Gli1, a potential regulator of esophageal cancer stem cell, is identified as an independent adverse prognostic factor in esophageal squamous cell carcinoma

Abstract

Purpose

The hedgehog (Hh) pathway is involved in cancer stem cell (CSC) maintenance in various tumors. Glioma-associated oncogene homolog 1 (Gli1) is a key mediator of the Hh pathway; however, its expression and clinical significance in esophageal squamous cell carcinoma (ESCC) have not been reported. In this study, we aimed to reveal clinical significance of Gli1 expression in ESCC and further investigate the potential of Gli1 as a CSC regulator of ESCC by comparing its expression with expressions of other stemness genes in ESCC.

Methods

We assessed the expressions of Gli1, Sox9, CD44, Sox2, LSD1, and Oct4 in 127 patients’ tissue specimens of ESCC using immunohistochemistry and in ESCC cell lines using Western blotting. The relationship of Gli1 expression with clinic–pathologic parameters as well as cell-cycle-regulating genes was investigated. We also investigated the biological pathways that are activated in Gli1-high ESCC using The Cancer Genome Atlas (TCGA) data.

Results

Gli1 expression was observed in 28.3 % of ESCC, and its expression was correlated with the expression of stemness genes, Sox9 (P = 0.003) and CD44 (P = 0.012). And Gli1, CD44, and Sox9 were highly expressed in more poorly differentiated ESCC cell lines such as TE8 and TE1 cells. Notably, Gli1 expression was positively associated with distant metastasis (P = 0.011), increased microvessel density (MVD) (P = 0.002), and expression of cell cycle regulators such as p21, cyclin D1, cyclin E1, and NF-κB (P < 0.05). Sox9 and CD44 expressions in ESCC were also significantly associated with unfavorable clinic–pathologic parameters such as increased MVD, advanced tumor (pT) stage, and higher TNM stage. Moreover, all three potential CSC markers such as Gli1, Sox9, and CD44 were strongly linked to worse clinical outcome and independent poor prognostic factors in overall survival and disease-free survival in ESCC. Gene set enrichment analysis revealed that the Gli1-high-expressing ESCC patients’ group was strongly enriched for gene expression signature of Hh signaling pathway, epithelial–mesenchymal transition, and cancer stem cell.

Conclusions

Targeting Gli1, a potential diagnostic marker of ESCC stem cells, will have a profound therapeutic and prognostic value.

Electronic supplementary material

The online version of this article (doi:10.1007/s00432-016-2273-6) contains supplementary material, which is available to authorized users.

Introduction

China is one of the countries with the highest incidence of esophageal squamous cell carcinoma (ESCC) worldwide; ESCC accounts for 90 % of esophageal cancers in China (Qi et al. 2012). Despite improvements in surgical techniques, enhanced imaging, and new chemotherapeutic agents, outcomes for patients remain extremely poor due to the poor prognosis of ESCC and the worst overall 5-year survival rate (Sun and Yu 2011). Therefore, it is necessary to identify effective diagnostic markers and prognostic factors for ESCC.

The glioma-associated oncogene homolog 1 (Gli1) family of zinc finger transcription factors is the nuclear mediator of the hedgehog (Hh) pathway that regulates genes essential for various stages of tumor development and progression (Min et al. 2013). In mammals, three Gli transcription factors, Gli1, Gli2, and Gli3, have been identified. Gli1 is the strongest transcriptional activator of the Hh pathway and is sufficient to induce tumor development (Guo et al. 2013). Weiwei Sheng et al. (2014) found that Gli1 was an independent adverse prognostic indicator in pancreatic carcinogenesis, glioblastoma, ERa-negative breast cancer, and head and neck squamous cell carcinoma.

In the present study, to identify Gli1 as a potential ESCC stem cell marker, Gli1 expression was examined using immunohistochemistry on tissue microarray slides from 127 human ESCC samples. To confirm the stem-cell-like characteristics of Gli1 positive cells, we analyzed and compared the expression of other stemness genes such as Sox9, CD44, LSD1, Sox2, and Oct4. In addition, we examined the gene expressional signature of Gli1-high-expressing ESCC patients’ group using The Cancer Genome Atlas (TCGA) dataset and gene set enrichment analysis (GSEA). We show here that Gli1 expression not only indicates poor prognosis for ESCC but is also a potential marker for ESCC stem cells.

Materials and methods

Tissue specimens

A total of 147 formalin-fixed and paraffin-embedded tumor tissue samples including 127 ESCC and 20 adjacent non-tumor esophageal mucosa were obtained from the Department of Pathology at Samsung Medical Center (Seoul, Korea) in accordance with protocols approved by the institutional review board (no. 2014-09-060-001). No patient received preoperative chemotherapy or radiotherapy. Clinical and pathological reports were reviewed for age, sex, tumor size, histological grade, invasion depth (pT), nodal status (pN), and distant metastasis (pM). The pTNM classification was applied according to guidelines from the 2010 American Joint Committee on Cancer staging manual (AJCC 7th edition).

Immunohistochemical staining procedure

Sections on microslides were deparaffinized with xylene, hydrated using a diluted alcohol series, and immersed in 3 % H₂O₂ in methanol to quench endogenous peroxidase activity. Sections were treated with TE buffer (10 mM Tris and 1 mM EDTA, pH 9.3) at 98 °C for 30 min. To reduce nonspecific staining, each section was blocked with 4 % bovine serum albumin in PBS with 0.1 % Tween 20 for 30 min. The sections were then incubated with anti-Gli1 monoclonal antibody (1:100, Abcam, UK), anti-Sox9 monoclonal antibody (1:100, Abnova, USA), anti-P21 (1:100, Big, USA), anti-cyclinD1 (1:100, Big, USA), anti-CyclinE (1:100, Big, USA), anti-pAkt-Thr308 (1:100, Abcam, UK), anti-pAkt-Ser473 (1:100, Abcam, UK), anti-CD44 (1:100, ZSGB-BIO, China), anti-LSD1 (1:250, Sigma, USA), anti-Sox2 (1:100, R&D, USA), NF-κB (1:100, CST, USA), and anti-Oct4 (1:100, Millipore, USA) in PBST containing 3 mg/ml goat globulin (Sigma, St. Louis, MO, USA) for 60 min at room temperature, followed by three successive washes with buffer. The chromogen used was ImmPACT AEC Peroxidase Substrate (VECTOR Laboratories) for 20 min. Sections were counterstained with Meyer’s hematoxylin. After reading and taking photograph of the slide, sections were then used stripping buffer (20 % SDS, 0.5 M Tris, and mercaptoethanol) to removing the original antibody for 1 h at the water bath of 56 °C and then for 10 min dehydrated alcohol to removing the red reaction, so that the sections can be used again. Omitting the primary antibody provided negative controls for immunostaining.

The double immunostaining procedure was performed using a two-step method with Gli1, Sox9, CD44 antibodies (developed with 3,3′-diaminobenzidine) (brown reaction product) and anti-CD105 antibodies (1:250, Abcam, Cambridge, UK) (developed with ImmPACT AEC Peroxidase Substrate) (red reaction product) to observe the relationship between the expression of Gli1/Sox9/CD105 and microvessel density (MVD) in ESCC. Primarily, for the Gli1, Sox9, and CD44 protocols, except that the chromogen with the 3,3′-diaminobenzidine (Dako) for 10 min (FLEX20), all steps are the same. Then, subsequent staining of the same section was performed after incubating the samples with an antibody to CD105 by ImmPACT AEC Peroxidase Substrate for 20 min.

Two pathologists (YHX and SHK) who did not possess knowledge of the clinical data examined and scored all tissue specimens. CD105-positive individual microvessel counts were made on a 200× fields, and three area microvessel numbers as MVD. In case of discrepancies, a final score was established by reassessment by both pathologists on a double-headed microscope. As described in detail previously, the staining results were semiquantitatively scored as negative and positive (Yang et al. 2016).

Cell lines

TE-1, TE-8, TE-10, TE-11, and HCE4 (ESCC cell lines) were maintained in DMEM with high glucose (Life Technologies, Grand Island, NY) supplemented with 10 % heat-inactivated fetal bovine serum (Life Technologies), 100 mg/ml penicillin G, and 50 mg/ml streptomycin (Life Technologies) at 37 °C in a humidified atmosphere containing 5 % CO₂. All cell lines were purchased from the RIEKN BRC Cell Bank (Tsukuba, Japan).

Western blotting analysis

Cell lysates were produced in RIPA lysis buffer (50 mM Tris pH 7.4, 150 mM NaCl, 1 mM EDTA, 1 % Triton x-100, 1 % Na-Doc, 0.1 % SDS) supplemented with protease inhibitor cocktail (Roche). Cell extracts were quantitated using a BCA Protein Assay Kit (Thermo). Western blot analysis was performed using standard techniques for anti-Gli1 (1:1000, Santa, US); anti-Sox9 (1:500, Abcam, UK); anti-CD44 (1:1000, ZSGB-BIO, China); and anti-β-actin (1:1000, Bioss, China).

Obtaining publicly available gene expression data and GSEA

For TCGA data, the following type of “Level 3” processed and normalized gene expression data of 185 esophageal cancer patients (illuminaHiseq RNAseq V2 gene expression, RSEM normalized) were downloaded from the TCGA Web site. We selected 40 Gli1-high-expressing group and 39 Gli1-low-expressing group from 185 esophageal cancer patients included in TCGA dataset. Then to identify gene-signature-based differences between Gli1-high/Gli1-low ESCC population, we performed GSEA using the Broad Institute’s GSEA tool (http://www.broadinstitute.org/gsea/index.jsp).

Statistical analysis

Correlations were examined using Pearson’s Chi-square test as appropriate; overall survival (OS) and disease-free survival (DFS) were determined using the Kaplan–Meier method and were compared using the log-rank test. Survival was measured from the date of surgery. The Cox proportional hazards model was used for multivariate analysis. Clinicopathologic factors, which were statistically significant in univariated analysis, were included as covariables in multivariate analysis. Hazard ratios (HRs) and 95 % confidence intervals (CIs) were assessed for each factor. All tests were two sided, and P value of <0.05 was considered statistically significant. These values were then corrected by Bonferroni methods according to the number of alleles analyzed. The statistical analysis was performed using SPSS statistical software (SPSS Inc, Chicago, IL, USA).

Results

The expression of Gli1 in ESCC and its relation with CSC markers and clinicopathologic characteristics

Immunohistochemical study revealed that Gli1 was abundantly expressed in both the nucleus and cytoplasm of esophageal cancer cells (28.3 %, 36/127) (Fig. 1). Because Gli1 has been known to be potential cancer stem cell (CSC) marker in other cancer (Bora-Singhal et al. 2015; Deng et al. 2015; Xu et al. 2015), we investigated the association of Gli1 with other stemness-related genes in ESCC using tissue array containing 127 ESCC patients’ tissue (Table 1). Sox9 was localized in the nuclei, whereas CD44 was present in the cell membrane of cancer cells (Fig. 1). The results were that Gli1 expression was significantly correlated with the expressions of stemness-related genes such as Sox9 (P = 0.003) and CD44 (P = 0.012), respectively, but not other CSC markers such as LSD1, Sox2, and Oct4 (Table 2). To verify the immunohistochemical results, we examined the protein levels of Gli1, CD44 and Sox9 in ESCC cell lines using Western blotting. As a result, Gli1, CD44, and Sox9 were more highly expressed in poorly differentiated ESCC cell lines such as TE8 and TE1 cells compared with the other cell lines (TE10, TE11, and HCE4) (Fig. 2).

Fig. 1.

Immunohistochemical staining of ESCC with Gli1/Sox9 (a) and Gli1/CD44 (b) in the same field, respectively

Table 1.

Comparison of clinicopathologic characteristics according to Gli1, Sox9, and CD44 expressions in ESCC tissues

Variable	n	Gli1 (+) n (%)	χ 2	P	Sox9 (+) n (%)	χ 2	P	CD44 (+) n (%)	χ 2	P
Age (years)			0.530	0.466		0.020	0.889		0.019	0.890
<65	34	8 (23.5)			17 (50.0)			10 (29.4)
≥65	93	28 (30.1)			46 (49.5)			26 (27.9)
Sex			0.000	0.989		0.112	0.738		2.871	0.090
Male	120	34 (28.3)			59 (49.2)			36 (30.0)
Female	7	2 (28.6)			4 (57.1)			0 (0)
Tumor size (cm)			1.298	0.255		0.000	0.983		1.066	0.302
<4	50	17 (34.0)			25 (50.0)			11 (22.0)
≥4	77	19 (24.7)			38 (49.4)			25 (32.5)
Grade			2.702	0.259		0.823	0.663		1.573	0.456
Well	10	2 (20.0)			5 (50.0)			2 (20.0)
Moderate	93	22 (23.7)			44 (47.3)			24 (25.8)
Poorly	24	12 (50.0)			14 (58.3)			10 (41.7)
T stage			1.887	0.169		4.966	0.026		5.632	0.018
T1	15	2 (13.3)			3 (20.0)			1 (6.7)
T2–T4	112	34 (30.4)			60 (53.6)			35 (31.3)
Lymph node metastasis			0.203	0.652		4.449	0.035		0.945	0.331
Negative	39	10 (25.6)			14 (35.9)			9 (23.1)
Positive	88	26 (29.5)			49 (55.7)			27 (30.7)
Distant metastasis			6.490	0.011		2.047	0.152		2.335	0.126
Negative	97	22 (22.7)			45 (46.4)			25 (25.6)
Positive	30	14 (46.7)			18 (60.0)			11 (36.7)
Clinical stage			2.198	0.138		4.691	0.030		4.241	0.039
1	11	1 (9.1)			2 (18.2)			1 (9.1)
2–4	116	35 (30.2)			61 (52.9)			35 (30.2)
Radiotherapy			1.970	0.160		0.338	0.533		0.117	0.732
Negative	78	16 (20.5)			37 (47.4)			22 (28.2)
Positive	49	20 (40.8)			26 (53.1)			14 (28.6)
Chemotherapy			0.335	0.563		0.036	0.850		0.000	0.994
Negative	12	5 (41.7)			6 (50.0)			3 (25.0)
Positive	115	31 (27.0)			57 (49.6)			33 (28.7)

Table 2.

Association between protein expression of Gli1 and cancer stem cell makers in esophageal squamous cancer tissues

Variable	n	Gli-1 (−) n (%)	Gli-1 (+) n (%)	χ 2	R	P	Pc
Sox9				8.850	0.263	0.003	0.002
Negative	64	53 (82.8)	11 (17.2)
Positive	63	38 (60.3)	25 (39.7)
CD44				6.261	0.227	0.012	0.006
Negative	91	70 (76.9)	21 (23.1)
Positive	36	21 (58.3)	15 (41.7)
LSD1				0.007	0.007	0.835	0.594
Negative	81	55 (67.9)	26 (32.1)
Positive	46	36 (78.3)	10 (21.7)
Sox2				1.182	0.107	0.277	0.150
Negative	92	65 (70.7)	27 (29.3)
Positive	35	26 (74.3)	9 (25.7)
Oct4				1.196	0.096	0.274	0.148
Negative	85	59 (69.4)	26 (30.6)
Positive	42	32 (76.2)	10 (23.8)

Pc P value correction

Fig. 2.

Western blot analysis of Gli1, Sox9 and CD44 in ESCC cells line. β-Actin was used as a loading control

Then, we studied the clinical relevance of Gli1 and stemness-related genes in ESCC. Gli1 expression determined by immunohistochemical study was significantly correlated with distant metastasis (P = 0.011). Sox9 expression was associated with higher pT stage (P = 0.026); lymph node metastasis (P = 0.035); and advanced clinical stage (P = 0.030). CD44 expression was also positively correlated with pT stage (P = 0.018) and clinical stage (P = 0.039) (Table 1). These results strongly suggest that Gli1 is frequently and highly expressed in ESCC with clinical significance. In addition, its expression is significantly correlated with stemness-related genes.

Gli1 and its expression-correlated CSC markers were independent prognostic factors in ESCC

The Kaplan–Meier survival analysis revealed that all three genes (Gli1, Sox9, and CD44) were strong prognostic factors in ESCC (Fig. 3). In detail, the expression of Gli1, Sox9, and CD44 in ESCC was all significantly associated with both poor overall survival (OS) and disease-free survival (DFS). (Gli1, OS: P = 0.004; DFS: P = 0.002; Sox9, OS: P < 0.001; DFS: P < 0.001; CD44, OS: P = 0.005; DFS: P = 0.003) (Fig. 3). And the other CSC markers such as LSD1 and Sox2 except for Oct4 were also positively correlated with both poor OS and DFS (LSD1, OS: P = 0.015; DFS: P = 0.480; Sox2, OS: P = 0.014; DFS: P < 0.001; Oct4, OS: P = 0.730; DFS: P = 0.536) (S1 Fig.). The univariate Cox regression analysis confirmed again that these all three genes were prognostic indicator of unfavorable clinical outcome of ESCC. In the univariate Cox regression analysis, the following were significant poor prognostic factors of both OS and DFS: pT stage (OS: P = 0.002; DFS: P = 0.002), lymph node metastasis (OS: P = 0.001; DFS: P = 0.001), distant metastasis (OS: P = 0.006; DFS: P < 0.001), Gli1 expression status in cancer cells (OS: P = 0.005; DFS: P = 0.003), Sox9 expression status (OS: P < 0.001; DFS: P = 0.001), CD44 expression status (OS: P = 0.006; DFS: P = 0.004), Sox2 expression status (OS: P = 0.008; DFS: P = 0.008), and LSD1 expression status (OS: P = 0.013; DFS: P = 0.017) (Table 3).

Fig. 3.

Kaplan–Meier analyses of overall and disease-free survival curves for Gli1 (a, d), Sox9 (b, e), and CD44 (c, f) expressions in ESCC patients

Table 3.

Univariate analyses for prognostic variables of overall survival and disease-free survival in ESCC patients using Cox proportional hazards regression

Characteristic	Overall survival			Disease-free survival
	HR	95 % CI	P value	HR	95 % CI	P value
Age (years)			0.110			0.106
<65	1.00		–	1.00		–
≥65	1.473	0.917–2.368		1.454	0.923–2.290
Tumor size (cm)			0.610			0.756
<4	1.00		–	1.00		–
≥4	1.111	0.742–1.662		1.063	0.723–1.563
T stage			0.002			0.002
1	1.00			1.00
2–4	9.041	2.227–36.699		4.760	1.751–12.937
Lymph node metastasis			0.001			0.001
Negative	1.00		–	1.00		–
Positive	2.254	1.390–3.656		2.207	1.389–3.505
Distant metastasis			0.006			<0.001
Negative	1.00		–	1.00		–
Positive	1.842	1.192–2.845		2.750	1.824–4.145
Gli1			0.005			0.003
Negative	1.00		–	1.00		–
Positive	2.001	1.233–3.248		2.006	1.273–3.163
Sox9			<0.001			0.001
Negative	1.00		–	1.00		–
Positive	2.139	1.404–3.260		1.998	1.338–2.984
CD44			0.006			0.004
Negative	1.00	1.198–3.030		1.00	1.232–2.974
Positive	1.906			1.914
LSD1			0.013			0.017
Negative	1.00		–	1.00		–
Positive	1.696	1.118–2.574		1.622	1.089–2.415
Sox2			0.008			0.008
Negative	1.00		–	1.00		–
Positive	1.920	1.181–3.121		1.873	1.175–2.986
Oct4			0.219			0.253
Negative	1.00		–	1.00		–
Positive	0.765	0.499–1.172		0.788	0.524–1.186

The multivariate survival analysis using Cox proportional hazards model revealed that these three genes were also independent poor prognostic factors. Specifically, patients’ age, pT stage, nodal metastasis, and distant metastasis were included as covariates in analysis. The results were that Gli1 expression status (HR = 1.825, P = 0.026 in OS), Sox9 expression status (HR = 1.833, P = 0.005 in OS; HR = 1.634, P = 0.019 in DFS), CD44 expression status (HR = 1.585, P = 0.046 in DFS), and Sox2 expression status (HR = 1.668, P = 0.044 in OS; HR = 1.662, P = 0.040 in DFS) were adverse independent prognostic predictors of ESCC in terms of either OS or DFS (Table 4).

Table 4.

Multivariate analyses for prognostic variables of overall survival and disease-free survival in ESCC patients using Cox proportional hazards regression

Characteristic	Overall survival			Disease-free survival
	HR	95 % CI	P value	HR	95 % CI	P value
Age (years)			0.099			0.050
<65	1.00		–	1.00		–
≥65	1.596	0.916–2.781		1.764	1.001–3.109
Tumor size (cm)			0.993			0. 961
<4	1.00		–	1.00		–
≥4	0.998	0.621–1.604		0.988	0.611–1.598
T stage			0.035			0.117
1	1.00			1.00
2–4	4.758	1.114–20.321		2.611	0.807–7.770
Lymph node metastasis			0.129			0.044
Negative	1.00		–	1.00		–
Positive	1.559	0.879–2.765		1.775	1.016–3.100
Distant metastasis			0.050			0.004
Negative	1.00		–	1.00		–
Positive	1.649	1.000–2.722		2.170	1.284–3.666
Gli1			0.026			0.057
Negative	1.00		–	1.00		–
Positive	1.825	1.076–3.096		1.623	0.986–2.670
Sox9			0.005			0.019
Negative	1.00		–	1.00		–
Positive	1.833	1.198–2.807		1.634	1.085–2.460
CD44			0.057			0.046
Negative	1.00	0.986–2.560		1.00	1.008–2.492
Positive	1.588			1.585
LSD1			0.197			0.229
Negative	1.00		–	1.00		–
Positive	1.328	0.863–2.042		1.286	0.853–1.936
Sox2			0.044			0.040
Negative	1.00		–	1.00		–
Positive	1.668	1.014–2.745		1.662	1.024–2.698
Oct4			0.332			0.427
Negative	1.00		–	1.00		–
Positive	0.808	0.524–1.244		0.846	0.559–1.279

Gli1 and its expression-correlated CSC markers were significantly associated with increased microvessel density and enhanced expression of cell-cycle-regulating genes in ESCC

To further understand how Gli1 and its expression-correlated stemness genes such as Sox9 and CD44 promote the progression of ESCC, we studied the association between these three genes and MVD as well as cell-cycle-regulating genes. To this end, immunohistological double stainings for Gli1/CD105, Sox9/CD105, and CD44/CD105 were performed. MVD was significantly higher in Gli1 (P = 0.002), Sox9 (P = 0.014), and CD44 (P = 0.037) expression-positive ESCC cases than in the negative ones (Fig. 4). To determine the association between these three genes and cell-cycle-regulating genes, the expression levels of p21, Cyclin D1, and Cyclin E were examined by IHC. Interestingly, Gli1, p21, Cyclin D1, and Cyclin E were colocalized in the same cancer cells (Fig. 5). Gli1 expression was associated with the expression of p21 (P = 0.042), Cyclin D1 (P = 0.034), Cyclin E (P = 0.008), and NF-κB (P = 0.038); however, Gli1 expression was not associated with pAKT-Thr308 and pAKT-Ser473 (Table 5).

Fig. 4.

Images of immunohistochemical double staining in ESCC. a Immunostaining of Gli1, Sox9, and CD44 (brown reaction product) in the nuclear and CD105 (red reaction product) around the cancer cells, respectively. b Gli1, Sox9, and CD44 in ESCC were significantly associated with increased MVD, respectively

Fig. 5.

Immunohistochemical staining of Gli1, p21, CyclinD1, and CyclinE in the same field of ESCC

Table 5.

Association between protein expression of Gli1 and the factors of cell cycle in esophageal squamous cancer tissues

Variable	n	Gli–1 (−) n (%)	Gli-1 (+) n (%)	χ 2	R	P	Pc
p21				4.138	0.181	0.042	0.021
Negative	83	64 (77.1)	19 (22.9)
Positive	44	27 (61.4)	17 (38.6)
Cyclin D1				4.473	0.189	0.034	0.017
Negative	80	63 (78.8)	17 (21.2)
Positive	47	28 (59.6)	19 (40.4)
Cyclin E				6.985	0.240	0.008	0.004
Negative	85	68 (80.0)	17 (20.0)
Positive	42	23 (54.8)	19 (45.2)
pAkt-Thr308				0.695	0.90	0.404	0.228
Negative	61	46 (75.4)	15 (24.6)				00
Positive	66	45 (68.2)	21 (31.8)
pAkt-Ser473				0.348	0.053	0.555	0.333
Negative	41	31 (75.6)	10 (24.4)
Positive	86	60 (69.8)	26 (30.2)
NF-κB				5.360	0.205	0.021	0.011
Negative	80	63 (78.8)	17 (21.3)
Positive	47	28 (59.6)	19 (40.4)

Pc P value correction

Sox9 was positively correlated with the expression of cell cycle genes such as p21 (P = 0.021), Cyclin D1 (P = 0.028), and pAKT-Thr308 (P = 0.032) (S2 Table). Similarly, the expression of CD44 was positively correlated with that of p21 (P = 0.003) (S3 Table).

Mesenchymal and CSC-related gene signatures are enriched in Gli1-high-expressing esophageal cancer

To identify potential functional implication of Gli1 in ESCC, we next investigated gene expression signatures of Gli1-high-expressing ESCC using GSEA (gene set enrichment assay) and public dataset. For this study, we selected Gli1-high- and Gli1-low-expressing ESCC groups from a representative pooled cohort of 185 primary esophageal cancer patients of TCGA dataset. Then, we performed GSEA using the Broad Institute’s GSEA tool (http://www.broadinstitute.org/gsea/index.jsp). GSEA revealed that the expression of Gli1 genes was strongly associated with Hh signaling pathway (NES = 2.03, FDR q-val = 0.09) as expected (Fig. 6). Gene sets involved in the process of extracellular cellular matrix secretion (NES = 1.91, FDR q-val = 0.005), epithelial mesenchymal transition (EMT) (NES = 1.70, FDR q-val = 0.22), and cancer mesenchymal signature (NES = 1.64, FDR q-val = 0.045) were upregulated in Gli1-high population (Fig. 6). Moreover, we found that mesenchymal stem cell (FDR = 1.91, FDR q-val = 0.24) and CSC gene signature (NES = 1.73, FDR q-val = 0.004) were markedly enriched in Gli1-high-expressing population (Fig. 6). Altogether, these results strongly suggest that Gli1-high-expressing ESCC is highly likely to be associated with more aggressive feature of cancers, such as EMT and CSC.

Fig. 6.

Gli1-high-expressing esophageal cancers are enriched in gene signatures for mesenchymal character and cancer stem cell. a Schematic diagram of study design. b GSEA results for the hedgehog signaling, epithelial mesenchymal transition, and cancer stem cell signatures in Gli1-high-expressing esophageal cancer; NES normalized enrichment score, FDR false discovery rate

Discussion

CSCs are considered the new targets of cancer therapy as they have the potential to form new tumors and are difficult to kill due to their multidrug resistance abilities (Du et al. 2008). Hedgehog (Hh) signaling pathway is reported to be involved in the maintenance of CSCs in various tumors including glioma, multiple myeloma, myeloid leukemia, colorectal cancer, gastric cancer, and melanoma (Senzer et al. 2012). Data from our present study suggest that Gli1 may also play an import role in cancer stem cells of ESCC. This is the first study to investigate the expression and significance of Gli1 and cancer stemness-related genes such as Sox9, CD44, LSD1, Sox2, and Oct4 in ESCC as far as we know.

Souzaki et al. (2011) studied 149 cases of Gli1 expression in breast cancer and found that Gli1 nuclear expression has a positive correlation with breast cancer invasion and metastasis. O’Toole et al. (2011) reported a similar finding that Gli1 and Hh ligand expressions were related to the risk of metastasis and poor prognosis. Our data showed that Gli1 expression in cancer cells was associated with distant metastasis, which provides further evidence that Gli1 might be involved in early invasion and metastasis in ESCC, leading to poor prognosis. Since dysregulation of the cell cycle could lead to malignant cellular proliferation (Lange and Yee 2011), we further investigated p21 and Cyclin D1/E, which participated in the regulation of P53-induced cell cycle arrest (Abbas and Dutta 2009; de Carcer et al. 2007), and they were positively associated with Gli1 expression in our data. Wang et al. (2010) have shown that Hh-Gli1 signaling blockade leads to glioma cell growth inhibition through cell cycle G0-/G1-phase arrest, as well as enhanced apoptosis. Therefore, our results and previous studies indicate that Gli1 would likely to promote tumor proliferation by inducing cell cycle arrest. Additionally, the association between Gli1 expression and MVD was studied as the angiogenesis drives tumor growth and malignant progression. Cui et al. (2012) revealed that the degree to which the Hh/Gli1 pathway was activated was positively correlated with MVD of gliomas. Our result suggests that the aberrantly active Hh/Gli1 pathway may contribute to angiogenesis as well as ESCC proliferation. Moreover, to elucidate the prognostic value of Gli1, we performed multivariate analysis of the survival rate, which showed that the overexpression of Gli1 was associated with poor OS in ESCC patients. In human gliomas, CSCs require the Hh/Gli1 pathway activity for proliferation, survival, self-renewal, and tumorigenicity (Clement et al. 2007). In breast cancer patients, a paracrine signature defined as high epithelial Hh ligand and high stromal Gli1 has been found to be an independent predictor for OS in multivariate analysis in a cohort of 279 patients (O’Toole et al. 2011). Highly consistent with these reports, our survival analysis exhibited that Gli1 overexpression in ESCC was significantly associated with poor prognosis. Overall, our results suggest that the upregulation of Gli1 expression in ESCC may play a key role in tumor growth and cancer cell proliferation, leading to poor prognosis.

Similarly, Sox9 has been established as a key regulator of tissue stem and progenitor cells during development in several organ systems; more recently, it has been linked to cancer stem cell function. Hong et al. (2015) found that Sox9 plays an important role in promoting the proliferation and tumorigenesis of ESCC and may represent a novel prognostic marker for the disease. Stem cells in the esophageal squamous epithelium which express CD44 act as tumor-initiating cancer stem cells as they showed higher tumorigenic potential in vivo (Zhao et al. 2011). Moreover, Gotoda’s et al. (2000) results indicate that CD44 variant 2 is a useful marker for clinical prognosis in patients with ESCC. In this study, we found that the Gli1+ cell population was not only overlapped with the CD44+ and Sox9+ cell population in the ESCC cell lines (TE8 and TE1), but Gli1 also co-localized with CSC markers in the same cancer cells. Moreover, Gli1 expression was found to be closely linked to two ESCC stem cell markers CD44 (Islam et al. 2015) and Sox9 (Hong et al. 2015), which in turn indicates that Gli1 is a potential marker for CSCs in ESCC. In addition, NF-κB has also been reported to activate the Hh signaling pathway and might induce Gli1 expression partially (Boopalan et al. 2015). Colavito et al. (2014) have shown the evidence of cross-talk between the Gli1 signaling and NF-κB pathways in EMT, indicating that activated Gli1 could be a mechanism that also operates in breast cancer stem cells. Interestingly, the expression of NF-κB was positively correlated with that of Gli1 in the current study. However, further study is required to elucidate the mechanism of regulation for Gli1 expression and CSCs in ESCC.

In conclusion, we have shown that Gli1 expression might be an effective diagnostic marker for ESCC stem cells, as the upregulation of Gli1 appears to be crucial for proliferation and tumor growth in ESCC and also indicates poor prognosis. Therefore, these results suggest a more profound therapeutic and prognostic value of Gli1 in ESCC.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Compliance with ethical standards

Conflict of interest

We declare that we have no conflict of interest.

Ethics statement

This research complied with the Helsinki Declaration and was approved by the Human Ethics Committee and the Research Ethics Committee of Samsung Medical Center. All patients provided written informed consent according to institutional guidelines. Patients were informed that the resected specimens were stored by the hospital and potentially used for scientific research and that their privacy would be maintained. Follow-up survival data were collected retrospectively through medical record analyses.