The inducible Cox-2 enzyme and its constitutively expressed isoform Cox-1 are the key regulators of human prostaglandin metabolism [5-7]. The end products of their enzymatic activity comprise a panel of prostaglandins and thromboxanes, which have been identified as critical regulators of fundamental physiological and pathological processes including platelet aggregation, parturition, T-cell development, inflammation and cancer [5-7]. Cox-2 enzymatic activity is largely regulated through de novo synthesis of Cox-2 protein [5,6].

In the stomach, enhanced Cox-2 expression has been found during H. pylori-triggered gastritis as well as in mucosal stress lesions, gastroduodenal ulcers and after ischemia/reperfusion damage [8-10]. Cox-2 and its related prostanoids also appear to contribute to the pathogenesis of gastric cancer. Gastric adenocarcinoma and premalignant mucosal lesions frequently over-express the Cox-2 gene [11-14], and elevated intratumoral Cox-2 levels seem to be associated with deeper tumor invasion [15] and an increased frequency of lymphatic metastasis [16]. In addition, Cox-2 inhibitors have been demonstrated to potently suppress proliferation of human gastric cancer cells in vitro [1,17,18] as well as experimental gastric adenocarcinomas in nude mice [17]. Recently however, a number of reports have challenged the notion that this anti- tumour activity is due to inhibition of Cox-2 itself [19]. Individuals taking Cox-inhibitors have been reported to display a reduced risk for development of gastric carcinoma [20], however the reported cardiovascular side effects associated with chronic coxib administration mean that clinical use of Cox- inhibitors for anti-carcinogenic treatment is controversial (Reviewed in[21]).

Expression of the Cox-2 gene therefore appears to be an important step in the pathogenesis of benign and malignant gastric diseases and therefore, clarification not only of its contribution to H. pylori-dependent pathogenesis, but also the downstream effects of Cox inhibiting drugs is of special clinical significance.

We have previously demonstrated that H. pylori can directly influence expression of Cox-2 in gastric epithelial cells through transcriptional mechanisms, and identified MAPK-ERK-dependent activation of a proximal cis-regulatory CRE-Ebox element as a key step in the H. pylori-response of the Cox-2 gene [22]. While these results further confirmed the pathophysiological link between the bacterium and Cox-2, molecular effectors located downstream of Cox-2 during gastric H. pylori infection remained unidentified.

Both H. pylori infection, and NS398 treatment elicited unique transcriptional signatures in the gastric mucosa of mice, despite the similarity in pathology scores and colonization density between the different groups (Figure (Figure1).1). This is of interest because in another report employing a model of mucosa-associated lymphoid tissue lymphoma (MALT lymphoma), transcriptional profiles in BALB/c mice infected with H. heilmannii were determined after 12 to 24 months, and the major changes in gene regulation occurred in the earlier stages of the disease (<12 months, mild to moderate pathology) [28]. After this time, clustering of the 300 most differentially expressed genes allowed segregation of infected mice into groups corresponding almost exactly with their pathological characterization. In view of our data, we would conclude that during the early development phase of chronic active gastritis, standard pathology analysis is not able to detect subtle, but important changes in the mucosa.

NS398 specifically inhibits the activity of the Cox-2 protein, and alterations in Cox-2 gene expression might be expected when infected mice were treated with a specific Cox-2 inhibitor, either as a result of a possible feedback mechanism involving PGE₂ [29], or as a compensatory mechanism to overcome the enzymatic inhibition. No significant change in Cox-2 expression was observed in vivo however, supporting the notion that the expression of the Cox-2 gene in the stomach is controlled by a variety of factors. Cox-2 is expressed by both inflammatory and gastric epithelial cells [22] and its expression may be controlled by different mechanisms in different cells types.

We were able to identify a subset of genes that were differentially expressed as a result of Cox-2 suppression, highlighting the scope of influence of Cox-inhibitors in gastric inflammation. Cox-2 dependent genes fell into numerous functional categories, chiefly those involved in gastric physiology (acid secretion, motility), epithelial repair and proliferation, and inflammatory mediators.

Gastrin (Gast) is an important mediator in the stomach [30-34] and expression in the mucosa was strongly influenced, not only by infection with H. pylori, but also by suppression of Cox-2 activity (Figure (Figure3,3, Additional File 1, Table S4). In addition to its role in regulating gastric acid secretion, gastrin has trophic effects and regulates proliferation and repair in the mucosa. Indeed INS-GAS transgenic mice which suffer from hypergastrinemia develop carcinoma after infection with H. pylori [35]. The development of carcinoma is however restricted to males in this model [35]. Other workers have also observed that Cox-2 inhibition influenced gastrin expression in an in vitro colorectal cancer model [34] and also in H. pylori positive gastric cancer patients [36].

Whereas expression of the apoptosis mediating growth differentiation factor 3 (Gdf3) and c-Myc (Mycs) genes peaked at week 13, the apoptosis inhibiting gene clusterin (Clu) strongly decreased at this time point. Taken together, the gene expression pattern is suggestive of a shift in the rate of proliferation/apoptosis of the epithelium after 13 weeks of infection as a result of NS398 treatment. Longer-term studies would be required to determine whether this effect continues or recurs. A number of genes which have been previously observed to be over-expressed in tumours (Mycs, Clu)[37], tumor suppressors (Patched, Ptch), or otherwise involved in metastasis or DNA repair: bikunin (Ambp)[38], ornithine decarboxylase gene (Odc), Trefoil factor 1 (Tff1)[39], insulin like growth factor (Igf2) [40] and DNA repair protein 1 (Ddb1), were also differentially expressed in NS398 treated infected mice (Figure (Figure3).3). The expression pattern was complex however, and some mediators tended to be enhanced by Cox-2 suppression, while the majority were down-regulated (Figure (Figure3).3). This probably reflects a regulatory role for Cox-2 as part of a network of control mechanisms for epithelial maintenance. The Odc gene, for example, encodes a key regulatory enzyme in the production of polyamines which are essential for cell proliferation [41] and has been shown to play a role alongside Cox-2 in the development of atrophic gastritis [42,43]. It has been proposed that Cox-2 inhibitors may inhibit Odc, and in this way be responsible for the observed anti-proliferative effects of Cox-2 inhibitors [44]. Our observation of decreases in Odc expression as a result of NS398 treated H. pylori infection in both in vivo and in vitro studies (Table 1 and Figure Figure4C)4C) is in keeping with this notion.

Helicobacter infection is strongly associated with the induction of a strong Th1-type inflammatory response, with high levels of Ifnγ, which induces expression of other inflammatory mediators such as iNOS and Cox-2, and also circulating growth factors such as gastrin [45]. Supporting the notion that Ifnγ plays the key role in attracting and activating lymphocytes [46,47], we observed that expression of T-cell surface markers Icam1 [48,49] and Cd86 and ligands (Sell) [50] peaked at week 13 in infected mice. The interferon dependent GTPases (Igtp, Iigp-pending) regulate the anti-microbial activities of Ifnγ in a STAT1 dependent manner [51,52], and their expression was dramatically reduced in NS398 treated mice by the end of the study. Overall, inhibition of Cox-2 activity led to a reduced expression of inflammatory mediators between weeks 13 and 19 (Figure (Figure3);3); interestingly this change was not reflected in the pathology scores. Cox-2 has been shown to modulate the Th1/Th2 balance in inflammatory responses and inhibition of Cox-2 using NS398 led to a polarization of the response of in vitro stimulated human PBMCs toward Th1[53]. The authors postulated that chronic expression of Cox-2 and production of PGE2 results in an inhibition of the effectiveness of the mucosal immune response by enhancing a state of tolerance. The gene expression pattern we observed here is indeed consistent with an effect of Cox-2 inhibition on the inflammatory response (See Figure Figure33 and Table 1) although the changes in expression of classical Th1/Th2 mediators such as IL-12, I-10 and IL-4 did not differ significantly in our study.

Infection with H. pylori has been reported to damage epithelial integrity and several potential mechanisms for this have been reported (reviewed in [54]). CagA, a major H. pylori pathogenicity factor, is translocated into epithelial cells via the type IV secretion apparatus [55,56]. Studies in a canine kidney cell model (MDCK) showed that CagA associates with the tight junction adaptor protein zona occludens 1 (zo-1, mouse homologue -tight junction protein 1, Tjp1) and the junctional adhesion molecule (Jcam or Jam), leading to a long term disruption of epithelial barrier function in vitro [57]. Whilst we observed increased expression of Tjp1 at the transcriptional level in NS398 treated infected mice, Jcam expression was not affected. In addition, an EST with homology to connexin 45 (AV148957, Gja7) and Aqp5 were influenced by H. pylori infection regardless of NS398 treatment (Figure (Figure3).3). As both connexin 45 and Aqp5 are know to play a role in intercellular transport of water and small molecules, and there is experimental evidence that connexin 45 interacts directly with Tjp1 [58,59], it would appear that Cox-2 also has a role in the maintenance of water balance in the gastric epithelium. This idea is supported by our in vitro observations of a cox-2 dependent increase in Zo-1 protein expression in MKN28 cells. In contrast to the reports by Amieva et al. [57] this effect was not related to the CagA status of H. pylori (Figure (Figure4).4). Barrier function effects in the mouse model are in any case unlikely to be due to the actions of CagA, as although we found that H. pylori SS1 expressed CagA protein, we were not able to detect translocation in either in vitro or in vivo experiments (data not shown). Therefore, it would appear that H. pylori infection has additional mechanisms to influence epithelial integrity. It is also of note that another H. pylori pathogenicity factor, vacuolating toxin (VacA) causes formation of fluid-filled vacuoles in epithelial cells and furthermore, this activity can be inhibited in vitro by NS398 treatment [60]. As the gastric aquaporin Aqp5 is expressed on the lateral and intercellular membranes in the gastric crypts [59], we speculate that this pore is influenced by changes to tight junction proteins, and that it plays a role in the development of oedema in the epithelium during infection.