Reciprocal activation of HEY1 and NOTCH4 under SOX2 control promotes EMT in head and neck squamous cell carcinoma.

Several comprehensive studies have demonstrated that the NOTCH pathway is altered in a bimodal manner in head and neck squamous cell carcinoma (HNSCC). In a previous study, it was found that the NOTCH4/HEY1 pathway was specifically upregulated in HNSCC and promoted epithelial‑mesenchymal transition (EMT), and that HEY1 activation supported SOX2 expression. However, the interactions in this pathway have not yet been fully elucidated. The present study investigated the NOTCH4/HEY1/SOX2 axis in HNSCC using in vitro models and the Cancer Genome Atlas (TCGA) database. To explore the association, reporter and ChIP RT‑qPCR assays using SOX2‑overexpressing (SOX2‑OE) cells were performed. The association between NOTCH4 and HEY1 was examined in the same manner using HEY1‑overexpressing (HEY1‑OE) cells. The results of the in vitro experiments indicated that HEY1 promoted EMT in the HNSCC cells. Furthermore, the overexpression of HEY1 also promoted sphere formation and increased murine xenograft tumorigenicity. Reporter assays and ChIP RT‑qPCR experiments indicated that SOX2 regulated HEY1 expression via direct binding of the HEY1 promoter. HEY1 expression significantly correlated with SOX2 expression in primary lung SCC and other SCCs using the TCGA database. HEY1 also regulated NOTCH4 expression to create a positive reciprocal feedback loop. On the whole, the present study demonstrates that HEY1 expression in HNSCC is regulated via the promotion of SOX2 and promotes EMT. The NOTCH4/HEY1 pathway is specifically upregulated via a positive reciprocal feedback loop mediated by the HEY1‑medaited regulation of NOTCH4 transcription, and SOX2 correlates with HEY1 expression in SCC from other primary sites.

Introduction

Head and neck squamous cell carcinoma (HNSCC) is the sixth most common malignancy, with >600,000 cases diagnosed annually worldwide (1). Half of the patients with HNSCC are diagnosed in an advanced stage at the first medical examination. In addition, >50% of recurrences occur within 3 years following treatment (2-4). Similar to other types of cancer, the accumulation of genetic and epigenetic alternations is considered to generate and promote HNSCC. Recently, several comprehensive analyses for HNSCC gene mutations were performed using high-throughput next generation sequencing defining NOTCH1 mutation at a 10-15% rate. This rate is the second most frequent following TP53 and higher than previously considered (5,6). Subsequently a previous study demonstrated a bimodal pattern of NOTCH pathway alterations in HNSCC, with a smaller subset of HNSCC exhibiting inactivating NOTCH1 receptor mutations, but a larger subset exhibiting NOTCH pathway activating alterations, resulting in downstream HES1/HEY1 pathway activation (7).

HES, HEY, CCND1, MYC, BCL-2 and p21 are NOTCH target genes. Among these genes, the HES and HEY families are considered prominent downstream effectors of the NOTCH pathway (8,9). HEY1 is known to promote epithelial-mesenchymal transition (EMT) in several normal tissues, such as the epidermis, kidney tubules, mammary gland and endocardia (10-12). HEY1 knockdown in glioblastoma cells has been shown to decrease colony formation and invasion (13). A previous study demonstrated that HEY1 expression in a skin human SCC cell line was increased under 3D culture and promoted an EMT phenotype (14). Man et al indicated that HNSCC exhibits a significantly higher HEY1 expression than normal epithelial cells (15). Recently, HEY1 has been shown to be associated with a poor prognosis, independent of NOTCH1 expression, indicating that other NOTCH members may drive HEY1 expression as a key pathway alteration in HNSCC (16). In a previous study, it was also found that the NOTCH4/HEY1 pathway was specifically upregulated in HNSCC and it was revealed that this pathway promoted EMT (17). However, the mechanisms that effect NOTCH4/HEY1 pathway activation in HNSCC remain unclear.

SOX2, as well as CD10 (18), CD44 (19) and ALDH1 (20) are HNSCC cancer stem cell (CSC) markers (21). SOX2 expression in HNSCC is significantly related to a worse prognosis (22) and SOX2 promotes migration, invasion and EMT in HNSCC (23). To define NOTCH downstream effectors, the present study examined the association between SOX2 and HEY1. The authors previously demonstrated that HEY1 knockdown significantly decreased NOTCH4 expression and decreased SOX2 expression in HNSCC cells (17). To further define these associations, the present study examined specific feedback loops between HEY1 and NOTCH4 and SOX2 in HNSCC.

Materials and methods

Cells and cell culture

Cal27, SCC61 and SCC090 HNSCC cell lines were used in the present study. Cal27 and SCC090 cells were obtained from the Gutkind Laboratory at the University of California San Diego, Moores Cancer Center. SCC61 cells were obtained from the Weichselbaum Laboratory at the University of Chicago. SCC090 cells were originally established from human papilloma virus (HPV)-positive HNSCC tissues. The other two cells were established from HPV-negative HNSCC tissues. These cells were cultured in Dulbecco's modified Eagle's medium (DMEM; Sigma-Aldrich; Merck KGaA) supplemented with 10% fetal bovine serum (FBS) and a penicillin (50 U/ml) and streptomycin (50 µg/ml) cocktail. All cells were cultured under an atmosphere of 5% CO2 at 37°C.

Vector transfection

A lenti-ORF clone of human SOX2 (#RC200757L3), HEY1 (#RC200257L3) and an empty vector control (#PS100092) were obtained from OriGene Technologies, Inc. The 293T cells obtained from the Gutkind Laboratory were seeded in 6-well plates one day prior to transfection, and each construct was transfected in Opti-MEM (#31985070, Thermo Fisher Scientific, Inc.) and Turbofect transfection reagent (#R0531, Thermo Fisher Scientific, Inc.). Viral supernatants were consisted of 10 µg lentiviral plasmid, 6.67 µg packaging vector and 3.33 µg envelope per well in 6 well plates, and collected at 48 and 72 h following transfection. Cal27, SCC61 and SCC090 cells were seeded one day prior to infection in a 6-well plate and allowed to reach 50-60% confluency. The virus supernatant and 2 µl of poly-brene (Sigma-Aldrich; Merck KGaA) were added to the cells. Cells were maintained under puromycin (#ant-pr-1, Invivogen; Thermo Fisher Scientific, Inc.) selection at a 1-µg/ml concentration. The following experiments using SOX2 and HEY1 overexpression cells were compared empty vector control transfected cells in Cal27, SCC61 and SCC090.

Reverse transcription-quantitative PCR (RT-qPCR)

To vali-date mRNA expression levels in each experiment, RT-qPCR was used. Briefly, total RNA was extracted from the cells using the RNeasy plus mini kit (Qiagen GmbH), and comple-mentary DNA was synthesized using a high-capacity cDNA reverse transcription kit (Thermo Fisher Scientific, Inc.). All primers were obtained from TaqMan Gene Expression assays (cat. no. 4331182. Thermo Fisher Scientific, Inc.). Each gene ID is described as follows: β-actin (ACTB): Hs01060665_g1; NOTCH4: Hs00965895_g1; HES1: Hs00172878_m1; HEY1: Hs01114113_m1; E-cadherin: Hs01023895_m1; fibronectin: Hs01549976_m1; Vimentin: Hs00958111_m1; TWIST1: Hs01675818_s1; and SOX2: Hs01053049_s1. The thermocycle program was set at 95°C for 10 min, followed by 50 cycles of denaturation at 95°C for 15 sec and annealing at 60°C for 60 sec. PCR quantification was conducted using the ΔΔCq method (24). qPCR was performed using the Quant Studio 6 Flex Real-Time PCR System (Thermo Fisher Scientific, Inc.).

Western blot analysis

Protein was obtained from the Cal27, SCC61 and SCC090 cells, and lysed with RIPA buffer (50 mm Tris-HCl pH 8.0, 150 mm NaCl, 1% IGE-PAL CA 630, 0.5% Na-DOC, and 0.1% SDS). Total protein concentrations were measured using Bio-Rad protein assay kit (Bio-Rad Laboratories, Inc.). 10 µl Equal amount of protein was set on Mini-PROTEAN TGX gels (Bio-Rad Laboratories, Inc.). The following primary antibodies were added to nitrocellulose membranes with 5% non-fat dry milk in Tris-buffered saline and 1% Tween-20, and incubated at 4°C overnight: NOTCH4 (1:500, #2423, Cell Signaling Technology, Inc.), HES1 (1:1,000, #sc-25392, Santa Cruz Biotechnology, Inc.), HEY1 (1:400, #ab22614, Abcam), E-cadherin (1:10,000, #610181, BD Biosciences), fibronectin (1:3,000, #ab2413, Abcam), Vimentin (1:500, #V6630, Sigma-Aldrich; Merck KGaA), TWIST1 (1:1,000, #sc-15393, Santa Cruz Biotechnology, Inc.) and SOX2 (1:1,000, #2748, Cell Signaling Technology, Inc.). HRP-conjugated goat anti-mouse (#1010-05, 1:20,000 dilution; SouthernBiotech) or anti-rabbit antibodies (#4010-05, 1:20,000 dilution; SouthernBiotech) were used as secondary antibodies. These secondary antibodies were incubated at room tempera-ture for 1 h. Western blots were developed using Pierce ECL Western Blotting Substate (Thermo Fisher Scientific).

Migration and invasion assays

Migration assays were performed in cell culture inserts (24-well, 8-µm pore size, #353097, Corning, Inc.). Cell concentrations ranged from 105 to 2×105 cells/ml. Invasion assays were also performed in Corning BioCoat Matrigel invasion chambers (24-well, 8-µm pore size, #353097, Corning, Inc.). Cell concentrations ranged from 2×105 to 4×105 cells/ml. Cells were seeded on uncoated or Matrigel-coated inserts in 500 ml of serum-free medium for migration and invasion assays, respectively. The lower chambers were filled with 750 µl of 10% FBS-supplemented medium. After 48 h, the cells on the lower surface of the insert were fixed and stained with crystal violet (Differential Quik Stain kit, Polysciences) at room temperature for 2 min. The number of stained cells was counted in >3 fields under an inverted microscope (Olympus CKX31).

Sphere formation assay

Cells were seeded in 96-well ultralow attachment culture dishes (Corning, Inc.) at 10-100 cells/well. Media consisted of serum-free DMEM/F12 Glutamax supplement medium (#10565042, Thermo Fisher Scientific, Inc.), basic fibroblast growth factor (bFGF: 20 ng/ml, #13256029, Thermo Fisher Scientific, Inc.), epithelial growth factor (EGF: 20 ng/ml, #PHG0313, Thermo Fisher Scientific, Inc.), B-27 (1:50 dilution, #17504044, Thermo Fisher Scientific, Inc.) and N2 supplement (1:100 dilution, #17502-048, Thermo Fisher Scientific, Inc.). Images were obtained at 10 days after seeding using a clinical upright microscope (Olympus, BX43) (Fig. 2A), and the numbers of sphere colonies in each well were counted using an inverted microscope (Olympus CKX31).

Figure 2

Sphere formation assay in HEY1-OE cells. (A) Representative images of sphere colony shapes in SCC090 cells. Scale bar, 100 µm. (B) The number of sphere colonies in each well. The differences between control and HEY1-OE cells are compared. P-values were calculated using Dunnett's t-test. **P<0.01. HEY1-OE cells, HEY1-overexpressing cells.

Cell viability assay

Cells were seeded in 96-well plates at 1,500 to 9,000 cells/well. Cell numbers were measured on day 3. Cell viabilities were measured using Vita Blue Cell Viability reagent (Bimake.com). Following a 1.5-h pre-incubation at 37°C in the assay solution, the viable cell number in each well was calculated using fluorescence (Ex=530-570 nm, Em=590-620 nm) in a microplate reader (BioTek Insturments, Inc.). The assays were performed ≥3 times.

Mouse xenograft models

Cells (2×106) were diluted in 200 ml and injected subcutaneously into nude mice (Charles River Laboratories, Inc.) using a 25-gauge needle. Mice were anesthetized with a mixture of oxygen and isoflurane (5% in air for induction and 2% for maintenance) prior to each experiment, such as cell injection and tumor size measurement. Mice were maintained under pathogen-free conditions and sacrificed 2 months later or when tumors exceeded 20 mm at the largest diameter or earlier if necessary [this was done if any animal was observed to be cachexic (weight loss >15% from starting weight), moribund, dehydrated, anorexic, or any tumor that was ulcerated or eroded]. Mice were euthanized using carbon dioxide gas for 10 min. The CO2 flow rate displaced 15-25% of the camber volume. Mice were euthanized in November, 2017. The confirmation of euthanasia was assured by verifying the absence of respiration, cardiac function and toe/tail pinch reflexes at least 10 min. Mice were handled in accordance with the procedures outlined in the Regulations on Animal Experiments at University of California San Diego. The Institutional Animal Care and Use Committee at the University of California San Diego approved the study.

Immunohistochemistry

Mouse xenograft tumors were stained overnight at 4°C with a HEY1 primary antibody (#ab22614, Abcam) diluted 1:100 in PBS with 2.5% BSA. Biotinylated IgG antibody (#BA-1000, Vector Laboratories, Inc.) were used at 1:400 as a secondary antibody for 30 min at room temperature. Staining was developed at room temperature for 2 min with DAB. Specimens were counterstained at room temperature for 1 min with hematoxylin and mounted with glycerol gelatin. A clinical upright microscope (Olympus, BX439) was used to examine these specimens.

TCGA dataset

The mRNA expression sequence data of patients with HNSCC were obtained from the firebrowse website (http://firebrowse.org/). These TCGA data included 522 HNSCC and 44 normal tissues. In total, 447 HNSCC cases were used for NOTCH analysis, excluding 73 tumors with NOTCH mutations. RNA expression was normalized by RSEM.

Reporter assay

Promoter reporter clone for human NOTCH4 (#HPRM45581-LvPG04), HEY1 (#HPRM10038-LvPG04) and negative control (#NEG-LvPG04) containing a 1,443 bp region of the NOTCH4 promoter regions and the HEY1 reporter assay based on a 1,455 bp region of the HEY1 promoter region, respectively, were obtained from GeneCopoeia, Inc. The Cal27, SCC61 and SCC090 cells were transfected using 2.0 µg of these clones and 6 µl of X-tremeGENE 9 (Roche) per well in 6 well plates, and incubated at 37°C for 24 h. The culture medium was collected 48 h after transfection. The Secrete-Pair Dual Luminescence Assay kit (#LF032, GeneCopoeia, Inc.) was used that was optimized using these promoter reporter clones to validate each promoter activity and the promoter activities were examined using the manufacturer's protocol.

Chromatin immunoprecipitation qPCR

For chromatin immu-noprecipitation (ChIP) qPCR assays, the SimpleChIP Plus Enzymatic Chromatin IP kit (#9005, Cell Signaling Technology, Inc.) was used. Chromatin was incubated overnight with anti-bodies for SOX2 (#2748, Cell Signaling Technology, Inc.) or HEY1 (#19929-1-AP, ProteinTech Group, Inc.) at 4°C under rotation. Chromatin was incubated with a polyclonal rabbit IgG as a negative control and Histone H3 (D2B12) XP-Rabbit mAb as positive control that were included in the ChIP kit. Primer sequences for qPCR were obtained from Integrated DNA Technologies, Inc. The following primer sequences were used: HEY1 promoter forward, 5′-CCC GCT GAG AGG ATC TG-3′ and reverse, 5′-CCC TGT GCA TCT CAT TTC C-3′; NOTCH4 promoter forward, 5′-AGT GGT GCT GGT GAA GTA-3′ and reverse, 5′-CCA CAC ACT GAG TTC CTT TAG-3′. The results were computed as percentage antibody bound per input DNA and normalized to the IgG controls using the Quant Studio 6 Flex Real-Time PCR System (Thermo Fisher Scientific, Inc.).

Statistical analysis

All in vitro experiments were performed at least in triplicate. Statistical comparisons of 2 groups were determined using the Student's t-test. For the comparisons of multiple group against the control group, Dunnett's test was used. The correlation between the expression of 2 genes was determined with Pearson's correlation analysis. Differences were considered significant at P<0.05. All statistical analyses were performed using JMP 12 software (SAS, Inc.).

Results

HEY1 promotes HNSCC EMT, migration and invasion

Previously, the authors demonstrated that HEY1 promoted EMT in HNSCC cells using cells in which HEY1 was knocked down (17). To validate this finding, the present study generated stable HEY1-overexpressing (HEY1-OE) and control cells using the Cal27, SCC61 and SCC090 cell lines (Figs. 1A and S1A). The results of RT-qPCR revealed that mesenchymal gene expression (fibronectin, Vimentin and TWIST1) in the HEY1-OE cells significantly increased in all HNSCC cells. However, E-cadherin expression was higher in the Cal27 HEY1-OE than in the control cells, with no significant differences observed between the SCC61 and SCC090 HEY1-OE and control cells (Fig. 1B). By contrast, western blot analysis revealed a decreased E-cadherin expression in all HEY1-OE cells. Furthermore, the expression of fibronectin, Vimentin and TWIST1 increased in all the HNSCC HEY1-OE cells (Fig. 1A). EMT is associated with increased cellular migration and invasion; therefore, migration and invasion assays were performed to determine the mechanisms through which the changes in mRNA and protein expression affected the cell phenotype in vitro. Increased migration and invasion were noted in the HEY1-OE cells compared to the control cells (Figs. 1C and D, and S1B). These results reveal that HEY1 promotes HNSCC EMT, migration and invasion.

Figure 1

EMT phenotypes in HEY1-OE cells. (A) Western blot analysis of EMT-related proteins (E-cadherin, Fibronectin, Vimentin and TWIST1), NOTCH4 and HEY1 in control and HEY1-OE cells. GAPDH was used as a control. (B) EMT-related gene expression in control and HEY1-OE cells was measured by RT-qPCR. The expression differences between control and HEY1-OE cells are compared. (C) Migration and (D) invasion assays in control, and HEY1-OE cells. The migration and invasion indexes are calculated by dividing the number of control cells through the chamber. The differences between control and HEY1-OE cells are compared. P-values were calculated using Dunnett's t-test. *P<0.05, **P<0.01; N.S., not significant. EMT, epithelial-mesenchymal transition; HEY1-OE cells, HEY1-overexpressing cells.

HEY1 promotes sphere formation ability

Spheroids were generated to define the increase in the expression of EMT-related genes associated with sphere formation (25). The present study compared the number of sphere colonies between the HEY1-OE and control cells. No evident differences in sphere shape were noted between the control and HEY1-OE cells (Fig. 2A); however, the HEY1-OE cells formed significantly more spheroids in all cell lines (Fig. 2B). The number of spheroids in the HEY1-OE group was several folds higher than that of the control group (Cal27 cells, 2.60; SCC61 cells, 1.61; SCC090 cells, 4.18) (Fig. 2B). These results indicate that HEY1 promotes spheroid formation.

HEY1 promotes HNSCC tumorigenicity

A proliferation assay was performed to assess phenotypic characteristics affected by HEY1. A statistically significant increase in proliferation was observed in all HEY1-OE cells compared to the control cells (Fig. S2A). The tumorigenicity of the Cal27 HEY1-OE cells was then examined using a nude mouse xenograft model. HEY1 expression in these tumors was confirmed to be markedly higher in the HEY1-OE cell tumors than in the control cell tumors using RT-qPCR and western blot analysis (Fig. 3A). Immunohistochemical staining for HEY1 also revealed that the xenograft tumors generated from HEY1-OE cells exhibited a higher HEY1 expression than those from the control cells (Fig. 3B). The HEY1-OE cells also generated significantly larger tumors than the control cells (Figs. S2B and 3C), confirmed by an increased tumor weight of the Cal27 HEY1-OE tumors compared to the control tumors following tumor excision. The maximum tumor diameter was 12.9 mm in the control group, and 14.9 mm in the HEY1-OE group (Fig. 3D).

Figure 3

Mouse xenograft tumorigenicity with HEY1-OE cells. (A) Analyses for HEY1 expression in xenograft tumors arising from HEY1-OE and control Cal27 cells by RT-qPCR and western blot analysis. A GAPDH antibody was used as a control. (B) HEY1 immunohistochemistry of xenograft tumors arising from control and HEY1-OE Cal27 cells. Scale bar, 80 µm. (C) Growth of mouse xenograft tumors. These mice were injected with 2×106 cells of HEY1-OE and control Cal27 cells into their flanks. (D) The tumor weight of HEY1-OE and control Cal27 cells. Maximum tumor volume and weight are shown on each graph. Whisker plots indicate the minimum and maximum values. P-values were calculated using a Student's t-test. *P<0.05, **P<0.01. HEY1-OE cells, HEY1-overexpressing cells.

SOX2 expression correlates with HEY1 expression

Several studies have indicated that SOX2 is associated with EMT and a CSC state (26-28). In the present study, to elucidate a potential SOX2 and HEY1 association in HNSCC, the correlation between HEY1 and SOX2 was examined using the TCGA mRNA sequence data from 522 HNSCC and 44 normal tissues samples. A significant positive correlation was noted between SOX2 and HEY1 mRNA expression in HNSCC (r=0.45, P<0.0001); however, no significant correlation between these genes was found in the normal tissues (Fig. 4A). Furthermore, other NOTCH downstream genes, such as HES1 and HES5 did not exhibit any significant positive correlations with SOX2 (Fig. S3A). Of note, this association was independent of the HPV status (Fig. S3B). To explore this in vitro, SOX2-overexpressing (SOX2-OE) and control cells were generated (Fig. S3C). RT-qPCR demonstrated a significant increase in HEY1 expression in all HNSCC SOX2-OE cells examined. The SOX2-OE cells exhibited an approximately 1.5- to 2.0-fold higher HEY1 expression in all cell lines (Fig. 4B). However, HES1 expression did not differ significantly between the SOX2-OE and control Cal27 and SCC090 cells. The SCC61 SOX2-OE cells exhibited a significantly lower HES1 expression compared to the control cells (Fig. 4C). All SOX2-OE cells exhibited a higher HEY1 expression, as shown by western blot analysis. However, HES1 expression between the SOX2-OE and control cells did not exhibit a marked difference (Fig. 4D). Among the 3 cell lines, it was found that both SOX2 and HEY1 expression was higher in the Cal27 control cells. This result indicated that there was an association between SOX2 and HEY1 expression in HNSCC wild-type cells (Fig. 4D). These results demonstrated that SOX2 regulated HEY1 expression in HNSCC.

Figure 4

SOX2 binds the HEY1 promoter region. (A) The correlations between SOX2 and HEY1 were examined using the TCGA HNSCC dataset, including HNSCC (n=522) and normal samples (n=44). 'r' indicates the Pearson's correlation coefficient. (B) HEY1 and (C) HES1 expression was compared between SOX2-OE and control cells by RT-qPCR. P-values were calculated using Dunnett's t-test. (D) Western blot analysis of SOX2, HES1 and HEY1 in SOX2-OE and control cells. GAPDH is used as a control. (E) Luciferase reporter assays are performed with vectors containing the HEY1 promoter region. (F) A SOX2 binding region of the human HEY1 promoter (shadow box) with primer pairs (underlined). Bp indicates the base pairs from the transcription starting site. (G) ChIP qPCR analysis using the HEY1 promoter primer in SCC61 SOX2-OE and parental control cells. Mouse IgG antibody is used as a negative control. P-values were calculated using a Student's t-test. *P<0.05, **P<0.01; N.S., not significant. HEY1-OE cells, HEY1-overexpressing cells.

SOX2 directly binds the HEY1 promoter in HNSCC

To assess whether SOX2 directly binds HEY1 promoter region, a luciferase vector with the HEY1 promoter region was transfected into SOX2-OE and control cells. A significant increase in luciferase activity was observed in all HNSCC SOX2-OE cells (Fig. 4E). A ChIP qPCR was then performed to validate this result. SOX2 is a transcription factor that binds to the DNA consensus sequence (T/A)(T/A)CAAAGA (29) or AACAA(A/T)(G/A)(G/A) (30). A candidate SOX2 binding sequence was found from -1,028 to -1,035 bp upstream of the HEY1 transcript starting site. Therefore, a ChIP qPCR primer pair was created that bound to this sequence (Fig. 4F). The results of ChIP qPCR revealed that the parental control and stable SOX2-OE cells incubated with a SOX2 antibody exhibited significantly higher enrichment compared to those incubated with IgG antibodies (Fig. 4G). These results demonstrated that SOX2 can bind and activate the HEY1 promoter in HNSCC.

SOX2 correlates with HEY1 expression in HNSCC and other SCCs

To examine whether SOX2 is related to HEY1 expression in other types of cancer, this association was explored using a TCGA dataset from multiple types of cancers (Table I). SCCs, including HNSCC, esophageal and lung cancer, exhibited a significantly higher HEY1 expression compared to normal tissues. By contrast, other cancer types, such as lung adenocarcinoma, colon, breast and prostate cancer, exhibited a significantly lower HEY1 expression compared to normal tissues. All types of SCC exhibited an approximately 1.5- to 1.8-fold higher HEY1 expression compared to normal cells (Table I). The correlation between HEY1 and SOX2 was also compared using the TCGA dataset for each type of cancer. Similar to HNSCC (Fig. 4A and Table I), significant positive correlations were found between SOX2 and HEY1 in esophageal and lung SCC. The SOX2-HEY1 correlation coefficients in lung adenocarcinoma, colon, breast and prostate were <0.20, indicating a weak correlation (Table I).

Table I

TCGA dataset analysis of HEY1 expression ratios between cancer and normal tissues and SOX2-HEY1 correlation coefficients in several cancer types.

Cancer types	Number of tumor samples	Number of normal samples	HEY1 expression ratio (tumor/normal)	P-value	SOX2-HEY1 correlation coefficient in tumor samples	P-value	SOX2-HEY1 correlation coefficient in normal samples	P-value
HNSCC	522	44	1.78	<0.0001	0.45	<0.0001	0.18	0.25
Esophageal carcinoma	185	11	1.80	0.082	0.57	<0.0001	-0.56	0.073
Lung SCC	501	51	1.46	0.0003	0.57	<0.0001	0.35	0.011
Lung adenocarcinoma	517	59	0.39	<0.0001	0.12	0.0076	0.024	0.85
Colon adenocarcinoma	459	41	0.58	<0.0001	0.20	0.0001	0.099	0.54
Breast carcinoma	1100	112	0.53	<0.0001	0.15	<0.0001	0.016	0.088
Prostate adenocarcinoma	498	52	0.72	0.011	0.12	0.0088	0.58	<0.0001

HNSCC, head and neck squamous cell carcinoma.

HEY1 correlates with NOTCH4 expression

HEY1 is a NOTCH target gene; however, a previous study by the authors demonstrated a significantly lower NOTCH4 expression in HNSCC cells in which HEY1 was knocked down (17). Based on this result, it was hypothesized that HEY1 also reciprocally increased NOTCH4 expression directly through the NOTCH4 promoter. Therefore, the correlation between NOTCH4 and NOTCH downstream genes, including HEY1, was examined using the head and neck TCGA dataset (Fig. 5A). A positive correlation was noted between NOTCH4 and HEY1 mRNA expression (r=0.39, P<0.0001). However, no significant positive correlation was found between NOTCH4 and HES1 and 5 (Fig. 5A). This correlation between NOTCH4 and HEY1 was also independent of the HPV status (Fig. S4A). When the association between all NOTCH receptors and HEY1 was compared, NOTCH4 exhibited the highest positive correlation to HEY1 of all the NOTCH receptors (Figs. S4B and 5A). RT-qPCR revealed a significantly increased NOTCH4 expression in the Cal27 and SCC61 HEY1-OE cells. NOTCH4 expression did not differ significantly between the SCC090 HEY1-OE and control cells (Fig. 5B). All the HEY1-OE cells exhibited an elevated NOTCH4 expression, as shown by western blot analysis (Fig. 1A). Therefore, the present study examined whether HEY1 directly binds the NOTCH4 promoter region and promotes its transcription, similar to the association of SOX2 and HEY1.

Figure 5

HEY1 binds the NOTCH4 promoter region. (A) The correlations between NOTCH4 and NOTCH downstream genes were examined using the TCGA HNSCC dataset, including NOTCH wild-type HNSCC (n=447). HNSCC samples with NOTCH mutations are excluded (n=73). 'r' indicates the Pearson's correlation coefficient. (B) NOTCH4 expression was compared between HEY1-OE and control cells by RT-qPCR. P-values were calculated using Dunnett's t-test. (C) Luciferase reporter assay performed with vectors containing the NOTCH4 promoter region. (D) HEY1 binding regions of the human NOTCH4 promoter (shadow box) with primer pairs (underlined). Bp indicates the base pairs from the transcription starting site. (E) ChIP qPCR analysis using the NOTCH4 promoter primer in SCC61 SOX2-OE and parental control cells. Mouse IgG antibody is used as a negative control. *P<0.05, **P<0.01; N.S., not significant. HEY1-OE cells, HEY1-overexpressing cells.

HEY1 directly binds the NOTCH4 promoter in HNSCC

To assess whether HEY1 directly binds NOTCH4 promoter region, a luciferase vector with a NOTCH4 promoter region was transfected into HEY1-OE and control cells. A significant increase in luciferase activity was observed in all HNSCC HEY1-OE cells (Fig. 5C). The HEY1 gene binds E-box (CANGTG) and N-box (CACNAG) sites (31,32). Three candidates of HEY1 binding sequences were found from -884 to -922 bp upstream of the NOTCH4 transcript starting site. Therefore, a ChIP qPCR primer pair that bound this region was generated (Fig. 5D). The ChIP qPCR results revealed that the parental control and stable HEY1-OE cells incubated with a HEY1 antibody exhibited significantly higher enrichment compared to those incubated with IgG antibodies (Fig. 5E). These data demonstrate that HEY1 can bind the NOTCH4 promoter in HNSCC and drive NOTCH4 expression.

The associations between these genes are summarized in Fig. 6; these data demonstrate that SOX2 regulates HEY1 and HEY1 creates a reciprocal loop with NOTCH4 to promote HNSCC EMT, sphere formation and tumorigenicity.

Figure 6

The scheme of the present study. SOX2 regulates HEY1 and HEY1 creates a reciprocal loop with NOTCH4 to promote HNSCC EMT, self renewal and tumorigenicity.

Discussion

Previous studies have revealed that the NOTCH pathway is upregulated in HNSCC and that NOTCH expression is related to an advanced clinical stage (33,34). In a previous study, the authors observed a specific upregulation of the NOTCH4-HEY1 pathway in HNSCC (17). In the present study, further analysis of a specific NOTCH pathway and a new mechanism of functional integration is reported between SOX2 and the NOTCH4-HEY1 axis.

It was found that the HEY1-OE cells increased proliferation and tumorigenicity in xenograft models. HEY1-OE cells also increased cell invasion and migration, known as EMT ability, and promoted sphere formation, reflecting a phenotype of cell stemness, self-renewal ability in vitro (35-37). As shown in Fig. 1B, E-cadherin expression was higher in the Cal27 HEY1-OE than the control cells, and did not differ significantly between the HEY1-OE and control SCC61 and SCC090 cells. By contrast, western blot analysis revealed a decreased E-cadherin expression in all HEY1-OE cells (Fig. 1A). This may be due to post-translational processing, as the post-translational E-cadherin modification is known to induce EMT in cancer (38). E-cadherin mRNA and protein expression differed in all HEY1-OE cells. Furthermore, the expression levels of mesenchymal genes, such as fibronectin, Vimentin and TWIST1 in the HEY1-OE cells were increased in all HNSCC cells (Fig. 1B). Western blot analysis of N-cadherin expression was also performed in these cells; however, no increase in N-cadherin increase expression was found in the HEY1-OE cells (Fig. S4C). The reason for this lack of change in N-cadherin expression is not clear. The present study did not validate N-cadherin and E-cadherin expression in control and HEY1-OE cells using immunocytochemistry. This is a limitation of the present study. However, the authors have previously demonstrated that N-cadherin expression was significantly increased in the HEY1 high expression group in HNSCC patients using a TCGA dataset (17). Thus, these in vitro assay results demonstrated that HEY1 promotes EMT in HNSCC.

HEY1 expression significantly correlated with SOX2 expression in the TCGA dataset and in in vitro experiments. SOX2 is an EMT inducer gene that promotes HNSCC cell invasion and migration (23,28,39). SOX2 and HEY1 are early sensory markers that exist in the same domain in mice inner ear development (40) and SOX2 is co-expressed with HEY1/HEY2 in the inner ear (41). In glioma CSC, both SOX2 and HEY1 expression is increased compared to non-CSC glioma cells (42). Chen et al demonstrated that JAG1 promoted HEY1 and SOX2 expression, and NOTCH inhibition by a gamma secretase inhibitor decreased SOX2 promoter activity in breast cancer cells (43). However, these studies did not examine which NOTCH related genes directly interacted with the SOX2 promoter (43). In this context, the present study examined the association between SOX2 and HEY1. As noted above, the reporter and ChIP assays indicated that SOX2 regulated HEY1 expression via binding to its promoter. There are no commercially available SOX2 blocking antibodies. Thus, the authors were not able to define HEY1 expression and an EMT phenotype using SOX2 blocking antibody. Of note, the present study did not perform SOX2 mutational analysis in reporter assays of HEY1 that would allow the precise localization of the SOX2 binding HEY1 promoter within a 1-2 kb segment. The authors have previously demonstrated that HEY1 expression was significantly increased in sphere cells of HNSCC cell lines; HEY1 expression in sphere cells was increased approximately 1.4- to 3.5-fold compared with parental cells (17). In the present study, it was demonstrated that HEY1 promotes sphere formation and tumorigenicity that are closely related to a cancer stem cell phenotype. These results indicate that SOX2 maintains HNSCC CSCs through HEY1 expression.

A previous study by the authors demonstrated that HEY1 knockdown HNSCC cells decreased SOX2 expression, as shown by RT-qPCR and western blot analysis (17). These two results indicate that SOX2 and HEY1 may exist in a reciprocal loop similar to that of NOTCH4-HEY1. On the other hand, Wang et al demonstrated that glioma stem cells made a NOTCH1-SOX2 positive feedback loop (44). The TCGA analysis in the previous study by the authors also indicated that NOTCH4 and SOX2 expression had a significant positive correlation (17). These results indicate that NOTCH4 may promote SOX2 expression in HNSCC similar to HEY1 expression (Fig. 6). If NOTCH4 promotes SOX2 expression, this may explain why cells in which HEY1 was knocked down had a decreased SOX2 expression, as HEY1 knockdown in cells decreased NOTCH4 expression (Fig. 6). Previous studies have indicated this connection in neural stem cells and brain endothelial cells; NOTCH increases SOX2 promoter activity and regulates SOX2 expression (45,46).

Furthermore, the current TCGA dataset analysis also revealed that SOX2 correlates with HEY1 expression in several SCCs, but not in non-SCC cancers or normal tissue (Table I). There are several studies on SOX2 function in SCC and its promotion of tumorigenesis, metastasis and EMT (21,47,48). As regards HEY1 function in SCC, Forghanifard et al examined NOTCH pathway gene expression in 50 patients with esophageal SCC and indicated that HEY1 and HEY2 expression were significantly associated with clinical stage and a poor prognosis (49). The results of the present study and these reports indicate that HEY1 upregulation promotes tumor progression, and that a SOX2-HEY1 expression correlation is found predominantly in SCC, reinforcing a context dependent setting for activation of the NOTCH4-HEY1 pathway. However, the TCGA dataset shows only each mRNA expression of individual patients. The present study performed a functional analysis for these molecules in HNSCC, but not in other SCC types, limiting functional confirmation in these systems.

In the present study, the reporter and ChIP assays also indicated that HEY1 reciprocally regulated NOTCH4 expression via binding to its promoter. James et al demonstrated that the NOTCH4 was ligand unresponsive in NOTCH signaling (50). These results may explain why the NOTCH4-HEY1 pathway was specifically upregulated in HNSCC. Of note, specific mutational analysis in NOTCH4 reporter assays to precisely localize the HEY1 binding site would add further depth to the current understanding of this interaction. Nevertheless, the results of additional mutational analysis would not substantively alter the conclusions based on these data, that SOX2 and HEY1 interact with these promoter regions. The present study did not explore the ability of specific NOTCH ligands, such as JAG1, JAG2, DLL1, DLL3 and DLL4 in the activation of the NOTCH4-HEY1 pathway, which is a limitation of the present study, as this may provide further data that support the model of constitutive activation in the absence of exogenous ligand. The present study did not assess whether HEY1 bound to the NOTCH1-3 promoter, and NOTCH1 and 2 expression was lower in HNSCC than normal samples; however, NOTCH3 expression was slightly higher in HNSCC than normal tissue, which is similar to NOTCH4 (51). NOTCH3 also significantly and positively correlated with HEY1 expression. NOTCH4 was more significantly and highly expressed in HNSCC and positively correlated with HEY1 than NOTCH3 (Figs. 4A and S4B) (51). Notably, a similar correlation was found between NOTCH4 and HEY1, HES1 and HES5 in head and neck normal samples of the TCGA dataset and HNSCC (Figs. 5A and S4D). A moderately positive correlation was noted only between NOTCH4 and HEY1 mRNA expression in normal samples (r=0.52, P=0.0003). It was hypothesized that the NOTCH4/HEY1 pathway is a specific pathway not only active in HNSCC, but potentially active in head and neck normal epithelial cells. Based on these findings, previous studies have reported trials of anti-NOTCH therapy for HNSCC (52,53).

In conclusion, the present study demonstrates that HEY1 expression in HNSCC is regulated by SOX2 and promotes EMT. The NOTCH4/HEY1/SOX2 pathway is specifically upregulated and creates a positive reciprocal loop in HNSCC, defining a pathway that may be a novel target for HNSCC therapy.