MicroRNA dysregulation and esophageal cancer development depend on the extent of zinc dietary deficiency

Zinc deficiency (ZD) increases the risk of esophageal squamous cell carcinoma (ESCC), and marginal ZD is prevalent in humans. In rats, marked-ZD (3 mg Zn/kg diet) induces a proliferative esophagus with a 5-microRNA signature (miR-31, -223, -21, -146b, -146a) and promotes ESCC. Here we report that moderate and mild-ZD (6 and 12 mg Zn/kg diet) also induced esophageal hyperplasia, albeit less pronounced than induced by marked-ZD, with a 2-microRNA signature (miR-31, -146a). On exposure to an environmental carcinogen, ∼16% of moderate/mild-ZD rats developed ESCC, a cancer incidence significantly greater than for Zn-sufficient rats (0%) (P ≤ 0.05), but lower than marked-ZD rats (68%) (P < 0.001). Importantly, the high ESCC, marked-ZD esophagus had a 15-microRNA signature, resembling the human ESCC miRNAome, with miR-223, miR-21, and miR-31 as the top-up-regulated species. This signature discriminated it from the low ESCC, moderate/mild-ZD esophagus, with a 2-microRNA signature (miR-31, miR-223). Additionally, Fbxw7, Pdcd4, and Stk40 (tumor-suppressor targets of miR-223, -21, and -31) were downregulated in marked-ZD cohort. Bioinformatics analysis predicted functional relationships of the 3 tumor-suppressors with other cancer-related genes. Thus, microRNA dysregulation and ESCC progression depend on the extent of dietary Zn deficiency. Our findings suggest that even moderate ZD may promote esophageal cancer and dietary Zn has preventive properties against ESCC. Additionally, the deficiency-associated miR-223, miR-21, and miR-31 may be useful therapeutic targets in ESCC.


INtrODUctION
Esophageal cancer, including esophageal squamous cell carcinoma (ESCC) and adenocarcinoma, is the eighth most common cancer worldwide and the sixth most common cause of death from cancer, with a 5-year survival rate of only 10%. In 2012, there were an estimated 456,000 new cases and 400,000 deaths [1]. Representing 80% of cases of esophageal cancer worldwide, ESCC is the predominant histological subtype [2]. ESCC is typically diagnosed at an advanced stage because early symptoms are usually absent. Thus, clarification of pathogenesis mechanisms and new methods for prevention, diagnosis, and treatment are urgently needed.
Risk factors for ESCC include alcohol and tobacco use, nutritional deficiencies, and exposure to environmental carcinogens, such as N-nitrosomethylbenzylamine (NMBA) [3]. In particular, zinc (Zn) deficiency (ZD) (defined as inadequate dietary Zn intake) is implicated in the etiology of ESCC in many populations [4][5][6][7][8][9], including people with heavy alcohol consumption [10]. In 2005 Abnet et al. [11] showed that high esophageal tissue Zn concentration was strongly associated with a reduced risk of developing ESCC as compared with low tissue Zn www.impactjournals.com/oncotarget concentration; their data provide the strongest evidence in humans of an association between dietary ZD and ESCC.
Our ZD rat model [12][13][14][36][37][38] recapitulates features of human ESCC, including ZD, miRNA dysregulation, and inflammation [4,15,16,39]. Thus, our model provides an opportunity to better define the relationship between dietary Zn intake and miRNA dysregulation in ESCC development. This ZD model with 3 mg Zn/kg diet is relevant to human health. The recommended dietary allowance (RDA) for Zn in males is 11 mg (NIH ODS). Assuming an adult male human on a 3 mg Zn/kg diet consumes about 1.2 kg (2.64 lb) of food/day, his daily Zn intake would be 3.6 mg Zn or 33% of RDA for Zn. Thus, this person would be considered as markedly-ZD. In rat studies by others [40,41], a "severely" ZD diet has less than 1 mg Zn/kg, a "marginally" ZD diet has 5 mg Zn/kg, and a "marginally "Zn-adequate diet has 10 mg Zn/kg. To extrapolate this to human Zn nutrition, a person on a 1, 5, and 10 mg Zn/kg experimental diet would have a daily Zn intake of about 11%, 54%, 108% of human RDA.
ZD is recognized as a major worldwide public health problem [42][43][44][45][46], affecting 31% of the global population (4-73%, depending on subregions), with higher rates in developing countries [45]. Whereas severe or clinical ZD is uncommon, mild-to-moderate ZD is prevalent throughout the world [47]. Using a well-characterized ZD rat esophageal cancer model [13,14,37,38], the current study asks whether moderate-ZD (6 mg Zn/kg diet, ~66% of human RDA) and mild-ZD (12 mg Zn/kg diet, ~132% of human RDA) might cause alterations in miRNA expression, as does a marked-ZD diet (3 mg Zn/kg, ~33% of human RDA) that provides a microenvironment conducive to ESCC development on exposure to low carcinogen doses [14]. For this, we conducted a long-term tumor bioassay by low doses of NMBA in rats fed diets with different amounts of Zn -3, 6, 12, or 60 mg Zn/kg to represent marked-ZD, moderate-ZD, mild-ZD, and Znsufficiency (ZS), respectively. In parallel, we performed miRNA profiling (nanoString platform) in esophageal mucosa from NMBA-treated rats at tumor endpoint and from NMBA-untreated rats at identical time point in order to correlate miRNA expression changes with ZD doses and esophageal tumor outcome.
At tumor endpoint, serum Zn levels in ZD cohorts were significantly lower than in ZS rats (P < 0.001) ( Figure 1C). Moderate-ZD and ZS rats had comparable body weight, because ZS group was paired-fed to moderate-ZD rats to match their relative reduced food intake ( Figure 1C). Moderate/mild-ZD rats also had higher body weight than marked-ZD rats, because of reduced food consumption in the latter group.
The difference in ESCC incidence between ZD12T and ZST groups was statistically significant (ZD12T vs ZST, P = 0.047) and that between ZD6T vs ZST group was close to statistical significance (P = 0.051). These data established for the first time that mild and moderate-ZD enhances esophageal tumorigenesis and promotes progression to ESCC.

Cellular localization of miR-223, miR-31 and miR-21 expression in human ESCC tissue
The cellular origins of miRNAs are of importance to their mechanistic roles in cancer development. Previously, we demonstrated an abundant miR-31 ISH signal in human ESCC tissue [60]. Whether miR-223 and miR-21 co-localize in the same ESCC tissue is not known. Thus, we evaluated the cellular localization of all three miRNAs in archived FFPE human ESCC tissues using in situ hybridization (ISH) (n = 12 cases). All 12 cases showed intense to moderate miR-31, miR-223, and miR-21 ISH signal in near serial sections of moderately to poorly differentiated ESCC tumor samples ( Figure 5). By contrast, the normal mucosa adjacent to the tumor cells had very weak staining (data not shown). These results represent the first simultaneous in situ detection of miR-223, -21, and -31 in human ESCC.

Esophagus-specific functional relationship prediction among tumor-suppressor targets
Functional Networks of Tissues in Mouse (FMTN) [65,66] is a prediction tool for tissue-specific protein interactions for the mouse that is based on the integration of a variety of genomic data and prior knowledge of gene function. To explore the esophagus-specific functional relationships for Fbxw7, Stk40, and Pdcd4 (tumorsuppressor targets of miR-223, -31, -21), we employed FNTM for the rat. The percentage of orthologous genes shared by mouse and rat is very high, and a similar tool is not available for the rat. We obtained a nine-gene network of functional relationship predictions in the esophagus ( Figure 7A) showing that Fbxw7, Stk40, and Pdcd4 were functionally related to several cancer-related genes, such as Pten, tumor suppressor of miR-21 [67,68], oncogene Bcl2 [69], a Wnt signaling pathway transcription factor Tcf4 [70], the dead box protein family of RNA helicases Ddx6 [71,72], fibroblast growth factor receptor 1 Fgfr1 [73][74][75], and Ppp3ca, a component of calcium/calcineurin signaling that includes apoptosis [76]. Enrichment Gene Ontology analysis showed that Pten, Bcl2, Ppp3Ca, and Fgfr1 were the genes most functionally related to Fbxw7, Stk40, and Pdcd4 and were statistically significantly enriched in biological processes related to cell cycle, growth, response to stress, and apoptosis regulation ( Figure 7B). We then showed that the expression of the most functionally related gene, Pten, was also down-regulated in the ESCC-bearing ZD3T esophagus that overexpressed miR-21 at the mRNA level by qPCR (P = 0.02, n = 6 rats/group) and at the protein level by immunohistochemistry compared to ZS counterpart ( Figure 7C). That the three tumor suppressor targets are predicted to interact to alter network of cancerrelated proteins [65,66,77] provide support that miR-223, miR-21, and miR-31 have an important role in ESCC and may be useful therapeutic targets in ESCC.

DIscUssION
In humans, low dietary Zn intake is associated with an increased risk of ESCC [7,9]. Although marginal ZD is prevalent in humans [47], the effect of moderate ZD on the etiology of ESCC has not been studied. Using a well-characterized ZD rat esophageal cancer model [13,14,37,38], the current study demonstrates for the first time that ESCC initiation and progression, as well as miRNA dysregulation, depend on the extent of deficiency of dietary Zn.
Sustained increased cellular proliferation is a hallmark of cancer [78]. Previously, we reported that marked-ZD (3 mg Zn/kg) induces prominent esophageal cellular proliferation, predisposing to tumor development [13,79]. We now show that a moderately-ZD (6 mg Zn/ kg) and a mildly-ZD (12 mg Zn/kg) diet also induced a sustained but less-pronounced, esophageal cellular proliferation. Importantly, in the presence of moderate or mild-ZD, low NMBA doses elicited statistically significantly higher tumor incidence/multiplicity, as well as ESCC progression than with Zn sufficiency ( Figure  1C). Although marked-ZD led to significantly higher tumor/ESCC incidence than moderate or mild-ZD, no statistically significant difference was obtained in tumor or ESCC outcome between moderate-ZD and mild-ZD, despite a two-fold difference in Zn content. These data show a dose-response relationship between the extent of ZD and ESCC development. Additionally, they provide the first evidence that moderate to mild-ZD, combined with low doses of the environmental carcinogen NMBA, produces ESCC.
MiRNA-expression profiling of human tumors has identified signatures associated with staging, progression, prognosis, and response to treatment [21]. Also, miRNA expression patterns have been shown to be potential classifiers for ESCC [29]. Using the nanoString platform, miRNA expression profiles distinguished the highly preneoplastic/proliferative marked-ZD esophageal phenotype with a 5-miRNA signature (miR-31, -223, -21, -146b, -146a), from the less proliferative, mild-ZD phenotype with a 3-miRNA signature (miR-146a, -31, -223). Importantly, the high ESCC-burden, marked-ZD esophagus showed a 15-miRNA signature (with miR-223, -21, and -31 as the top-up-regulated species), thus differentiating it from the low ESCC-burden, mild-ZD esophagus with a 2-miRNA signature (miR-223, -31). In addition, our data show that these miRNA signatures not only differentiate esophageal preneoplasia from normal esophagus and stages of ESCC progression, but also highlight the molecular impact of dietary ZD on miRNA dysregulation in the pathogenesis of ESCC.
A limitation of this study is the fact that the underlying biological mechanisms of the key dysregulated miRNAs in ESCC development, namely, miR-223, miR-21, and miR-31, were not investigated. Studies are in progress to specifically address this issue.
In summary, this dose-response study demonstrates that ESCC development and the underlying miRNA dysregulation are dependent on the extent of deficiency of the nutrient Zn. Although it remains to be determined if the results in this ZD dose-response study in the rat will translate to human ESCC, our findings suggest that dietary Zn may have preventive properties against ESCC and provide a mechanistic rationale for exploring the therapeutic use of Zn against ESCC. In addition, our study has identified Zn deficiency-associated miRNA signatures that may underlie the molecular pathogenesis of ESCC in Zn-deficient populations. Our study suggests that miR-223, miR-31 and miR-21 alone or in combination could be used as therapeutic targets for treatment of ESCC.

Experimental design
Animal protocols were approved by the Thomas Jefferson University Institutional Animal Care and Use Committee. Weanling male rats were randomly divided into 4 dietary groups (ZD3, ZD6, ZD12, and ZS, n = 47-49 rats/group) and were tail-tattooed for identification. ZD rats were fed ad libitum and ZS rats were pair-fed to ZD6 animals to match the decreased food consumption of ZD6 rats [38]. After 5 weeks, 12 rats per group were killed for evaluation of esophageal cell proliferation [38]. The remaining animals were divided into NMBAtreated groups (n = 25-27 rats/dietary group) and NMBAuntreated groups (10 rats/group). Carcinogen-treated rats were administered intragastrically 4 NMBA doses (2 mg/ kg body weight), once a week for 4 consecutive weeks. NMBA-untreated groups received saline. The animals were weighed weekly and monitored daily. The study was concluded at 17 weeks after the 1st NMBA dose (22 weeks of ZD). At sacrifice, the animals were anesthetized by delivering isoflurane (GE Healthcare) to the respiratory tract of the rat using a vaporizer at 3% concentration. Blood was obtained from the retro-orbital venous plexus for serum preparation and subsequent Zn analysis. Whole esophagus was excised and longitudinally slit open. Tumors greater than 0.5 mm in diameter were mapped.

Esophageal epithelia preparation
Esophagi were isolated and cut into two equal portions. Esophageal epithelium was prepared from a portion by using a blade to remove the submucosal and muscularis layers, snap-frozen in liquid nitrogen and stored at -80°C [13]. The remaining portion was fixed in 10% buffered formalin and paraffin embedded.

rNA isolation
Esophageal epithelial samples frozen in liquid nitrogen were pulverized to a fine powder using a chilled hammer. Total RNA was extracted from the pulverized samples using an animal tissue RNA extraction Kit (#25700, Norgen Biotek, Ontario, Canada). RNA concentration of each sample was determined using a NanoDrop 1000 (Thermo Scientific). All RNA samples displayed a 260:280 ratio >1.8, and a 260:230 ratio > 1.8.

nanoString rat miRNA expression assay
The nanoString rat miRNA expression assay kit that profiles 423 rat miRNAs was employed (n = 6 rats/group). This assay was performed at the Ohio State University Comprehensive Cancer Center Genomics Shared Resource according to manufacturer's instruction. Briefly, 100 ng of total RNA was used as input material. Small RNA samples were prepared by ligating a specific DNA tag onto the 3' end of each mature miRNA. These tags normalized the melting temperatures (Tms) of the miRNAs and provided identification for each miRNA species in the sample. Excess tags were then removed, and the resulting material was hybridized with a panel of miRNA:tagspecific nCounter capture and barcoded reporter probes. Hybridization reactions were incubated at 64 0 C for 18 h. Hybridized probes were purified and immobilized on a streptavidin-coated cartridge using the nCounter Prep Station. nCounter Digital Analyzer was used to count individual fluorescent barcodes and quantify target RNA molecules present in each sample. For each assay, a highdensity scan (600 fields of view) was performed.

nanostring data analysis
Abundances of miRNAs were quantified using the nanoString nCounter gene expression system [56]. Each sample was normalized using the global sum method that uses the entire miRNA content. The nanoString nSolver software tool was used to facilitate normalization. Student's t-test was used to calculate statistical significances of pair-wise comparisons. Calculations were performed using the R statistical computing environment (http://www.r-project.org/).

TaqMan miRNA assay
Reverse transcription of miRNAs was performed according to the manufacturer's instructions (Applied Biosystems, Foster City, CA) with a reaction volume of 15 μl containing 350 ng of total RNA. The real-time PCR was performed using the 7300 Real-Time PCR Systems (Applied Biosystems). Each miRNA and endogenous control (snoRNA and U87) was measured in triplicates.
As an overall quality control, CT values above 35 were excluded from analysis.

Quantitative real time PCR
cDNA was reverse transcribed using the High-Capacity cDNA Archive Kit (Applied Biosystems, Foster City, CA) using reverse transcription reaction volumes of 20 μl containing 1 μg of total RNA for each sample according to the manufacturer's protocol. qPCR was performed using pre-designed probes (Applied Biosystems), Psmb6 and Oaz1 as the normalizers, and the comparative Ct method .

Microscopy
IHC and ISH analyses were performed by light microscopy using an Olympus BX51 microscope and photographs taken with a Spot RT3 camera and Spot software v. 4.6.

Zn measurement
Serum Zn content was determined using Atomic Absorption Spectrometer Analyst 400 (PerkinElmer, Waltham, MA).

Statistical analysis
Dietary Zn effects on continuous data (tumor multiplicity, serum Zn, weight) were analyzed by analysis of variance (ANOVA). Differences among the groups were assessed using the Tukey-HSD multiple comparisons post hoc t-tests. When the data exhibited heteroscedasticity (tested by Levene's homogeneity of variance test), the Welch ANOVA test was used to detect an overall difference in the dietary groups and the Games-Howell pairwise comparison test was used for detecting differences among the groups. For data where only 2 groups were analyzed or for the inflammation genes where we were only interested in detecting differences between the zinc sufficient group and each zinc deficient group, the students t-test was used to compare the groups. Dietary group effects in tumor/ESCC incidence were assessed by an overall chi-square test. Pairwise Fisher's exact test was used to compare the individual dietary groups. Statistical tests were 2-sided and were considered significant at P < 0.05. Statistical analysis was performed using R (http:// www.R-project.org).

AcKNOWLEDGMENts
We thank K. Huebner for critical reading of the manuscript and helpful suggestions.

FUNDING
This work was supported by grants from the National Institutes of Health (CA118560, and R21CA152505 to LYF; U01CA152758 to CMC), American Institute for Cancer Research (Grant #207232 to LYF), as well as a grant from Dr. Lit H. Leung (to LYF).

cONFLIcts OF INtErEst
The authors have no conflicts of interest to declare.