Endotoxin/Lipopolysaccharide Activates NF-κB and Enhances Tumor Cell Adhesion and Invasion Through a β₁ Integrin-Dependent Mechanism Free

The establishment of metastases in distant organs is dependent upon the adherence of shed circulating tumor cells to the endothelium lining the microvessels in target organs, attachment to the subendothelial basement membrane, invasion into the extracellular matrix by local proteolysis, and, finally, proliferation and angiogenesis (1, 2). During these steps it is necessary for tumor cells to interact with endothelial cells and extracellular matrix in distant organs, and this interaction is dependent on the various adhesion molecules expressed on the tumor cell surface. Among these, the integrin-mediated interaction between tumor cells and the subendothelial extracellular matrix is one of the most important determinants for organ-specific metastasis (3, 4).

Integrins are the most important family of cell surface adhesion molecules that mediate interactions between cells and the extracellular matrix (5). They are heterodimeric transmembrane receptors consisting of α and β subunits. Each subunit is a glycoprotein with a large extracellular domain and a relatively small cytoplasmic domain. The different integrin subfamilies are determined by the β subunit; for example, the β₁ subunit associates with different α subunits to form the β₁ integrin subfamily. To date >20 different integrin heterodimers are known, which bind to specific ligands present in the extracellular matrix or expressed on target cells. Members of the β₁ integrin subfamily primarily bind to components of the extracellular matrix, such as fibronectin, collagens, and laminins, but some of them also participate in direct cell-to-cell adhesion (5, 6). Among the different integrin heterodimers, β₁ and β₃ integrins appear to play a crucial role in regulating tumor cell proliferation, differentiation, adhesion, motility, and invasion (7, 8). Increased expression of different members of the β₁ integrin has been found to be associated with tumor cell survival, invasion, and tumor progression. Up-regulation of α₂β₁ integrin expression enhances tumor cell adhesion and prevents high dose epidermal growth factor-induced cell death (9). Overexpression of the α₃ subunit of β₁ integrin in the MDA-MB-231 breast carcinoma cell line is associated with the potent migratory and invasive properties of these cells, which is strongly inhibited by a specific function-blocking anti-α₃ mAb (10). Using a targeted elimination technique, it has been found that the expression of integrin α₆β₁ in MDA-MB-435 breast carcinoma cell line is essential for facilitating tumorigenesis and promoting tumor cell survival in distant organs in mice (11). Overexpression of α₆ subunit of the β₁ integrin has been found in human hepatocellular carcinoma with aggressive phenotypes (12). Recently, it has been shown that organ-specific sites of metastatic lesions are determined at least in part by β₁ integrin-mediated adhesion to and invasion into the subendothelial extracellular matrix, and furthermore, different metastatic behaviors of tumors correlate with β₁ integrin-mediated adhesive properties (4, 13).

LPS or endotoxin, a predominant glycolipid in the outer membrane of Gram-negative bacteria, provides a highly potent stimulus to cells of the immune system. LPS stimulates monocytes, macrophages, and neutrophils through the activation of transcription factors and protein kinases such as NF-κB and p38 kinase, resulting in an increased production of proinflammatory cytokines and overexpression of cell adhesion molecules (14, 15). LPS-induced cellular activation is mediated by its complexing with circulating LPS-binding protein (LBP)² and subsequent binding to CD14, a 55-kDa glycosylphosphatidylinositol (GPI) membrane-anchored glycoprotein (16, 17), which, in turn, facilitates intracellular transduction of LPS signaling through Toll-like receptor 4 (TLR4) (18, 19). It has been noted that rapid growth of previously dormant metastases occurs following surgical removal of a primary tumor (20). The mechanisms underlying this phenomenon are not fully elucidated. Two recent studies have shown a crucial role for LPS contamination following surgery in tumor growth and metastases (21, 22). Both laparotomy and air laparoscopy result in endotoxin contamination of the peritoneal cavity and systemic endotoxemia via bacterial translocation across the gut (21). In a murine model of metastatic disease, animals subjected to laparotomy or air laparoscopy with high levels of circulating LPS had increased lung metastatic burdens, whereas animals subjected to CO₂ laparoscopy with very low levels of plasma LPS had metastatic tumor growth similar to that in controls. Furthermore, there were significantly increased lung metastases in animals that received an equivalent LPS injection (22). These results indicate that LPS entering the peritoneal cavity or systemic circulation during surgery is associated with enhanced growth of metastases following surgical trauma. However, the direct effect of LPS on invasive and metastatic behavior of tumor cells is unclear.

The transcription factor NF-κB is involved in the regulation of multiple cellular processes, including proinflammatory cytokine gene expression, cellular adhesion, cell cycle activation, apoptosis, and oncogenesis. Currently known subunit members of the NF-κB family in mammals are five proteins related by the Rel homology domain, namely p50, p65 (Rel A), the proto-oncogene c-rel (c-Rel), p52, and Rel B (23). In most cells the NF-κB heterodimer is sequestered in the cytoplasm as an inactive form through interaction with an inhibitory κB (IκB) protein that inhibits nuclear translocation of NF-κB. Several IκB isoforms have been identified, but the most extensively characterized isoforms are IκB-α, IκB-β, and IκB-ε (24). When IκB is degraded through the phosphorylation of IκB-α induced by activated IκB kinases, the NF-κB heterodimer will enter the nucleus, bind to the promoter regions of the target genes, and stimulate transcription (23, 25). It has been shown that constitutive activation of NF-κB is present in a number of tumor cells, including hepatocellular carcinoma cells (26), breast cancer cells (27), and non-small cell lung cancer cells (28). Elevated NF-κB activity in tumor cells may be related to an overproduction of cytokines, such as vascular endothelial growth factor (VEGF), GM-CSF, and IL-8, by tumor cells (29, 30), which may provide continued positive growth stimuli. Constitutive activation of NF-κB has been shown to protect tumor cells against apoptosis (31), and inhibition of NF-κB can sensitize tumor cells to undergo chemotherapy-induced apoptosis (28, 32), indicating a role for NF-κB in the development of tumor resistance to cancer treatment. A recent study has shown that blockade of NF-κB signal in a murine lung alveolar carcinoma cell line by transfection of a dominant negative mutant form of IκB-α that cannot be phosphorylated prevents intravasation of tumor cells in an in vivo chick embryo metastasis model and lung metastasis in a murine model (33), suggesting that activation of NF-κB plays a central and specific role in the regulation of tumor metastasis.

In the present study we examined the effects of LPS on β₁ integrin expression and subsequent tumor cell adhesion to and invasion into the extracellular matrix. We also investigated the effects of LPS on NF-κB activation and its relation to tumor cell invasive behavior. Here we provide evidence for the first time that LPS directly enhances tumor cell invasive and metastatic potentials. Using two human colorectal tumor cell lines, SW480 and SW620, cultured in an in vitro serum-free culture system, this report shows that LPS in the presence of soluble CD14 (sCD14) can up-regulate β₁ integrin expression and promote tumor cell adhesion and invasion through a β₁ integrin-dependent mechanism. Furthermore, LPS can directly activate NF-κB in SW480 and SW620 tumor cells. Blockade of NF-κB activation prevents LPS-induced β₁ integrin overexpression and attenuates tumor cell adhesion and invasion, indicating a key role for NF-κB activation in transducing LPS signaling and subsequent LPS-promoted tumor cell adhesion and invasion.

Medium 199, medium l-15, HBSS, PBS without Ca²⁺ and Mg²⁺, FCS, penicillin, streptomycin sulfate, glutamine, and 0.05% trypsin/0.02% EDTA solution were purchased from Life Technologies (Paisley, Scotland). Human plasma fibronectin, collagen I from calf skin, laminin from human placenta, LPS (Escherichia coli O55B5), HEPES, MgCl₂, KCl, NaCl, EDTA, Tris-HCl, glycerol, nuclease-free BSA, Nonidet P-40, PMSF, and DTT were purchased from Life Technologies and Sigma-Aldrich (St. Louis, MO), respectively. [γ-³²P]ATP (3000 Ci/mmol) and poly(dI-dC) were obtained from Amersham International (Little Chalfont, U.K.) and Amersham Pharmacia Biotech (Milton Keynes, U.K.), respectively. SN50, a cell-permeable peptide inhibitor of NF-κB, was purchased from Calbiochem (San Diego, CA). Recombinant human sCD14 was obtained from Biometec (Greifswald, Germany). Mouse anti-human CD14 mAb was obtained from BD Biosciences (Mountain View, CA). Mouse anti-human integrin β₁, α1, α₂, α₃, α₄, α₅, α₆, and α_vβ₃ mAbs were purchased from Chemicon (Temecula, CA), BD PharMingen (San Diego, CA), and Serotec (Oxford, U.K.), respectively. β₁ integrin-blocking mAbs 4B4 and 6S6 were purchased from Coulter Clone (Miami, FL) and Chemicon, respectively.

Human colorectal tumor cell lines SW480 and SW620 were obtained from American Type Culture Collection (Manassas, VA). SW480 and SW620 cells were grown in medium L-15 supplemented with 10% FCS, penicillin (100 U/ml), streptomycin sulfate (100 μg/ml), and glutamine (2.0 mM). Cells were maintained at 37°C in a humidified 5% CO₂ atmosphere and subcultured by trypsinization with 0.05% trypsin/0.02% EDTA when cells became confluent. After the second passage, SW480 and SW620 cells were incubated in an in vitro serum-free culture system in all experiments conducted in this study.

SW480 and SW620 cells cultured in serum-free medium were exposed to LPS alone (0.1 μg/ml) or a combination of LPS (0.1 μg/ml) and sCD14 (0.2 μg/ml) for 4 h at 37°C in humidified 5% CO₂ conditions. The expression of different subunits of β₁ integrins, α_vβ₃, and membrane-bound CD14 (mCD14) on SW480 and SW620 cells was determined using direct and indirect immunofluorescent staining. For direct immunofluorescent staining, 20 μl of FITC-conjugated anti-integrin α₅ (anti-CD49e), FITC-conjugated anti-integrin α_vβ₃ (anti-CD51/61), FITC-conjugated anti-Leu M3 (anti-CD14), PE-conjugated anti-integrin α₄ (anti-CD49d), and PE-conjugated anti-integrin β₁ (anti-CD29) mAbs were added to 100 μl of cell suspension (1 × 10⁶ cells/ml) and incubated at 4°C for 30 min. FITC- and PE-conjugated isotype IgG1 and IgG2b mAbs were used as negative controls. For indirect immunofluorescent staining, 100 μl of cell suspension (1 × 10⁶ cells/ml) was incubated with 20 μl of pure mAbs against integrin α₁ (anti-CD49a), integrin α₂ (anti-CD49b), integrin α₃ (anti-CD49c), and integrin α₆ (anti-CD49f) at 4°C for 30 min and further stained with secondary FITC-conjugated mAb at 4°C for 30 min. Pure isotype IgG1 mAb was used as a negative control. Different β₁ integrins, α_vβ₃, and mCD14 expression on SW480 and SW620 cells were analyzed on a FACScan flow cytometer (BD Biosciences) for detecting the log of the mean channel fluorescence intensity with an acquisition of FL1 and FL2, respectively. A minimum of 10,000 events were collected and analyzed on CellQuest software (BD Biosciences).

Tumor cell attachment to fibronectin, collagen I, and laminin was performed as previously described (34) with some modifications. Briefly, fibronectin (2.0 μg/ml), collagen I (1.5 μg/ml), and laminin (3.5 μg/ml) were coated onto 96-well, flat-bottom, microtiter plates (Falcon, Lincoln Park, NJ). SW480 and SW620 cells cultured in serum-free medium were treated with LPS alone (0.1 μg/ml) or LPS plus sCD14 (0.1 and 0.2 μg/ml) for 4 h at 37°C in humidified 5% CO₂ conditions. For β₁ integrin and NF-κB blocking experiments, cells were treated with different mAbs or SN50 for 30 min before being exposed to LPS or LPS plus sCD14. Cells (5 × 10⁴ cells/ml) were then added to the fibronectin-, collagen I-, and laminin-coated 96-well plates and incubated at 37°C in humidified 5% CO₂ conditions for 1 h. The plate was washed twice with HBSS to remove unbound cells. Tumor cell adhesion to fibronectin, collagen I, and laminin was assessed using CellTite 96 Aqueous One Solution Assay (Promega, Madison, WI) on a Microtiter Plate Reader (Dynex Technologies, Chantilly, VA). The specific absorbance at 490 nm generated from CellTite 96 Aqueous One Solution Assay is directly proportional to the number of adherent cells in the plate. The ratio of the percentage of adherent tumor cells to that of total added tumor cells was calculated according to the following formula:

Background binding was <5% of the total.

In vitro tumor cell invasion was assessed using a Biocoat Matrigel invasion chamber (BD Biosciences) with cell culture inserts containing an 8-μm pore size positron emission tomography membrane with a thin layer of Matrigel basement membrane matrix as previously described (35, 36). Briefly, 0.5 ml of tumor cells (1 × 10⁵ cells/ml) resuspended in serum-free medium containing either LPS alone (0.1 μg/ml) or LPS plus sCD14 (0.1 and 0.2 μg/ml) was added to the cell culture insert of a Biocoat Matrigel invasion chamber. Fibronectin (20 μg/ml) was added in the outer chamber as a chemoattractant. The cells were then incubated at 37°C in humidified 5% CO₂ conditions for 18 h. To quantitative tumor cell invasion, noninvading cells were removed from the upper surface of the membrane by scrubbing gently with a cotton-tipped swab. The cells on the lower surface of the membrane were fixed with Rapi-Diff II (DiaChem, Lancashire, U.K.). Five random microscope fields of the lower surface of the membrane were counted for numbers of cells that had invaded through the Matrigel layer and the membrane. The results were expressed as numbers of invaded cells per microscope field. To examine whether β₁ integrins and NF-κB activation are involved in LPS-induced tumor cell invasion, SW480 and SW620 cells were treated with different blocking mAbs or SN50 for 30 min before being exposed to LPS alone or LPS plus sCD14.

Dual transfection of SW480 and SW620 cells was accomplished using 16 μl/ml of Lipofectamine 2000 reagent (Life Technologies) and 4.0 μg/ml of pNF-κB-Luc vector DNA (Clontech Laboratories, Palo Alto, CA) that contains the firefly luciferase gene as the reporter. The pRL-CMV vector (Promega) containing the Renilla luciferase gene was used as an internal control. Briefly, SW480 and SW620 cells were plated in 96-well plates (Falcon; 2 × 10⁴ cells/well) to grow at 37°C in humidified 5% CO₂ conditions until they reached 90–95% confluence. Reporter DNA (0.2 μg) was mixed with 0.8 μl of Lipofectamine in 50 μl of serum-free Opti-MEM I (Life Technologies), and incubated at room temperature for 20 min to form Lipofectamine-DNA complexes. Cells in 96-well plates were transfected with the complexes for 6 h and cultured for an additional 18-h period after replacement of the medium. Each transfection was performed in duplicate. Transfected cells were exposed to LPS alone (0.1 μg/ml) or LPS plus sCD14 (0.1 and 0.2 μg/ml) for 6 h. Cell extract was prepared using the Passive lysis buffer (Promega), and protein content in each sample was determined using a Micro BCA protein assay reagent kit (Pierce, Rockford, IL). Firefly and Renilla luciferase activities were measured using the Dual luciferase reporter assay system (Promega) to assess promoter activity and transfection efficiency.

SW480 cells were also transfected with a dominant negative IκB-α vector or an empty expression vector pcDNA3.1 as a negative control (provided by Dr. A. G. Bowie, Trinity College, Dublin, Ireland) (37). Briefly, SW480 cells were plated in 24-well plates (Falcon; 1 × 10⁵ cells/well) and grown until they reached 90–95% confluence. The vector DNA-Plus Reagent-Lipofectamine complexes were produced by mixing dominant negative IκB-α vector DNA (0.12 μg) or empty vector DNA (0.12 μg/ml) with 4 μl of Plus Reagent (Life Technologies) and 1 μl of Lipofectamine (Life Technologies). Cells in 24-well plates were transfected with the complexes for 3 h and cultured for an additional 15-h period after replacement of the medium. Each transfection was performed in duplicate. Expression of β₁ integrin, cell adhesion, and invasion in response to LPS plus sCD14 stimulation were assessed in the transfected cells. Transfection efficiency, and inhibition of NF-κB activation were assessed by dual transfection of dominant negative IκB-α vector (0.04–0.16 μg) in combination with pSV-β-galactosidase vector (Promega) or pNF-κB-Luc vector (Clontech Laboratories).

SW480 and SW620 cells were cultured in serum-free medium in six-well plates (Falcon; 1 × 10⁶ cells/well) and treated with LPS alone (0.1 μg/ml) or LPS plus sCD14 (0.1 and 0.2 μg/ml) for 30 min. Nuclear and cytoplasmic extracts were prepared as previously described (38). Briefly, cells were lysed in a hypotonic solution (10 mM HEPES, 1.5 mM MgCl₂, 10 mM KCl, and 0.1% Nonidet P-40, pH 7.9) on ice for 10 min and centrifuged at 13,000 rpm to pellet nuclei. Cytoplasmic supernatants were removed, and nuclei were resuspended in nuclear extract buffer (20 mM HEPES, 25% glycerol, 420 mM NaCl, 1.5 mM MgCl₂, and 0.2 mM EDTA, pH 8.0) on ice for 15 min. The lysates were centrifuged at 13,000 rpm, and supernatants containing the nuclear proteins were collected. All buffers contained freshly added 0.5 mM DTT, 0.5 mM PMSF, and protease inhibitor cocktail (Roche, Mannheim, Germany). Protein concentrations were determined using a Micro BCA protein assay reagent kit (Pierce). All extracts were stored at −70°C until analyzed.

EMSAs were performed as previously described (39). Briefly, 2.0–4.0 μg of nuclear extracts were incubated with 30,000 cpm of double-stranded oligonucleotide 5′-AGT TGA GGG GAC TTT CCC AGG C-3′ containing the NF-κB consensus sequence (underlined; Promega) that had been previously labeled with [γ-³²P]ATP (10 mCi/mmol) by T4 polynucleotide kinase (Promega). The DNA binding reactions were performed in the presence of 2.0 μg of poly(dI-dC) as a nonspecific competitor in binding buffer (10 mM Tris-HCl (pH 7.5), 100 mM NaCl, 1.0 mM EDTA, 5.0 mM DTT, 4% glycerol, and 100 μg/ml of nuclease-free BSA) at room temperature for 30 min. For competition experiments, unlabeled double-stranded oligonucleotide 5′-AGT TGA GGC GAC TTT CCC AGG C-3′ containing the mutated NF-κB consensus sequence (underlined; Promega) was added to the nuclear extracts 30 min before the addition of the radiolabeled probe. All reaction mixtures were subjected to electrophoresis on native 5% (w/v) polyacrylamide gels, which were subsequently dried and autoradiographed.

All data are presented as the mean ± SD. Statistical analysis was performed using ANOVA. Differences were judged statistically significant at p < 0.05.

As determined by FACScan analysis, human colorectal tumor cell lines SW480 and SW620 constitutively expressed high levels of integrin β₁ subunit, whereas various low levels of integrin α₁, α₂, and α₄ expression on SW480 and of α₂, α₄, and α₆ expression on SW620 were also detected (Fig. 1). However, the expressions of integrin α₃ and α₅ subunits, and integrin α_vβ₃ were almost absent in both SW480 and SW620 cells (data not shown). Furthermore, these two tumor cell lines did not contain mCD14 (Fig. 2), the cell surface receptor of a 55-kDa GPI membrane-anchored glycoprotein for LPS recognition and binding.

When tumor cells were incubated in a serum-free culture system, LPS alone at 0.1 μg/ml had no effect on modulating β₁ integrin expression. However, LPS (0.1 μg/ml) in the presence of sCD14 (0.2 μg/ml) significantly up-regulated the expression of integrin β₁, α₂, and α₄ subunits on SW480 cells and the expression of integrin β₁, α₄, and α₆ subunits on SW620 cells (Fig. 3). LPS plus sCD14 did not enhance integrin α₁ expression on SW480 and integrin α₂ expression on SW620 cells (Fig. 3). Soluble CD14 alone (0.2 μg/ml) did not affect the expression of different β₁ integrin subunits on these two tumor cell lines (data not shown).

In the absence of exogenous stimulation, both SW480 and SW620 cells showed various levels of spontaneous adhesion to extracellular matrix proteins such as fibronectin, collagen I, and laminin (Fig. 4). Under an in vitro serum-free culture condition, SW480 and SW620 cells stimulated with LPS alone (0.1 μg/ml) did not significantly affect the attachment of tumor cells to extracellular matrix proteins (Fig. 4, A and B). When tumor cells were treated with a combination of LPS (0.1 μg/ml) and sCD14 (0.2 μg/ml), however, there were significant increases in SW480 and SW620 cell adhesion to fibronectin, collagen I, and laminin by 2- to 3-fold (p < 0.05 vs tumor cells treated with either culture medium or LPS alone; Fig. 4, A and B). To evaluate LPS-dependent tumor cell invasion in an in vitro model, we used Biocoat Matrigel chambers. As shown in Fig. 5, LPS (0.1 μg/ml) in the presence of sCD14 (0.2 μg/ml) significantly promoted SW480 and SW620 cell invasion through the Matrigel (p < 0.05 vs tumor cells treated with either culture medium or LPS alone). In contrast, when SW480 and SW620 cells were treated with LPS alone (0.1 μg/ml), no significant effects on tumor cell invasion were observed (Fig. 5).

β₁ integrins play a crucial role in supporting tumor cell attachment to, migration, and invasion into extracellular matrix proteins. To determine whether β₁ integrin is involved in LPS-induced tumor cell adhesion and invasion, we used anti-β₁ integrin mAbs to selectively block integrin β₁ subunits. Pretreatment of SW480 and SW620 cells with 4B4 and 6S6, two specific integrin β₁ subunit blocking mAbs, almost totally prevented LPS plus sCD14-induced tumor cell adhesion to fibronectin (p < 0.05 vs tumor cells treated with LPS plus sCD14; Fig. 6). Furthermore, 4B4 and 6S6 significantly attenuated LPS plus sCD14-promoted SW480 and SW620 cell invasion (p < 0.05 vs tumor cells treated with LPS plus sCD14; Fig. 6). An isotype-matched mouse IgG1 used as the control for 4B4 and 6S6 had no effect on LPS- plus sCD14-induced tumor cell adhesion and invasion (Fig. 6). These results indicate that LPS-induced SW480 and SW620 cell adhesion and invasion through extracellular matrix proteins are β₁ integrin dependent.

NF-κB activation induced by a combination of LPS and sCD14 was assessed by transfection of SW480 cells with pNF-κB-Luc reporter vector. As shown in Fig. 7,A, LPS in the presence of sCD14 resulted in a significant increase in luciferase activity (p < 0.05), indicating the activation of NF-κB following LPS and sCD14 stimulation. NF-κB-DNA binding activity was also assayed by EMSA using nuclear extracts that were prepared from SW480 cells incubated with serum-free culture medium, LPS alone, sCD14 alone, or LPS plus sCD14. There was a low level of constitutive activation of NF-κB in unstimulated cells; however, LPS in the presence of sCD14 caused a marked activation of NF-κB (Fig. 7 B). LPS alone or sCD14 alone had no effect on NF-κB activation. Similar levels of NF-κB activation were also found in SW620 cells treated with LPS plus sCD14 (data not shown).

SN50 is a synthetic peptide containing a cell membrane-permeable motif and nuclear trans-locating hydrophobic sequence that inhibits nuclear translocation of NF-κB in a dose-dependent manner in cultured endothelial and monocytic cells stimulated with LPS or TNF-α (40). In a dose-response experiment we found that inhibition of NF-κB activation, as represented by luciferase activity in the transfected SW480 cells by SN50 was 18% at 25 μg/ml, 62% at 50 μg/ml, 94% at 100 μg/ml, and 92% at 200 μg/ml. As shown in Fig. 8, when SW480 cells were pretreated with SN50 at 100 μg/ml, there was a total abrogation of up-regulation of integrin β₁, α₂, and α₄ expression induced by LPS plus sCD14. Furthermore, pretreatment of SW480 cells with SN50 at 100 μg/ml significantly attenuated LPS-induced tumor cell adhesion and invasion (p < 0.05 vs tumor cells treated with LPS plus sCD14; Fig. 8). SN50 at 100 μg/ml was not toxic to the cells as determined by measurement of lactate dehydrogenase release and did not affect cell viability as determined by trypan blue exclusion and propidium iodide staining on flow cytometry (data not shown).

The dominant negative IκB-α vector lacks both constitutive and inducible phosphorylation sites and does not dissociate from NF-κB in response to stimulation of the IκB kinase pathways (41). Thus, dominant negative IκB-α vector can be used to eliminate NF-κB activation and to block the NF-κB signal transduction pathway in the transfected cells. As shown in Fig. 9, FACScan analysis of β₁ integrin expression on SW480 cells demonstrated a significant attenuation of LPS plus sCD14-induced integrin β₁, α₂, and α₄ overexpression in dominant negative IκB-α transfected cells. Transfection of SW480 cells with dominant negative IκB-α vector also prevented the increased tumor cell attachment to fibronectin and invasion in Matrigel in response to LPS plus sCD14 stimulation. In contrast, transfection with empty vector showed no effect on LPS- plus sCD14-stimulated β₁ integrin expression, tumor cell adhesion, and invasion (Fig. 9).

Before investigating the direct effect of LPS on tumor cell attachment to and invasion through the extracellular matrix, we have examined how LPS binds to and activates SW480 and SW620 tumor cells. Over the years many LPS-binding proteins have been identified (42). CD14, a 55-kDa GPI-linked glycoprotein that recognizes and binds to LPS with high affinity, exists in two forms: mCD14, expressed on monocytes, macrophages, and neutrophils, and sCD14, found as a plasma protein at a concentration of 2–6 μg/ml (17, 43). LPS can directly activate monocytes and macrophages through its binding to mCD14. However, the cells that do not possess mCD14, such as endothelial cells and epithelial cells, also respond to LPS stimulation in a serum-dependent fashion or in the presence of sCD14 (44, 45). LBP, an acute phase serum protein that interacts and forms complexes with LPS molecules, transports and facilitates LPS binding to CD14 (16). Since CD14 lacks a transmembrane region and is incapable of transducing LPS signaling, the mechanism by which it confers LPS responsiveness to target cells has remained a long-standing question. Recently, the highly conserved family of TLR proteins were discovered, each TLR being a type I transmembrane protein possessing an extracellular lucine-rich repeat and a cytoplasmic Toll/IL-1 receptor homology domain (46, 47). Several studies have demonstrated that TLR4, in combination with an accessory molecule MD-2 and an adaptor protein MyD-88, is the most likely LPS signal-transducing molecule for CD14 (18, 19, 48, 49).

In the present study SW480 and SW620 cells did not express mCD14, as confirmed by FACScan analysis. To identify a critical role for CD14 in LPS-induced cellular responses in tumor cells, we have employed an in vitro serum-free culture system to eliminate the influences of sCD14 and other soluble factors, such as LBP, that are present in the serum. Results obtained under serum-free culture conditions demonstrated that SW480 and SW620 cells did not respond to LPS stimulation. In contrast, LPS in the presence of recombinant sCD14 activated NF-κB, up-regulated β₁ integrin expression, and promoted tumor cell adhesion and invasion. Thus, the apparent function of sCD14 is to enable tumor cells that lack mCD14 to recognize and respond to LPS. Furthermore, LPS at 0.1 μg/ml, an optimal dose according to the preliminary dose-response experiments (data not shown), was sufficient to cause SW480 and SW620 cell activation. These results are consistent with previous studies in which endothelial cells can be activated by high concentrations of LPS (50–500 ng/ml) in the presence of sCD14, whereas low concentrations of LPS (5–10 ng/ml) require the presence of LBP for sCD14-dependent activation (45).

Different subunits of β₁ integrin and various levels of their expression have been found in different types of tumor cells. Forexample, human hepatocellular carcinoma cell lines, PLC/PRF/5, Hep3B, HepG2, HLE, HuH7, and C3A cells, constitutively express high levels of β₁ and α₆ subunits and various low levels of α1, α₂, α₃, and α₅ subunits, whereas expression of the α₄ subunit is absent in each cell line (50). We found that the predominant adhesion receptors in SW480 cells were the α1β₁, α₂β₁, and α₄β₁ integrins; that α₂β₁, α₄β₁, and α₆β₁ integrins were highly expressed in SW620 cells; and that the α₃β₁ and α₅β₁ integrins were undetectable in either cell line. It has been found that overexpression of integrin α_vβ₃ correlates with aggressive phenotypes in different malignancies and with tumor angiogenesis (51, 52). However, neither SW480 nor SW620 cells expressed α_vβ₃. The various levels of β₁ integrin expression on SW480 and SW620 cells may represent a low constitutive activity of different β₁ integrin subunits, which may account for spontaneous adhesion of tumor cells to fibronectin, collagen I, and laminin, and invasion in Matrigel as observed in the present study.

It is notable that LPS stimulation in the presence of sCD14 resulted in the up-regulation of β₁ integrin expression in SW480 and SW620 cells. LPS-modulated β₁ integrin expression appears to be both subunit and cell specific, as LPS significantly enhanced the expression of β₁, α₂, α₄, and α₆ subunits, but had no effect on α₁ expression on SW480 cells or α₂ expression on SW620 cells. In contrast, a stimulatory anti-β₁ mAb TS2/16 has been found to induce a rapid activation of β₁ integrin as confirmed by an enhanced cell adhesion to collagen, but it does not change the expression of different β₁ integrin subunits (50, 53). Increased expression of β₁ integrin is thought to be associated with the ability of tumor cells to interact with the extracellular matrix and to subsequently form an organ-specific metastatic colonization through β₁ integrin-mediated cell adhesion and invasion. For example, TGF-β₁ stimulates the hepatocellular carcinoma cell line SMMC-7721 cell adhesion to extracellular matrix through up-regulation of α₅β₁ integrin expression (54). Furthermore, blockade of the α₃ subunit prevents breast carcinoma cell line MDA-MB-231 cell migration and invasion (10). β₄-δ CYT-transfected MDA-MB-435 cells that are deficient in forming the α₆β₁ heterodimer are unable to establish metastatic foci in lungs (11).

Importantly, the present study has demonstrated that LPS stimulation significant enhances tumor cell attachment to extracellular matrix proteins and promotes tumor cell invasion, indicating a direct and distinct effect of LPS on the invasive and metastatic ability of tumor cells. By using β₁ integrin function-blocking mAbs, we further demonstrated that LPS-mediated tumor cell adhesion and invasion correlated with an increased expression of β₁ integrins, as 4B4 and 6S6 strongly inhibited LPS-induced tumor cell adhesion and invasion. The process of tumor cell invasion involves cell attachment to the subendothelial extracellular matrix and subsequent unidirectionally cell migration coupled with local proteolysis induced by a number of degradative enzymes, particularly matrix metalloproteinases (MMP). It has been shown that increased activity of β₁ integrin subunits results in MMP production and the activation of MMP proenzymes (10, 55). Although we did not measure MMP directly in the present study, the enhanced tumor cell invasion may correlate with the release and activation of MMP that could be mediated by increased β₁ integrin activity or by LPS stimulation directly. We have previously shown that endotoxin or LPS contamination of the peritoneal cavity and systemic endotoxemia are involved in surgically induced lung metastases of 4T1 mammary adenocarcinoma cells in mice, which correlates with an elevation in circulating levels of VEGF and an increased cell mitosis/apoptosis ratio in metastatic tumor burden (22). Results from this study provide further evidence for a direct effect of LPS on tumor cell adhesion to and invasion into the extracellular matrix.

Constitutive activation of NF-κB has been described in a number of tumor cells, and appears to be associated with continued cytokine production, inhibition of apoptosis, activation of cell cycle, and possibly tumor progression (56). A low level of constitutive activation of NF-κB was also found in SW480 and SW620 cells, which may account for an overproduction of VEGF by these cells in standardized culture conditions (data not shown). In the present study it is interesting to find that LPS stimulation of tumor cells resulted in a marked increase in NF-κB-DNA binding activity. To examine the correlation between NF-κB activation and LPS-induced tumor cell metastatic potential, SN50, a synthetic peptide that blocks NF-κB signaling by inhibition of nuclear translocation of NF-κB (40), and a dominant negative IκB-α vector, which prevents NF-κB activation in transfected cells (41), were used in additional experiments. Blockade of NF-κB activation by either SN50 or IκB-α transfection abrogated LPS-induced β₁ integrin overexpression and attenuated tumor cell adhesion and invasion, indicating that NF-κB activation is a prerequisite not only for the transduction of LPS signals in tumor cells, but also for the enhanced tumor cell metastatic ability induced by LPS stimulation. LPS-mediated NF-κB activation in the tumor cells switches the target gene to transcribe and synthesize new adhesion molecules, including β₁ integrins. Increased surface expression of β₁ integrins mediates tumor cell adhesion to ligands in the extracellular matrix, which elicits a variety of intracellular signals, including NF-κB activation (57, 58), a positive feedback loop that could occur to sustain enhanced levels of NF-κB activity. Activated NF-κB has been found to stimulate the production of a number of degradative enzymes, such as MMP and urokinase-like plasminogen activator (59, 60), which, together with β₁ integrin, results in tumor cell invasion. Blockade of NF-κB signaling has been shown to result in the down-regulation of MMP9 and heparanase and the up-regulation of tissue inhibitors of matrix metalloproteinases, thus preventing tumor cell intravasation and lung metastases (33). Taken together, these reports suggest NF-κB to be an important regulator of the metastatic phenotype, and the up-regulation of NF-κB in tumor cells induced by LPS in this study could have significant effects on metastatic potential.

Two recent studies have reported that commercially available LPS is contaminated by microbial proteins such as bacterial lipoprotein, as repurification of commercial preparations of LPS results in TLR4-mediated, but not TLR2-mediated, cellular activation (61, 62). Although we cannot completely exclude the possibility that LPS used in this study was contaminated with non-LPS bacterial cell wall components, it must be pointed out that LPS from any bacterial species is a mixture of different LPS, which may elicit distinct biological effects (63, 64), and that repurification of LPS by repeated phenol extraction may preferentially concentrate certain LPS subsets more than others. Further work will be required to confirm the effect of pure LPS on tumor cell activation. Furthermore, although this study has demonstrated the activation of NF-κB by LPS through a CD14-dependent mechanism, the signal transduction pathways of LPS in tumor cells are largely unexplored.

In conclusion, LPS or endotoxin released by Gram-negative bacteria may have a direct effect on tumor progression by promoting tumor cell adhesion and invasion. This effect is mediated by LPS-induced up-regulation of β₁ integrin and activation of NF-κB. These findings further implicate LPS, introduced by surgical procedures such as laparotomy, as a causative factor in surgically induced tumor metastatic growth. Therefore, neutralization of LPS and modulation of NF-κB may be considered therapeutic strategies for the prevention of tumor relapse and metastasis in perioperative surgical practice.