Helicobacter-induced intestinal metaplasia in the stomach correlates with Elk-1 and serum response factor induction of villin.

Chronic Helicobacter pylori infection results in serious sequelae, including atrophy, intestinal metaplasia, and gastric cancer. Intestinal metaplasia in the stomach is defined by the presence of intestine-like cells expressing enterocyte-specific markers, such as villin. In this study, we demonstrate that villin is expressed in intestine-like cells that develop after chronic infection with H. pylori in both human stomach and in a mouse model. Transfection studies were used to identify specific regions of the villin promoter that are inducible by exposure of the cells to H. pylori. We demonstrated that induction of the villin promoter by H. pylori in a human gastric adenocarcinoma cell line (AGS) required activation of the Erk pathway. Elk-1 and the serum response factor (SRF) are downstream transcriptional targets of the Erk pathway. We observed inducible binding of Elk-1 and the SRF after 3 and 24 h of treatment with H. pylori, suggesting that the bacteria alone are sufficient to initiate a cascade of signaling events responsible for villin expression. Thus, H. pylori induction of villin in the stomach correlates with activation and cooperative binding of Elk-1 and the SRF to the proximal promoter of villin.


Chronic Helicobacter pylori infection results in serious sequelae, including atrophy, intestinal metaplasia, and gastric cancer. Intestinal metaplasia in the stomach is defined by the presence of intestine-like cells expressing enterocyte-specific markers, such as villin. In this study, we demonstrate that villin is expressed in intestine-like cells that develop after chronic infection with H. pylori in both human stomach and in a mouse model. Transfection studies were used to identify specific regions of the villin promoter that are inducible by exposure of the cells to H. pylori. We demonstrated that induction of the villin promoter by H. pylori in a human gastric adenocarcinoma cell line (AGS) required activation of the Erk pathway. Elk-1 and the serum response factor (SRF) are downstream transcriptional targets of the Erk pathway. We observed inducible binding of Elk-1 and the SRF after 3 and 24 h of treatment with
H. pylori, suggesting that the bacteria alone are sufficient to initiate a cascade of signaling events responsible for villin expression. Thus, H. pylori induction of villin in the stomach correlates with activation and cooperative binding of Elk-1 and the SRF to the proximal promoter of villin.
Chronic inflammation of the gastric mucosa (chronic gastritis) develops in response to Helicobacter pylori infection or bacterial overgrowth in the hypochlorhydric stomach (1,2). Over time, the inflammatory process progresses, and alteration of the epithelial cell population occurs, which includes gradual loss of parietal cells coinciding with an increase in the number of mucous cells. Proliferation of mucous cell types with evidence of an intestinal phenotype (intestinal metaplasia) is a major precursor lesion in gastric cancer (3). Moreover, intestinal metaplasia is a lesion that develops in a variety of cancers derived from organs of the forestomach (4 -6). Interestingly, intestinal metaplasia in different organs all express protein markers commonly found in normal intestinal enterocytes, e.g. TFF3, Cdx2, villin (7)(8)(9)(10). The gastric epithelium develops from intestinal endoderm by 16 days post-coitum when the pyloric border is formed. At that time, fetal intestinal markers cease to be expressed in the gastric epithelium. Re-expression of intestine-specific genes in the adult stomach represents a shift to a metaplastic phenotype that correlates with increased gastric proliferation.
Villin is expressed in primitive endoderm by embryonic day 5 and coalesces at the apical surface by day 8.5 postcoitum as the microvilli of the gut are being formed (11)(12)(13). By 16 days postcoitum, there is a one-cell distinction between villin expression in the duodenum and minimal expression in the antral glands (14). By postnatal day 1, villin expression in the stomach has completely receded. Thus, there is a complete absence of villin expression in the corpus of the mouse stomach (15). Villin is a structural protein regulated by increased intracellular calcium that in turn binds to actin and contributes to the formation of microvilli in the small bowel (16). Mice deficient for the villin gene are viable (17). Microvilli develop but do not respond normally to calcium-dependent signals, implicating the protein in the response to cell injury (17,18). In addition, villin is expressed in intestinal metaplasia observed in Barrett's esophagus and in chronic atrophic gastritis (19). Therefore villin is an important marker of the pre-neoplastic cell type that forms in the gut in response to chronic injury (20).
Whether true villin-positive intestinal metaplasia is a feature of the altered pattern of differentiation observed with Helicobacter colonization and inflammation has not been examined. In this report, we established that villin expression emerges in the infected stomachs of human subjects and a mouse model of Helicobacter infection. Further, we showed that the first 554 bp of the villin promoter contain elements capable of responding to H. pylori in culture. Within this proximal promoter region, we found that a serum response element (SRE) 1 confers inducible regulation of the villin promoter by H. pylori and that Elk-1 and the serum response factor (SRF) form a ternary complex at this element.

MATERIALS AND METHODS
Bacteria-The mouse-adapted H. pylori SS1 strain was used to inoculate mice for up to 14 months and was a gift from Dr. K. Eaton, University of Michigan. The H. pylori J99 strain (American Type Culture Collection 700824) is a human isolate that was used in the cell culture experiments because of its ability to strongly induce human cell lines. The bacteria were cultivated on blood agar plates containing Campylobacter base agar (Difco) supplemented with 5% horse blood (Colorado Serum, Denver, CO), 5 g/ml vancomycin, 10 g/ml trimethoprim lactate, and 2 g/ml nystatin (Sigma). The plates were incubated under microaerophilic conditions (CampyPak Plus, BBL; BD Biosciences) at 37°C for 1-2 days.
Human Tissue-Institutional Review Board approval from the University of Michigan was obtained prior to the acquisition of human tissue by Dr. Nguyen T. Vinh and Dr. Nguyen N. Thanh during endoscopy at Friendship Hospital and Tran Hung Dao Central Hospital, Hanoi, Vietnam. The biopsies were fixed in formalin and then paraffin-embedded.
Plasmids-The Ϫ554 Villin P/Intron reporter construct was prepared as described previously (21) and consists of 554 bp of the mouse 5Ј-flanking sequence, the first exon, and the entire first intron. Dominant negative Erk-1 and Erk-2 constructs were gifts from Dr. Melanie Cobb (University of Texas Southwestern, Dallas, TX) (22). Mutagenesis of the Ets and SRE sites within the villin reporter construct was accomplished using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA) according to the manufacturer's protocol using the following primer: 5Ј-GGGCCCTATCTAACCTTATAAAGT-GAAG-GAAGTAACT-3Ј. Deletion of the SRE/EBS binding site was created the same way using the following QuikChange primer: 5Ј-TCTGGGGCCCTATCTGGGGGGGGGGT-GGTGGTGAGGACC-3Ј. All constructs were confirmed by sequence analysis. The dominant negative Elk construct in pCMV5 (dnElk-1) was a gift from Dr. Peter E. Shaw (Queen's Medical Center, Nottingham, UK). The Elk-1 coding sequence contains a deleted C-terminal Apa fragment rendering the expressed protein kinase-deficient (23,24).
Animals and Helicobacter Challenge-C57BL/6 mice were purchased from Jackson Laboratories (Bar Harbor, ME) and maintained in individual, sterile microisolators for up to 14 months. Five days before infection, all mice were treated for 3 days with streptomycin (5 mg/kg) in their drinking water to reduce the natural bacterial flora. SS1 bacteria were harvested in Brucella broth (BBL), the concentration was adjusted according to the OD at 550 nm, and a 0.2-ml suspension containing 10 8 viable bacteria was used for oral infection. Each animal was challenged with bacteria three times over 5 days. All mice were fasted overnight with access to water ad libitum before analysis. The study protocol used was approved by the University of Michigan Animal Care and Use Committee, which maintains an American Association of Assessment and Accreditation of Laboratory Animal Care facility. Infection by H. pylori was verified by testing re-isolates for urease (using a drop of urea broth containing 10 g of urea, 0.5% w/v phenol red, 0.22 g of NaH 2 PO 4 H 2 O, 0.51 g of Na 2 HPO 4 , 100 mg of NaN 3 , and 500 ml of distilled water at pH 6.2), catalase (using 3% H 2 O 2 solution) and oxidase activity (DrySlide, BBL), as well as by PCR of the H. pylori-specific 16 S rRNA according to Fox et al. (25).
Histology-Paraffin sections of the human biopsies or mouse stomach were stained with hematoxylin and eosin for grading the intensity of inflammation and metaplasia. The presence of metaplasia was confirmed by a periodic acid-Schiff procedure/Alcian blue stain.
Immunofluorescence-Paraffin-embedded tissue sections of H. pylori SS1-infected and -non-infected mice were used. Deparaffinized sections were used for citrate buffer retrieval (10 mM citric acid for 10 min at 89 -95°C) and blocked with 15% donkey serum. To document villin expression in the stomach mucosa, primary mouse anti-proton pump IgG antibody (2B6, Medical & Biological Laboratories Co., Ltd., Naka-ku Nagoya, Japan), detecting the ␤ subunit of the mouse H ϩ , K ϩ -ATPase, and goat anti-villin antibody (Santa Cruz Biotechnology) were added in a dilution of 1:500 and 1:100, respectively, and then incubated with the sections at room temperature for 1 h. After rinsing the slides in phosphate-buffered saline, secondary antibodies were applied using Texas Red-conjugated anti-mouse IgG from donkey and fluorescein isothiocyanate-conjugated anti-goat IgG from donkey (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) in a dilution of 1:75 at room temperature for 1 h. DAPI (5 ng/slide, Sigma) was used to counterstain the nuclei. Slides were mounted with an anti-fade aqueous mount (Biomeda Corp, Foster City, CA). The cells were visualized with an Olympus BX60 fluorescence microscope and photographed with the digital SPOT camera (Diagnostic Instruments).
H. pylori Co-culture with Human Gastric Cell Lines-NCI-N87 cells were cultured to 60% confluency in Dulbecco's modified Eagle's medium, supplemented with 5% fetal bovine serum, 5% horse serum, 100 g/ml penicillin, and 100 g/ml streptomycin (Invitrogen) in 5% CO 2 , 95% air-humidified atmosphere. The cells were then starved for 48 h in F-12 medium supplemented with 100 g/ml penicillin and 100 g/ml streptomycin. Unlike AGS cells, the NCI-N87 cell line expresses a variety of gastric peptides (26). Endogenous regulation of villin protein was examined in NCI-N87 cells using 1:100 m.o.i. (ratio of 100 bacteria/ eukaryotic cell) of either Campylobacter jejuni, SS1, or J99 H. pylori strains for 24 h. Due to the higher efficiency of transfection, all trans-fections, co-culture studies, and nuclear extract preparation were performed using AGS cells.
Transient Transfections and Reporter Assays-AGS cells were cultured to 60% confluency in medium as described above. The cells were grown on 12-well plates and transfected using calcium phosphate coprecipitation without glycerol shock (5 Prime-3 Prime, Inc., Boulder, CO) or FuGENE 6 transfection reagent (Roche Applied Science) according to the manufacturer's protocol. Forty-eight hours after transfection, the cells were washed twice with phosphate-buffered saline and the medium replaced with unsupplemented F-12 containing H. pylori J99 strain at 100 m.o.i. Dominant negative expression constructs were co-transfected and kinase inhibitors were added 1 h before stimulation of the transfected cells. Luciferase assays were performed 36 h after transfection using a Berthold AutoLumat luminometer (LB953; EG & G, Gaithersburg, MD) or Wallac Victor3 1420 multilabel counter (PerkinElmer Life Sciences). Luciferase activity was normalized to the protein concentration determined by the Bradford method (27).
To detect phospho-Elk-1, nuclear protein extracts were prepared by a two-buffer detergent method (28) supplemented with 50 mM NaF to inhibit phosphatase activity. The extracts were heat-denatured in Laemmli sample buffer and resolved on a 4 -20% SDS-polyacrylamide gel. The proteins were electroblotted onto a Hybond-C Extra nitrocellulose membrane (Amersham Biosciences). The membrane was washed three times with NaF-supplemented TTBS (add 50 mM NaF), blocked for 1 h in 5% nonfat dry milk in NaF-supplemented TTBS, and then incubated with phospho-Elk-1 antibody (Santa Cruz) for 2 days at 4°C. After three rinses with NaF-supplemented TTBS, the membrane was incubated with a secondary horseradish peroxidase antibody (Cell Signaling Technology, Beverly, MA) at 1:1000 dilution for 1 h at room temperature. Proteinantibody complexes were detected with the LumiGlo reagent and peroxide chemiluminescent detection kit (Cell Signaling Technology).
Gel Shift Assays-For nuclear extract preparation, the cells were co-cultured with the bacteria for either 3 or 24 h. They were then washed twice with cold Tris-buffered saline, harvested, and placed on ice. Uninfected cells were starved for 48 h but received no H. pylori. Nuclear protein extracts were prepared by a two-buffer detergent method (28). A double-stranded 28-bp oligonucleotide probe 5Ј-TCCCT-TATATGGTGAAGGAAGTTCCTGG-3Ј (sense strand) was hybridized and then end-labeled with [␥-32 P]dATP using polynucleotide kinase. Gel shift reactions were carried out at 25°C in a total volume of 20 l containing 10 mM Tris-HCl (pH 7.9), 100 mM KCl, 5 mM MgCl 2 , 0.5 mM EDTA, 0.5 mM dithiothreitol, 10% glycerol, 600 ng of poly(dI-dC), 6 g of nuclear extract, and the corresponding antibody or unlabeled competing oligonucleotides (100ϫ the molar concentration of the probe). After a 10-min preincubation, 30,000 cycles/min of labeled probe (10,000 -15,000 cycles/min/0.1 ng) was added to each reaction followed by the loading buffer containing bromphenol blue. Supershifts were performed by adding 2 g of anti-Elk-1, anti-SRF, anti-YY1, or anti-GATA-4 antibodies (Santa Cruz Biotechnology) 20 min before adding the probe. The reactions were run on a 4% non-denaturing polyacrylamide gel containing 45 mM Tris, 45 mM boric acid, and 1 mM EDTA. The gels were dried and then used to expose x-ray film.
Statistical Analysis-The results were statistically tested by unpaired t test or one-way analysis of variance as appropriate, using commercially available software (GraphPad Prism, GraphPad Software, San Diego, CA). A p Ͻ 0.05 was considered significant.

RESULTS
Chronic Helicobacter infection in both human subjects and mice induces gastritis, atrophy, and metaplasia. In human stomach, intestinal metaplasia is correlated with the presence of Alcian blue-positive cells and an increase in intestine-specific markers, e.g. villin (29). The metaplasia appears prior to the development of gastric cancer (3). However, the expression of villin as a marker for intestinal metaplasia developing in response to Helicobacter infection has not been studied. Therefore, we examined biopsies from human subjects infected with H. pylori. The gastric mucosa of infected patients showed Alcian blue-positive cells consistent with the presence of intestinal-type acidic mucins and goblet cells (Fig. 1A). Using immunohistochemical staining, we found that these areas of intestinal metaplasia expressed villin protein (Fig. 1B). To examine whether villin also correlated with intestinal metaplastic changes in a H. pylori-infected animal model, mice were infected for 14 months with the mouse-adapted SS1 H. pylori strain. Using immunofluorescence, we showed that the villinpositive cells were located in the corpus (Fig. 2). Small intestine from the same mouse was used as a positive control for the villin antibody ( Fig. 2A). Because normal corpus does not express villin (14), this acid-secreting portion of the stomach was used as a negative control for the immunofluorescence (Fig.  2B). The parietal cells were detected using an antibody to the ␤ subunit of the H ϩ , K ϩ -ATPase. When compared with the agematched control, there was a decrease in the number of parietal cells in the 14-month-old H. pylori-infected animal coincident with an increase in villin-positive cells (Fig. 2C). We concluded that there was an increase in villin expression with intestinal metaplastic changes in both human and mouse H. pylori-infected stomachs.
Because villin expression directly correlated with the expression of intestinal metaplasia in both human and mouse during chronic atrophic gastritis, we examined whether endogenous villin was regulated by H. pylori in a human gastric cell line, NCI-N87. The NCI-N87 cell line expresses several gastric peptides, e.g. intrinsic factors, pepsinogens (26). We found that these cells also express villin (Fig. 3). These cells were co-cultured with H. pylori strains SS1, J99, and an organism not known to induce gastritis, i.e. C. jejuni (Fig. 3). The results demonstrated that villin protein increased with the J99 H. pylori strain but not with C. jejuni or the mouseadapted SS1 strain. Because the NCI-N87 cells transfect poorly, the AGS cell line was used in the subsequent transient transfection studies.
Prior studies have demonstrated that there are critical tissue-specific regulatory elements in the villin promoter between Ϫ1236 and Ϫ554 and between Ϫ554 and Ϫ446 (21). In addition, the first intron synergizes with the first 554 bp of the 5Ј-flanking sequence. We used the reporter construct containing 554 bp of the human villin promoter and the first intron to study regulation by H. pylori. We examined induction of the villin promoter for 24 h by three different bacterial species, H. pylori, Acinetobacter lwoffi, and C. jejuni (Fig. 4). The results showed greater induction with the H. pylori J99 strain compared with C. jejuni or A. lwoffi at 100 m.o.i. C. jejuni is a luminal pathogen that does not cause gastritis, whereas, A. lwoffi does induce gastritis (30).
We found that co-culturing the transient transformants with H. pylori stimulated the villin reporter over 6-fold at a bacterial concentration of 200 m.o.i. (Fig. 5A). Maximal stimulation occurred within 24 -30 h (data not shown). To assess whether activation of the villin promoter was Mek-1-Erk-dependent, Erk kinase-deficient mutants and the Mek-1 inhibitor PD98059 were used to block villin induction by H. pylori. Dominant negative Erk-1 and -2 constructs were co-transfected with the villin reporter prior to co-culturing with H. pylori (Fig.  5B). Consistent with dominant negative inhibition, PD98059

FIG. 2. Expression of villin in mouse stomach after H. pylori infection.
Paraffin-embedded sections from a 14-month-old H. pylori SS1-infected mouse. Fluorescein isothiocyanate-conjugated IgG was used to detect villin antibody staining (green), and a Texas Red-conjugated IgG was used to detect antibody raised against the ␤ subunit of H ϩ , K ϩ -ATPase (red). DAPI was used to stain nuclei (blue). A, enterocytes of the mouse small intestine used as a positive control stained positive for villin (block arrowhead). Magnification, 100ϫ. B, no villin expression was detected in the corpus of a non-infected 14-month-old animal. Shown is a typical region of the corpus glands with abundant parietal cells (red). Magnification, 100ϫ. C, the corpus of a 14-month-old infected C57BL/6 mouse revealed increased villin expression (green, white arrowhead) and a decrease in number of parietal cells (red, white arrows). Magnification, 400ϫ. Representative of three uninfected and three H. pylori-infected mice. FIG. 1. Villin expression in H. pylori-infected human stomach. Paraffinembedded sections from human stomach were used to detect villin expression. A, periodic acid-Schiff procedure/Alcian blue stain of stomach from a H. pylori-infected subject. Goblet cells indicative of intestinal metaplasia are indicated with an arrowhead. B, the same section was stained with villin antibody. The arrow indicates villin staining of intestinal metaplastic cells. Goblet cells are indicated with an arrowhead. Magnification, 400ϫ. also blocked villin induction by H. pylori (Fig. 5C). The p38 inhibitor SB203580 did not block villin induction (Fig. 5D). This indicated that the effect of H. pylori on the villin promoter specifically used the Erk pathway (Fig. 5D). To further document the role of the Mek-1-Erk pathway, an immunoblot was performed with phosphorylated Erk antibody and confirmed that co-culturing with H. pylori resulted in an increase in activated Erk protein (Fig. 6). We also found that H. pylori also stimulated p38 phosphorylation as reported previously (31). However, when coupled with the fact that SB203580 did not affect induction by H. pylori, we concluded that the p38 pathway was not upstream of villin activation. These results were consistent with prior studies implicating the Ras-Erk kinase pathway in H. pylori induction of the chemokine interleukin-8 (32). Moreover, it has recently been confirmed that H. pylori activates both the Erk-1 and p38 Map kinase pathways but that only the Erk-1 pathway regulates the transcription factor Egr-1 and its downstream targets, e.g. ICAM-1 (33).
We examined the villin promoter for likely targets of the Mek-1-Erk pathway and identified a cluster of Elk-1 sites adjacent to an SRE at Ϫ85 bp upstream from the cap site (Fig.  6A). Two adjacent elements consisting of the CArG box and an Ets binding site (EBS) form the SRE (34). The constitutive SRF forms a homodimer and binds the CArG box, which in turn stabilizes the binding of inducible Ets proteins, e.g. Elk-1, SAP-1, Net, collectively known as ternary complex factors. One of the Ets proteins will bind the short Ets site and make contact with the constitutively bound SRF homodimer to form the ternary complex (35,36). To determine whether Elk-1 and SRF binding to the Ϫ85 villin SRE was required for H. pylori induction, we introduced point mutations into the element (Fig. 7A). Dominant negative Elk-1 was used to examine the contribution of Elk-1 to H. pylori induction of the promoter. The results showed that point mutations within the Ϫ85 SRE/EBS site diminished induction by H. pylori (Fig. 7B). Moreover, co-transfection with dnElk-1 reduced both basal and inducible activation of the promoter by H. pylori, confirming that this factor is required for the induction.
It is known that Elk-1 binding is activated by translocation of Erk kinase to the nucleus and phosphorylation of this transcription factor (37). Therefore, to determine whether Elk-1 was phosphorylated during co-culture with H. pylori, an immunoblot was performed. The results showed that Elk-1 was phosphorylated at 3 and 24 h of treatment with H. pylori (Fig. 8). As a control for H. pylori induction, we examined the induction of Elk-1 by interleukin-1␤ (Fig. 8B). In both cases, induction of phosphorylated Erk accompanied Elk-1 protein induction, indicating that the increase was due not only to an increase in the phosphorylated form but also because of an increase in the amount of protein. Thus, we concluded that the mechanism of H. pylori activation of the villin promoter was by Mek-1 activation of Erk that, in turn, phosphorylates Elk-1.
Phosphorylation of Elk-1 facilitates its binding to DNA and cooperation with SRF (38). Therefore, electrophoretic mobility shift assays were performed to determine whether co-culturing with H. pylori increased Elk-1 binding to the inducible villin element. AGS nuclear extracts were prepared after incubating the cells with the bacteria for 0, 3, and 24 h. In untreated extracts, there was binding of YY1 and SRF and no binding of Elk-1 (Fig. 9). An increase in DNA binding to the inducible villin element was observed after 3 h of incubation with the bacteria. By 24 h, there was increased SRF binding in addition to the binding by Elk-1. Anti-GATA-4 antibody was used as a control and did not shift any of the complexes. The identity of the bound complexes was confirmed in the untreated and 24h-treated samples by supershifting with specific antibody (Fig.  9). Moreover, the specificity of the complexes at each time point was confirmed by competition with the unlabeled element, which competed for all three of the major complexes. Collectively, the electrophoretic mobility shift assays revealed that YY1 and SRF bind to the element in untreated extracts (Fig.  10). After exposure to bacteria for 3 h, Elk-1 binds and YY1 binding is reduced. There is also a decrease in SRF, which correlates with greater occupancy of the element by the induced binding of Elk. By 24 h, both SRF and Elk-1 binding is prominent and YY1 binding remains low, suggesting sustained occupancy of this element by these inducible factors (Fig. 10).

DISCUSSION
In this study, we showed that induction of endogenous villin in the stomach occurs in atrophic human and mouse stomach in response to chronic H. pylori infection. These results are consistent with the concept that villin is an indicator of this preneoplastic lesion in the stomach and esophagus (10). Despite numerous reports that villin expression develops in tissues responding to chronic injury, little is known regarding the cis-acting elements that control its inducible expression. In the small intestine, villin is expressed during fetal development. Specific cis-acting transcriptional domains within the 6.7-kb promoter are responsible for regulating intestinal villin expression vertically (crypt versus villus) and horizontally (duodenum versus cecum) prior to birth (21). Based on the results reported here, this promoter also contains the elements capable of mediating regulation of villin expression in gastric cells.
In this study, we found that villin was expressed in vivo in both human subjects and in a mouse model infected with H. pylori. The expression clearly correlated with the development of intestinal metaplasia in the human stomach. Although true goblet cells are not consistently observed in mouse models of H. pylori infection, we show here that villin expression was clearly induced with the chronic infection. Nevertheless, in vivo studies cannot distinguish between regulation of villin expression due to the organism alone, due to inflammation, or the atrophic epithelium. To address the effect of the organism on villin expression, we performed a series of co-culture experiments. These studies revealed that indeed H. pylori was sufficient to stimulate villin expression and that the inducible expression required a SRE. These results are consistent with other reports indicating that SREs mediate regulation by the H. pylori CagA protein (39). However, the results shown here are the first to document regulation of an SRE by H. pylori within a native promoter and more specifically a reporter that correlates with a pre-neoplastic state. Although both Elk-1 and SRF inducibly bind to the SRE sites, multiple point mutations of the Ϫ85 element as well as a deletion did not completely abolish induction of the promoter. Therefore, we expressed kinase-deficient Elk-1 to abolish any contribution of Elk to the induction. This result was quite striking in that the dominant negative construct reduced the basal activity of the Ϫ554 reporter construct and completely blocked induction by H. pylori. Despite inducible binding of SRF to this element, we concluded that inducible activation and binding of Elk-1 plays a critical role in the regulated expression of villin and, presumably, the intestinal metaplastic phenotype of the stomach.
Interestingly, the binding of both Elk-1 and SRF was sustained and did not return to the base line even after 24 h of contact with the bacteria, suggesting continued activation of the promoter or possibly setting the stage for further cooperativity with other factors involved in chronic expression of villin. These other factors may be involved histone modifications through the recruitment of histone deacetylases or histone acetyltransferases. In fact, it has been shown that the ternary complex interacts with repressor complexes that include histone deacetylases and activator complexes, which include histone acetyltransferases (34). The organization of the villin SRE is also flanked by at least three EBSs, which may add to the complexity of how these factors bind and regulate villin. Given that Elk-1 and SRF binding sites are known to overlap and exhibit cooperativity, the combination of the expression and DNA binding studies suggest that SRF may act as a scaffold protein. This notion is consistent with other promoters exhibiting cooperative binding between Elk-1 and SRF (34). YY1 acts as both a repressor and activator through the recruitment of transcriptional co-regulators and chromatin interactions (40). Given the decrease in YY1 binding with H. pylori, it would appear that the predominant activity of YY1 on the villin promoter is as a repressor.
An important aspect of these results is the connection between Elk-1 activation and the pre-neoplastic condition of the stomach. It is clear that Elk-1 activation is regulated by the Mek-1-Erk pathway. Interestingly, Tarnawski and co-workers (41) showed that non-steroidal anti-inflammatory agents (NSAIDS), which inhibit Cox 2, also inhibit Erk activation in FIG. 9. H. pylori stimulates nuclear protein binding. After 48 h of serum starvation, H. pylori J99 bacteria were co-cultured with AGS cells. Electrophoretic mobility shift assays were performed with 6 g of nuclear extracts after no treatment (Un) and 3 and 24 h of co-culture with the bacteria (lanes 1-3). Lane 4 contains the untreated extracts competed with the wild type element (WT). Lanes 5-8 contain extract from the 3-h-treated cells incubated with Elk-1, SRF, YY1, and GATA-4 antibody. The major complexes Elk-1, SRF, YY1 and supershifted SRF (SS) are indicated. FIG. 10. H. pylori stimulates Elk-1 and SRE binding. After 48 h of serum starvation, H. pylori J99 bacteria were co-cultured with AGS cells. Electrophoretic mobility shift assays were performed with 6 g of nuclear extracts after no treatment (Un) and 24 h of co-culture with the bacteria. A, the extract from the untreated cells (lane 1) was incubated with unlabeled oligonucleotide (WT)(lane 2) and YY1, SRF, Elk, or GATA-4 antibodies (lanes 3-6). B, similarly, extract from cells co-cultured with bacteria for 24 h (lane 1) was competed using the wild type oligonucleotide (lane 2), YY1, SRF, Elk-1, and GATA-4 antibodies (lanes 3-6). the stomach (41). This is consistent with the knowledge that NSAIDS exhibit potent effects in the stomach, i.e. ulcer formation and erosion. Moreover, prostaglandin E2, a product of elevated Cox 1,2 activity, phosphorylates the epidermal growth factor receptor and activates the Erk-2 signaling pathway in the colon (42). Thus, it is reasonable to consider that increased Erk-2 signaling in the stomach correlates with heightened cell proliferation. Because the results here demonstrate that induction of villin lies downstream of the Erk-Elk-1 pathway, its expression in the stomach signals is a marker of increased cell growth. Although Helicobacter may be only one of several triggers that increase epithelial proliferation through the Erk-Elk-1 pathway, clearly there is the potential for activating a variety of genes involved in increasing the proliferative rate in the stomach.
Villin, as a proliferative marker in the stomach, will be useful in identifying relevant signaling pathways and transcription factor networks during the early stages before gastric transformation. Despite microarray studies of the human stomach, specific transcription factors have yet to be discovered (43). Thus, the signaling networks revealed here will, hopefully, become important tools to employ in dissecting the molecular steps leading to gastric transformation and, eventually, uncovering more effective and specific suppressive therapy.