Initiation of esophageal squamous cell carcinoma (ESCC) in a murine 4-nitroquinoline-1-oxide and alcohol carcinogenesis model

Esophageal squamous cell carcinomas (ESCCs) are very common, aggressive tumors, and are often associated with alcohol and tobacco abuse. Because ESCCs exhibit high recurrence rates and are diagnosed at late stages, identification of prognostic and drug targets for prevention and treatment is critical. We used the 4-nitroquinoline-1-oxide (4-NQO) murine model of oral carcinogenesis and the Meadows-Cook model of alcohol abuse to assess changes in the expression of molecular markers during the initial stages of ESCC. Combining these two models, which mimic chronic alcohol and tobacco abuse in humans, we detected increased cellular proliferation (EGFR and Ki67 expression), increased canonical Wnt signaling and downstream elements (β-catenin, FoxM1, and S100a4 protein levels), changes in cellular adhesive properties (reduced E-cadherin in the basal layer of the esophageal epithelium), and increased levels of phosphorylated ERK1/2 and p38. Additionally, we found that treatment with ethanol alone increased the numbers of epithelial cells expressing solute carrier family 2 (facilitated glucose transporter, member 1) (SLC2A1) and carbonic anhydrase IX (CAIX), and increased the phosphorylation of p38. Thus, we identified both 4-NQO- and ethanol-specific targets in the initial stages of esophageal carcinogenesis, which should lead to the development of potential markers and therapeutic targets for human ESCC.


INTRODUCTION
Cancers of the esophagus, which can be divided into squamous cell carcinomas (ESCCs) and adenocarcinomas, are the eighth most common malignancy worldwide, affecting over 450, 000 people [1]. Esophageal cancers typically are classified as a subgroup of cancers of the upper aerodigestive tract (UADT), which also include malignancies of the oral cavity, pharynx, and larynx [2]. Additionally, esophageal cancers can be grouped with malignancies that affect the digestive system as the fifth most common in terms of new cases and the fourth in terms of estimated deaths in the United States during 2012 [3]. Moreover, ESCC represents one of the top 10 types of cancer-related deaths in males between the ages of 40 and 59 years [3]. Although there have been advances in the diagnosis, operative techniques, and prognosis, the 5 year relative survival rate (5-YRSR) still remains low at 19% [3]. The low 5-YRSR for ESCC can be attributed primarily to diagnosis at an advanced stage, characterized by invasion and metastasis to the lymphatic system and remote organs at the time of diagnosis [4]. Identification of predictive biomarkers to identify and diagnose ESCC at earlier stages, before the observation of frank tumors, would be beneficial for patient survival.
Alcohol and tobacco abuse are major risk factors for the development of ESCC and these two social risk factors may have a synergistic effect on the initiation of ESCC [5,6]. The odds ratio (OR) for oral cavity cancers of chronic tobacco smokers and alcohol drinkers is 50 to 1, relative to never smokers and drinkers [5]. The International Agency for Research on Cancer (IARC) has determined that chronic alcohol consumption can lead to the development of cancers of the oral cavity, esophagus, and liver [7]. The role of alcohol in the initiation and development of ESCC is closely related to the metabolism of ethanol. Alcohol dehydrogenases (ADHs) oxidize ethanol to acetaldehyde, whose concentrations in alcoholic beverages can reach approximately 200 mM [8]. Acetaldehyde is then metabolized into acetate by aldehyde dehydrogenase-2 (ALDH2) [7]. Acetaldehyde, a genotoxic compound, acts as a carcinogen in humans by inducing mutations, promoting sister chromatid exchange, and interfering with DNA synthesis and repair [7,9]. The formation of acetaldehyde-derived DNA adducts, such as N 2 -(3hydroxybutyl)-dG, α-methyl-ϒ-OH-propano-dG, and N 2 -(4-hydroxybutyl)-dG, induces DNA polymerase errors, thus initiating mutations that silence tumor suppressors and/or activate oncogenes [2,10].
Regarding tobacco abuse, the IARC has determined that cigarette smoke contains over 60 putative carcinogens, with 15 of these, including polycyclic aromatic hydrocarbons (PAHs) and N-nitrosamines, confirmed as carcinogenic in humans [11,12]. The carcinogens in cigarette smoke can be metabolized by cytochrome P450 enzymes into water-soluble forms, generating DNA adducts [13]. For instance, the major adduct of the PAH, Benzo[a]pyrene (BaP), generates G-to-T transversions in DNA. These G-to-T transversions often are preferred sites for the formation of CpG islands, which have been linked to the silencing of tumor suppressor genes [14][15][16].
To understand the roles of chronic tobacco and alcohol abuse in the initiation of ESCC, we have combined the Meadows-Cook and 4-Nitroquinoline-1-oxide (4-NQO) carcinogenesis models. The Meadows-Cook murine model simulates the effects of chronic alcohol abuse of humans in the oral cavity, the esophagus, and the liver [17]. The 4-NQO murine model of oral and esophageal carcinogenesis in which the carcinogen, 4-NQO, is a surrogate for tobacco [18], has been extensively used to analyze cancers of the oral cavity and esophagus in animal models [19][20][21]. We and others have used these two models to characterize the molecular effects of chronic alcohol and tobacco abuse in the oral cavity [22][23][24]. We combined these two models to simulate chronic alcohol and tobacco abuse in humans, to investigate the molecular alterations present during the early stages of ESCC, and to identify ethanol-specific targets in ESCC carcinogenesis.
Here we show that changes in the expression of molecular markers involved in cellular proliferation, oncogenic signaling, cellular adhesion, and cellular metabolism occur during the initial stages of ESCC.

RESULTS
Histopathological analyses of esophagi from mice subjected to the 4-NQO murine model for oral carcinogenesis and the Meadows-Cook model for chronic alcohol abuse.

Figure 3: 4-NQO and 4-NQO followed by ethanol administration increase E-cadherin in the suprabasal layers, decrease E-cadherin in the basal layer, and increase canonical Wnt signaling during the initiation of ESCC. (A) Densitometry
analyses, performed by ImageJ analysis software, were used to determine the percentages of cells in the epithelial layer stained by the signaling targets described in Figure 2 -E-cadherin (suprabasal layer expression and basal layer expression), β-catenin, FoxM1, and S100a4. (B) QRT-PCR analysis of transcript levels of a canonical Wnt ligand (Wnt3a), noncanonical Wnt ligands (Wnt5a and Wnt7a), noncanonical Frizzled receptors (Fzd2 and Fzd6), and a canonical downstream target (S100a4). In B, the fold change for each target was determined by normalizing the ratios of the target mRNAs to 36B4 and then to the V.C./Untr. group (set at 1.0). For panels A and B, ANOVA combined with the Tukey post-hoc tests determined statistical significance, where each bar represents mean±s.d. of 3-5 mice and *p<0.05, **p<0.01, and ***p<0.001. Primer pairs for the targets in B can be found in Supplementary Table 1.
To understand further the role of canonical and noncanonical Wnt signaling in the initiation of ESCC, we measured the transcript levels of Wnt ligands and Fzd receptors (Fig. 3B). As expected, we detected increases in Wnt3a transcript levels in mice treated with 4-NQO (4-NQO/Untr.) (~3.1 fold, p<0.01) and 4-NQO followed by ethanol (4-NQO/EtOH) (~3.5 fold, p<0.01; Fig. 3B). Also, we observed a significant decrease in transcripts of Wnt5a, a noncanonical Wnt ligand, in the 4-NQO/Untr. and 4-NQO/EtOH groups (Fig. 3B). However, we did not observe any changes in Wnt7a (another noncanonical Wnt ligand) transcript levels among the four groups (Fig. 3B). Interestingly, there were large increases in the transcript levels of Fzd2 and Fzd6, noncanonical Fzd receptors, in the esophagi of mice from the 4-NQO/Untr. and 4-NQO/ EtOH treatment groups (Fig. 3B). These results show that the administration of 4-NQO can modify several markers of the canonical and noncanonical Wnt signaling pathways.

Chronic 4-NQO and 4-NQO/EtOH treatments induce changes in cell proliferation in the esophagus.
By combining the 4-NQO carcinogenesis and Meadows-Cook models for chronic alcohol abuse we delineated changes in various signaling pathways during the initiation of ESCC. We harvested esophageal tissue from mice at only 11 weeks after the termination of 10 showing the levels of phosphorylated β-catenin, phosphorylated ERK 1/2, and phosphorylated p38. The GAPDH blots, which represent loading controls, are located directly below the corresponding phosphorylated targets. (B) Quantitative analysis of phosphorylated levels of β-catenin, ERK 1/2, and p38. The fold change for each target was determined by normalizing the ratios of the target phosphorylated proteins levels to GAPDH protein levels and then to the V.C./Untr. group. Statistical significance was determined by repeating the Western blotting analyses three times with the same samples to ensure the reproducibility of the blots, where **p<0.01 and ***p<0.001. Note that the y-axes are different for each graph in B. weeks of 4-NQO administration so that we could analyze the early events in esophageal carcinogenesis. We show that the 4-NQO/Untr. and 4-NQO/EtOH treatments induce major changes in the architecture of the epithelium (Table 1 and Fig. 1A). Compared to the V.C./Untr. group, we measured increased levels of low-grade dysplasia in the 4-NQO/Untr. and 4-NQO/EtOH groups and increased levels of high-grade dysplasia in the 4-NQO/ EtOH group (Table 1). Thus, ethanol enhanced the early steps (hyperplasia and dysplasia) of 4-NQO-induced carcinogenesis in the esophagus.
To analyze cell proliferation during the initiation of ESCC, we used quantitative IHC to measure the staining of EGFR and Ki67 in the esophagus (Fig. 1B-E). Both EGFR [25] and Ki67 [26] have been used extensively in IHC analyses as markers to characterize proliferation in cancers of the oral cavity and esophagus in both humans and mice. In both the V.C./Untr. and V.C./EtOH groups, EGFR(+) (Fig. 1B, 1D) and Ki67(+) (Fig. 1C, 1E) cells were limited to the basal layer. In the 4-NQO/Untr. and 4-NQO/EtOH groups, in contrast, we detected suprabasal EGFR staining (Fig. 1B). We previously reported increases in these markers in the suprabasal layers of the tongues of mice in both the 4-NQO/Untr. and 4-NQO/ EtOH experimental groups [24]. Thus, ESCC induced by 4-NQO administration could involve horizontal expansion of epithelial basal stem cells, which has been demonstrated in the tongue [23] and the skin [36].

4-NQO and 4-NQO followed by ethanol treatments modify E-cadherin expression in the esophagus.
In normal tissues E-cadherin mediates cell-cell adhesion and functions as a tumor suppressor; however, its deregulation can initiate epithelial-to-mesenchymal transition (EMT) and tumor metastasis [37]. Often, reduced E-cadherin expression occurs at the genetic level through mutations and gene hypermethylation [38]. Here, we detected approximately a 50% reduction in the number of epithelial basal cells that are E-cadherin(+) in the 4-NQO/Untr. and 4-NQO/EtOH groups compared to the V.C./Untr. group (Figs. 2A, 3A). In the 4-NQO/Untr. and 4-NQO/EtOH groups we detected increased numbers of E-cadherin stained cells in the suprabasal layers ( Figs.  2A, 3A).

Administration of 4-NQO activates the canonical Wnt signaling pathway during the initial stages of ESCC.
Increased activity of the canonical Wnt signaling pathway, indicated by the nuclear localization of β-catenin and its interaction with Wnt responsive elements, has been characterized in ESCC via increased expression of downstream factors involved in tumorigenesis, such as Cyclin D1 and c-Jun [39]. We measured increased numbers of murine esophageal cells that were β-catenin, FoxM1, and S100A4 positive by IHC (Figs. 2, 3), and we found reduced levels of phosphorylated β-catenin protein by Western blotting (Fig. 6), similar to what we demonstrated in the tongue [24]. Additionally, we detected increased numbers of cells that express β-catenin(+) (Fig.  5A, B) and FoxM1(+) cells (Fig. 5C, D) in human patients diagnosed with ESCC.
Increased FoxM1 expression has been implicated in many types of cancers, including HNSCC and ESCC [40]. More specifically, high expression of FoxM1 induces changes in DNA methylation and epigenetic remodeling programs to maintain stem/progenitor cell renewal and to antagonize pathways inducing differentiation [41,42]. Modifications in the proliferative nature of stem/progenitor cells have been implicated in the initiation and recurrence of several cancers, including HNSCC and OSCC [43,44]. FoxM1 also directly interacts with β-catenin during glioma tumorigenesis to induce nuclear localization of β-catenin, a clear indication of canonical Wnt signaling activation [31]. Moreover, FoxM1 has been linked to the development of chemotherapeutic drug resistance. Additionally, Oncomine microarray analyses show that FoxM1, along with other Forkhead box transcript levels (e.g. FoxK1 and FoxK2) are greatly increased in human ESCC (Supplementary Table 3). Further analysis of the Forkhead box transcript and protein levels could identify additional roles of this family of transcription factors in esophageal tumorigenesis.
Metastasis is a major characteristic of late stage tumorigenesis in ESCC and HNSCC. In nontransformed cells, the S100A subfamily of proteins influences calcium homeostasis, which controls cell survival, differentiation, and metabolism [45]. In the esophagus, S100A4 may contribute to the metastatic stage by inducing the phosphorylation of Akt, mTOR, and p70S6K, and by interactions with myosin IIA and various matrix metalloproteinases (MMPs) [46]. Although we did not observe metastasis at the time point assayed in our model, we did detect increased numbers of cells expressing S100A4, both at the protein (Figs 2D, 3A) and mRNA levels (Fig. 3B), in the esophagi of the 4-NQO/Untr. and 4-NQO/EtOH treated mice.
Activation of the canonical Wnt signaling pathway can induce neoplastic transformation [47]. However, the definitive role for the noncanonical Wnt signaling pathway in cancer remains controversial. Reports have demonstrated both over-and underexpression in Wnt5a transcripts, a key noncanonical Wnt ligand, in tumorigenesis [48]. We detected reduced levels of Wnt5a transcripts, but increased Fzd2 and Fzd6 noncanonical Wnt receptors, in the 4-NQO/Untr. and 4-NQO/EtOH groups compared to the V.C./Untr. group (Fig. 3B). Also, we did not detect any changes in the transcript levels of Wnt7a (Fig. 3B), another noncanonical Wnt ligand. Oncomine data from human ESCC specimens show that Wnt5a (Supplementary Table 4) mRNA is reduced and Fzd2 and Fzd6 (Supplementary Table 3) transcripts are increased, similar to our data in this murine model. The noncanonical Wnt signaling pathway may prevent early transformation of cells at the expense of canonical Wnt signaling; however, the noncanonical pathway can enhance cellular invasiveness after transformation [49]. Additional investigations will have to be conducted to delineate the role of noncanonical Wnt signaling in carcinogen-induced ESCC.

Ethanol increases SLC2A1-and CAIX-positive epithelial cells in the esophagus.
In many types of the cancer, hypoxia (characterized by a low concentration of oxygen) can induce changes in the tumor microenvironment that switch cellular metabolism from oxidative phosphorylation to glycolysis, thus increasing glycogen synthesis and the use of glutamine instead of glucose for energy production [50]. In addition, hypoxia is often associated with low survival rates, poor responses to chemotherapy, and higher resistance to anticancer drugs in many cancers [50]. Although there is substantial information regarding SLC2A1 (also known as GLUT1) and CAIX expression in HNSCC and other malignancies [51,52], the information is more limited for animal models of ESCC.
We detected increased percentages of SLC2A1(+) cells in the esophagi of mice after the administration of ethanol, 4-NQO, and 4-NQO followed by ethanol (Fig.  4A, 4C). Even though there are very few reports linking elevations in SLC2A1 to ESCC, published microarray data show that SLC2A1 is in the top 10% of overexpressed transcripts in two separate data sets generated from ESCC patients (Supplementary Table 5). Transcripts of other SLC family members involved in cell metabolism, eg. SLC2A2 (GLUT2), SLC2A3 (GLUT3), and SLC16A1 (monocarboxylic acid transporter 1; MCT), are also overexpressed in ESCC patients (Supplementary Table 5). However, the levels and duration of alcohol and/or tobacco consumption by these patients were not reported in these data sets. Finally, a similar increase in SLC2A1 protein levels in the esophagi of 4-NQO/Untr. and 4-NQO/EtOH treated mice also was observed in the malignant tissue from a human ESCC tissue microarray (Fig. 5E, 5F).
We also identified CAIX as a possible target of alcohol exposure in ESCC, as we measured increases in the percentages of CAIX(+) epithelial cells expressing CAIX protein in the esophagi of mice in the V.C./EtOH, 4-NQO/Untr., and 4-NQO/EtOH groups compared to the V.C./Untr. group (Fig. 4B, 4D). These data implicate glucose metabolism (i.e. SLC2A1) and carbonic anhydrases (i.e. CAIX) as potential biomarkers for the contribution of chronic alcohol consumption to the initial stages of ESCC.

The administration of 4-NQO followed by ethanol is associated with induction of MAPK pathways through the phosphorylation of ERK 1/2 and p38 .
The members of the mitogen-activated protein kinase (MAPK) family can mediate various cellular functions and responses that are induced by growth factors, hormones, and cellular stress. Aberrant regulation of MAPK signaling is seen in many cancer types, including ESCC. The four major MAPK pathways include the extracellular signal-regulated kinase (ERK), Big MAP kinase-1 (BMK1), p38, and c-Jun N terminal kinase (JNK), where the ERK and BMK1 pathways respond to various growth factors and hormone signals and the JNK and p38 pathways respond to extracellular stress signals [53]. Here, we detected higher levels of phosphorylated ERK 1/2 after the administration of 4-NQO and 4-NQO/ EtOH (Fig. 6A, 6B), similar to what we and others reported in the oral cavity and esophagus [24,[54][55][56].
We measured increased levels of phosphorylated p38 in the V.C./EtOH group compared to the V.C./Untr. group (Fig. 6A, 6B). We also observed increased levels of total p38 in the oral cavities of the V.C./EtOH group [24]. The activation of p38 through phosphorylation on Thr180 and Tyr182 residues has been associated with the angiogenic, invasive, and migratory properties of ESCC [57]. p38 contributes to immune and inflammatory responses by causing the release of tumor-related cytokines and chemokines such as interleukin-6 (IL-6), platelet-derived growth factor (PDGF), and vascular endothelial growth factor (VEGF) in ESCC [57,58]. Also, activation of the p38 pathway may influence the activation of the canonical Wnt signaling pathway in transformed cell lines and embryonic stem cells [59,60]. More specifically, p38 inhibits the phosphorylating capacity of GSK-3β, which participates in the degradation of β-catenin by phosphorylating the Ser33/37 and Thr42 residues of β-catenin [60]. Our study shows that the activation of the p38 MAPK pathway, possibly through a response to extracellular stress, may be an ethanol-specific effect in the initiation of ESCC.

CONCLUDING REMARKS
We have combined the 4-NQO model of oral carcinogenesis and the Meadows-Cook model of chronic alcohol abuse to investigate the molecular changes associated with the initial stages of esophageal carcinogenesis. In both the tongue [23,24] and in the esophagus, we saw increased cell proliferation, increased expression of canonical Wnt signaling markers, and changes in the expression of cell-cell adhesion molecules.
Additionally, we detected similar expression patterns of makers involved in the canonical Wnt signaling pathway and glycolysis in malignant esophageal tissue derived from human ESCC patients. Because of the similarities of the esophageal epithelial tissue between humans and mice the combination of the 4-NQO and Meadows-Cook models serves as an excellent model to determine prognostic markers for early ESCC and to analyze potential drugs for prevention or treatment of ESCC.

Esophagus tissue processing
Hematoxylin and eosin (H&E)-stained, paraffinembedded cross sections of the esophagi were analyzed by a trained pathologist (T.S.) blinded to the four experimental groups. The sections of the esophagi from the H&E-stained slides were classified as normal, hyperplasic, dysplastic (low-or high-grade), or squamous cell carcinoma (SCC).

RNA isolation, reverse transcription, and QRT-PCR
Esophagi were homogenized in TRIzol (Cat# 155596-026, Life Technologies, Norwalk, CT) and total RNA extraction was performed as described by the manufacturer. QRT-PCR was performed after reverse transcription of the total RNA (1 µg). Primer pairs used for the QRT-PCR can be found in Supplementary Table 1.

Semiquantitative IHC analysis
Formalin-fixed, paraffin-embedded esophageal sections (3-5 from each experimental group) were photographed at 200x magnification. Using ImageJ 1.48v (http://rsp.info.nih.gov/ij), the photographs were resolved into separate RGB channels, where the green channel was used for analysis. Within each photo, only the epithelial layer was selected for analysis by setting the threshold levels to a point were DAB(+) cells were selected. Then ImageJ calculated the total area of the DAB(+) cells compared to the total area of the epithelial layer. The results are shown as the fold changes in percent of the epithelial layer positive for the stain by comparing to the photographs taken from the V.C./Untr. experimental groups, where the percent positive was set at 1.0.

Western blotting analysis
Total protein lysates (30 µg) were extracted from the esophagi of three mice from each experimental group and Western blotting was performed three times as described in [24]. Details regarding the antibodies used for Western blotting can be found in the Supplementary Materials and Methods. www.impactjournals.com/oncotarget

Oncomine gene expression data analysis
The top 10% under and overexpressed transcripts of various genes in human ESCC-derived tissue were obtained from the Oncomine Cancer Microarray database analysis (http://www.oncomine.org). The following data sets were used: Hu Esophagus (RNA data set comprised of 12, 603 genes from 34 patients) and Su Esophagus 2 (RNA data set comprised of 17, 779 genes from 106 patients).

Statistical analyses
The Fisher's exact probability test was used to determine significance among the distribution of dysplasia in each treatment group. Statistical significance for IHC, QRT-PCR, and Western blotting was determined by using analysis of variation (ANOVA) and the Tukey posthoc tests (Prism 4, GraphPad Software, Inc.). ANOVA and Tukey posthoc tests were performed by comparing the V.C./Untr. control group to the V.C./EtOH, 4-NQO/Untr., and 4-NQO/EtOH experimental groups, where statistical significance was set at p<0.05. For these analyses, the fold changes in protein levels were assessed by normalizing the means of each group to the V.C./Untr. group. Note, only the average staining intensities were used to determine statistical significance for the IHC analyses.