IL-8-Mediated Cell Migration in Endothelial Cells Depends on Cathepsin B Activity and Transactivation of the Epidermal Growth Factor Receptor¹

Chemokine receptors, which are best known for their role in leukocyte trafficking, are expressed on a variety of adherent cells including endothelial, epithelial, and cancer cells, where their function is poorly understood. Specifically, expression of IL-8Rs has been shown on neurons and microvascular endothelial cells (1, 2), in both normal breast epithelium and breast carcinomas as well as in melanomas (3), and on ovarian (4) and pancreatic cancer cells, all cancers with a high metastatic index.

The angiogenic (5) and growth-promoting property of CXC chemokines (IL-8, growth-related oncogene-α (gro-α),³ neutrophil-activating peptide-2, epithelial neutrophil-activating peptide-78, and others) depends on the presence of an ELR sequence in the N terminus of the protein (6). Chemokines which do not express this sequence—or in which this sequence was mutated—are devoid of angiogenic and growth-promoting activity (7). Recent investigations have shown that activation of the CXCR2 expressed on microvascular endothelial cells is responsible for the angiogenic response to IL-8 (1, 2). In accordance, the CXCR2 is activated by all ELR-containing chemokines in contrast to the CXCR1, which is specific for IL-8 (8).

In previous work, we showed that expression of the CXCR2 in NIH 3T3 cells caused transformation as assessed by focus formation and growth in soft agar (9). In addition, NIH 3T3 cells expressing the CXCR2 caused tumors and metastases in nude mice (our unpublished results). This oncogenic behavior of the receptor—in the absence of added ligand—could be explained by autocrine stimulation by mouse KC, the murine equivalent of gro-α, which stimulates the human CXCR2. NIH 3T3 cells expressing a point mutation of the CXCR2, which is poorly G-protein coupled (D138Q-CXCR2), behaved like untransfected cells (9), indicating that cell transformation depended on signal transduction.

In leukocytes, chemotaxis is largely mediated by the CXCR1 (10, 11) despite similar expression levels and affinities of the two IL-8Rs in these cells. In contrast, migration of microvascular endothelial cells, which similarly express both IL-8Rs, was largely a function of the CXCR2 (1, 2). Migration of microvascular endothelial cells is an important component of the angiogenic response, which is a known in vivo function of IL-8 (5). Similarly, cell migration is a prerequisite for cancer cell invasion and metastasis, and is more readily manipulated than transformation assays. For this reason, the effect of inhibitors on cell migration was used to screen for possible signal transduction pathways in endothelial cells and NIH 3T3 cells expressing the CXCR1 or CXCR2.

The CXCR1 and CXCR2 are G-protein-coupled receptors (GPCRs), which couple primarily to G_i (12), the G-protein generally associated with chemotactic behavior (13). However, cell migration, especially of adherent cell types can also be mediated by activation of receptor tyrosine kinases (14, 15).

It has become apparent over the past few years that there is extensive cross talk between GPCRs and receptor tyrosine kinases such as the epidermal growth factor (EGF) receptor and the platelet-derived growth factor receptor (16, 17). In particular, transactivation of the EGFR has been observed with a number of GPCR ligands including thrombin (16), lysophosphatidic acid (18), angiotensin, and, in one report, IL-8 (4). However, this report did not investigate which of the two IL-8Rs was involved or define the cell biological consequences of this activation. It was believed initially that transactivation of the EGFR was independent of EGFR ligands, based on the rapid onset of EGFR phosphorylation and the failure to detect EGF in the conditioned medium. More recently, it has been shown that EGFR transactivation can be mediated by heparin-binding EGF-like growth factor (HB-EGF) (19, 20). HB-EGF is synthesized as a transmembrane precursor protein that is proteolytically cleaved into the mature, soluble growth factor responsible for transactivation of the EGFR (reviewed in Ref. 21). The signal transduction cascade that leads from the activation of GPCRs to ectodomain shedding of HB-EGF is still poorly defined, but appears to involve activation of the Ras/mitogen-activated protein kinase (MAPK) cascade, of rac (22, 23, 24), and of src (25, 26). In the case of the thrombin receptor, it has been shown that the proteolytic cleavage and ectodomain shedding of HB-EGF is mediated by matrix metalloproteases (MMPs) and can be blocked by hydroxamate MMP inhibitors (19, 20). The intermediate steps in this activation are still obscure and may vary for different GPCRs or different cell types.

In this study, we report that activation of the CXCR2-mediated cell migration of microvascular endothelial cells or of transfected NIH 3T3 cells depends on transactivation of the EGFR through HB-EGF, and that this pathway is essential for the in vitro angiogenic and tumorigenic behavior of IL-8. However, IL-8-mediated HB-EGF activation appears to involve cathepsin B and only to a lesser degree MMP activity.

Immortalized dermal microvascular endothelial cells (HMECs) were grown in endothelial cell growth medium (EGM; Clonetics, San Diego, CA), NIH 3T3 murine fibroblasts and HEK293 human embryonic kidney cells were grown in DMEM, and RBL2H3 rat basophilic leukemia cells were grown in RPMI 1640. NIH 3T3 and RBL2H3 cell lines stably expressing the CXCR1 or the CXCR2 cloned into the pSFFV.neo expression vector have been described previously (27, 9). To create a green fluorescent CXCR2, the CXCR2 construct was subcloned into the EcoRI and KpnI sites of pEGFP-N3 (Clontech, Palo Alto, CA), transfected into HEK293 using Lipofectamine Plus (Invitrogen, Carlsbad, CA), and selected with G418.

For inhibitor studies, cells were preincubated for 30 min with 10 μM LY294002, 100 nM wortmannin, 0.5 μM PP2, 1 μM tyrphostin AG 1478, 1–50 μM GM 6001, 10 μM CA 074 Me (all Calbiochem, La Jolla, CA), 10 μM E-64 (Roche Molecular Biochemicals, Indianapolis, IN), 10 μg/ml leupeptin, or 10 μg/ml aprotinin (both Sigma-Aldrich, St. Louis, MO). For Ab inhibition studies, 10 μg/ml each of the following neutralizing Abs were used: anti-EGFR Ab (clone LA1; Upstate Biotechnology, Waltham, MA), anti-HB-EGF Ab (goat polyclonal; R&D Systems, Minneapolis, MN), and anti-human EGF Ab (MAB236; R&D Systems).

Polymerized actin was visualized as previously described (2). In short, HMECs were seeded at low density on collagen-coated coverslips and grown in EGM (Clonetics) containing 5% FCS. On day 5, when the cells had reached confluence, they were serum starved for 3–4 h and stimulated with 10 nM IL-8 for the time indicated for each experiment. All experiments were performed at 37°C in a tissue culture incubator. For F-actin localization, cells were fixed for 20 min in 3% paraformaldehyde in PBS, permeabilized for 5 min in 0.2% Triton X-100, incubated with 25 mU/ml of Alexa 488 phalloidin (Molecular Probes, Eugene, OR) for 30 min, washed three times with PBS, and mounted with Antifade (Molecular Probes). Fluorescence microscopy was performed on a Leica (Deerfield, IL) DM ERBE microscope equipped with a Hamamatsu (Hamamatsu City, Japan) digital camera and Openlab 3.1 software (Improvision, Lexington, MA) to obtain and analyze images. Actin polymerization in NIH 3T3 cells expressing the CXCR2 was performed in the same fashion except that the cells were serum starved for 16 h.

The bottom side of Transwell filters (Corning Costar, Acton, MA) was coated with 230 μl of bovine collagen (100 μg/ml; Cohesive Technologies, Franklin, MA) in PBS and blocked with 1% BSA (Sigma-Aldrich). To determine cell migration in HMECs, endothelial cell basal medium (EBM; 500 μl/well; Clonetics) containing 0.1% BSA, 0.5 μg/ml hydrocortisone, and 50 μg/ml gentamicin was pipetted into the bottom well, and 5 × 10⁴ HMECs in the same medium were added to 8-μm pore size inserts. For Ab inhibition studies, the cells were preincubated for 30 min with 10 μg/ml the respective Ab. Endothelial cells were stained for 10 min with 1 μM calcein-AM (Molecular Probes), and following the addition of IL-8 to the bottom wells, the cells were incubated in a tissue culture incubator for 4 h at 37°C. Cells in the upper well were carefully removed with a cotton swab, and transmigrated cells were counted with a ×10 objective on a Leica DM ERBE microscope using FITC excitation and emission. Results are expressed as percentage of cells transmigrated in the absence of IL-8 (means ± SEM of three to four experiments in triplicate).

For NIH 3T3 cells, the conditions were the same except that 10⁵ cells were used and allowed to migrate for 16 h before counting (28). For RBL2H3 cells, 5-μm² pore size filters were used, and chemotaxis was assessed by counting transmigrated cells after a 4-h incubation. Results represent means ± SEM of three experiments.

Cells were grown to confluence on 100-mm tissue culture plates, serum starved for 18 h (HMECs) or 24 h (NIH 3T3 cells), stimulated with 400 ng/ml IL-8, 400 ng/ml gro-α, or 50 ng/ml EGF for the indicated times at 37°C, washed with ice-cold PBS, and lysed in 400 μl of lysis buffer (150 mM NaCl, 25 mM Tris (pH 7.5), 1 mM EDTA, 2 mM sodium vanadate, 10 mM NaF, 2 mM sodium pyrophosphate, 1% Nonidet P-40, 2 μg/ml aprotinin, 2 μg/ml leupeptin, 2 mM PMSF, and 10% glycerol). Following centrifugation for 10 min at 10,000 × g, rabbit anti-EGFR Ab (5 μg/0.8 mg protein; Upstate Biotechnology) or mouse anti-src (clone GD11; 2 μg/0.8 mg protein; Upstate Biotechnology) was added to each supernatant, and the samples were rotated for 2 h at 4°C. Protein A-Sepharose (Amersham Pharmacia, Piscataway, NJ) (30 μl of a 50% slurry) was added for 30 min followed by three washes with lysis buffer and one wash with PBS. The resulting pellets were suspended in 30 μl of SDS sample buffer, boiled for 3 min, and loaded onto 7% SDS polyacrylamide gels. Western transfers and blots were performed according to standard protocols, using 5% nonfat dry milk to block nonspecific binding. A 1/2000 dilution of anti-phospho-tyrosine Ab (PY20; BD Transduction Laboratories, Lexington, KY) was added followed by incubation with goat-anti-mouse-HRP-IgG conjugate (1:8000; BioSource International, Camarillo, CA) and detection by ECL (ECL Plus reagent; Amersham Pharmacia). The blots were stripped with Re-Probe (Geno Technology, St. Louis, MO), and redeveloped with anti-EGFR or anti-src Ab to assure equal loading. One blot representative of five is shown. Films were digitized and quantified using UN-SCAN-IT software (Silk Scientific, Oreu, UT).

Src activity was measured using the in vitro kinase assay kit from Upstate Biotechnology following the supplier’s manual.

Endothelial cell spheroids of defined monolayers of HMECs were generated as described (29). In brief, confluent monolayers of HMECs were trypsinized. Cells were suspended in corresponding culture medium containing 20% methocel, seeded into nonadhesive round-bottom 96-well plates (Greiner, Frickenhausen, Germany). Under these conditions, all suspended cells contribute to the formation of a single spheroid per well of defined size and cell number (750–1000 cells/spheroid). The methocel used for these experiments was diluted from a stock solution that was generated by dissolving 6 g of carboxymethylcellulose (Sigma-Aldrich) in 500 ml of medium (EGM; Clonetics). The spheroids were harvested within 24 h, centrifuged at 300–500 × g, and embedded into collagen gels. A collagen stock was prepared before use by mixing 8 vol of acidic rat tail collagen type I (Upstate Biotechnology) (equilibrated to 2 mg/ml at 4°C) with 1 vol of 10× HBSS (Life Technologies) and 1 vol of 0.2 N NaOH to adjust the pH to 7.4. This stock solution (0.5 ml) was mixed with 0.5 ml of room temperature medium (EBM; Clonetics) with 2% FCS containing 0.5% (w/v) carboxymethylcellulose to prevent sedimentation of spheroids before polymerization of the collagen gel. The spheroid-containing gel was rapidly transferred into 24-well plates and allowed to polymerize for at least 30 min. EBM (0.1 ml) containing stimulants and inhibitors was pipetted on top of the gel. The gels were incubated in a 37°C tissue culture incubator, and sprouting of the cells was photographed at 24 h.

For the focus formation assay, 200 stably transfected NIH 3T3 cells were seeded on a layer of 10⁵ untransfected cells, as described (9), and cell foci were counted after 2 wk in culture. A concentration of 1 μM AG 1478 was added daily where indicated.

For the detection of cathepsin B activity, HMECs were grown to confluency in 96-well culture plates (Corning Costar), either in normal tissue culture plates or in plates that had been collagen-coated as described above for the cell migration assay. RBL2H3 cells were used in suspension at a concentration of 50,000 cells/well. Cells were incubated in 200 μl of freshly prepared assay buffer (137 mM NaCl, 5 mM KCl, 0.6 mM CaCl₂, 0.6 mM MgSO₄, 0.7 mM Na₂HPO₄, 5.6 mM glucose, 2 mM l-cysteine, and 25 mm PIPES (pH 7.0)) in the presence or absence of inhibitors for 30 min. RBL2H3 cells were treated with 5 μM cytochalasin B (Sigma-Aldrich) for the last 15 min. Following stimulation with IL-8 for 10 min at 37°C, 100 μM Z-Arg-Arg-AMC (Bachem, Torrance, CA) (30) was added to 100-μl aliquots of supernatants (RBL2H3 cells) or directly to the cells in the well (HMECs), and fluorescent product formation was recorded for 20 min using a fluorescent plate reader (Fluoroscan; Packard, Meriden, CT) with a 360-nm excitation and 460-nm emission filter setting. To determine the total cellular activity of cathepsin B, cells were lysed with 0.1% Triton X-100. Nonspecific activity (0–5%) determined in the presence of CA-074, a specific inhibitor of cathepsin B, was subtracted from all samples.

To detect intracellular cathepsin B activity, a kit (Biomol (Plymouth Meeting, PA) CV-cathepsin B detection kit) was used according to the supplier’s instruction. In this assay, cells are incubated with cell-permeable cresyl violet-Arg-Arg (CVRR), in which the fluorescence of the cresyl violet is quenched by the adjacent dipeptide. In the presence of activated cathepsin B, the arginines are hydrolyzed, allowing detection of the red fluorescent cresyl violet within the organelles of the cell (31). Specifically, cells were incubated at 37°C for 20 min in the presence of ligand and a 500-fold dilution of substrate, washed with PBS, fixed with 4% paraformaldehyde, and examined with a Leica DM ERBE microscope. In the case of HEK293, cells were incubated with CVRR for the last 5 min of incubation.

Release of β-hexoseaminidase from RBL2H3 cells was determined using p-nitrophenyl-N-acetyl-β-d-glucosamide (Sigma-Aldrich) as the substrate as previously described (32).

Cytoskeletal reorganization is a prerequisite for chemotaxis. Because actin polymerization can be easily visualized with phalloidin fluorophores, F-actin formation was used to screen for inhibitors of IL-8 stimulation in adherent cells expressing the CXCR2. NIH 3T3 cells stably transfected with the CXCR2 show prominent stress fiber formation following stimulation with IL-8 (Fig. 1,A). This was attenuated by the phosphatidylinositol 3-kinase (PI3K) inhibitors LY294002 (Fig. 1 B) and wortmannin (results not shown), by the src kinase inhibitor PP2 (C), and by the specific EGFR kinase inhibitor tyrphostin AG1478 (D).

This behavior was not restricted to this transfected cell line, but also applied to cells that express the receptor constitutively. HMECs, immortalized human dermal microvascular endothelial cells, express both the CXCR1 and the CXCR2 (1, 2). Activation of either of the two receptors causes an increase in F-actin content, but the cytoskeletal response for the two receptors can be distinguished temporarily and morphologically (2). Activation of the CXCR1 is transient, maximal by 1 min, and characterized by stress fiber formation, whereas activation of the CXCR2 is prolonged, maximal between 5 and 30 min, and associated with cell contraction. All of the inhibitors that blocked stress fiber formation in NIH 3T3 cells expressing the CXCR2 also prevented the late CXCR2-mediated cytoskeletal response in HMECs (Fig. 1, middle panel). To confirm that the late-phase response was mediated by the CXCR2, the experiments were repeated with gro-α with the same results (Fig. 1, right panel). The early, CXCR1-mediated response was attenuated by inhibition of PI3K or src, but only minimally affected by inhibition of the EGFR (Fig. 1, middle and bottom panels). gro-α shows practically no cytoskeletal response at this time (2) and could therefore not be used as a control in this case.

Next inhibitors that blocked actin polymerization were used in chemotaxis assays of HMECs, NIH 3T3 cells expressing the CXCR1 or the CXCR2, and RBL2H3, rat basophilic leukemia cells, similarly transfected with the CXCR1 or CXCR2 (9, 27). Previous work has shown that, in neutrophils, chemotaxis is primarily a function of the CXCR1 as shown with blocking Abs against the CXCR1 or CXCR2 (10, 11). Similarly, IL-8 was chemotactic for RBL2H3 cells expressing the CXCR1, but not for those expressing the CXCR2 (see Fig. 2,C). In contrast, in endothelial cells, migration was almost entirely mediated by the CXCR2 as shown previously with blocking Abs against the CXCR1 or CXCR2 (1, 2). In addition, gro-α, which only activates the CXCR2, was as active as IL-8 in mediating cell migration in HMECs (Fig. 3,A). Similarly, when cell migration toward IL-8 was determined in NIH 3T3 cells expressing the CXCR1 or the CXCR2, it was found that cell migration was a function of the CXCR2 (Fig. 2 A).

As expected, inhibition of PI3K with LY294002 blocked chemotaxis in all cell types tested, shown only for HMECs in Fig. 3,A. Inhibition of EGFR function in contrast had no effect on the CXCR1-dependent chemotactic response in RBL2H3 cells (Fig. 2,D), but prevented CXCR2-mediated cell migration of HMECs (Fig. 3,A) and of NIH 3T3 cells transfected with the CXCR2 (Fig. 2,B). This was initially shown using an inhibitor of the tyrosine kinase activity of the EGFR (AG1478), and could not be ascribed to any nonspecific drug effect on cell motility, because migration of RBL2H3 cells, which do not express an EGFR, was not influenced by AG1478 (Fig. 2,D). In addition, in the case of the human-derived HMECs, the specificity of this effect could be confirmed independently by use of a neutralizing Ab against the human EGFR, which similarly blocked IL-8- or gro-α-mediated migration of these cells (Fig. 3 B).

Inhibition of src kinase activity with PP2 similarly blocked IL-8-mediated cell migration in HMECs (Fig. 3 A). Because PP2 also blocked EGF-mediated cell migration, src activation appeared to occur downstream of the transactivation of the EGFR.

Because Ab against the EGFR added into the medium could prevent the cellular response, an inside-out signaling event had to have been initiated by stimulation with IL-8. Because previous reports with other ligands of GPCRs had shown that transactivation of the EGFR is mediated by metalloprotease-dependent cleavage of pro-HB-EGF to HB-EGF (19), the effect of anti-HB-EGF and anti-EGF Ab was tested next. It was found that anti-HB-EGF Ab, but not anti-EGF Ab, blocked the IL-8- or gro-α-mediated cell migration of HMECs (Fig. 3,B). In contrast, when EGF was used as chemoattractant, the anti-EGF, but not the anti-HB-EGF Ab, blocked the endothelial cell response (Fig. 3 B).

To verify the IL-8-mediated transactivation pathway biochemically, the phosphorylation state of the EGFR was assessed in HMECs and NIH 3T3 cells expressing the CXCR2 and activated with IL-8 or gro-α. IL-8 and gro-α caused phosphorylation of the EGFR in both cell types (Fig. 4), which lasted for ∼20 min. In RBL2H3 cells, neither phosphorylated nor unphosphorylated EGFR could be detected. These results confirm that activation of the CXCR2 causes transactivation of the EGFR in cells that express this receptor (4).

Because cell migration is a prerequisite for the sprouting of vessels during angiogenesis, the effect of various inhibitors of the EGFR transactivation pathway was determined in an in vitro angiogenesis assay, in which the formation of capillary-like structures in a collagen gel is determined (29). IL-8 induced the sprouting of HMECs from a central spheroid into branching-out structures (Fig. 5,B). Inhibitors that had blocked endothelial cell migration (AG1478, anti-EFGR, and anti-HB-EGF) were found to block the sprouting of capillary tubules as well (Fig. 5, C–E, left panel), indicating that transactivation of the EGFR is essential for IL-8-mediated in vitro angiogenesis.

To determine whether this pathway was important for cell invasion of cancer cells that express the CXCR2, we next asked whether inhibition of the EGFR inhibited CXCR2/mouse KC-mediated focus formation, and found attenuation, but not complete inhibition, of focus formation in AG1478-treated cells (Fig. 5 G). The Abs used in the endothelial cell assays are human specific and could not be used with NIH 3T3 cells.

The intermediate steps between activation of a GPCR and cleavage of pro-HB-EGF to HB-EGF are not well understood, but seem to include src recruitment (33, 25) and activation of membrane metalloproteinases that cleave pro-HB-EGF to HB-EGF (19, 20). To determine involvement of membrane metalloproteinases, IL-8-dependent cell migration of HMECs was determined in the presence of the membrane metalloprotease inhibitor GM 6001 (34). Contrary to the expectation, this inhibitor blocked IL-8-mediated chemotaxis by only ∼40% (Fig. 6 A). A second membrane metalloprotease inhibitor, TAPI-2 (N-[2-hydroxyaminocarbonyl-methyl]-4-pentanoyl-l-(tert-butyl)-analyl-l-alanine, 2-aminoethyl amide; Calbiochem) used at up to 100 μM showed similar effects (results not shown).

Unless the proform of HB-EGF becomes activated directly, e.g., by complex formation of different components of the CXCR2 signaling cascade, cleavage of pro-HB-EGF by IL-8 appears to be mediated by a protease that is not metal dependent. To address this possibility, HMEC migration toward IL-8 was measured in the presence of various protease inhibitors. The broad-spectrum inhibition of cysteine proteases and some serine proteases with leupeptin (10 μg/ml) blocked IL-8-mediated chemotaxis (Fig. 6,B) in HMECs. Inhibition of serine proteases with 10 μg/ml aprotinin showed no effect (Fig. 6,B). To further define which protease was involved, several cysteine protease inhibitors were tested. The general inhibitor of cysteine proteases, E-64 (10 μg/ml; Roche Molecular Biochemicals), which blocks cathepsins and calpain, abrogated IL-8-mediated chemotaxis in HMECs (Fig. 6,B). Inhibition of calpain with 100 μM Z-Val-Phe-CHO/MDL 28170 had no effect (results not shown). In contrast, a specific inhibitor of cathepsin B (10 μg/ml CA-074 Me) (35) blocked cell migration (Fig. 6,B). Neither leupeptin, E-64, nor CA-074 blocked EGF-mediated cell migration (Fig. 6 B), placing the mode of action of these three inhibitors between activation of the CXCR2 and transactivation of the EGFR, and excluding nonspecific effects of the inhibitors on cell locomotion.

Because leupeptin and E-64, which are not cell permeable, blocked the migratory response to IL-8, the question arose whether IL-8 caused release of cathepsin B following cell activation. Using Z-Arg-Arg-AMC as a fluorogenic substrate (30), it was found that IL-8 induced release of cathepsin B from RBL2H3 cells transfected with the CXCR2 (Fig. 7 A). The enzyme activity was in the cell supernatants and not cell associated. Formation of the fluorescent Z-Arg-Arg-AMC product was prevented by preincubation of the cells with CA 074 Me, indicating the specificity of the reaction for cathepsin B. Exocytosis of cathepsin B was paralleled by release of β-hexoseaminidase, 12% of which was released following stimulation with 800 ng/ml IL-8 in agreement with previous reports (36).

In the case of HMECs, the situation was more complicated. Cathepsin B could be detected in the supernatants of IL-8-stimulated cells in only two of seven experiments. Even if cathepsin B activity was determined directly in the wells, in which the cells had been cultured, cathepsin B activity increased in some, but not all experiments following stimulation with IL-8 (Fig. 7,B). If the cells were grown on collagen, a marginally significant statistical difference (p < 0.05) between unstimulated and IL-8-stimulated cells could be detected (Fig. 7,B). The activity was specific for cathepsin B, because it was completely blocked when cells were preincubated with CA-074 Me. However, although inhibition of cathepsin B blocked cell migration in all cases, cell surface-associated cathepsin B activity could be detected in only about half of the experiments, which suggested that cathepsin B activation rather than its release was the important factor. Indeed, the presence of intracellular cathepsin B activity could be verified with a cell-permeable substrate (CVRR), which becomes fluorescent when it is cleaved by cathepsin B. This compound stained unstimulated HMECs only weakly, but red fluorescence accumulated within vesicular structures in cells stimulated with IL-8 or gro-α (Fig. 7,C). Similar red staining of the endosomal/lysosomal compartment was observed in IL-8-stimulated HEK293 cells stably expressing a CXCR2-green fluorescent protein (GFP) construct (Fig. 7,D). At the 15- to 20-min time point, when there was the largest difference in CVRR fluorescence in stimulated vs unstimulated cells, there was little evidence that this fluorescence was associated with the plasma membrane, nor was there colocalization between active cathepsin B and the CXCR2 receptor in the CXCR2-GFP-expressing cells. However, at earlier time points, between 2 and 10 min, cathepsin B activity and the CXCR2 stained identical compartments (Fig. 7 D), and in the initial phase (between 2 and 5 min), cathepsin B activity was associated with the plasma membrane (D). Results are shown for HEK293 cells to show colocalization between the CXCR2-GFP and cathepsin B activity, but similar results for CVRR fluorescence were seen with HMECs and NIH 3T3 cells (results not shown). This transient association of cathepsin B with the plasma membrane explains that nonpermeable protease inhibitors can block the IL-8-mediated response, but also illustrates why it was difficult to measure extracellular cathepsin B activity. Although the activation of cathepsin B following IL-8 stimulation deserves further biochemical analysis, it is clear that cathepsin B activity was important for EGFR-dependent cell migration in the presence of IL-8.

IL-8 mediated cell migration in endothelial cells and NIH 3T3 cells in a CXCR2-dependent fashion. This contrasts with the situation in neutrophils and other leukocytes, where chemotaxis is primarily mediated by the CXCR1. Therefore, it appears that chemotaxis is not determined simply by the receptor sequence, but depends on the specific downstream effector interplay that differs in different cell types. Although inhibition of PI3K blocked cell migration in all cells tested, inhibition by AG1478 was limited to cells that express an EGFR (HMECs and NIH 3T3 cells). In these cells, IL-8 causes transactivation of the EGFR and activates its downstream signaling cascade. This pathway is at least partially responsible for the transformation observed in NIH 3T3 cells expressing the CXCR2 (9), and for the angiogenic response seen in microvascular endothelial cells (1, 2). This difference in migratory behavior of leukocytes, which do not express EGFRs, and endothelial cells and cancer cells, which express both CXCR2 and EGFRs, is of interest for possible therapeutic intervention. Specific inhibition of the CXCR2-dependent functions, which determine the angiogenic and tumorigenic property of IL-8R activation may be feasible without interfering with the leukocytic immune surveillance role of IL-8, which is primarily a function of the CXCR1 (10, 11).

Cell migration is a prerequisite for tumor cell invasion and metastasis. In tumors, which express a CXCR2, the cascade from ELR chemokine activation of the CXCR2 to transactivation of the EGFR to increased IL-8/gro-α production may serve as an autocrine loop that augments the motility and anchorage-independent growth of the cancer cells. The paired appearance of increased EGFR and gro-α levels in a subset of human breast cancers detected by gene array technology (37) supports that this CXCR2/EGFR cooperation may indeed play a role in human malignancies and deserves to be analyzed systematically in other cancers. This appears particularly relevant, because the expression of high levels of EGFRs (38) and of IL-8 family chemokines (39, 40) have both been associated with poor outcome in a variety of human cancers. It remains to be seen whether there is a positive correlation between these two families of proteins in various cancers.

Stimulation of the CXCR1 or CXCR2, like that of other GPCRs, causes dissociation of the α and the β,γ subunits of the G-protein, which is pertussis toxin sensitive in the case of the IL-8Rs (12). The β,γ subunits of G_iα have multiple functions, including the activation of ras and the MAPK (41), and are essential for GPCR-mediated chemotaxis (13). Receptor kinases quickly phosphorylate activated IL-8Rs (27, 42), which leads to association of the receptor with β-arrestin and uncoupling from its G-protein. It has been shown for several GPCRs, among them the β₂-adrenergic receptor (25) and the CXCR1 (36) that β-arrestin recruits src family kinases to the receptor, which in turn activate tyrosine kinases.

However, we found no evidence of CXCR2-mediated src phosphorylation and only negligible (30%) increase in src activity using a commercial assay kit (Upstate Biotechnology). A similar lack of src activation following IL-8 stimulation of ovarian cancer cells has been reported previously (4). Finally, a mutant CXCR2, which lacks the last 12 aa and cannot be phosphorylated (27) and which poorly recruits arrestin to the plasma membrane (our unpublished observation), still causes migration and focus formation in NIH 3T3 cells. In contrast, PP2, a specific inhibitor of the kinase activity of src, which is often used to show src involvement (43), inhibited actin polymerization and endothelial cell migration. Because it also blocked EGF-mediated cell migration, src involvement appears to occur downstream of activation of the EGFR, which is known to activate src.

The sequence of events leading from stimulation of GPCRs to transactivation of the EGFR appears to involve activation of the Ras/MAPK pathway (44) and of rac (24). In agreement with this, it has been shown previously that activation of the CXCR2 causes activation of both the MAPK pathway (45) and of rac (2).

Fig. 8 depicts the sequence of events that occurs following stimulation with IL-8 in cells that express the EGFR. Initially activation of G_i leads to activation of cathepsin B, which can transiently be detected at the plasma membrane. However, it cannot be excluded that part of the signaling process occurs intracellularly, because it has been shown that the EGFR can be activated within endosomes (46), and similarly, that pro-HB-EGF can be cleaved to HB-EGF within endocytic vesicles (47). The early focal fusion of the intracellular vesicle membrane with the plasma membrane may contribute to the polarized membrane extension observed during chemotaxis (48). Although the specifics of this activation pathway are not understood at this point, it is clear that cathepsin B activity is necessary for CXCR2-mediated cell migration and involves activation of HB-EGF. The activation steps downstream of activation of the EGFR, including activation of MAPK, have been described in detail previously (reviewed in Ref. 49). In addition to the pathway depicted in this study, it is also known that both IL-8 (50) and EGFR activation (51) induce activation of NF-κB, which in turn leads to increased production of IL-8, thus causing a vicious cycle of autocrine stimulation.

It has long been known that leukocyte activation causes the release of granular enzymes in general, and specifically that IL-8 stimulates enzyme release from neutrophils (12). In particular, IL-8 causes release of β-hexoseaminidase (36), which, like cathepsin B, is stored in primary granules. Furthermore, release of active cathepsins including cathepsin B has been observed in monocytes/macrophages (52). Therefore, it does not surprise that IL-8 caused release of cathepsin B from RBL2H3 cells. The presence of cathepsin B on the cell surface of HMECs was less expected, because it is an endosomal and lysosomal protein in these cells (53). However, it has been suggested that insertion of recycling endosome membrane at the leading edge of a cell is not only required for phagocytosis, but also for cell motility (48). Thus, cell surface appearance of endosomal proteins and receptor internalization may be two different aspects of the same process, and receptor endocytosis appears to be necessary for CXCR2-dependent chemotaxis of adherent cells (54). In addition, it cannot be excluded that part of the cathepsin B-mediated response occurred within endosomes, which were transiently accessible to E-64 as they fused with the plasma membrane, and then were endocytosed together with ligand-bound CXCR2 and EGFRs. Signaling from such internalized receptors has found some attention lately (55).

We had not anticipated that the proteolytic activity of cathepsin B was involved in the HB-EGF-dependent cell migration initiated by IL-8. In previous reports in which GPCR activation caused transactivation of the EGFR, membrane metalloproteinases mediated the cleavage of pro-HB-EGF to HB-EGF (19, 44). Our results do not contradict these results. Although membrane metalloprotease inhibitors only partially blocked IL-8-mediated cell migration, this could be due to a different signaling cascade used by the CXCR2 or due to the use of different cell lines. Finally, cathepsin B and metalloproteases may form a proteolytic cascade. For instance, it has been shown that cathepsin B inactivates tissue inhibitors of membrane metalloprotease, thus leading to increased metalloprotease activity (56). Furthermore, there is precedence for growth factor activation as a result of cleavage by cathepsin B (57), and cathepsin B has been shown to cause peptide cleavage at the same sites as membrane metalloproteases (58). Clearly, the biochemistry involved in the proteolytic activation of HB-EGF deserves further investigation.

It has been described previously that HB-EGF-mediated cell migration depends on the interaction of HB-EGF with cell surface heparan sulfate (59). IL-8 similarly binds to heparin and heparan sulfates (60), which both increase IL-8 binding and enhance IL-8-dependent chemotaxis of leukocytes (60). Future investigation will have to show whether there is indeed a complex formation between these different components and their receptors.

Shed HB-EGF is able to stimulate EGFRs in an autocrine or paracrine fashion (19, 44) so that activation of the CXCR2 in cells that express HB-EGF, but no EGFR, such as monocytes, can activate the EGFR on surrounding endothelial or tumor cells. These cells will then show the effects of EGFR activation including accelerated proliferation (44), migratory capacity, and up-regulation of IL-8 and other NF-κB-responsive gene products. This mechanism may underlie the reported importance of monocyte activation by IL-8 in early atherosclerosis (61).

In summary, our results indicate that activation of the CXCR2 causes cathepsin B-dependent activation of HB-EGF, which in turn is responsible for cell migration. This signaling cascade may be important in an autocrine fashion in angiogenesis and in cancer cells that express the CXCR2, but may also function in a paracrine fashion.

We thank Vassily Portnoy for his technical assistance.