Lipopolysaccharide Initiates a Positive Feedback of Epidermal Growth Factor Receptor Signaling by Prostaglandin E2 in Human Biliary Carcinoma Cells1

Bacterial products (e.g., LPS) are viewed as critical stimuli in inflammation-associated cancer. Cyclooxygenase 2 (COX-2), a major effector of LPS, and EGFR, are key to carcinogenesis, notably in the hepatobiliary tract. In this study, we tested the hypothesis that LPS can initiate an interaction between the epidermal growth factor receptor (EGFR) and COX-2 pathways. We examined the effect of LPS in biliary carcinoma cells that displayed constitutive COX-2 expression and PGE2 production and in normal human biliary epithelial cells in which COX-2/PGE2 expression was virtually absent. LPS induced early phosphorylation of EGFR and ERK1/2 in both types of cells, which reached maximum levels within 30 min (first phase). However, only the carcinoma cells showed a second significant rise in both EGFR and ERK phosphorylation 6 h after exposure to LPS (second phase). Inhibition of COX-2/PGE2 production prevented the second, but not the first, phase of EGFR and ERK1/2 phosphorylation, implicating COX-2/PGE2 in the second phase of phosphorylation. LPS induced COX-2-derived PGE2 production at 4 h, which was before the rise in the second phosphorylation that occurred at 6 h. Exogenous PGE2 also caused EGFR activation via a signaling pathway involving TACE-dependent TGF-α release. Inhibition of the second phase of EGFR phosphorylation with EGFR or COX-2 inhibitor prevented LPS-induced cell invasion in vitro, demonstrating the biological importance of this COX-2 feedback signaling in cancer cells. We conclude that LPS triggers a positive feedback loop involving COX-2/PGE2 in biliary carcinoma cells and that this second phase of EGFR phosphorylation is implicated in cell invasion by LPS.

I nflammation and cancer are often associated and share some common signaling pathways (1). Cyclooxygenase 2 (COX-2), 4 the inducible form of cyclooxygenase, is important in both conditions (2). COX-2 is highly expressed in several types of carcinomas but not in normal biliary and other epithelial cells. Cyclooxygenase enzymes catalyze the rate-limiting reaction in the conversion of arachidonic acid to PGs (3). The production of COX-2-derived PGE 2 may occur in response to the stimulation of the epidermal growth factor receptor (EGFR) and ERK cascade (4). In contrast, COX-2-derived PGE 2 is capable of causing EGFR activation (5)(6)(7)(8). Despite evident communication between the two signaling pathways, the precise mechanism(s) underlying this communication and the pathophysiological implications remain incompletely explored. In this study, we investigated the signaling pathways involving COX-2/PGE 2 and EGFR activated by LPS and their interactions in human biliary epithelial cells.
LPS induces COX-2 and PGE 2 production in a variety of epithelia, notably in the digestive and hepatobiliary tracts (9 -12). LPS is a major integral component of the outer membrane of Gram-negative bacteria and is one of the most potent stimuli of inflammation. Epithelia provide the first line of defense against invading pathogens and display expression of TLR4, the cognate receptor of LPS (13,14). Among different signaling pathways induced by LPS, activation of an EGFR cascade has been identified previously in the respiratory epithelium (15) and recently in biliary epithelial cells (12). Thus, LPS has the ability to stimulate both the EGFR and COX-2/PGE 2 pathways in epithelial cells.
In inflammatory biliary tract diseases and associated carcinomas, the epithelium is exposed to high concentrations of LPS (14,16,17). We hypothesized that the exposure of some biliary epithelial cells to LPS initiates a signaling cascade leading to EGFR activation, which subsequently induces COX-2-dependent PGE 2 production, resulting in a second phase of EGFR activation, reinforcing and amplifying LPS-induced signaling (a positive feedback loop). The experiments were conducted in both normal human biliary epithelial cells and biliary carcinoma cell lines (Mz-ChA-1 and KMBC cells). The novel feedback loop and the finding that this loop is important in cell invasion in response to an inflammatory stimulus (LPS) suggest that this feedback signaling pathway may play an important role in the pathophysiology of biliary carcinomas and inflammatory biliary diseases.

Immunoblotting and EGFR phosphorylation assay
After various treatments, cells were lysed on ice in PBS lysis buffer containing 25 mmol/L Tris-HCl, 300 mmol/L NaCl, 1 mmol/L CaCl 2 , 1% Triton X-100 (pH 7.4), and protease inhibitor (Complete Mini; Roche). Lysates were precleared by centrifugation at 14,000 rpm for 20 min at 4°C. Protein concentration was determined by the bicinchoninic acid-based BCA Protein Assay kit (Pierce/Perbio Science France). Fifty micrograms of proteins was subjected to 7.5% SDS-PAGE electrophoresis and transferred to a polyvinylidene difluoride membrane (Bio-Rad), which was blocked with 5% BSA, probed with primary Abs, washed with PBS, and then probed with secondary Abs conjugated to HRP. Immunoreactive bands were visualized by ECL using an ECL kit (Amersham Biosciences). The detected bands were quantified with the National Institutes of Heath image software.
EGFR phosphorylation in some studies was also analyzed using the PhosphoDetect EGFR phosphorylation ELISA kit (Calbiochem) according to the manufacturer's instructions.

PGE 2 production assay
Cells were incubated in serum-free medium (SFM) and then exposed or not to LPS (10 g/ml) for various times. In inhibitory studies, the cells were preincubated with inhibitors for 30 min to 2 h before LPS was added. At the end of the experiments, cell supernatants were collected and the PGE 2 concentration was determined by ELISA kit (Cayman Enzyme Immunoassay Kit) according to the manufacturer's instructions. The amount of PGE 2 production in each sample was related to total protein in cell lysate and was expressed as ng of PGE 2 per mg of total cellular protein.

TGF-␣ shedding assay
The shedding (cleavage and release) of TGF-␣ was analyzed in human biliary epithelial cells incubated with PGE 2 (10 ⌴) or LPS (10 g/ml) for 2 h. To prevent soluble TGF-␣ from binding to EGFR, an EGFR-neutralizing Ab (4 g/ml) was added 2 h before PGE 2 or LPS. In inhibitory studies, the cells were preincubated with inhibitors for 30 min to 2 h before PGE 2 or LPS was added. At the end of the experiments, cell supernatants were collected and TGF-␣ was measured using the TGF-␣ ELISA kit (R&D Systems) according to the manufacturer's instructions.

Cell invasion assay
Cell invasion through a three-dimensional extracellular matrix was assessed by a Matrigel invasion assay using BD Matrigel Invasion Chambers (BD Biosciences Biocoat) with 8.0-m filter membranes. Cells (2.5 ϫ 10 4 ) resuspended in 200 l of SFM or SFM supplemented with 10 g/ml of LPS were plated onto each filter, and 750 l of SFM or SFM supplemented with 10 g/ml of LPS was placed in the lower chamber. After 24 h, filters were washed, fixed, and stained with Coomassie brilliant blue. Cells on the upper surface of the filters were removed with cotton swabs. Cells that had invaded to the lower surface of the filter were counted under the microscope.

Statistical analyses
All statistical analyses were performed with the StatView 5.01 software (SAS Institute). The paired Student t test was used to analyze statistical differences between groups. In the time course analysis, comparison among groups was also performed using Newman-Keuls multiple range test. Differences of p Ͻ 0.05 were considered statistically significant.

Expression of COX-2 and production of PGE 2 in biliary epithelial cells
The expression of COX-2 was examined in cell cultures of the human biliary carcinoma cell line Mz-ChA-1 and compared with normal human biliary epithelial (NHBE) cells. Although COX-2 was not detected by Western blot in NHBE cells, strong expression was demonstrated in Mz-ChA-1 carcinoma cells (Fig. 1A). Whereas COX-2derived PGE 2 protein production was barely detectable in the supernatants of NHBE cells, a high level of PGE 2 was produced by Mz-ChA-1 carcinoma cells (Fig. 1B). Thus, biliary cancer cells overexpress COX-2 and up-regulate PGE 2 production.

LPS triggers different temporal patterns of EGFR and ERK1/2 activation in biliary carcinoma vs normal epithelial cells
LPS is known to stimulate both the EGFR and COX-2/PGE 2 signaling pathways (15,20). Because NHBE and Mz-ChA-1 cells display different expression patterns of COX-2/PGE 2 , we examined the effect of LPS on the temporal activation of an EGFR cascade: incubation of NHBE cells with LPS increased EGFR phosphorylation, which occurred as early as 15 min, was maximal at ϳ30 min, and decreased by 2 h (Fig. 2A). Because ERK1/2 plays important roles in multiple cellular functions and is one of the most common effectors downstream of EGFR, we also FIGURE 1. Expression of COX-2 and production of PGE 2 in normal biliary epithelial cells and in biliary carcinoma cells. A, NHBE cells (NHBEC) or a biliary carcinoma cell line, Mz-ChA-1 cells, were grown to confluence. Cell lysates were collected to detect COX-2 protein expression by Western blot using anti-COX-2 Ab. ␤-Actin was used as an internal control, showing equal loading of protein samples. Results are representative of three different preparations. B, NHBE cells and Mz-ChA-1 cells were grown to confluence and were then serum (or serum substitute) starved for 24 h. Cell supernatants were collected to measure concentrations of PGE 2 protein using PGE 2 ELISA as described Materials and Methods. PGE 2 protein was normalized to total protein in cell lysates. Values represent means Ϯ SEM of three different experiments performed in duplicate. ‫,ء‬ p Ͻ 0.05, compared with NHBE cells. examined ERK1/2 activation by LPS. ERK1/2 displayed a pattern of activation similar to EGFR, although after 30 min, ERK1/2 phosphorylation showed more variability before returning to baseline (Fig. 2B). However, LPS stimulation of Mz-ChA-1 carcinoma cells induced a biphasic temporal pattern of EGFR (Fig. 2C) and ERK1/2 ( Fig. 2D) activation, with a rapid increase in phosphorylation within 30 min, followed by a second rise at ϳ6 h. Western blot analyses with Abs that recognize total EGFR and ERK proteins showed that EGFR and ERK1/2 protein levels remained unchanged in both normal epithelial cells and carcinoma cells throughout the experiments. Next, we examined whether LPS induced the first phase of EGFR activation via binding to its putative receptor, TLR-4, in the plasma membrane. We preincubated cancer cells with neutralizing Abs that block the ligand binding site on the extracellular domain of TLR4 and compared with TLR5 (used as a control). Preincubation of the cells with TLR4 Ab but not TLR5 Ab inhibited LPS-induced EGFR phosphorylation (Fig. 2E), confirming the role of TLR4 in LPS-induced responses. To investi-gate whether this LPS-induced biphasic phosphorylation of EGFR is limited to a specific cancer cell line (i.e., Mz-ChA-1), we examined two other cancer cell lines, biliary carcinoma KMBC cells and lung adenocarcinoma A549 cells, which were previously reported to express COX-2 (21,22). Both cancer cell lines showed a similar pattern of biphasic EGFR phosphorylation in response to LPS stimulation ( Fig. 2F; data not shown for A549 cells). Thus, these studies show that LPS induces different temporal patterns of EGFR and ERK1/2 activation in these cancer vs normal biliary epithelial cells.

The second phase of LPS-induced EGFR and ERK1/2 activation is dependent on COX-2 signaling in biliary carcinoma cells
LPS can induce EGFR activation in lung cancer cells via TNF-␣converting enzyme (TACE)-dependent release of the EGFR ligand TGF-␣ (15). However, the mechanism(s) of the second phase of EGFR activation by LPS is unknown. Given that LPS can induce PGE 2 production via COX-2 and that PGE 2 can trigger EGFR activation (5), we hypothesized that the second phase of EGFR FIGURE 2. Effects of LPS on the activation of EGFR and ERK1/2 in biliary epithelial cells. A and B, NHBE cells were incubated with LPS (10 g/ml) for various times (0 -10 h), and the phosphorylations of EGFR and ERK1/2 was analyzed using Western blotting. Phosphorylations of EGFR and ERK1/2 were detected by immunoblotting with an anti-phospho-EGFR Ab (Tyr 992 , A) and with an anti-phospho-ERK1/2 Ab (B), respectively. Total EGFR and ERK1/2 protein at each time point were detected by Western blot using anti-EGFR or anti-ERK1/2, respectively. Quantification of phospho-protein is expressed as the ratio between the phosphorylated forms and total protein levels of EGFR or ERK1/2. Results are expressed as percentage above the basal levels (without LPS). Data represent means Ϯ SEM (n ϭ 5). ‫,ء‬ p Ͻ 0.05, compared with the basal levels (without LPS). C and D, Mz-ChA-1 cells were incubated with LPS (10 g/ml) for various times (0 -10 h), and the phosphorylation of EGFR (C) and ERK1/2 (D) was analyzed using Western blot as described above. Quantification of phospho-protein is expressed as the ratio between the phosphorylated forms and total protein levels of EGFR or ERK1/2. Results are expressed as percentage above the basal levels (without LPS). Data represent means Ϯ SEM (n ϭ 5). ‫,ء‬ p Ͻ 0.05, compared with the basal levels (without LPS). ‫,ءء‬ p Ͻ 0.05, compared with the levels of phospho-proteins detected at 2 h after LPS stimulation. E, Mz-ChA-1 cells were preincubated with or without Abs blocking TLR4 or TLR5 (4 g/ml) for 1 h and then stimulated with LPS (10 g/ml) for 30 min. EGFR phosphorylation was analyzed using Western blotting as described above. F, Biliary carcinoma KMBC cells were incubated with LPS (10 g/ml) for various times (0 -10 h), and the phosphorylation of EGFR was detected using the Active EGFR Assay, an ELISA measuring the phospho-EGFR as described in Materials and Methods. Data are expressed as means Ϯ SEM (n ϭ 3). ‫,ء‬ p Ͻ 0.05, compared with the basal levels (time 0). activation in the carcinoma cells is caused by COX-2-derived PGE 2 production induced by LPS. According to this hypothesis, LPS stimulates EGFR phosphorylation (first phase, within 30 min), leading to COX-2 up-regulation and increased production of PGE 2 , which then induces EGFR activation (second phase, at and after 6 h). To test this hypothesis, we examined the temporal production of PGE 2 by LPS. In Mz-ChA-1 cells, LPS induced a significant increase in PGE 2 production at 4 h (Fig. 3A), which preceded the second phase of EGFR activation. Then, we examined the pathway of PGE 2 -induced EGFR activation, because NHBE cells contain receptors for PGE 2 but do not produce PGE 2 constitutively. We examined the effect of exogenously delivered PGE 2 on EGFR signaling: stimulation of NHBE cells with PGE 2 induced rapid EGFR phosphorylation (Fig. 3B). These results indicate that NHBE cells activate EGFR. They also suggest that the second phase of EGFR activation induced by LPS in cancer cells could be due to PGE 2 generation via COX-2. To test this hypothesis, we pretreated cancer cells with a selective COX-2 inhibitor (NS-398) that prevented PGE 2 production in the cancer cells (Fig. 3C). As expected, NS-398 prevented the second, but not the first, phase of LPS-induced activation of EGFR (Fig. 3D) and ERK1/2 (Fig. 3E). From these results, we conclude that LPS induces the second phase of EGFR activation via COX-2-dependent PGE 2 production in biliary cancer cells.

LPS induces PGE 2 production via an EGFR cascade in biliary carcinoma cells
Having determined that PGE 2 production is responsible for the LPS-induced second phase of EGFR and ERK1/2 activation, we examined the mechanism involved in LPS-induced PGE 2 production. Previously, we have shown that an EGFR cascade is involved in LPS-induced MUC5AC mucin and IL-8 production in human airway epithelial cells (15,23). Whether an EGFR cascade also mediates LPS-induced PGE 2 production in biliary cancer cells is presently unknown. Pretreatment of the cells with a selective EGFR tyrosine kinase inhibitor (AG1478) prevented LPS-induced PGE 2 production (Fig. 4), implicating EGFR activation in LPSinduced PGE 2 production. Preincubation of the cells with an EGFR-blocking Ab, which binds and occupies the ligand binding sites on the extracellular domains of EGFR, also prevented LPSinduced PGE 2 production, implicating ligand-dependent EGFR phosphorylation in the PGE 2 response (Fig. 4). PGE 2 production by LPS was also inhibited by the ERK1/2 inhibitor (PD98059), implicating ERK1/2 (Fig. 4). From these results, we conclude that LPS induces PGE 2 production in biliary cancer cells via EGFR ligand-dependent EGFR and ERK1/2 activation.

PGE 2 induces EGFR activation via a TACE-TGF-␣ signaling pathway
Next, we examined the epithelial pathway utilized by PGE 2 in signaling EGFR phosphorylation. Mz-ChA-1 carcinoma cells express COX-2 and PGE 2 constitutively, whereas the expressions are minimal or absent in normal biliary epithelial cells. To avoid constitutive COX-2 and PGE 2 effects, we chose normal biliary epithelial cells for these experiments and we stimulated the cells exogenously with PGE 2 . To determine whether PGE 2 -induced EGFR activation involves ligand binding to EGFR, we pretreated the cells with an EGFR-blocking Ab. This pretreatment suppressed PGE 2induced EGFR phosphorylation (Fig. 5A), implicating ligand-dependent EGFR activation in the response to PGE 2 . Pretreatment of the cells with a neutralizing Ab to the EGFR ligand TGF-␣ also suppressed PGE 2 -induced EGFR phosphorylation (Fig. 5A). Preincubation with an EGF-neutralizing Ab also reduced PGE 2 -induced EGFR phosphorylation slightly, but the inhibition was not significant (Fig. 5A). These results suggest that TGF-␣ plays a major role in PGE 2 -induced EGFR phosphorylation in biliary epithelial cells.

COX2/PGE 2 -mediated second phase of EGFR phosphorylation by LPS is involved in cancer cell invasion
Having determined the two phases of LPS-induced EGFR phosphorylation, we proceeded to examine the biological significance of the biphasic EGFR phosphorylation. We chose to perform in vitro cell invasion studies for the following reasons: 1) LPS exposure is implicated in tumor progression in animal models (25,26) and 2) cell invasion is a critical step in tumor progression. First, we examined whether LPS can increase cancer cell invasion in vitro: Exposure of Mz-ChA-1 cancer cells to LPS for 24 h induced cell invasion significantly (Fig. 6A). Then, we examined the role of EGFR phosphorylation in LPS-induced cell invasion. We treated cells with AG1478 at various time points before and after LPS stimulation. Pretreatment with AG1478 for 1 h prevented LPS-induced cell invasion significantly; treatment starting at 4 h after LPS stimulation, which preserved the first phase but prevented the second phase of the EGFR phosphorylation, also inhibited LPS-induced cell invasion significantly. Treatment with AG1478 starting at 12 h after LPS stimulation (2-4 h after the second phase EGFR phosphorylation) did not have significant inhibition of LPS-induced cell invasion (Fig. 6A). These results suggest that the second (COX-2/PGE 2 ) phase of EGFR phosphorylation is important in LPS-induced cell invasion. For inhibition studies, the cells were incubated with LPS (10 g/ml) and then treated with or without the EGFR inhibitor AG1478 (AG; 10 M) added at various time points (0, 4, and 12 h after LPS treatment) or the COX-2 inhibitor NS-398 (NS; 10 M, added with LPS). Cell invasion was determined after 24 h of LPS treatment. Data are means Ϯ SEM (n ϭ 3). ‫,ء‬ p Ͻ 0.05, compared with LPS alone. B, A proposed model of LPS-induced positive feedback loop between the EGFR and the COX-2/PGE 2 signaling pathways in biliary carcinoma cells. LPS induces 1) a ligand-dependent EGFR phosphorylation, initiating MAPK signaling to the nucleus that subsequently causes COX-2-derived PGE 2 production. The proposed surface signaling pathway responsible for this initial activation of EGFR is based on its demonstration in airway epithelial cells (15). 2) The secreted PGE 2 binds to its cognate receptor, resulting in the activation of an epithelial cell surface signal, leading to the activation of the cell surface metalloprotease TACE and subsequent cleavage of pro-TGF-␣ into soluble TGF-␣. TGF-␣ binds to and activates EGFR, resulting in a second wave of EGFR and ERK1/2 phosphorylation.
To further examine the role of the second phase of EGFR phosphorylation on cell invasion, we pretreated the cells with the COX-2 inhibitor NS-398 to prevent the second (COX-2/PGE 2dependent) phase. This pretreatment also inhibited LPS-induced cell invasion significantly (Fig. 6A), implicating COX-2/PGE 2 in LPS-induced cell invasion. Together, these results show that in addition to the effects of the first phase of EGFR phosphorylation, subsequent activation of the COX-2/PGE 2 -dependent second phase of EGFR phosphorylation is important in cell invasion in response to LPS.

Discussion
In the present study, we show that in biliary carcinoma cells, LPS, a pathophysiological stimulus, induces biphasic EGFR phosphorylation: the early phase occurs at ϳ30 min after stimulation, involving TACE-dependent EGFR ligand release; the delayed phase occurs at ϳ6 h after stimulation via a signaling pathway involving COX-2/PGE 2 -dependent TACE activation and EGFR ligand release. However, in normal biliary epithelial cells, LPS induces only the early phase of EGFR activation due to the lack of COX-2 expression and PGE 2 production. We propose a novel model showing the pathways mediating this biphasic EGFR activation (Fig. 6B).
Our current studies extend the previous discovery that LPS can induce EGFR phosphorylation (30 min) via a TACE-TGF-␣ cascade in human airway epithelial cells (15) to biliary epithelial cells. In this study, we report that LPS can induce a second phase of EGFR phosphorylation, which occurs at ϳ6 h after the stimulation. We found that COX-2-dependent PGE 2 production is responsible for the second but not the first phase of LPS-induced EGFR phosphorylation.
LPS induces COX-2-derived PGE2 production in various epithelia, including enterocytes, hepatocytes, and biliary epithelial cells (9 -12). Recently, Fukata et al. (12) showed that LPS induced EGFR phosphorylation (at 30 min) in a human intestinal epithelial cell line (SW480) and suggested that this is caused by LPS-induced COX-2 up-regulation and PGE 2 production that did not require EGFR activation. However, they also showed that LPS did not induce COX-2 up-regulation significantly until 4 h after stimulation, suggesting that COX-2/PGE 2 is unlikely to be responsible for the EGFR phosphorylation that they found at 30 min. Our studies found that the COX-2 inhibitor NS-398 inhibited the second but not the first phase of EGFR phosphorylation, confirming our hypothesis. In this study, we also found that LPS enhanced PGE 2 production in biliary carcinoma (Mz-ChA-1) cells, which express COX-2 constitutively, via an EGFR signaling cascade. In Mz-ChA-1 cells, LPS-induced PGE 2 production was prevented completely by inhibitors of EGFR and ERK1/2. These findings indicate that COX-2-derived PGE 2 synthesis caused by LPS is mediated by the activation of an EGFR signaling cascade. It was previously reported that EGFR activation induces COX-2 expression and PGE 2 production (4,8). In this study, we show that activation of EGFR and subsequent ERK1/2 phosphorylation occurs within 30 min after LPS stimulation, in agreement with the temporal pattern of LPS-induced EGFR phosphorylation observed in other epithelial cells (15). This activation in carcinoma cells results in PGE 2 production, which is maximal 4 h after exposure to LPS, preceding the second phase of EGFR activation.
In relation to the second phase of EGFR activation, we investigated the response to PGE 2 in normal biliary epithelial cells. These cells do not express COX-2 constitutively, which allowed us to ascertain specifically the effect of exogenous PGE 2 . It was previously established that PGE 2 exerts its biological actions via specific G protein-coupled receptors (GPCRs) (EP-1, -2, -3, -4) (27).
It was initially proposed that EGFR transactivation was mediated via an intracellular signaling pathway (ligand-independent EGFR phosphorylation), based on the rapid kinetics of the transactivation signal and the absence of detectable levels of soluble EGFR ligands (28). However, recent studies have shown unequivocally that GPCR activation leads to the cleavage of EGFR proligand, allowing the secreted soluble ligand to bind EGFR (5,29), although the metalloprotease responsible for the proteolytic cleavage event was not determined. In this study, we show that PGE 2 , a GPCR stimulus, activates the EGFR ligand by causing the activation of the metalloprotease TACE, which cleaves EGFR ligands, shown here to consist, at least in part, of TGF-␣.
The EGFR ligand TGF-␣ is synthesized as a transmembrane precursor molecule that requires proteolytic cleavage by transmembrane metalloproteases known as ADAMs. TACE, an ADAM family member, causes the ectodomain shedding of TGF-␣ (30,31). We found that PGE 2 -induced TGF-␣ release was blocked by a metalloprotease inhibitor, GM6001, and by a TACE inhibitor, TAPI-1 (24). Together, these data indicate that PGE 2 initiates an epithelial cell surface signaling cascade involving TACE/TGF-␣-EGFR phosphorylation.
The major contribution of the present work is the demonstration that the two phases of EGFR activation described above are integrated in a positive feedback loop initiated by the bacterial product LPS. We further document that this loop is present in carcinoma cells (probably a major mechanism making these cancer cells different from the normal cells). Our experiments were designed to analyze EGFR phosphorylation over a prolonged time period in two types of cells (i.e., normal cells vs carcinoma cells) incubated with LPS. They allowed us to show that in biliary carcinoma cells, LPS caused two phases of EGFR phosphorylation. Because the initial transient activation of an EGFR cascade occurred before (and was responsible for) the activation of the COX-2/PGE 2 pathway, we suggested that the second phase of EGFR phosphorylation could be the consequence of COX-2/PGE 2 activation. This assumption is based on the fact that this late phase response was prevented by a selective COX-2 inhibitor in the carcinoma cells. That PGE 2 release was the cause of the second phase of EGFR phosphorylation was further supported by the observation that the normal cells, which do not produce PGE 2 constitutively, lack a second phase of EGFR phosphorylation. The temporal correlation between LPS-induced PGE 2 release and the occurrence of the second phase of EGFR activation is also fully consistent with this mechanism. Importantly, LPS induced an EGFR-mediated positive feedback loop involving COX-2/PGE 2 production, causing rephosphorylation of EGFR, resulting in a significant stimulation of cell invasion in these cancer cells, a critical aspect of tumor progression. Future studies will address a variety of potential outcomes of this important feedback pathway, including carcinogenic effects and effects on cell differentiation. Thereby, this signaling cascade can provide further insights into the pathophysiology of epithelial inflammation and tumorogenesis.
The present results are in keeping with recent computational analyses showing that positive feedback loops allow cells to modulate the amplitude and the duration of signaling responses (32). They are of particular interest with respect to chronic inflammation and its relationship to cancer. Thus, LPS triggers a positive feedback loop involving COX-2/PGE 2 and the EGFR signaling cascade, which could contribute to the promotion of carcinogenesis, notably in the biliary tract where carcinoma generally arises in a background of chronic inflammation.

Disclosures
The authors have no financial conflict of interest.