Dysregulation of miR-31 and miR-21 induced by zinc deficiency promotes esophageal cancer

Zinc deficiency (ZD) increases the risk of esophageal squamous cell carcinoma (ESCC). In a rat model, chronic ZD induces an inflammatory gene signature that fuels ESCC development. microRNAs regulate gene expression and are aberrantly expressed in cancers. Here we investigated whether chronic ZD (23 weeks) also induces a protumorigenic microRNA signature. Using the nanoString technology, we evaluated microRNA profiles in ZD esophagus and six additional tissues (skin, lung, pancreas, liver, prostate and peripheral blood mononuclear cells [PBMC]). ZD caused overexpression of inflammation genes and altered microRNA expression across all tissues analyzed, predictive of disease development. Importantly, the inflammatory ZD esophagus had a distinct microRNA signature resembling human ESCC or tongue SCC miRNAomes with miR-31 and miR-21 as the top-up-regulated species. Circulating miR-31 was also the top-up-regulated species in PBMCs. In ZD esophagus and tongue, oncogenic miR-31 and miR-21 overexpression was accompanied by down-regulation of their respective tumor-suppressor targets PPP2R2A and PDCD4. Importantly, esophageal miR-31 and miR-21 levels were directly associated with the appearance of ESCC in ZD rats, as compared with their cancer-free Zn-sufficient or Zn-replenished counterparts. In situ hybridization analysis in rat and human tongue SCCs localized miR-31 to tumor cells and miR-21 to stromal cells. In regressing tongue SCCs from Zn-supplemented rats, miR-31 and miR-21 expression was concomitantly reduced, establishing their responsiveness to Zn therapy. A search for puta- tive microRNA targets revealed a bias toward genes in inflammatory pathways. Our finding that ZD causes miR-31 and miR-21 dysregulation associated with inflammation provides insight into mechanisms whereby ZD promotes ESCC.


Introduction
Oral-esophageal squamous cell carcinomas (OSCC, ESCC) are major causes of cancer deaths worldwide. Because of absence of early symptoms, ESCC is frequently diagnosed at an advanced stage and has a 5-year survival of 10%. Patients with OSCC (major site, tongue) have a high mortality rate, because of the appearance of second cancers in the esophagus through field cancerization effects (1). Despite the decline in worldwide cancer-mortality rates since the mid 1980s, ESCC and OSCC remain deadly diseases. Thus, clarification of their pathogenesis and development of new prevention strategies are urgently needed.
The major risk factors for developing oral-esophageal carcinomas include alcohol consumption, tobacco use and nutritional deficiencies. Epidemiological and clinical studies have implicated dietary zinc (Zn) deficiency (ZD) in the pathogenesis of ESCC and OSCC (2,3). Zn is an essential trace element required for the activity of >300 enzymes, for proper immune function and for the conformation of >2000 transcription factors that control cell proliferation, apoptosis and signaling pathways (4). ZD predisposes to diseases by adversely affecting these processes.
Our ZD rat oral-esophageal cancer model reproduces the ZD link to human ESCC and OSCC (5) and provides a unique opportunity to decipher the mechanism by which ZD promotes oral-esophageal carcinogenesis. Previously, we showed that weanling rats on a ZD diet for 6 weeks developed a precancerous condition in the upper aerodigestive tract (tongue, esophagus and forestomach [expanded lower esophagus]), with increased cellular proliferation (5) and extensive gene-expression changes, including overexpression of proinflammation genes S100 calcium binding protein a8 and a9 (S100a8 and S100a9) (6). Subsequently, prolonged ZD (21 weeks) was shown to amplify this inflammation in the esophagus by causing overexpression of numerous inflammation genes in addition to S100a8/S100a9, thereby providing an inflammatory microenvironment conducive to ESCC development on repeated exposure to the environmental carcinogen N-nitrosomethylbenzylamine (NMBA). Zn replenishment (ZR) reversed the inflammatory signature and prevented cancer formation (7). In addition, Zn supplementation in nutritionally complete rats suppressed tongue SCC development by attenuating the inflammation (8). microRNAs (miRNAs) are a family of short noncoding RNAs that have emerged as powerful posttranscriptional regulators of gene expression (9). The ability of individual miRNAs to regulate a large number of mRNA species allows them to coordinate complex programs of gene expression that alter a range of biological processes, including cellular proliferation, apoptosis, immune response and signaling events (10). miRNAs expression levels are altered in many types of human cancer, including ESCC and OSCC (11). Furthermore, miRNA expression patterns are associated with chronic inflammatory diseases and other inflammatory conditions (12).
We asked whether overexpression of cancer-related inflammation genes induced by ZD in the rat esophagus (7) is a consequence of a protumorigenic miRNA signature. For this, we examined miRNA profiles in esophagus and six additional tissues (skin, lung, pancreas, liver, prostate and peripheral blood mononuclear cells [PBMC]) after chronic ZD (23 weeks) using the novel nanoString nCounter technology (Seattle, WA) (13,14). The deficient esophagus exhibited a prooncogenic miRNA signature with miR-31 and miR-21 as the top-up-regulated species. The role of these two oncomiRs in esophageal and oral neoplasia was further explored. ZD also induced distinct miRNA expression patterns across all tissues profiled.

Animals
Male weanling male Sprague-Dawley rats were obtained from Taconic Laboratory (Germantown, NY). Custom-formulated egg-white-based diets were from Harlan Teklad (Madison, WI). ZD-and Zn-sufficient (ZS) diets were identical except for the Zn content, which was 3-4 ppm and 65 ppm, respectively.

Experimental design
The animal study was conducted according to approved Institutional animal protocols. Weanling rats (n = 19) were fed ad libitum a ZD diet for 23 weeks. Control rats (n = 15) were pair fed a ZS diet to match the decreased food consumption of ZD animals (7). At study conclusion, blood was collected from the retro-orbital venous plexus after anesthesia with isoflurane. PBMCs were prepared from whole blood using Histopaque-1077 (Sigma-Aldrich, St Louis, MO). Esophagus, tongue, skin, lung, liver, prostate and pancreas were isolated. Esophagus and tongue were cut into two portions. One portion was formalin fixed and paraffin embedded (FFPE) for histological, immunohistochemical (IHC), and in situ hybridization (ISH) studies. Esophageal and tongue mucosa were prepared as previously described (15). The prostate was microdissected into dorsal, lateral, ventral and anterior lobes. Tissues were snap frozen in liquid nitrogen and stored at -80°C.

miRNA expression profiling
The multiplexed nanoString nCounter miRNA system (nanoString Technologies) (13,14) was used for miRNA expression profiling in esophagus and six additional tissues (skin, lung, pancreas, liver, prostate and PBMC; n = 6 rats/tissue/dietary group). This assay was performed at the Ohio State University Comprehensive Cancer Center Nucleic Acid Facility.
Total RNA (100 ng) was used as input material. Small RNA samples were prepared by ligating a specific DNA tag onto the 3' end of each mature miRNA according to manufacturer's instruction (nanoString Technologies). 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:tag-specific nCounter capture and barcoded reporter probes. Hybridization reactions were incubated at 64°C for 18 h. Hybridized probes were purified and immobilized on a streptavidin-coated cartridge using the nCounter Prep Station (nanoString Technologies). nCounter Digital Analyzer was used to count individual fluorescent barcodes and quantify target RNA molecules present in each sample. For each assay, a high-density scan (600 fields of view) was performed.
nanoString data analysis Abundances of miRNAs were quantified using the nanoString nCounter gene-expression system (13). Each sample was normalized to the geometric mean of the top 50 most highly expressed miRNAs. 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/).

Gene ontology (GO) analysis
GO analysis was performed using miRNAs with P < 0.05 and fold change of >1.3. For each miRNA, the common target genes found in two or more databases (RNAhybrid, TargetScan, miRDB and microRNA.org) were used. GO analysis was then performed for each tissue using DAVID (Database for Annotation, Visualization and Integrated Discovery) (16).
TaqMan miRNA assay TaqMan miRNA assays was performed to quantify mature miRNAs using predesigned probes, rat U87 as the normalizer (Applied Biosystems, Foster City, CA), and the comparative C t method. Real-time PCR was performed on the ABI Prism 7900HT Sequence detection system (Applied Biosystems).
Quantitative real time PCR qRT-PCR was performed using predesigned probes (Applied Biosystems), PSMB6 as the normalizer and the comparative C t method (7).

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

Rat and human tongue SCC samples
Archived rat tongue SCC tissues (FFPE) were derived from a previous study (8). Human tongue SCC tissues (FFPE) were from Thomas Jefferson University Hospital.

Rat ESCC samples
Archived rat ESCC tissues (FFPE) and RNAs were derived from a previous study (7).

Statistical analysis
Differences in Zn, gene expression and protein levels were assessed with Student's t-test. Statistical tests were two-sided and were considered significant at P < 0.05.
Chronic ZD induced up-regulation of numerous inflammatory genes in the esophagus (7). Whether it also causes inflammation in the other tissue types was evaluated by analyzing the expression of eight cancer-associated inflammation genes (S100a8, S100a9, Cxcl5, Il1b, Cxcl2, Cxcr4, Ptgs2 and Tlr4) (7), using qRT-PCR. The ZD tongue is highly susceptible to chemically induced carcinogenesis (5) and was the only tissue that showed overexpression of all eight inflammatory genes ( Figure 1B). S100a8 and S100a9 were overexpressed in all tissues studied ( Figure 1B). In PBMC, the difference between ZD and ZS samples was suggestive but not statistically significant (S100a8, P = 0.07; S100a9, P = 0.08). Immunoblot and enzyme-linked immunosorbent assay (ELISA) analyses showed that the ZD effect observed at the transcript levels for S100a8, S100a9, Cxcl5 and Il1b were also reflected at the protein level in tongue, lung and pancreas ( Figure 1C and 1D). These data establish that prolonged ZD reduces Zn levels across tissue types and up-regulates expression of key inflammation genes.

Normal ZS tissues show high abundance of tissue-specific miRNAs
The nanoString platform directly measures miRNA expression levels without reverse transcription or PCR amplification, thereby eliminating enzymatic bias (13,14). At the time we initiated this study, the nanoString rat miRNA kit was unavailable. Initially, we used the human miRNA assay to profile PBMC and then the mouse miRNA assay to profile esophagus, skin, lung, pancreas, liver and Zn deficit induces oncogenic miRNAs in esophagus prostate. Because of a high degree of sequence conservation of miRNAs across human, mouse and rat, 178 and 236 miRNAs on the human and mouse panels, respectively, have sequences identical to the rat miRNAs. Thus, we profiled 178 rat miRNAs in PBMC and 236 rat miRNAs in the other six tissues (n = 6 rats/tissue/ dietary group).
The nanoString profiling platform provided an estimate of miRNA abundance in normal ZS and ZD organs. ZS organs showed high abundance of many tissue-specific miRNAs (see Supplementary Table I online). As shown in Figure 2B, the liverspecific miR-122 (19) was abundantly expressed in the liver; skinspecific miR-203 (20) in the skin; epithelial tissue-specific miR-205 in skin and esophagus; pancreatic islet-specific miR-375 (21) and pancreas-specific miR-216 and miR-217 (22) in the pancreas; and the hematopoietic markers miR-142-3p (23), miR-150, miR-16, miR191, miR-26a, miR-19b (24) in PBMC. Conversely, the ubiquitous oncogenic miR-21 was expressed with moderate abundance across all seven tissues, and oncogenic miR-31 in low abundance in these same tissues. These miRNA abundance data from normal rat tissues establish that the nanoString technology is both sensitive and specific.

ZD tissues have distinct miRNA profiles
In addition to esophagus, hierarchical clustering of miRNAs revealed that the other six ZD tissues had distinct miRNA profiles (Figure 2A), indicating that ZD alters miRNA expression in a tissue-specific manner. Using P-values <0.05 and cutoff of ≥1.3 fold difference; groups of differentially expressed miRNAs were identified in these ZD tissues ( Figure 2C, see Supplementary Table II online). Skin. Several of the up-regulated miRNAs in the inflammatory ZD skin, namely, miR-31 (3.3 fold), miR-21 (1.5 fold) and miR-142-3p (1.3 fold), are hallmarks of the miRNAome of human psoriasis, a chronic inflammatory skin disease (34). Interestingly, ZD is implicated in the pathogenesis of psoriasis (35). As in the ZD skin ( Figure 1B,  1C), psoriasis overexpresses the inflammatory mediators S100A8 and S100A9 (36).

PBMC. PMBC are leukocytes that have a round nucleus such as lymphocytes (T cells, B cells), natural killer cells and monocytes.
Circulating miR-31 was the top-up-regulated species (1.8 fold) in PBMC from ZD versus ZS rats. Because miR-31 was also overexpressed in ZD esophagus, tongue and skin, this result suggests an association between circulating and tissue miR-31. miR-150 was the most down-regulated species in PBMC of ZD rats (-5.1 fold). In patients with sepsis, miR-150 levels in PBMC and plasma were significantly reduced (43). In this regard, patients with sepsis are frequently ZD (44).
To summarize, dietary ZD induces dysregulation of key oncogenic and tumor-suppressor miRNAs in the esophagus and across all tissues profiled. The most shared miRNAs in signatures of the ZD tissues were oncogenic miR-21 and miR-31 (see Supplementary Table III online), with miR-21 up-regulated in six (esophagus, pancreas, liver, skin, lung, tongue) and miR-31 in four tissues (esophagus, skin, PBMC, tongue).

Validation of nanoString profiling data
To validate the nanoString results, we performed Taqman miRNA assays on 14 selected miRNAs (n = 6 rats/tissue/dietary group). In all 14 cases, the same pattern and fold changes response to dietary ZD was observed using Taqman as with the nanoString assay ( Figure 3A).

Cellular localization of miR-31 and miR-21 in ZD esophageal and tongue epithelia
We focused our study on the top-up-regulated miR-31 and miR-21 because of their roles in human ESCC and OSCC (27,(29)(30)(31)(32). Their cellular localizations in esophageal and tongue preneoplastic tissues after chronic dietary ZD (23 weeks) was defined by performing ISH on FFPE tissues using high affinity double Dig-labeled LNA probes (Exiqon, Vedbaek, Denmark; n = 10 rats/dietary group). The hyperplastic ZD esophagus and tongue epithelial cells showed abundant miR-31 ISH signal, as compared with nonproliferative ZS tissues ( Figure 3B). In contrast, miR-21 ISH signals were infrequent and found mostly in stromal calls of ZD tissues and were generally scarce in ZS tissues ( Figure 3B). These data establish that miR-31 and miR-21 expression is cell-type specific with miR-31 localized to epithelial and miR-21 mostly in stromal cells.

miR-31 and miR-21 overexpression correlates with down-regulation of their tumor-suppressor targets
In human tongue SCC, a correlation of miR-21 overexpression with loss of its tumor-suppressor targets PDCD4 (30) and TPM1 (45) mRNA and protein expression is documented. We determined if up-regulation of oncogenic miR-21 in ZD tissues is accompanied by down-regulation of these two known tumor-suppressor targets. Using IHC, PDCD4 was strongly expressed in the nuclei of epithelial cells of ZS esophagus and tongue. This nuclear staining of PDCD4 was reduced or absent in the hyperplastic ZD esophagus and tongue ( Figure 3C). In addition, immunoblotting analysis showed that both PDCD4 and TPM1 protein expression was markedly reduced in ZD esophagus and tongue versus ZS tissues ( Figure 3D).
Using immunoblotting analysis ( Figure 3D), we showed that overexpression of miR-31 in ZD esophagus and tongue as compared with ZS tissues was associated with down-regulation of its known tumor-suppressor target PPP2R2A (46) and putative targets RHOBTB1 and STAU2 (identified by TargetScan algorithm). The putative target RHOBTB1 is a candidate tumor-suppressor gene in head and neck SCC (47) and STAU2 mRNA down-regulation was previously noted in ZD versus ZS esophagus by microarray analysis (6). Whether RHOBTB1 and STAU2 are bona fide targets of miR-31 remain to be investigated.

Overexpression of miR-31 and miR-21 is associated with ESCC in ZD rats
ZD induced an inflammatory gene-expression profile that fueled ESCC in rats with repeated exposure to NMBA (7). Using qRT-PCR on archived RNA samples (7), miR-31 and miR-21 exhibited 5.4 and 3.2-fold up-regulation in ESCC-bearing ZD rats versus cancer-free ZS controls ( Figure 4A). ZR, a regimen that prevented ESCC development (7) reduced their expression to ZS levels ( Figure 4A). ISH on archived FFPE tissues (7) (n = 15/dietary group) from ZD rats showed abundant miR-31 signal in tumors/dysplastic epithelium and miR-21 signal in tumor stroma ( Figure 4B). Conversely, ISH signals of both oncomiRs were weak/absent in ZS or ZR samples (see Supplementary  Figure 1 online). Thus, dietary Zn levels regulate miR-31 and miR-21 expression in esophageal cancer development and prevention; and their dysregulation by ZD is associated with ESCC.

Zn therapy reduces miR-31 and miR-21 expression in rat tongue SCC
Zn supplementation in rats decreased 4-nitroquinoline1-oxide (NQO)induced tongue SCC incidence and elicited a reduced proliferative/ inflammatory cancer phenotype (8). We asked if Zn supplementation that suppresses tongue cancer development also attenuates miR-31 and miR-21 expression. ISH analysis was performed on FFPE tissues from NQO-induced rat tongue SCCs with or without Zn therapy (8) (n = 20 rats/Zn treatment group). In near serial sections of these tumor samples ( Figure 5, SCC rat tongue), miR-31 expression was localized to tumor cells, whereas miR-21 was predominantly found in stromal cells. Importantly, ISH signals of both oncomiRs were reduced in intensity and scope following Zn therapy ( Figure 5, ZnSupp versus ZS), thereby showing that their expression is responsive to Zn therapy.

Expression of miR-31 and miR-21 in human tongue SCC
The cellular localization of miR-31 and miR-21 in human tongue SCC was determined using ISH (n = 5 cases). As in rat tongue SCC, all five cases of human tongue SCC showed prevalent expression of miR-31 in tumor cells and miR-21 expression in stromal cells ( Figure 5, SCC human tongue). The ISH data in human and rat tongue SCC ( Figure 5) and rat ESCC ( Figure 4B) establish that miR-31 and miR-21 expression in SCC is cell-type specific, with miR-31 localized to tumor cells and miR-21 to tumor stroma.

GO analysis
To gain an understanding of the molecular pathways that are governed by the predicted targets of dysregulated miRNAs in ZD tissues, we performed biological processes enrichment using GO (48). Using target genes found in common in more than two databases (RNAhybrid, TargetScan, miRDB and microRNA.org), we performed GO analysis using DAVID (16) for each of the seven tissues (cutoff  (7). miR-31 signal (arrow heads: blue, NBT/ BCIP; counterstain, nuclear fast red) was strong/abundant in tumor cells of ESCC from a ZS rat; whereas miR-21 signal (blue, arrows) was mostly in tumor stroma.
Zn deficit induces oncogenic miRNAs in esophagus of ≥1.3 fold-change and P-value < 0.05). A survey of overrepresented GO terms documented a bias toward genes involved in Zn binding and gene regulatory processes (Table I). In particular, ZD esophagus showed overrepresented processes important in inflammation and cellular proliferation, including serine/threonine-protein kinase activity, MAPKK and JNK cascade and epithelial cell proliferation.
In addition, we performed functional classification of miRNA targets of up-and down-regulated miRNAs in ZD esophagus (see Supplementary Table IV online). The biological processes targeted by the 23 up-regulated miRNAs include processes in inflammation and cell proliferation, namely, serine/threonine-protein kinase activation (40 genes, including tumor suppressors, LATS2 [target of miR -31] and MAP2K4 [target of miR-31], P = 2.27E-08); regulation of cell motion (27 genes, P = 3.53E-07); and inflammatory response (seven genes, P = 0.028); negative regulation of gene expression (41 genes, including tumor suppressor PDCD4; P = 1.10E-04). Importantly, our present study documented a correlation between overexpression of miR-21 ISH signal ( Figure 3B) in the inflammatory ZD esophagus/ tongue and loss of expression of its tumor-suppressor target PDCD4 in the same tissues ( Figure 3C). In addition, functional classification of targets of the seven down-regulated miRNAs showed overrepresentation of biological processes such as transcription regulation, negative regulation of gene expression and Zn ion binding.
To determine if these predicted targets of dysregulated miRNAs are relevant to human ESCC development, we retrieved differentially expressed genes (tumor versus normal tissue, P-value < 0.05) from human ESCC/OSCC gene-expression data using ArrayExpress and Gene Expression Omnibus databases. All the selected genes (112 down-regulated and 38 up-regulated, see Supplementary Table V online) belong to the functional families of the predicted targets of dysregulated miRNAs (see Supplementary Table IV online). Interestingly, 47.5% (112 of 236) of our predicted targets for the overexpressed miRNAs are down-regulated in the public ESCC/OSCC gene expression database as compared to 29% (38 of 133) of our predicted targets for the down-regulated miRNAs. Together, our bioinformatics analysis provides a repertoire of putative target genes (see Supplementary Table V online) related to the altered miRNAome of the inflammatory ZD esophagus that may prove to be highly relevant to ESCC development.
In particular, miR-31 and miR-21, the top-up-regulated species in the ZD esophagus, are reported to be elevated in many cancers and inflammation-associated diseases. miR-21 is one of the most consistently up-regulated oncomiRs in human cancers, and is thought to contribute to inflammation-associated diseases, including cancers (12). miR-31 is a pleiotropically acting miRNA that is up-regulated in many types of human cancers, including ESCC, tongue SCC, liver cancer and colorectal cancer, but down-regulated in breast metastasis (49). Interestingly, miR-31 expression levels are reported to increase during inflammatory bowel disease-associated neoplastic transformation (50). Our data show that miR-21 overexpression in inflammatory ZD esophagus and tongue is correlated with down-regulation of its tumor-suppressor targets PDCD4 and TPM1 (Figure 3, B-D). In human ESCC, miR-21 negatively regulates PDCD4 and knockdown of miR-21 inhibits cellular proliferation and invasion (51). In ESCC tissues, loss of PDCD4 protein expression is correlated with cancer aggressiveness (tumor stage, nodal metastasis) (52). In tongue SCC tissues, miR-21 may act as an antiapoptotic factor by blocking TPM1 expression (45). Although miR-31 overexpression in ZD esophagus and tongue ( Figure 3B) is accompanied by down-regulation of its known tumor-suppressor target PPP2R2A (46) (Figure 3D), the role of PPP2R2A down-regulation has not been reported in human ESCC or OSCC. We speculate that the down-regulation of such tumor-suppressor targets may relate to the increased cell proliferation and inflammation associated with ZD, as suggested by our GO enrichment analysis of targets of up-regulated miRNAs (see Supplementary  Table IV online). The biological role of miR-21 and miR-31 in ZD-induced inflammation-associated esophageal carcinogenesis remains to be determined.
Significantly, our present Taqman and ISH data show that miR-31 and miR-21 levels were substantially higher in ZD esophagus bearing ESCC versus cancer-free ZS controls ( The tumor-suppressive miR-375 is known to circulate at low levels in ESCC patients (25), and Ryu et al. (53) showed its acute reversal in serum of Zn-depleted human subjects following a Zn-repletion regimen. The present ISH data that miR-31 and miR-21 expression was concomitantly reduced in regressing tongue SCCs from Zn-supplemented rats (8) indicate that these oncomiRs are responsive to Zn therapy ( Figure 5). Thus, Zn supplementation may be therapeutic in combating ESCC or OSCC by correcting aberrant miRNA expression. Although Zn therapy inhibits tongue cancer development by suppressing inflammation and cellular proliferation (8), the . ISH localization of miR-31 and miR-21 in archived rat and human tongue SCCs. Rat tongue SCC: miR-31 signal (arrow heads: blue, NBT/ BCIP; counterstain, nuclear fast red) was strong/abundant in tumor cells of proliferative tongue SCC from ZS rats but moderate/infrequent in tumor cells of regressing tongue SCC from Zn-supplemented (ZnSuppl) rats. miR-21 signal (blue, arrows) was strong/abundant in stromal cells of ZS tongue SCC but reduced/infrequent in stromal cells of ZnSuppl tongue SCC. Two cases of human tongue SCC (H&E) are shown: miR-31 signal was predominantly in tumor cells (blue, arrow heads), and miR-21 signal in stroma (blue, arrows). No ISH signal was seen with mm-miR-31 or mm-miR-21 (2 mismatches). miRNA was localized by 4-nitro-blue tetrazolium (NBT) and 5-brom-4chloro-3′-indolylphosphate (BCIP). Scale bars: 100 µm (H&E, rat tongue SCC; mm-miR-31 and mm-miR-21); 50 µm (miR-31 and miR-21; H&E, human tongue SCC).

H. Alder et al.
mechanism whereby Zn decreases miR-31 and miR-21 expression in tongue cancer inhibition needs to be further investigated.
The cellular origins of individual miRNAs are of critical importance in the understanding of the mechanistic roles of miRNAs in tumor development. Whereas miR-21 presence in stromal cells of human colorectal (17) and breast cancer (54) was previously reported, the localization of miR-31 to esophageal/tongue preneoplastic epithelia ( Figure 3B) and tumor cells ( Figure 4B and Figure 5) is reported here for the first time. Thus, miR-31 and miR-21 expression is cell-type specific and may serve as markers in tumor diagnosis and prognosis. In addition, the profiling result that miR-31 was up-regulated in ZD esophagus and tongue, as well as in PBMC from ZD animals, indicates that miR-31 may serve as a noninvasive diagnostic marker for a premalignant esophageal/lingual condition. In fact, patients with ESCC or OSCC display high levels of miR-31 (31,32) as well as miR-21 (25,30) in tumor tissues and in serum/plasma. Conversely, miR-21 that was overexpressed in most ZD tissues was not up-regulated in PBMC from ZD animals. This result requires further exploration to determine whether the miR-21 overexpression in PBMC could arise at later stages of ESCC or OSCC development due to elevation of miR-21 expression in the tumor stroma ( Figure 4B and Figure 5), as compared to stroma of hyperplastic tissues ( Figure 3B).
The miRNAomes of the other six inflammatory ZD tissues ( Figure 1, B-D) showed dysregulation of specific oncogenic and tumor-suppressor miRNAs ( Figure 2C). These changes resemble the hallmarks of the miRNAomes of various human conditions, such as pancreatitis and pancreatic ductal adenocarcinoma (22,38), liver cancer (39), inflammatory COPD and lung cancer (37,38), prostate cancer (40)(41)(42), inflammatory human psoriatic skin (34) and PBMC of patients with sepsis (43). Interestingly, patients with these inflammation conditions, namely psoriasis (35), COPD and sepsis (44) are frequently ZD. Thus, ZD modulation of expression of specific miR-NAs has implications with regard to the role of ZD in the pathogenesis of these diseases.
With the exception of red meat and seafood, the Zn content in most foods is low. Individuals subsisting largely on a cereal diet are likely to be ZD. Global dietary ZD is estimated to affect 30% of the population, or 4% to 73% across subregions, with higher frequencies occurring in developing countries (55). The mechanistic link to ZD in most disease conditions has remained undefined.
In summary, our present data show that prolonged dietary ZD causes aberrant miRNA expression in a wide variety of tissues associated with inflammation, suggesting a likely mechanism contributing to the burden of human diseases associated with dietary ZD. Importantly, the demonstration of the dysregulation of miR-31 and miR-21 by ZD in inflammatory esophageal/lingual neoplasia provides new insight into the mechanisms whereby ZD promotes human ESCC and OSCC.

Supplementary material
Supplementary Tables I, II, III