EMT is an evolutionary conserved, reversible process in which polarized epithelial cells acquire a mesenchymal phenotype through phenotypical and biochemical changes, thereby resulting in increased capacities of migration, invasion, and apoptosis resistance as well as ECM production and remodeling^[38]. Transcription factors such as Snail, Slug, zinc-finger E-box binding (ZEB1/2) and FOXC2 are activated at the beginning of the process. This is accompanied by the expression of specific microRNAs (miRs), for instance the miR-200 family, changes in the expression of particular cell surface proteins, cytoskeletal reorganization, and activation of Wnt/β-catenin and Notch signaling^[38,46,47].

A critical feature of EMT is the down-regulation of E-cadherin^[48], a surface glycoprotein expressed in epithelial cells that is a key component of the adherent junctions in epithelial tissues^[49]. Expression of E-cadherin can be repressed directly or indirectly by multiple transcription factors, including ZEB1/2, Snail, Slug, nuclear factor-kappa B (NF-κB), E47 and KLF8, but also by the proteins SIX1 and FOXC2^[50-52]. Furthermore, various signaling pathways can influence the expression of E-cadherin, including TGFβ, hypoxia-induced response, Wnt/β-catenin, Notch and PI3K/Akt, and therefore play a role in EMT^[53,54]. Although EMT is usually depicted as a binary switch that shifts cells from a fully epithelial to a fully mesenchymal state, this is a misrepresentation of this process. Frequently, the EMT program drives cells from a fully epithelial state to a partially mesenchymal one in which some epithelial markers are retained. Nonetheless, this subset of mesenchymal traits has profound effects on the cell biology^[55].

Activation of EMT programs in neoplastic cells is usually connected to their dedifferentiation and acquisition of stem cell-like properties^[56]. The existence of cancer cells with stem-like properties, the so-called cancer stem cells (CSCs), was first described in breast cancer and subsequently documented in various malignancies, including GC. One the first studies about CSCs in GC showed that these cells have enhanced capability of invasion and tumorsphere formation. Also, it was found that CSCs have distinctive features of the EMT, such as reduced expression of E-cadherin and increased levels of vimentin and MMP2^[57]. In primary GC tissue, it was demonstrated by immunohistochemistry that the combination of Snail-1, vimentin, E-cadherin and CD44 predicts tumor aggressiveness^[58]. Furthermore, it has been reported that MKN7 GC cells undergoing Wnt5a-induced EMT acquire CSC properties^[59], similar to what has been observed in hypoxia-driven EMT in vitro models with the BGC823 and SGC7901 GC cell lines^[60].

Using in vitro systems, it was revealed that H. pylori infection results in the activation of EMT programs and the emergence of CD44^high cell populations with CSC properties in the AGS, MKN45 and MKN74 GC cell lines^[61]. These cells acquire elongated shape and show enhanced expression of mesenchymal markers (i.e., Snail1, ZEB1, and vimentin). Compared to the CD44^low cells, the CD44^high GC cell population gained the ability to migrate and invade and was better at forming tumorspheres in vitro and tumors in immunodeficient mice. According to that study, the induction of the EMT and CD44^high cell population was dependent on the CagA oncoprotein. CagA induces the EMT in a number of ways, as exemplified by a recent in vitro study showing that this bacterial protein up-regulates MMP3, which is also part of the EMT program, through EPIYA motifs in a phosphorylation-dependent manner^[62]. Immunohistochemistry staining of human and murine gastric tissue have confirmed that H. pylori infection is correlated with high expression of CD44 and EMT markers^[61]. Presumably, ERK and JNK are involved in the described EMT-like changes, CD44 overexpression and the ability to form tumorspheres in vitro that is triggered by H. pylori cagA-positive strains^[61].

Other studies addressing the potential induction of CSC-like properties by H. pylori concluded that Wnt/β-catenin activation in response to this bacterial infection is a necessary event for the acquisition of such properties in GC cell lines^[63]. Finally, observations on patients with gastric dysplasia and early GC, before and after eradication of H. pylori, showed a connection between the mRNA expression levels of TGF-β1, EMT markers, and immunohistochemical expression of CD44, suggesting that H. pylori infection may trigger a TGF-β1-induced EMT and the emergence of CSCs^[64].

Epithelial cells in the gastrointestinal tract are normally found in organized layers or epithelia. Their polygonal shape and functional organization in an apico-basal polarized manner allow them to lay in an orderly fashion, and the unions they form with each other and with the basal membrane give the epithelium a barrier function. As already mentioned, invasiveness is enhanced when cells lose their polarity due to alterations in the cytoskeleton and intercellular unions. H. pylori has been implicated in changing the polarity of the gastric epithelial cells, which may have important consequences in the context of gastric carcinogenesis. Of the three types of filaments that compose the cytoskeleton, the actin microfilaments are the most affected during H. pylori infection.

The actin cytoskeleton is a very dynamic structure whose assembly is finely tuned by complex signaling networks and involves numerous regulatory proteins. Actin microfilaments form a wide variety of structures: contractile rings, phagocytosis- and endocytosis-related structures, microvilli, cortex, adherens belts (associated with adherens junctions), filopodia, lamellipodia, and stress fibers. The ability of H. pylori to promote rearrangements of the actin cytoskeleton is well-established^[65-67]. The most evident demonstration of this is the so-called hummingbird phenotype, comprising a change in the epithelial cell shape to the characteristic elongated morphology of H. pylori-infected cells in vitro. This phenotype is thought to be linked to cancer cell migration and invasive growth in vivo^[68]. The hummingbird phenotype involves the formation of stress fibers and protrusions, the disruption of cell-to-cell adhesions, and the deregulation of focal adhesions between the cell and the ECM.

The basic mechanisms by which H. pylori changes the dynamics of the actin cytoskeleton during cell migration have been reviewed in depth by Wessler et al^[69]. Briefly, H. pylori, via cag-PAI type-4 secretion system (T4SS; especially CagL) and CagA, is able to modify the host cell’s signaling networks. On the one hand, CagL binds β1 integrins, thereby stimulating the focal adhesion kinases (FAKs) and the Src-family kinases (SFKs); meanwhile, CagA activates the Abl-kinase. FAK, SFK, and Abl activate Crk, which in turn activates Rac1, which then promotes the assembly of actin filaments via activation of the Arp2/3 complex, contributing to cell motility. On the other hand, upon injection into the host-cell cytosol, CagA is phosphorylated by SFK (c-Src) and binds Shp-2 and Csk, which then inhibit SFK in a negative feedback loop. Inhibition of SFK induces dephosphorylation of actin regulatory proteins, such as ezrin, vinculin, and cortactin. Cortactin stimulates the actin nucleation activity of Arp2/3 and, upon H. pylori-induced dephosphorylation, accumulates at the tip of the cellular protrusions and colocalizes with F-actin^[70].

The serine/threonine kinase polarity-regulating kinase partitioning-defective 1b (PAR1) participates in the CagA-mediated remodeling of the actin cytoskeleton. PAR1 inhibits the formation of stress fibers and cortical actin in the cell periphery. Kikuchi et al^[71] showed that the physical interaction between the CagA multimerization sequence and PAR1b, the isoform present in gastric epithelial cells, is crucial for the stable binding of CagA and Shp-2. In fact, a second study found that CagA indirectly activates RhoA-dependent formation of stress fibers by impairing PAR1b-mediated inhibition of RhoA^[72]. These results were elegantly combined in a model that proposes a link among cell polarity regulation, the hummingbird phenotype, and actin cytoskeleton^[73]. More specifically, upon cell polarity loss of the epithelial cell, PAR1b and aPKC are relocated, resulting in the establishment of a front-to-rear polarity in which these two molecules are asymmetrically distributed, with PAR1b localized in the rear part of the migrating cell.

The binding of CagA to PAR1b modifies this program by perturbing PAR1b localization, which translates into loss of its kinase activity, lifting of the repression of RhoA, and formation of stress fibers; the salient manifestation of this is the hummingbird phenotype^[73]. The affinity of CagA for PAR1b and formation stress fibers increases proportionally to the number copies of the CagA-multimerization (CM) domain present in CagA, which is seemingly higher in East Asian CM than in Western CM and differs in five amino acid residues^[74].

Podosomes are dot-like structures of densely packed F-actin and serve as regulatory proteins by their capacity to degrade ECM components due to the presence of MMPs within. It has been shown in a model of primary hepatocytes and hepatoma cell lines that H. pylori can enhance the formation of podosomes by the induction of inflammatory cytokines such as TGFβ^[75], thus providing additional evidence of the capacity of H. pylori to modify actin structures. Actin-remodeling activity has also been described in another Helicobacter species, H. pullorum. Its cytolethal distending toxin, responsible for the cytopathological effects observed upon infection, induces actin cytoskeleton remodeling that is accompanied by delocalization of vinculin and up-regulation of cortactin in large, cortical actin-rich lamellipodia^[76].

Another important component of the cellular cytoskeleton is the microtubules. Structurally, they are formed by tubulins and regulated by microtubule-associated proteins. The microtubular network organizes the cell movement of organelles and is part of specialized structures such as cilia, flagella, mitotic spindles, centrosomes, and basal bodies. During cell migration, the small Rho-GTPase protein Cdc42 controls the actin microfilaments in the cell migration front and binds PAR6. The Cdc42-PAR6 dimer recruits the microtubule-regulating dynein/dynactin complex, which directs the machinery of the secretory pathway to the migration front. Importantly, this facilitates the delivery of integrins and other proteins that mediate the interaction with the ECM to the sites of migration.

Although not many, some studies have addressed the role of H. pylori infection in microtubule regulation in the context of cell migration. Slomiany and colleague^[77], for example, concluded that H. pylori lipopolysaccharide induced the secretion of MMP9 in a primary culture of murine gastric mucosal cells. In that study, the authors also found an accompanying increase of microtubule stabilization. Presumably, those changes are modulated by ghrelin and involve the activation of PKCδ and SFK.

Focal adhesions provide the structural link between the stress fibers and the ECM. They need to be dynamically assembled and disassembled in order to allow cell migration. H. pylori impairs focal adhesion release during cell migration, which leads to the characteristic elongation of infected cells^[68]. Paxillin, a multidomain protein that acts as an adaptor between the cytoplasmic tail of integrins and the actin cytoskeleton, has been designated as the convergent point of the epithelial growth factor receptor, FAK/Src, and PI3K/Akt signaling pathways in the context of H. pylori infection^[78]. Paxillin phosphorylation is dependent on the presence of a functional Cag-PAI or OipA. The phosphorylated paxillin was localized along the elongations, suggesting a role in the formation of stress fibers^[78].

The PA system comprises a few proteins that, by acting in sequence, lead to the conversion of zymogenic plasminogen into its active enzymatic form, plasmin. Extravascular activation of plasminogen is controlled by the urokinase-type plasminogen activator (uPA), its receptor (uPAR), its inhibitor PAI-1, and α₂-antiplasmin. Besides degrading major ECM proteins (e.g., fibrin fibronectin, laminins, and vitronectin), the generated plasmin also releases latent growth and angiogenic factors sequestered in the matrix^[79,80]. The expression of uPA, uPAR, and PAI-1 under normal homeostatic conditions is almost undetectable; however, in cancer and other pathologies, their expressions increase significantly^[81]. An important body of evidence correlates uPAR expression in cancer lesions with invasive and metastatic disease. Accordingly, high levels of uPAR in tissue and plasma are associated with poor patient survival in various types of cancer, including GC^[82-84]. Most of these reports have focused on uPAR, since this receptor is crucial for the initiation of the sequential series of events that ultimately result in the activation of plasminogen.

As already mentioned, H. pylori has been linked to the induction of members of the PA system, primarily uPAR. Part of the experimental evidence supporting this phenomenon comes from in vitro studies, for example global gene-expression analyses ranking uPAR among the top up-regulated genes in AGS and T84 cell lines, when co-cultured with H. pylori^[85-87]. This has been confirmed by more specific in vitro studies, showing that in co-cultures, the bacterium rapidly induces uPAR expression in GC cell lines^[88-91]. A few of these reports indicate that uPAR induction is predominantly linked to CagA-positive strains^[87,89]. The potential connection between H. pylori and uPAR induction has also been documented in non-neoplastic tissue adjacent to GC lesions^[92] and in gastric biopsies from healthy patients who are infected with the bacterium^[93]. Interestingly, it has been reported that the expression of uPAR in neoplastic tissue may be correlated with the presence of H. pylori in adjacent non-neoplastic tissue^[94].

The link between H. pylori and uPAR has been systematically investigated in a mouse model of H. pylori-induced gastritis (Figure 3)^[95]. In this model, uPAR expression is up-regulated very early in response to the infection and increases progressively during the course of infection, and this is reverted to its physiological baseline levels if H. pylori is eradicated by antimicrobial therapy^[95]. Additional experiments in this model suggest that uPAR expression is directly induced by the bacterium (and Alpízar-Alpízar, unpublished results)^[95]. It is not possible to rule out, however, that uPAR induction in murine gastric epithelium is a consequence of the inflammatory reaction against H. pylori.

Figure 3 Plasminogen activator receptor induction in gastric epithelial cells in response to Helicobacter pylori infection. Stomach tissue sections of a mouse infected with H. pylori and euthanized 14 wk after inoculation processed for immunohistochemistry against H. pylori (A) and uPAR (C), and with H&E staining (B). Clusters of H. pylori bacteria (arrows) are observed in the upper third of the gastric glands along the gastric epithelium of the mouse stomach (A). Histopathological alterations are seen, including inflammation and mucous metaplasia (B). uPAR expression becomes evident at the apical membrane of foveolar epithelial cells in the corpus epithelium of H. pylori-colonized mice, such as the representative immunohistochemistry staining shown here (C). uPAR-positive scattered neutrophils are seen in the microphotograph (C) since they constitutively express this molecule. Scale bars: A: 100 μm; B and C: 200 μm. H. pylori: Helicobacter pylori; H&E: Hematoxylin and eosin; uPAR: Plasminogen activator receptor.

A few signaling pathways and transcription factors have been proposed as potential inducers of uPAR in cancer; however, much less is known about the mechanisms of induction in response to H. pylori. Studies in cancer cell lines have found that the NF-κB can drive uPAR expression by direct binding to specific sequences within the regulatory region of the gene encoding uPAR^[96] or indirectly via HIF1α activation^[97]. It is well known that H. pylori infection can lead to the activation of NF-κB^[36,98]. Therefore, NF-κB is a likely transcriptional inducer of uPAR in epithelial cells of H. pylori-colonized mucosa, both in human and mouse. In vitro evidence supports this idea^[88], but no experimental data have been generated in vivo.

AP-1 is another transcriptional regulator that is activated by H. pylori infection^[36] and has been implicated in the induction of uPAR in cancer^[99,100]. Thus, AP-1 may explain the potential connection between these two parameters^[90]. Both NF-κB and AP-1 can be activated via the Ras–ERK MAPK signaling pathway^[99,101]. This pathway is often manipulated by H. pylori^[36], which makes it an interesting study target to gain further insight about the mechanism of induction of uPAR in the gastric epithelium colonized with H. pylori.

The MMP family comprises more than 23 zinc-dependent endopeptidases, subdivided into eight groups according to structural characteristics^[42,102]. MMPs are synthesized in the form of zymogens (pro-MMPs) by several cell types of the tumor microenvironment; and, when released to the extracellular space, they become activated by other proteases, including MMPs themselves and plasmin^[42]. Besides their role in invasion and metastasis, MMPs are involved in other aspects of tumor biology. The degradation of ECM constituents results in the liberation of sequestered growth, proliferative and angiogenic factors but also in the generation of ECM-derived peptides with similar biological properties to those factors. Some MMPs can cleave membrane-bound growth factor precursors, thus releasing their active form, for example TGFβ^[42]. Elevated expression of several MMPs has been consistently correlated with poor cancer patient survival in several types of cancer, including GC. Of note, a few MMPs actually inhibit malignant transformation and tumor growth, including MMP8, MMP12, and MMP26^[103-107].

The possible connection between H. pylori infection and induction of MMPs (e.g., MMP2, MMP3 and MMP9) in gastric epithelial cells has been suggested; however, the most compelling evidence is probably for MMP7. MMP7 enhances tumor formation in rodents^[108], and it is particularly interesting in the context of the gastric carcinogenesis because its expression is increased in human GC lesions^[109,110]. In human gastric cell lines co-cultured with H. pylori, it was found that cag-PAI-positive strains augment the levels of MMP7 up to 7-fold compared to uninfected controls or to cells incubated with specific isogenic mutant strains^[111]. According to that report, the induction of MMP7 in the in vitro system was dependent on the activation of ERK 1/2 and required an active interplay between viable bacteria and epithelial cells^[111]. That study also evaluated the expression of MMP7 in gastric biopsies of human patients and found that it was over-expressed in epithelial cells of gastritis-affected individuals infected with CagA-positive strains^[111], which has also been previously documented by an independent report^[112].

These observations served as the driving force for conducting subsequent investigations in MMP7 knockout mouse models. Such studies concluded that gastric inflammation and epithelial cellular turnover are substantially increased in MMP7-deficient mice infected with H. pylori, compared to their wild-type counterparts^[113,114]. It is speculated that over-expression of MMP7 in response to H. pylori colonization may be a mechanism to protect the gastric mucosa from damage and development of lesions that could ultimately result in GC^[112,113]. Nevertheless, it is also proposed that sustained expression of MMP7 in the gastric epithelium could lead to malignant transformation^[112]. In fact, a few studies suggest that MMP7 proteolytically cleaves specific pro-apoptotic molecules, such as Fas ligand, from tumors cells, thus promoting tumor survival^[115,116].