Zinc deficiency activates S100A8 inflammation in the absence of COX-2 and promotes murine oral-esophageal tumor progression

Zinc (Zn)-deficiency (ZD) is implicated in the pathogenesis of human oral-esophageal cancers. Previously, we showed that in ZD mice genetic deletion of cyclooxygenase-2 (Cox-2) enhances N-nitrosomethylbenzylamine-induced forestomach carcinogenesis. By contrast, Cox-2 deletion offers protection in Zn-sufficient (ZS) mice. We hypothesize that ZD activates pathways insensitive to COX-2 inhibition, thereby promoting carcinogenesis. This hypothesis is tested in a Cox-2−/− mouse tongue cancer model that mimics pharmacologic blockade of COX-2 by firstly examining transcriptome profiles of forestomach mucosa from Cox-2−/− and wild-type mice on a ZD vs. ZS diet, and secondly investigating the roles of identified markers in mouse forestomach/tongue preneoplasia and carcinomas. In Cox-2−/− mice exposed to the tongue carcinogen 4-nitroquinoline 1-oxide, dietary ZD elicited tongue/esophagus/forestomach carcinomas that were prevented by ZS. The precancerous ZD:Cox-2−/−vs. ZS:Cox-2−/− forestomach had an inflammatory signature with upregulation of the proinflammation genes S100a8 and S100a9. Bioinformatics analysis revealed overrepresentation of inflammation processes comprising S100a8/a9 and an nuclear factor (NF)-κB network with connectivity to S100A8. Immunohistochemistry revealed co-overexpression of S100A8, its heterodimeric partner S100A9, the receptor for advanced glycation end-products (RAGE), NF-κB p65, and cyclin D1, in ZD:Cox-2−/− forestomach/tongue preneoplasia and carcinomas, evidence for the activation of a RAGE-S100A8/A9 inflammatory pathway. Accumulation of p53 in these carcinomas indicated activation of additional inflammatory pathways. Zn-replenishment in ZD:Cox-2−/−mice reversed the inflammation and inhibited carcinogenesis. Thus, ZD activates alternative inflammation-associated cancer pathways that fuel tumor progression and bypass the antitumor effect of Cox-2 ablation. These findings have important clinical implications, as combination cancer therapy that includes Zn may improve efficacy.

factors. 3 Epidemiologic and clinical studies have long implicated zinc (Zn)-deficiency (ZD) in the pathogenesis of oralesophageal cancers in many populations. [4][5][6] ZD is associated with increased tumor size and poor disease prognosis. 4,7 Zn is required for the activity of many enzymes, for proper immune function, and for the conformation of many transcription factors that control cell proliferation, apoptosis, and signaling pathways. 8,9 Zn is known to undergo rapid ligand exchange reactions and is used as an information carrier in signal transduction pathways similar to calcium. 10 Consequently, ZD predisposes to disease by adversely affecting immune system, by increasing oxidative stress, and by increasing the generation of inflammatory cytokines. 11 Although the role of ZD as a causative factor of disease and as a determinant in disease progression is gaining attention, 12 the mechanisms underlying its protumorigenic effect, however, remain unclear.
In the rat, a ZD diet creates a precancerous condition in the upper digestive tract, including tongue, esophagus and forestomach (an expanded lower esophagus), by inducing proliferation 13 and gene expression changes, including overexpression of cyclooxygenase-2 (Cox-2) and the proinflammation-genes S100 calcium binding protein a8 (S100a8) and a9 (S100a9). 13,14 ZD rats rapidly develop esophageal tumors after a single exposure to the environmental carcinogen N-nitrosomethylbenzylamine (NMBA) 15 and concurrent tongue, esophageal and forestomach tumors with exposure to the tongue carcinogen 4-nitroquinoline 1-oxide (NQO). 13 Znreplenishment (ZR) reverses cell proliferation, corrects gene expression and inhibits carcinogenesis. [14][15][16] Targeted therapies that block molecules crucial to tumor growth are being explored in attempts to prevent or cure cancer. 17 The rationale for targeting the COX-2 pathway is supported by numerous studies. COX-2 is overexpressed in many human cancers, including esophageal and tongue SCC. 18,19 COX-2 catalyzes the formation of prostaglandins and is induced by factors implicated in carcinogenesis, including growth factors, inflammatory stimuli, oncogenes and tumor promoters. 20 The report that deletion of the Cox-2 gene in Apc knockout mice greatly reduces intestinal polyp formation provides genetic evidence that COX-2 plays a key role in tumorigenesis. 21 COX-2 selective inhibitors, celecoxib in particular, are being tested in clinical trials for the prevention of several cancers, 22 including esophageal cancer. 23 Although such targeted therapies have shown promising results in several cancers, their efficacy in oral-esophageal cancer has been limited. 24 Our previous work showed that in ZD rats pharmacologic COX-2 inhibition by the drug celecoxib did not prevent tongue carcinogenesis, and in ZD mice genetic Cox-2 deletion actually enhanced NMBA-induced forestomach tumorigenesis. 16 Aside from the result that ZD:Cox-2 À/À mouse forestomach overexpressed leukotriene A 4 hydrolase protein, indicating a shift of arachidonic acid to the 5-lipoxygenase pathway, mechanisms underlying this effect of ZD were not elucidated. We hypothesized that ZD adversely affects treatment outcome by stimulating pathways not inhibited by the pharmacologic blockade of COX-2. We tested this hypothesis in a ZD:Cox-2 À/À mouse oral-esophageal cancer model that mimics pharmacologic COX-2 blockade, using techniques that included transcriptome profiling, bioinformatics analyses, and investigation of the pathobiological roles of identified markers in murine tongue/forestomach preneoplasia and neoplasia.

Expression profiling and related studies
Weanling Cox-2 À/À and Cox-2 þ/þ mice were fed ZD or ZS diets to form four groups, namely, ZD:Cox-2 À/À (n ¼ 20), ZS:Cox-2 À/À (n ¼ 12), ZD:WT (n ¼ 12) and ZS:WT (n ¼ 12). After 9 weeks, 8 ZD:Cox-2 À/À mice were switched to a ZS diet to form the ZR:Cox-2 À/À group. After a week, all mice were sacrificed. This experimental regimen produced unbridled cell proliferation in ZD:Cox-2 À/À forestomach. 16 Tongue and forestomach were isolated and cut into two portions. One portion was formalin-fixed and paraffin-embedded for immunohistochemical (IHC) studies. Forestomach epithelia for expression profiling studies were prepared from the remaining portion by using a blade to strip off the submucosal layers and snap-frozen in liquid nitrogen.
We performed expression profiling of forestomach mucosa from ZD:Cox-2 À/À , ZS:Cox-2 À/À , ZD:WT mice and ZS:WT mice after 10 weeks of ZD or ZS diets (n ¼ 4 mice/group), using GeneChipV R Mouse Genome 430 2.0 Array (Affymetrix, Santa Clara, CA). Total RNA was extracted from forestomach mucosa using TRIZOL reagent (Invitrogen, Carlsbad, CA). Five micrograms of total RNA was reverse transcribed into cDNA followed by in vitro transcription and labeling to produce biotin-labeled cRNA. The cRNA was hybridized to the arrays as described. 14

Expression data analysis
We used the Class Comparison analysis of BRB-Array Tools software version 3.7.0 (Biometric Research Branch, NCI) to identify differentially expressed mRNAs. The Robust Multichip Average method was performed. The array data were submitted to ArrayExpress (Accession number: E-TABM-778).

Gene ontology and pathway analyses
We used DAVID (Database for Annotation, Visualization and Integrated Discovery) 25 bioinformatics to identify relevant biological processes/functions from expression data captured by transcriptome analysis. Based on gene ontology, differentially expressed genes were grouped by scoring the statistical significance of predefined functional gene groups according to their functional similarity.
We used Ingenuity Pathway Analysis software (IPA, http://www.ingenuity.com) to analyze probable network/pathway and functional group enrichment. For each data set, the selected genes were uploaded into the IPA application. Networks were then algorithmically generated based on genegene connectivity.
ZR and forestomach carcinogenesis in ZD:Cox-2 2/2 mice This mouse study was approved by the Thomas Jefferson University Animal Use Committee. Thirty-nine 4-week old Cox-2 À/À mice were fed a ZD diet to form the ZD:Cox-2 À/À group. After 4 weeks, the mice received three intragastric doses of NMBA (2 mg/kg body weight, twice weekly), a regimen that produced a high tumor outcome in ZD:Cox-2 À/À mice. 16 A day after the 3rd dose, 18 mice were switched to a ZS diet to form the ZR:Cox-2 À/À group, which were given an intragastric dose of Zn gluconate weekly for 14 weeks (0.04 mg Zn). The remaining ZD:Cox-2 À/À mice continued on ZD diet. All mice were sacrificed for tumor outcome analysis at 14 weeks of Zn intervention.

Tumor analysis
At autopsy, tongue, esophagus and forestomach were excised. Tumors greater than 0.5 mm were mapped. Tissues were formalin-fixed and paraffin-embedded for histopathologic/IHC studies.

Quantitative reverse transcriptase-polymerase chain reaction
Quantitative reverse transcriptase-polymerase chain reaction (qRT-PCR) was performed using the comparative C t method and predesigned probes on 7300 Real-time PCR System (Applied Biosystems, Foster City, CA). GAPDH was used to normalize RNA samples. 14

Statistical analysis
Tumor multiplicity was analyzed by two-way analysis of variance (ANOVA). Differences among the groups were assessed using the Tukey-HSD post hoc t-tests for multiple comparisons. Tumor and carcinoma incidence rates were assessed by Fisher's exact test. CIs for the differences in incidence rates were calculated using the Wilson Score Method. 26 Statistical tests were two-sided and considered significant at p < 0.05.
Among mice of the same genetic background, tongue tumor incidence/multiplicity and carcinoma incidence were significantly higher in homozygous ZD:Cox-2 À/À vs. ZS:Cox-2 À/À or heterozygous ZD:Cox-2 þ/À vs. ZS:Cox-2 þ/À mice, but not in ZD:WT vs. ZS:WT mice ( Fig. 1a and b, Supporting Information Table 1), demonstrating that combined ZD and Cox-2 ablation led to a worse tumor outcome. These results are consistent with and extend our previous study in NMBAinduced forestomach carcinogenesis. 16 ZD per se induces an inflammatory gene signature in ZD:Cox-2 2/2 forestomach To test the hypothesis that ZD promotes carcinogenesis by activating cancer pathways not inhibited by genetic Cox-2 ablation, we performed transcriptome profiling of forestomach mucosa from ZD:Cox-2 À/À , ZS:Cox-2 À/À , ZD:WT and ZS:WT mice (n ¼ 4/group). We used forestomach rather than tongue because its epithelia can be readily separated from the muscularis layers without enzymatic digestion.
First, we examined the effect of ZD on gene expression changes in Cox-2 À/À forestomach and WT forestomach. By using a cutoff of p 0.05 and 2-fold difference in expression levels, we found 314 dysregulated probe sets in ZD:Cox-2 À/À vs. ZS:Cox-2 À/À forestomach (Supporting Information Table  2) but only 67 in ZD:WT vs. ZS:WT forestomach (Supporting Information Table 3). Thus, dietary ZD causes more extensive changes in gene expression in Cox-2 À/À than WT forestomach. A cohort of 36 genes, including the proinflammation mediators S100a8/a9, small proline-rich protein 2 Sprr2f/2h, and keratins Krt6a/8/19, was common to both class comparisons, indicating that these genes were induced by ZD regardless of genotype.
Next, we compared the effect of Cox-2 deletion on gene expression changes in ZD forestomach and in ZS forestomach. With a cutoff of 2-fold difference, we found 90 dysregulated genes in ZD:Cox-2 À/À vs. ZD:WT forestomach (Supporting Information Table 4) but only 17 in ZS:Cox-2 À/À vs. ZS:WT forestomach (Supporting Information Table 5). There are no common changes in gene expression between these two class comparisons, and Cox-2 deletion causes fewer changes in ZS than ZD forestomach.

IPA reveals a NF-jB-centric network
To understand gene expression interactions in ZD:Cox-2 À/À vs. ZS:Cox-2 À/À forestomach (Table 1) in the context of signaling pathways, we performed pathway analysis using IPA. We identified a nuclear factor (NF)-jB centric network of 35 genes, with 60% of the genes (21 of 35) from the upregulated genes that included S100a8 (Fig. 2c). Because NF-jB is a transcription factor that regulates immune responses/cell proliferation and it is a link between inflammation and cancer development/progression, 31 our result that NF-jB showed connectivity to S100a8 predicted activation of a S100A8-NF-jB inflammatory pathway in ZD:Cox-2 À/À forestomach.
ZD activates S100A8 inflammatory signaling in preneoplastic ZD:Cox-2 2/2 forestomach We focused our study on S100A8 and its heterodimeric partner S100A9 because of their role in inflammation and cancer, 14,32 and their prominence among ZD-induced
protein expression was strong in both ZD:Cox-2 À/À tongue and forestomach but weak or absent in similar tissues of other mouse groups (Fig. 3e, bottom). These data suggests that in tongue and forestomach ZD activates similar inflammatory pathways that are not affected by COX-2 inhibition.

ZR attenuates the inflammation and restores the antitumor effect of COX-2 blockade in cancer prevention
Finally, we investigated whether replenishing Zn can restore the antitumor effect of COX-2 blockade in tumor prevention.

Discussion
Increasingly cancers are treated with drugs that target specific pathways shown to be of pathogenetic significance. Our study shows that the antitumor effect of genetic disruption of Cox-2 in tongue cancer prevention is bypassed by Zn depletion (Fig. 1), owing to activation of an alternative proneoplastic pathway that is not affected by COX-2 inhibition. Using a combination of techniques that included expression profiling, bioinformatics and investigation of identified markers in ZD:Cox-2 À/À mouse models of oral-esophageal cancers, our data document a mechanism for the inability of COX-2 blockade to prevent tumor growth under ZD conditions. The hyperplastic ZD:Cox-2 À/À vs. ZS:Cox-2 À/À forestomach has a distinct signature ( Table 1). The pro-inflammation mediators S100a8 and S100a9 are upregulated 24-fold and 2.2-fold. In addition, the typical genes of the cornified envelope Sprr2h/2f and Krt6A/16/17/8/20 are upregulated 64-to 5.5-fold. Because simultaneous upregulation of S100A8/A9, SPRR2 and KRT6A/16/17 is a common feature of human inflammatory skin diseases such as psoriasis 38 and atopic dermatitis, 39 this same signature in ZD:Cox-2 À/À forestomach indicates an association between inflammation and its highly hyperplastic phenotype. Our conclusion that S100a8/a9 are relevant ZD-induced markers belonging to an inflammatory pathway that drives forestomach cell proliferation rather than an epiphenomenon of this process or of dietary Zn-deficit is supported by DAVID bioinformatics (Table 7a). S100A8/A9 have emerged as important markers for inflammation-associated cancers. 32,40 They are overexpressed in many human cancers, 32 including lung, colorectal, prostate, skin cancer, as well as HPV18-infected oral SCC. 41 The mechanistic role of S100A8/A9 in tumor biology is emerging. In a mouse skin cancer model, Gebhardt et al. 42 provided genetic evidence that S100A8/A9 binds to RAGE, and RAGE signaling sustains skin inflammation and promotes tumorigenesis. In the lung, S100A8/A9 induces the activation of serum amyloid A that activates NF-jB inflammatory signaling and facilitates metastasis. 43 In a colitis-induced mouse cancer model, S100A8/A9 and RAGE augment carcinogenesis 44 and in an inflammation-associated liver cancer model, S100A8/A9 are identified as NF-jB target genes and their overexpression promotes malignant progression. 45 Conversely, blockade of RAGE suppresses tumor growth and metastasis. 42,46 Our IHC data in ZD:Cox-2 À/À tongue and ZD:Cox-2 þ/À forestomach carcinomas (Fig. 4) demonstrate that under complete or partial blockade of COX-2 pathway dietary ZD activates an alternative cancer-associated RAGE-S100A8 inflammatory pathway. The finding that these same carcinomas showed high PCNA proliferative activity and prominent accumulation of p53 protein indicates that additional inflammation-associated cancer pathways are activated. The p53 tumor suppressor gene is mutated in human oral and esophageal cancer. 36 Mutated p53 protein has a prolonged half-life that leads to its accumulation in the nucleus. In this regard, human head and neck squamous cell cancer (HNSCC), which is a highly inflammatory, proliferative and aggressive cancer, 47 exhibits high levels of p53 expression, abundant cell proliferative activity, 34,35 as well as divergent carcinogenic pathways mediated separately by NF-jB and p53. 37 Chronic inflammation contributes to the development of $20% of all human cancers. The causes of inflammation are often unknown. 48 Our recent report in rat esophagus that dietary Zn regulates S100A8 expression and modulates the link between S100A8-RAGE and downstream NF-jB/COX-2 provides the first evidence that Zn has an inflammation-modulating role in essophageal cancer initiation/reversal. 14 Here we demonstrate that with COX-2 pathway blockade prolonged dietary ZD causes chronic inflammation in the tongue/forestomach by activating alternative inflammatory RAGE-S100A8/A9 and p53 response pathways, thereby fueling tumor progression and bypassing the antitumor effect of Cox-2 deletion. These new data provide a likely mechanism to explain the inefficacy of such targeted cancer therapy in oral-cancer patients, since many of these patients are frequently Zn-deficient. [4][5][6][7] Recent studies reported that Zn supplementation improves clinical outcomes in patients receiving radiotherapy for HNSCC, 49 as well as concomitant chemotherapy and radiotherapy for advanced nasopharyngeal carcinoma. 50 The present finding that ZR attenuates the inflammatory response and restores the antitumor effect of COX-2 blockade has important clinical implications. Thus, stratification of patients by Zn status would be useful, and a personalized cancer therapeutic paradigm that includes Zn may improve efficacy.