarkers, and human papillomavirus status has not proven beneficial in predicting patient outcomes. A gene expression pathway analysis was conducted using OSCC patient specimens to identify molecular targets that may improve management of this disease. RNA was isolated from 19 OSCCs treated surgically at the University of Alabama at Birmingham (UAB; Birmingham, AL) and evaluated using the NanoString nCounter system. Results were confirmed using the oral cavity subdivision of the Head and Neck Squamous Cell Carcinoma Cancer (HNSCC) study generated by The Cancer Genome Atlas (TCGA) Research Network. Further characterization of the in vitro phenotype produced by Notch pathway activation in HNSCC cell lines included gene expression, proliferation, cell cycle, migration, invasion, and radiosensitivity. In both UAB and TCGA samples, Notch pathway upregulation was significantly correlated with patient mortality status and with expression of the proinvasive gene FGF1. In vitro Notch activation in HNSCC cells increased transcription of FGF1 and induced a marked increase in cell migration and invasion, which was fully abrogated by FGF1 knockdown. These results reveal that increased Notch pathway signaling plays a role in cancer progression and patient outcomes in OSCC. Accordingly, the Notch–FGF interaction should be further studied as a prognostic biomarker and potential therapeutic target for OSCC.

Implications: Patients with squamous cell carcinoma of the oral cavity who succumb to their disease are more likely to have upregulated Notch signaling, which may mediate a more invasive phenotype through increased FGF1 transcription. Mol Cancer Res; 14(9); 883–91. ©2016 AACR.

Head and neck squamous cell carcinoma (HNSCC) is a multifactorial and heterogeneous disease encompassing multiple anatomic sites but most commonly arising from the oral cavity [oral squamous cell carcinoma (OSCC)] or oropharynx [oropharyngeal squamous cell carcinoma (OPSCC)]. The majority of HNSCC patients present with locally advanced disease, requiring intense multimodality therapy associated with significant acute and long-term toxicities. Increasing incidence of human papillomavirus (HPV) in OPSCCs has corresponded with decreased recurrence and improved overall survival in this subtype. In contrast, relapse is still common in patients with OSCC and survival remains below 70% (1, 2). One reason for this limited progress may be the lack of reliable biomarkers for OSCCs. HPV is not associated with therapeutic sensitivity or outcomes in tumors from the oral cavity, and therapies targeting EGFR, which is commonly activated in this disease, produce only a modest response (3, 4). In the absence of effective treatments, there is a strong need to identify molecular targets and predictive markers to improve the management of OSCC.

Recently, the Notch signaling pathway has emerged as a candidate to fill this void. Notch signaling plays an important role in cell proliferation, differentiation, and apoptosis during embryogenesis (5, 6). The canonical pathway is activated by the binding of Delta or Jag ligands expressed on the cell surface to Notch receptors located on the surface of a neighboring cell. This trans interaction induces γ-secretase–mediated cleavage of the Notch intracellular domain (NICD), which translocates into the nucleus (7). NICD complexes with and activates CBF1-SU(H)-LAG1 (CSL) and recruits other coactivators such as Mastermind-like (MAML) proteins (8). The resulting complex activates transcription of Notch target genes, including HEY1/2, HES1/5, NFκB, MYC, CCND1, BCL2, and CCR7, in a cell type-specific manner.

Aberrant Notch signaling has been found in a number of solid and hematologic malignancies and can function in either an oncogenic or tumor-suppressive capacity depending on cell type and context (9). For example, Notch signaling activates transcription of cell-cycle regulators p21Cip1 and p27Kip1 in epithelial cells but not in other tissue types, suggesting the Notch pathway may function as a tumor suppressor in squamous cell carcinomas (10, 11). In support of this theory, two separate exome sequencing studies of HNSCC revealed 10% to 15% of tumors had inactivating mutations in NOTCH1 (12, 13). However, this model has been challenged by evidence that HNSCCs display a bimodal pattern of Notch alterations, including a subset of tumors with increased expression of ligands JAG1 and JAG2, as well as the NOTCH1 and NOTCH3 receptors (14–18). In addition, in vitro work has shown a relationship between Notch1 activity and clinically relevant characteristics, such as cancer stem cell–like properties and TNFα-mediated inflammation (15, 16, 18).

HNSCC candidate biomarker studies typically focus on identifying individual genes or proteins that correlated with patient outcomes and tumor biology. We hypothesized that dysregulation at the pathway level may significantly alter signaling activity, tumor behavior, and patient outcomes, which may not be detectable in individual biomarker studies. Thus, we conducted a gene expression pathway analysis to identify cell signaling abnormalities associated with cancer-specific mortality in OSCC. We report a significant association between Notch pathway upregulation and mortality in two independent OSCC datasets. A positive association between Notch activation and upregulation of the proinvasive gene FGF1 was also observed in OSCC samples from the University of Alabama at Birmingham (UAB; Birmingham, AL) and validated using The Cancer Genome Atlas (TCGA) dataset. In vitro, activation of Notch signaling in human head and neck cancer cells altered the expression of Notch pathway genes in a pattern comparable with that observed in patients with a high Notch pathway score and produced a similar increase in transcription of FGF1. These in vitro gene expression changes were accompanied by significantly increased cell migration and invasion in Notch-activated cells, which was reversed upon FGF1 knockdown. Our results suggest that upregulation of the Notch pathway may be a marker of invasive phenotype in OSCC.

Human tissue samples

Tissue samples were obtained from the UAB surgical pathology archives (2000–2009) with Institutional Review Board approval and represent histologically confirmed OSCC. All patients were treated surgically in the Department of Otolaryngology-Head and Neck Surgery at the UAB Hospital. Selected patients had previously untreated primary disease. Formalin-fixed paraffin-embedded tissue samples from a cohort of 19 primary OSCC tumors were used (Table 1).

Table 1.

Patient demographics

OSCC patients w/o CSMOSCC patients w/ CSM
CharacteristicsN = 14%N = 5%P
Gender 
 M 11 79 80 1.00 
 F 21 20  
Average age, y 60.57 64.80 0.61 
Site of origin 
 Mandibular alveolar ridge 14 0.90 
 Gingivobuccal sulcus  
 Floor of mouth 36 40  
 Buccal mucosa  
 Mandible 14 20  
 Floor of mouth + Mandible 14 20  
 Unknown 20  
Nodal disease 
 Yes 13 93 80 1.00 
 No 20  
Average size, cm 3.67 3.60 0.92 
T stage (AJCC) 
 T2 57 0.04 
 T4 43 100  
Bone invasion 
 Yes 43 80 0.30 
 No 57 20  
Perineural invasion 
 Yes 40 0.15 
 No 13 93 60  
Margins positive 
 Yes 36 20 0.61 
 No 64 80  
Adjuvant XRT 
 Yes 50 80 0.34 
 No 50 20  
Average follow-up (mo) 27.09 28.44 0.92 
Average PFS (mo) 26.44 25.44 0.94 
Average OS (mo) 60.92 37.98 0.11 
OSCC patients w/o CSMOSCC patients w/ CSM
CharacteristicsN = 14%N = 5%P
Gender 
 M 11 79 80 1.00 
 F 21 20  
Average age, y 60.57 64.80 0.61 
Site of origin 
 Mandibular alveolar ridge 14 0.90 
 Gingivobuccal sulcus  
 Floor of mouth 36 40  
 Buccal mucosa  
 Mandible 14 20  
 Floor of mouth + Mandible 14 20  
 Unknown 20  
Nodal disease 
 Yes 13 93 80 1.00 
 No 20  
Average size, cm 3.67 3.60 0.92 
T stage (AJCC) 
 T2 57 0.04 
 T4 43 100  
Bone invasion 
 Yes 43 80 0.30 
 No 57 20  
Perineural invasion 
 Yes 40 0.15 
 No 13 93 60  
Margins positive 
 Yes 36 20 0.61 
 No 64 80  
Adjuvant XRT 
 Yes 50 80 0.34 
 No 50 20  
Average follow-up (mo) 27.09 28.44 0.92 
Average PFS (mo) 26.44 25.44 0.94 
Average OS (mo) 60.92 37.98 0.11 

Abbreviations: AJCC, American Joint Committee on Cancer; F, female; M, male; OS, overall survival; PFS, progression free survival; w/, with; w/o, without; XRT, radiotherapy.

Cell culture and reagents

HNSCC cell lines UM-SCC1 and UM-SCC6 were obtained courtesy of Thomas E. Carey (University of Michigan, Ann Arbor, MI; 1993). The HNSCC cell line FaDu (HTB-43) was obtained from ATCC (2011). Cell lines have not been authenticated by our laboratory. All cell lines were maintained in DMEM growth medium (Sigma) supplemented with 10% FBS (SAFC Biosciences) and 1% penicillin/streptomycin (Gibco). Notch signaling was activated by coating cell culture dishes with 1.0 μg/mL recombinant human DLL4 (R&D Systems, 1506-D4-050) for 18 hours at 4°C prior to plating. FGF1 expression was inhibited by transfection with FGF1 siRNA (Santa Cruz Biotechnology, sc-39444) using Lipofectamine RNAiMAX (Invitrogen, REF 13778-150) as per manufacturer's instructions. Control siRNA-A (Santa Cruz Biotechnology, sc-37007) was used for comparison. Cells were transfected for 48 hours, then trypsinized, centrifuged, and replated for use in experiments. Where indicated, FGF1 knockdown was complemented by recombinant FGF1 (R&D Systems, 232-FA-025), which was added to growth media at 100 ng/mL.

RNA isolation and NanoString PanCancer Pathways analysis

For patient samples, RNA was harvested from areas of >70% tumor, as identified by a pathologist, using the High Pure FFPET RNA Isolation Kit (Roche, REF 06 650 775 001) as per the manufacturer's instructions. For cell lines, RNA was harvested using the PureLink RNA Mini Kit (Ambion, no. 12183018A). RNA samples had a concentration ≥12.5 ng/μL and an A260/280 ratio between 1.7 and 2.3 as determined by spectrophotometer. mRNA was analyzed by the UAB NanoString Laboratory (www.uab.edu/medicine/radonc/en/nanostring; ref. 19). Samples were processed for analysis on the NanoString nCounter Flex system using the PanCancer Pathways Plus panel as per the manufacturer's instructions (NanoString Technologies). RCC data files were imported into NanoString nSolver 2.5 and further analyzed using the PanCancer Pathways Advanced Analysis Module with the dynamic housekeeping gene selection option.

Analysis of TCGA OSCC data

The results shown here are in part based upon data generated by the TCGA Research Network (http://cancergenome.nih.gov/) viewed in the cBioPortal for Cancer Genomics (http://www.cbioportal.org/; refs. 20, 21). Gene expression results were obtained from RNASeq V2 RSEM normalized results for samples from the oral cavity anatomic subdivision of the Head and Neck Squamous Cell Carcinoma study (22) with at least 12 months of survival data (n = 71). Notch pathway gene results were multiplied by the NanoString pathway score gene weights to derive a Notch pathway score for TCGA samples.

Protein expression

Protein was analyzed via SDS-PAGE as described previously (23). Cleaved Notch1 (Val1744; Cell Signaling Technology, #4147) primary antibody was used at a 1:750 dilution for immunoblotting. β-Actin (Santa Cruz Biotechnology, catalog #sc-47778) levels were analyzed as a loading control. Knockdown of FGF1 was confirmed by ELISA using the Thermo Scientific Pierce Human FGF1 ELISA Kit (Thermo Scientific, EHFGF1) as per the manufacturer's instructions.

Scratch assay

Cells were plated at high density in 6-well plates and reached monolayer confluence within 10 hours. A 200-μL pipette tip was used to create a scratch. Cells were washed twice with PBS and then covered with normal DMEM media. Scratches were imaged at 0, 6, 12, and 24 hours with an EVOSfl digital inverted microscope using the transmitted light application at 2× magnification. Wound closure % = (scratch width in pixels at time point/average scratch width in pixels at time 0) × 100.

Transwell invasion assay

Matrigel basement membrane matrix (40 μL; Corning, catalog #354234) was added to 24-well cell culture inserts (Falcon, REF 353097) and solidified at 37°C for 20 minutes. A total of 40 μL of either DLL4 or PBS vehicle was added to the top of the membrane and incubated at 37°C for another 10 minutes. HNSCC cells were washed, trypsinized, centrifuged, and resuspended at a density of 1 × 106 cells/mL in serum-free DMEM. A total of 200,000 cells were added to the top of the Matrigel-coated membrane. DMEM (600 μL) with 20% FBS was added to the well below the membrane. Cells were incubated at 37°C for 24 hours. After 24 hours, Matrigel and media were removed from the top of the membrane using a cotton-tipped applicator. Inserts were dipped into distilled water to remove remaining media, then stained in 0.5% crystal violet/25% glutaraldehyde/H2O for 20 minutes. Inserts were then dipped into distilled water to remove remaining stain and allowed to dry. Membranes were imaged using a Zeiss Axio Observer.A1 manual inverted microscope with AxioCam MRc5 camera and AxioVs40 software (V4.8.2.0, 2006–2010, Carl Zeiss MicroImaging GmbH) at 10× magnification. Five fields were imaged per insert, and the average number of invaded cells per field was calculated.

qRT-PCR

RNA was harvested from cell lines using the PureLink RNA Mini Kit (Ambion, no. 12183018A). One microgram of RNA was converted into cDNA using SuperScript III First-Strand Synthesis System for RT-PCR (Invitrogen, REF18080-051) as per the manufacturer's recommendations. Expression was quantified by qRT-PCR using 2× TaqMan Universal PCR Master Mix (Applied Biosystems, REF 4304437) and primers (ThermoFisher #4331182) for FGF1 (Hs01092738_m1) and GAPDH (Hs02758991_g1). FGF1 expression = 2−(CtFGF1−CtGAPDH).

Cell proliferation

Cells were seeded in 24-well plates and harvested at 0, 48, and 96 hours. Cells were washed with PBS, trypsinized, and diluted 1:20 in isotonic saline solution (RICCA Chemical, catalog #7210-5). Diluted cells were counted using a Beckman Z1 Coulter particle counter. Cell counts represent cells/mL.

Colony Formation Assay

Clonogenic survival was determined by the colony formation assay as described previously (23). Briefly, cells were treated with the indicated dose of radiation and left undisturbed for 2 weeks. Cells were then fixed and stained (25% glutaraldehyde, 12 mmol/L crystal violet), and the number of colonies (>50 cells) was counted. Survival fraction was calculated using the following formula:

formula

Experiments were performed at least in triplicate.

Cell Cycle Analysis

HNSCC cells were plated in 6-well plates and incubated for 24, 48, or 72 hours. At the indicated time point, cells were washed with PBS, trypsinized, and centrifuged. Pellets were resuspended in 1 mL cold PBS, fixed with 4 mL cold 70% ethanol, and incubated at −20°C for 18 hours. After centrifugation, solvent was removed and cells were resuspended in 300 μL PBS with 1 μL RNase. Samples were incubated at 37°C for 30 minutes and labeled with 30 μL propidium iodide (in PBS at a final concentration of 100 μg/mL) at room temperature for 10 minutes. Samples were then assessed for cell-cycle phase using flow cytometry to detect DNA content.

Statistical analysis

Patient demographics were assessed by t test using GraphPad Prism version 4.02 (GraphPad Software). NanoString data analysis is described above. Pathway scores and differential gene expression was evaluated in GraphPad by t test or Mann–Whitney test, depending on results from the D'Agostino and Pearson omnibus normality test. Cluster analysis of NanoString and RNA sequencing (RNA-seq) data was completed using GENE-E version 3.0.204 (http://www.broadinstitute.org/cancer/software/GENE-E/; Broad Institute, Inc., Cambridge, MA), followed by KEGG gene set enrichment analysis using GSEA v2.2.2 (http://www.broad.mit.edu/gsea/; Broad Institute, Inc.; ref. 24). In vitro phenotypic data were assessed by two-way ANOVA with Bonferroni posttest in GraphPad.

OSCC patient characteristics

Nineteen patients with previously untreated, surgically resectable OSCC were included in this study. Patients were stratified into two cohorts based on disease status at the conclusion of outcomes collection in 2013 (25). The median follow-up for all patients was 27.69 months. The first cohort consisted of 5 patients (26%) who had cancer-specific mortality (CSM) with a median overall survival of 30.55 months (range 16.68–71.78) and median follow-up period of 17.42 months (range 0–70.49). The second cohort consisted of 14 patients (74%) who had a status of alive with disease or no evidence of disease at their last known follow-up (no CSM) with a median follow-up period of 28.56 months (range 0.30–67.04). Demographics and tumor characteristics did not differ significantly between the two cohorts with the exception of increased T stage in the CSM cohort (Table 1).

The Notch signaling pathway is upregulated in OSCC patients with mortality

To investigate the relationship between regulation of signaling pathways and OSCC patient outcomes, we first conducted a pathway analysis using the NanoString platform and the PanCancer Pathways Advanced Analysis Module. The TGFβ and Notch signaling pathways were identified as having the greatest degree of differential regulation between the CSM and no CSM cohorts (Fig. 1A). However, the association between CSM and signaling pathway score reached statistical significance only for the Notch pathway (P = 0.032; Fig. 1B). We further observed that Notch pathway scores were significantly higher in the CSM cohort than in the no CSM cohort, indicating pathway-level upregulation (P = 0.036; Fig. 1C). To validate the association between Notch upregulation and mortality, we examined the OSCC subgroup from TCGA Head and Neck Squamous Cell Carcinoma study and identified 71 patients with at least one year of survival data (22). When the Notch pathway gene weights used in NanoString analysis were applied to normalized RNA-seq data for each sample, the resulting Notch pathway scores were significantly higher in patients with all-cause mortality (ACM; n = 44) than in patients with no ACM (n = 27; P = 0.037; Fig. 1D). We also examined mRNA expression of individual Notch genes to determine which pathway components, if any, drive the association between Notch pathway score and mortality. In OSCC samples from UAB, we identified only four Notch pathway genes that were significantly differentially expressed between the CSM and no CSM cohorts: ligand DLL1, negative regulator HDAC2, transcription target HES5, and ligand JAG1 (Supplementary Table S1). In OSCC samples from TCGA, we found a modest increase in expression of ligand JAG2 and receptor NOTCH3 in the ACM cohort compared with the no ACM cohort (Supplementary Table S1).

Figure 1.

Notch pathway upregulation is associated with mortality in OSCC. Gene expression pathway analysis was performed by NanoString in UAB OSCC samples (AC) and validated using RNA-seq data from TCGA OSCC samples (D). A, heatmap of all pathway scores. Red, low pathway score; yellow, high pathway score. Apop, apotosis; CC, cell cycle; TXmisREG, transcriptional misregulation; ChromMod, chromatin modification; HH, hedgehog. B, pathway significance plot for CSM. The global significance statistic indicates the extent of differential expression of pathway genes with respect to CSM. The x-axis indicates the statistical significance of the association between pathway score and CSM. Dotted line, P = 0.05. C, comparison of Notch pathway scores in UAB OSCC samples with and without CSM using normalized NanoString transcript counts. Lines, median with interquartile range. D, comparison of Notch pathway scores in TCGA OSCC samples with and without ACM using normalized RNA-seq transcript counts. Lines, median with interquartile range. *, P < 0.05.

Figure 1.

Notch pathway upregulation is associated with mortality in OSCC. Gene expression pathway analysis was performed by NanoString in UAB OSCC samples (AC) and validated using RNA-seq data from TCGA OSCC samples (D). A, heatmap of all pathway scores. Red, low pathway score; yellow, high pathway score. Apop, apotosis; CC, cell cycle; TXmisREG, transcriptional misregulation; ChromMod, chromatin modification; HH, hedgehog. B, pathway significance plot for CSM. The global significance statistic indicates the extent of differential expression of pathway genes with respect to CSM. The x-axis indicates the statistical significance of the association between pathway score and CSM. Dotted line, P = 0.05. C, comparison of Notch pathway scores in UAB OSCC samples with and without CSM using normalized NanoString transcript counts. Lines, median with interquartile range. D, comparison of Notch pathway scores in TCGA OSCC samples with and without ACM using normalized RNA-seq transcript counts. Lines, median with interquartile range. *, P < 0.05.

Close modal

Finally, we explored patterns in gene expression data using unsupervised hierarchical cluster and gene set enrichment analyses. In UAB patients, we identified two primary sample clusters that were not driven by CSM status (Supplementary Fig. S1A). Using comparative biomarker analysis, we observed a nonsignificant trend toward enrichment of Notch pathway genes in samples from the CSM cohort with a nominal P value of 0.523 (Supplementary Fig. S1B). Among the most highly ranked upregulated genes in samples from the CSM group were the Notch pathway members DTX3, NOTCH3, NUMBL, HES1, MAML2, NOTCH1, EP300, and NOTCH2 (Supplementary Fig. S1C). Interestingly, the Notch suppressors HDAC1 and HDAC2 were among the highest ranked downregulated genes in this cohort. We did not find any gene sets significantly enriched at nominal P < 0.05 in the UAB data. In contrast, TCGA samples did not form distinct clusters, possibly indicating a high degree of heterogeneity (Supplementary Fig. S2A). Comparative biomarker analysis revealed significant enrichment of the lysine degradation, butanoate metabolism, glutathione metabolism, and glycerophospholipid metabolism pathways in the ACM cohort, but the Notch pathway was neither up- nor downregulated (Supplementary Fig. S2B). On the basis of these findings, the activation of Notch signaling identified by NanoString pathway analysis is not the result of large changes in individual Notch genes. Instead, small alterations in the expression of Notch ligands, receptors, and regulatory factors combine to produce a significant change in Notch signaling at the pathway level. In addition, our results support a previous study indicating the Pathifier algorithm for pathway analysis may detect more subtle changes in gene expression as compared with the other methods (26).

In vitro activation of Notch signaling increases expression of FGF1

The Notch pathway mediates multiple cancer-associated cell phenotypes through transcriptional regulation, which occurs in a context-specific manner via activation of the Notch transcription complex, repression of CREB-mediated transcription, and cross-talk with the JAK-STAT pathway (27). To identify transcriptional changes induced by Notch activation in HNSCCs, we exposed the HNSCC cell lines UM-SCC1, UM-SCC6, and FaDu to recombinant Notch ligand DLL4 in vitro, resulting in increased Notch receptor cleavage (Fig. 2A), an indicator of canonical Notch activation, and subsequently measured gene expression using NanoString. We first examined HES1 and HES5, two well-known Notch targets, as markers of signaling activity. A statistically significant 1.7-fold increase in expression of HES1 and a 3.3-fold increase in HES5 was observed in Notch activated (Notch-a) compared with control cells (Fig. 2B). These changes were consistent across all three cell lines with the exception of UM-SCC6, which lacked detectable HES5 expression, and are congruous with an increase in Notch-mediated transcription. We also identified a 2-fold decrease in the expression of DLL1 and pathway regulator LFNG in Notch-a compared with control HNSCC cells (data not shown). These trends are in agreement UAB OSCC samples, in which the average expression of DLL1 and LFNG was substantially lower in the CSM cohort compared with the no CSM cohort (6.5-fold and 1.5-fold difference, respectively). Collectively, the expression of Notch genes in HNSCC cells validated the approach of DLL4-mediated pathway activation, including a possible negative feedback loop involving DLL1.

Figure 2.

Notch activation increases expression of FGF1 in vitro and in clinical samples. HNSCC cell lines were treated with the Notch ligand DLL4 to activate Notch signaling in vitro. A, representative image showing expression of the cleaved NICD in cells with and without DLL4 treatment. Blots were cropped for clarity. B, normalized NanoString transcript counts of the Notch transcription targets HES1 and HES5 in cells treated with PBS vehicle [control (Ctrl)] or DLL4 (Notch-a) for 8 hours. Each bar represents one cell line. C, normalized NanoString transcript counts of the fibroblast growth factor 1 gene (FGF1) in control and Notch-a cells. D, correlation between normalized NanoString FGF1 transcript counts and Notch pathway score in UAB OSCC samples. E, correlation between normalized RNA-seq FGF1 transcript counts and Notch pathway score in TCGA OSCC samples. F, combined comparison of normalized FGF1 transcript counts in UAB and TCGA OSCC samples with and without mortality. Lines, median with interquartile range. **, P < 0.01. n.s., not significant.

Figure 2.

Notch activation increases expression of FGF1 in vitro and in clinical samples. HNSCC cell lines were treated with the Notch ligand DLL4 to activate Notch signaling in vitro. A, representative image showing expression of the cleaved NICD in cells with and without DLL4 treatment. Blots were cropped for clarity. B, normalized NanoString transcript counts of the Notch transcription targets HES1 and HES5 in cells treated with PBS vehicle [control (Ctrl)] or DLL4 (Notch-a) for 8 hours. Each bar represents one cell line. C, normalized NanoString transcript counts of the fibroblast growth factor 1 gene (FGF1) in control and Notch-a cells. D, correlation between normalized NanoString FGF1 transcript counts and Notch pathway score in UAB OSCC samples. E, correlation between normalized RNA-seq FGF1 transcript counts and Notch pathway score in TCGA OSCC samples. F, combined comparison of normalized FGF1 transcript counts in UAB and TCGA OSCC samples with and without mortality. Lines, median with interquartile range. **, P < 0.01. n.s., not significant.

Close modal

Along with these classical targets of Notch signaling, we also identified increased expression of several genes associated with tumor progression and invasion in Notch-a compared with control HNSCC cells, including cyclin A1 (CCNA1, 1.6-fold increase), IL1β (2.1-fold increase), SIX homeobox 1 (SIX1, 5.5-fold increase), and FGF1 (1.8-fold increase; Fig. 2C). To determine which of these genes play a role in Notch-mediated patient outcomes, we examined their expression in OSCC samples from UAB and TCGA. Although patient mortality and Notch pathway score were not consistently associated with expression of CCNA1, IL1B, or SIX1, FGF1 emerged as a potential target of Notch activation in clinical samples. In UAB and TCGA tumors with detectable FGF1 expression, we observed a statistically significant positive correlation between normalized FGF1 transcript copy number and Notch pathway score (Fig. 2D and E). The Pearson r value for the correlation was 0.4734 in the UAB dataset (P = 0.0203, R2 = 0.2241) and 0.3286 in the TCGA dataset (P = 0.0121, R2 = 0.1080). Some OSCC samples, however, did not express the FGF1 transcript (n = 0 in UAB OSCC NanoString data; n = 24, 27% in TCGA OSCC RNA-seq data). In addition, FGF1 was not significantly associated with mortality in either the UAB or TCGA datasets, although there was a trend toward increased FGF1 in the ACM TCGA cohort (Fig. 2F). These results suggest that activation of Notch signaling enhances, but may not induce, FGF1 expression in OSCCs, although a relationship between upregulation of FGF1 and patient mortality was not evident in these data.

In vitro activation of Notch signaling promotes FGF1-mediated cell migration and invasion in HNSCC cell lines

As Notch activation was associated with poor patient outcomes in our clinical data and with increased expression of the proinvasive gene FGF1 in vitro, we began our phenotypic characterization by investigating cell migration and invasion as indicators of tumor aggressiveness (28). In vitro cell migration was assessed in control and Notch-a cells using the wound healing (scratch) assay (29). Percent wound closure was significantly higher in Notch-a cells as compared with control cells for all three cell lines at 12 hours (Fig. 3A and B). Difference in wound closure ranged from approximately 20% in UM-SCC1 to 40% or more in UM-SCC6 and FaDu (Fig. 3A), indicating notably increased migration in Notch-a cells. Next, we used the Transwell invasion assay to measure the ability of cells to invade through basement membrane. Cell invasion was significantly increased in Notch-a compared with control cells for all three cell lines at 24 hours (Fig. 3C and D). UM-SCC1 demonstrated a 2.9-fold difference between Notch-a and control cells, UM-SCC6 a 3.6-fold difference, and FaDu a 2.5-fold difference. Notch activation also altered cell morphology as detected by light microscopy; we observed substantial spreading in invaded cells from the Notch-a group, whereas invaded control cells remained more closely associated with membrane pores (Fig. 3D). To determine the role of FGF1 in Notch-mediated cell migration and invasion, we used siRNA to silence FGF1 in UM-SCC1 cells, resulting in an 80% decrease in both FGF1 transcript expression (Fig. 3E) and secreted FGF1 protein levels (Fig. 3F), and repeated the scratch and Transwell invasion assays. Loss of FGF1 completely abrogated the effects of Notch activation on wound closure and cell invasion as compared with control siRNA-transfected cells (Fig. 3G and H). When recombinant FGF1 protein was added to the growth media, the Notch-induced migratory and invasive phenotype was restored in FGF1-silenced cells (Fig. 3G and H). These results were replicated in the FaDu cell line by a repeat experiment, including both FGF1 knockdown and rescue (Supplementary Fig. S3).

Figure 3.

Notch activation increases HNSCC cell migration and invasion in a FGF1-dependent manner. A, HNSCC cells treated with PBS vehicle [control (Ctrl)] or DLL4 (Notch-a) for 8 hours were assessed for migration by the scratch assay. Percent wound closure was calculated at 12 hours postscratch using scratch width measured in pixels. B, representative images of the scratch assay in control and Notch-a FaDu cells, taken at 2× magnification. C, cell invasion was determined by the Transwell invasion assay. Cells (200,000) were plated per insert and incubated for 24 hours, at which point Transwell membranes were removed, stained, and imaged. Shown is the average number of invaded cells per field at 10× magnification. D, representative images of stained Transwell membranes in control and Notch-a cells at 24 hours, taken at 10× magnification. UM-SCC1 cells were transfected with control or FGF1 siRNA for 72 hours, and knockdown of FGF1 expression was validated by qRT-PCR (E) and ELISA (F). Transfected cells were treated with PBS vehicle (Ctrl), DLL4 (Notch-a), and/or recombinant FGF1 and assessed for migration by the scratch assay (G) and cell invasion by the Transwell invasion assay (H). Shown is the mean ± SEM from at least three independent experiments performed in triplicate, with all treatment groups compared with control for each cell line. ***, P < 0.001; *, P < 0.05.

Figure 3.

Notch activation increases HNSCC cell migration and invasion in a FGF1-dependent manner. A, HNSCC cells treated with PBS vehicle [control (Ctrl)] or DLL4 (Notch-a) for 8 hours were assessed for migration by the scratch assay. Percent wound closure was calculated at 12 hours postscratch using scratch width measured in pixels. B, representative images of the scratch assay in control and Notch-a FaDu cells, taken at 2× magnification. C, cell invasion was determined by the Transwell invasion assay. Cells (200,000) were plated per insert and incubated for 24 hours, at which point Transwell membranes were removed, stained, and imaged. Shown is the average number of invaded cells per field at 10× magnification. D, representative images of stained Transwell membranes in control and Notch-a cells at 24 hours, taken at 10× magnification. UM-SCC1 cells were transfected with control or FGF1 siRNA for 72 hours, and knockdown of FGF1 expression was validated by qRT-PCR (E) and ELISA (F). Transfected cells were treated with PBS vehicle (Ctrl), DLL4 (Notch-a), and/or recombinant FGF1 and assessed for migration by the scratch assay (G) and cell invasion by the Transwell invasion assay (H). Shown is the mean ± SEM from at least three independent experiments performed in triplicate, with all treatment groups compared with control for each cell line. ***, P < 0.001; *, P < 0.05.

Close modal

In addition to migration and invasion, we further characterized the Notch-activated phenotype in HNSCCs with respect to cell proliferation, radioresponse, and cell cycle, important cancer-related characteristics influenced by the Notch pathway in other cancer types (30). Cell proliferation rate varied slightly between cell lines but did not differ between Notch-a and control groups (Fig. 4A). Similarly, survival fraction following 0 to 10 Gy radiation did not change with Notch activation, and the dose enhancement ratios for Notch-a cells were not significantly altered, ranging from 1.040 to 1.132 (Fig. 4B). Finally, a comparison of cell-cycle profile between Notch-a and control cells did not reveal differences in any of the three HNSCC cell lines tested (Fig. 4C–E). Taken together, these findings indicate that Notch activation promotes cell migration and invasion in an FGF1-dependent manner without affecting cell proliferation, radiosensitivity, or cell cycle in HNSCC cells in vitro.

Figure 4.

Notch activation does not affect cell proliferation, radiosensitivity, or cell cycle. HNSCC cells treated with PBS vehicle [control (Ctrl)] or DLL4 (Notch-a) for 8 hours were assessed for cell proliferation, radiosensitivity, and cell-cycle profile. A, cells (2,500) were plated per well and collected by trypsinization at the indicated time points. Shown is the average cell count/mL as measured by Coulter counter. B, cells were treated with increasing doses of radiation (RT), left undisturbed for 2 weeks, then fixed and stained for colony counting. Dose enhancement ratio (DER) for Notch-a was calculated at 50% survival. CE, cells were collected at the indicated time points, fixed, and labeled with propidium iodide. DNA content was evaluated by flow cytometry. Shown is the mean ± SEM from at least three independent experiments performed in triplicate.

Figure 4.

Notch activation does not affect cell proliferation, radiosensitivity, or cell cycle. HNSCC cells treated with PBS vehicle [control (Ctrl)] or DLL4 (Notch-a) for 8 hours were assessed for cell proliferation, radiosensitivity, and cell-cycle profile. A, cells (2,500) were plated per well and collected by trypsinization at the indicated time points. Shown is the average cell count/mL as measured by Coulter counter. B, cells were treated with increasing doses of radiation (RT), left undisturbed for 2 weeks, then fixed and stained for colony counting. Dose enhancement ratio (DER) for Notch-a was calculated at 50% survival. CE, cells were collected at the indicated time points, fixed, and labeled with propidium iodide. DNA content was evaluated by flow cytometry. Shown is the mean ± SEM from at least three independent experiments performed in triplicate.

Close modal

In this study, we report an association between the Notch signaling pathway, patient outcomes, and molecular phenotype in OSCC. Expression of Notch genes and weighted Notch pathway score were significantly increased in OSCC patients with mortality as compared with those without mortality in independent datasets from UAB and TCGA. In vitro modeling of Notch activation resulted in increased expression of FGF1, and the positive relationship between Notch activity and FGF1 expression was validated in both OSCC sample sets. Finally, activation of Notch signaling produced a significant, FGF1-dependent increase in cell migration and invasion. These findings, which were consistent across multiple cell lines and two sets of patients, validate the presence of Notch activation at the pathway level in a subset of HNSCCs and clarify the clinical relevance of the Notch-a phenotype. In addition, our results suggest Notch mediates HNSCC behavior through the activity of FGF1. Accordingly, the interplay between Notch and FGF signaling may be a biomarker for invasive phenotype and mortality risk in a subset of HNSCCs. Overall, these results confirm our original hypothesis that pathway level dysregulation is detectable in HNSCC tissue and can inform the biology of this disease.

Along with differential Notch pathway expression, we also identified Notch-mediated dysregulation of FGF1 in both HNSCC cell lines and OSCC patient samples. Although FGF1 is not a classical target of Notch transcription, ours is not the first evidence of a regulatory relationship between Notch and FGF signaling. In several previous studies, the constitutive activation of canonical (CBF1 dependent) Notch-mediated transcription was found to suppress FGF1 expression, while inhibition of CBF1 and its coactivator MAML induced FGF1 expression and transport into the extracellular compartment (31–33). Surprisingly, we observed evidence of the opposite, as in vitro Notch activation increased FGF1 transcript levels and produced an FGF1-dependent invasive phenotype. This discrepancy may be attributable to cell type-specific or ligand-specific Notch function. Another possibility is that downstream Notch signaling in HNSCC cells occurs through the noncanonical (CBF1 independent) pathway, which is not well characterized but has been associated with differences in pathway output (34). Importantly, our in vitro findings were validated in patient samples, in which Notch pathway score and FGF1 expression were also positively correlated. And although we did not identify a direct relationship between FGF1 transcript expression and mortality in this study, which may be limited by small sample size, Koole and colleagues recently reported a significant association between overexpression of FGF receptor 1 protein and decreased overall and disease-free survival in HNSCC (35). Additional studies of the mechanisms by which Notch recognizes cell context and regulates FGF signaling in a larger patient cohort are needed to better understand pathway function in this disease.

Another important objective of the current study was to identify characteristics of cell behavior associated with increased Notch activity in HNSCC, as multiple hallmarks of cancer have been associated with Notch signaling in other models (36). Our characterization of Notch activation included increased cell migration and invasion in vitro, as well as worsened OSCC patient outcomes. In addition, we observed structural changes in Notch-a cells, which could be indicative of cell membrane ruffling (Fig. 3B), a feature of actively migrating cells, although this effect was not further characterized in the current study. Overall, these findings are consistent with studies in breast and prostate cancer, in which high expression of Notch genes was associated with more aggressive disease, as well as in melanoma where constitutive Notch activation enhanced tumor progression (37–39). Our data also support a previous report in HNSCC, which identified a role for Notch in determining angiogenic potential following treatment with human growth factor (40). We further identified FGF1 as a possible mediator of the Notch-induced migratory and invasive phenotype in HNSCC cells. Interestingly, treatment with recombinant FGF1 alone was not sufficient to increase cell migration and invasion, indicating the potential role for FGF1 in altering cell behavior is restricted to the context of Notch activation. Again, this may be consistent with the studies cited above, which demonstrated that Wnt, PI3K-Akt, and MAPK signaling contributed significantly to downstream Notch effects. Although we did not detect changes in these pathways in our pathway expression analysis, both PI3K-Akt and MAPK are activated by FGF receptor signaling. Thus, the mechanism by which Notch and FGF1 combine to produce increased migration and invasion may include one of these interconnected signaling pathways.

Reports of dysregulated Notch signaling in the development and progression of solid tumors, including HNSCC, have increased interest in Notch inhibition as a potential anticancer therapeutic strategy. Small-molecule inhibitors of γ-secretase, the enzyme responsible for cleavage of Notch receptors and downstream signaling, are the most clinically advanced option, having progressed to early-stage clinical trials (36). However, these agents are associated with significant side effects, including gastrointestinal cytotoxicity, which may limit their use. Additional Notch-targeted therapies, including antisense Notch, RNA interference, soluble receptor decoys, and dominant negative variants of transcription complex members MAML and CSL, are still in the early stages of development. The excitement surrounding Notch inhibition should also be tempered by the lack of clarity with regard to Notch function in HNSCCs and other solid tumors. Although we describe a role for Notch in invasion and mortality, this applies only to a subset of OSCC patients. We did not detect sensitivity to γ-secretase inhibition as a single agent in HNSCC cell lines, and combination treatment with radiotherapy actually decreased radiosensitivity (data not shown). It is probable that a fine balance of Notch signaling is maintained in normal cells and that dysregulation in either direction can be detrimental for cancer management. For these reasons, targeting downstream Notch effectors, including FGF1 and FGF receptors, rather than upstream signaling may be a safer and more effective therapeutic strategy for Notch-a OSCC.

E.S. Yang has received speakers bureau honoraria from and is a consultant/advisory board member for NanoString. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A.N. Weaver, E.S. Yang

Development of methodology: A.N. Weaver, M.B. Burch, E.S. Yang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.N. Weaver, M.B. Burch, D.L. Della Manna, A.I. Ojesina, E.S. Yang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.N. Weaver, T.S. Cooper, S. Wei, A.I. Ojesina, E.L. Rosenthal, E.S. Yang

Writing, review, and/or revision of the manuscript: A.N. Weaver, T.S. Cooper, D.L. Della Manna, A.I. Ojesina, E.L. Rosenthal, E.S. Yang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.N. Weaver, M.B. Burch, T.S. Cooper, E.S. Yang

Study supervision: E.S. Yang

We would like to acknowledge Lisa Clemmons, Yolanda Hartman, and Anthony Morlandt for their assistance with procuring patient tissue and clinical data.

This work was supported by funding from the University of Alabama at Birmingham Department of Radiation Oncology and the ROAR Southeast Cancer Foundation.

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.

1.
Pulte
D
,
Brenner
H
. 
Changes in survival in head and neck cancers in the late 20th and early 21st century: a period analysis
.
Oncologist
2010
;
15
:
994
1001
.
2.
Suh
Y
,
Amelio
I
,
Guerrero Urbano
T
,
Tavassoli
M
. 
Clinical update on cancer: molecular oncology of head and neck cancer
.
Cell Death Dis
2014
;
5
:
e1018
.
3.
Hansen
AR
,
Siu
LL
. 
Epidermal growth factor receptor targeting in head and neck cancer: have we been just skimming the surface?
J Clin Oncol
2013
;
31
:
1381
3
.
4.
Egloff
AM
,
Grandis
JR
. 
Improving response rates to EGFR-targeted therapies for head and neck squamous cell carcinoma: candidate predictive biomarkers and combination treatment with Src inhibitors
.
J Oncol
2009
;
2009
:
896407
.
5.
Ntziachristos
P
,
Lim Jing
S
,
Sage
J
,
Aifantis
I
. 
From fly wings to targeted cancer therapies: a centennial for Notch signaling
.
Cancer Cell
2014
;
25
:
318
34
.
6.
Yin
L
,
Velazquez
OC
,
Liu
Z-J
. 
Notch signaling: emerging molecular targets for cancer therapy
.
Biochem Pharmacol
2010
;
80
:
690
701
.
7.
De Strooper
B
,
Annaert
W
,
Cupers
P
,
Saftig
P
,
Craessaerts
K
,
Mumm
JS
, et al
A presenilin-1-dependent [gamma]-secretase-like protease mediates release of Notch intracellular domain
.
Nature
1999
;
398
:
518
22
.
8.
Wilson
JJ
,
Kovall
RA
. 
Crystal structure of the CSL-Notch-Mastermind ternary complex bound to DNA
.
Cell
2006
;
124
:
985
96
.
9.
Previs
RA
,
Coleman
RL
,
Harris
AL
,
Sood
AK
. 
Molecular pathways: translational and therapeutic implications of the Notch signaling pathway in cancer
.
Clin Cancer Res
2015
;
21
:
955
61
.
10.
Leong
KG
,
Karsan
A
. 
Recent insights into the role of Notch signaling in tumorigenesis
.
Blood
2006
;
107
:
2223
33
.
11.
Rangarajan
A
,
Talora
C
,
Okuyama
R
,
Nicolas
M
,
Mammucari
C
,
Oh
H
, et al
Notch signaling is a direct determinant of keratinocyte growth arrest and entry into differentiation
.
EMBO J
2001
;
20
:
3427
36
.
12.
Agrawal
N
,
Frederick
MJ
,
Pickering
CR
,
Bettegowda
C
,
Chang
K
,
Li
RJ
, et al
Exome sequencing of head and neck squamous cell carcinoma reveals inactivating mutations in NOTCH1
.
Science
2011
;
333
:
1154
7
.
13.
Stransky
N
,
Egloff
AM
,
Tward
AD
,
Kostic
AD
,
Cibulskis
K
,
Sivachenko
A
, et al
The mutational landscape of head and neck squamous cell carcinoma
.
Science
2011
;
333
:
1157
60
.
14.
Sun
W
,
Gaykalova
DA
,
Ochs
MF
,
Mambo
E
,
Arnaoutakis
D
,
Liu
Y
, et al
Activation of the NOTCH pathway in head and neck cancer
.
Cancer Res
2014
;
74
:
1091
104
.
15.
Yoshida
R
,
Nagata
M
,
Nakayama
H
,
Niimori-Kita
K
,
Hassan
W
,
Tanaka
T
, et al
The pathological significance of Notch1 in oral squamous cell carcinoma
.
Lab Invest
2013
;
93
:
1068
81
.
16.
Lee
SH
,
Hong
HS
,
Liu
ZX
,
Kim
RH
,
Kang
MK
,
Park
NH
, et al
TNFalpha enhances cancer stem cell-like phenotype via Notch-Hes1 activation in oral squamous cell carcinoma cells
.
Biochem Biophys Res Commun
2012
;
424
:
58
64
.
17.
Leethanakul
C
,
Patel
V
,
Gillespie
J
,
Pallente
M
,
Ensley
JF
,
Koontongkaew
S
, et al
Distinct pattern of expression of differentiation and growth-related genes in squamous cell carcinomas of the head and neck revealed by the use of laser capture microdissection and cDNA arrays
.
Oncogene
2000
;
19
:
3220
4
.
18.
Lee
SH
,
Do
SI
,
Lee
HJ
,
Kang
HJ
,
Koo
BS
,
Lim
YC
. 
Notch1 signaling contributes to stemness in head and neck squamous cell carcinoma
.
Lab Invest
2016
;
96
:
508
16
.
19.
Sun
Z
,
Asmann
YW
,
Kalari
KR
,
Bot
B
,
Eckel-Passow
JE
,
Baker
TR
, et al
Integrated analysis of gene expression, CpG island methylation, and gene copy number in breast cancer cells by deep sequencing
.
PLoS One
2011
;
6
:
e17490
.
20.
Gao
J
,
Aksoy
BA
,
Dogrusoz
U
,
Dresdner
G
,
Gross
B
,
Sumer
SO
, et al
Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal
.
Sci Signal
2013
;
6
:
pl1
.
21.
Cerami
E
,
Gao
J
,
Dogrusoz
U
,
Gross
BE
,
Sumer
SO
,
Aksoy
BA
, et al
The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data
.
Cancer Discov
2012
;
2
:
401
4
.
22.
The Cancer Genome Atlas Network
. 
Comprehensive genomic characterization of head and neck squamous cell carcinomas
.
Nature
2015
;
517
:
576
82
.
23.
Nowsheen
S
,
Cooper
T
,
Stanley
JA
,
Yang
ES
. 
Synthetic lethal interactions between EGFR and PARP inhibition in human triple negative breast cancer cells
.
PLoS One
2012
;
7
:
e46614
.
24.
Subramanian
A
,
Tamayo
P
,
Mootha
VK
,
Mukherjee
S
,
Ebert
BL
,
Gillette
MA
, et al
Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles
.
Proc Natl Acad Sci U S A
2005
;
102
:
15545
50
.
25.
Kejner
AE
,
Burch
MB
,
Sweeny
L
,
Rosenthal
EL
. 
Bone morphogenetic protein 6 expression in oral cavity squamous cell cancer is associated with bone invasion
.
Laryngoscope
2013
;
123
:
3061
5
.
26.
Drier
Y
,
Sheffer
M
,
Domany
E
. 
Pathway-based personalized analysis of cancer
.
Proc Natl Acad Sci U S A
2013
;
110
:
6388
93
.
27.
Ranganathan
P
,
Weaver
KL
,
Capobianco
AJ
. 
Notch signalling in solid tumours: a little bit of everything but not all the time
.
Nat Rev Cancer
2011
;
11
:
338
51
.
28.
Li
J
,
Wei
Z
,
Li
H
,
Dang
Q
,
Zhang
Z
,
Wang
L
, et al
Clinicopathological significance of fibroblast growth factor 1 in non–small cell lung cancer
.
Hum Pathol
2015
;
46
:
1821
8
.
29.
Liang
C-C
,
Park
AY
,
Guan
J-L
. 
In vitro scratch assay: a convenient and inexpensive method for analysis of cell migration in vitro
.
Nat Protoc
2007
;
2
:
329
33
.
30.
Roy
M
,
Pear
WS
,
Aster
JC
. 
The multifaceted role of Notch in cancer
.
Curr Opin Gen Dev
2007
;
17
:
52
9
.
31.
Kacer
D
,
McIntire
C
,
Kirov
A
,
Kany
E
,
Roth
J
,
Liaw
L
, et al
Regulation of nonclassical FGF1 release and FGF-dependent cell transformation by CBF1-mediated Notch signaling
.
J Cell Physiol
2011
;
226
:
3064
75
.
32.
Duarte
M
,
Kolev
V
,
Kacer
D
,
Mouta-Bellum
C
,
Soldi
R
,
Graziani
I
, et al
Novel cross-talk between three cardiovascular regulators thrombin cleavage fragment of Jagged1 induces fibroblast growth factor 1 expression and release
.
Mol Biol Cell
2008
;
19
:
4863
74
.
33.
Small
D
,
Kovalenko
D
,
Soldi
R
,
Mandinova
A
,
Kolev
V
,
Trifonova
R
, et al
Notch activation suppresses fibroblast growth factor-dependent cellular transformation
.
J Biol Chem
2003
;
278
:
16405
13
.
34.
Nofziger
D
,
Miyamoto
A
,
Lyons
KM
,
Weinmaster
G
. 
Notch signaling imposes two distinct blocks in the differentiation of C2C12 myoblasts
.
Development
1999
;
126
:
1689
702
.
35.
Koole
K
,
Brunen
D
,
van Kempen
PM
,
Noorlag
R
,
de Bree
R
,
Lieftink
C
, et al
FGFR1 is a potential prognostic biomarker and therapeutic target in head and neck squamous cell carcinoma
.
Clin Cancer Res
2016
Mar 2.
[Epub ahead of print]
.
36.
Koch
U
,
Radtke
F
. 
Notch and cancer: a double-edged sword
.
Cell Mol Life Sci
2007
;
64
:
2746
62
.
37.
Reedijk
M
,
Odorcic
S
,
Chang
L
,
Zhang
H
,
Miller
N
,
McCready
DR
, et al
High-level coexpression of JAG1 and NOTCH1 is observed in human breast cancer and is associated with poor overall survival
.
Cancer Res
2005
;
65
:
8530
7
.
38.
Santagata
S
,
Demichelis
F
,
Riva
A
,
Varambally
S
,
Hofer
MD
,
Kutok
JL
, et al
JAGGED1 expression is associated with prostate cancer metastasis and recurrence
.
Cancer Res
2004
;
64
:
6854
7
.
39.
Balint
K
,
Xiao
M
,
Pinnix
CC
,
Soma
A
,
Veres
I
,
Juhasz
I
, et al
Activation of Notch1 signaling is required for β-catenin–mediated human primary melanoma progression
.
J Clin Invest
2005
;
115
:
3166
76
.
40.
Zeng
Q
,
Li
S
,
Chepeha
DB
,
Giordano
TJ
,
Li
J
,
Zhang
H
, et al
Crosstalk between tumor and endothelial cells promotes tumor angiogenesis by MAPK activation of Notch signaling
.
Cancer Cell
2005
;
8
:
13
23
.