Arrestin Regulates MAPK Activation and Prevents NADPH Oxidase-dependent Death of Cells Expressing CXCR2*

Activation of CXCR2 IL-8 receptor leads to activation of extracellular signal-regulated kinases 1 and 2 (ERK1/2) and rapid receptor endocytosis. Co-immuno-precipitation and co-localization experiments showed that arrestin and CXCR2 form complexes with components of the ERK1/2 cascade following ligand stimulation. However, in contrast to the activation of the (cid:1) 2 - adrenergic receptor, arrestin was not necessary for ERK1/2 phosphorylation or receptor endocytosis. In contrast, (cid:1) -arrestin 1/2 double knockout cells showed greatly enhanced phosphorylation of ERK1/2, as well as phosphorylation of the stress kinases p38 and c-Jun N-terminal protein kinase. The stimulation of stress kinases in arrestin double knockout cells could be attenuated in the presence of diphenylene iodonium (DPI), an inhibitor of the NADPH oxidase, suggesting that reactive oxidant species (ROS) participated in mitogen-acti-vated protein kinase (MAPK) activation. ROS could indeed be detected in IL-8-stimulated (cid:1) -arrestin 1/2 knockout cells, and cytoplasmic Rac was translocated to the membrane fraction, which is a prerequisite for oxidant formation. The oxidative burst induced cell death withi n 6 h ofIL-8 stimulation of these cells, which could be prevented in the presence of DPI. These results indicate

show that CXCR2 forms complexes with arrestin, but that these complexes are not necessary for receptor internalization or ERK1/2 phosphorylation. In contrast, MAPK activation was enhanced in ␤-arrestin 1/2 double knockout cells to the point, where cells responded with cell death rather than cell proliferation.
CXCR2 is one of a number of G i -coupled GPCRs on neutrophils, which include CXCR1, the formyl peptide receptor, and C5a receptor. These last three receptors are all important for neutrophil chemotaxis in vitro and in vivo. The physiological role of CXCR2 on these cells is, however not clear. Neutrophils clearly express functional CXCR2 receptors, which respond to stimulation with calcium mobilization, actin polymerization, and moderate enzyme release (19), but show little effect on activation pathways important for the immune surveillance functions of the neutrophil. Stimulation of neutrophilic CXCR2 neither supports chemotaxis (20), nor does it cause a respiratory burst (21). There is no good explanation for this behavior other than that the rate of phosphorylation, which can be detected within seconds of stimulation, and consequent uncoupling from the G-protein may supercede the rate of assembly of the NADPH oxidase complexes.
While the NADPH oxidase system is best described in neutrophils (22), adherent cell types including endothelial and epithelial cells, smooth muscle cells and fibroblasts can produce intracellular oxidants following receptor activation (23)(24)(25). Although some of the components of these NADPH oxidase systems differ from those of the neutrophilic complexes (26), activation of all these systems depends on translocation of Rac from the cytoplasm to the cell membrane (27), and activation of a flavin-dependent membrane-bound cytochrome, which is inhibited by diphenylene iodonium (DPI) (28).
Stimulation of various cell types expressing CXCR2 (neutrophils and microvascular endothelial cells, which express CXCR2 constitutively, NIH3T3 cells and HEK293 cells transfected with CXCR2) failed to respond with a respiratory burst (21). 2 However, in ␤-arrestin1/2 double knockout fibroblasts stimulation with IL-8 caused a pronounced intracellular respiratory burst, which was responsible for later cell death. It therefore appears that arrestin is instrumental in suppressing or terminating activation of the respiratory burst in cells expressing CXCR2.
Cell Culture and Transfection-HEK293, human embryonic kidney epithelial cells, and SYF cells, mouse embryonic fibroblasts, which are immortalized triple knockouts for Src/Yes/Fyn (31) (both from ATCC), were grown in Dulbecco's modified Eagle's medium containing 10% fetal calf serum, and transfected with LipofectAMINE 2000 (Invitrogen, Carlsbad, CA). Stable cells lines were selected with 800 g/ml of G418. RBL2H3 cells were grown in RPMI containing 10% fetal calf serum and transfected by electroporation as described (30) and similarly selected with G418. MEF (mouse embryonic fibroblasts) ␤-arrestin 1/2 knockout cells as well as ␤-arrestin 1 or ␤-arrestin 2 single knockout MEFs and the wild-type parental cell line, were obtained from Dr. R. Lefkowitz (Duke University) and grown in Dulbecco's modified Eagle's medium containing 10% fetal calf serum and transiently transfected with Lipo-fectAMINE 2000. All cells were serum-starved overnight prior to IL-8 or SDF-1 stimulation.
Immunoprecipitation and Immunoblotting-For immunoprecipitation, monolayer cells were stimulated as described in the figure legends, then lysed with modified radioimmune precipitation assay buffer (50 mM Tris-HCl, pH 7.4, 10% glycerol, 1% Nonidet P-40, 150 mM NaCl, 5 mM MgCl 2 , 2 mM EDTA, 2 mM PMSF, 2 g/ml leupeptin, 2 g/ml aprotinin, 2 mM sodium pyrophosphate, 2 mM sodium vanadate, and 10 mM NaF) and clarified by centrifugation. The supernatants were incubated with primary antibodies for 2 h at 4°C, followed by capture of the immunocomplexes with protein A beads for 1 h at 4°C. The immunoprecipitates were washed twice with lysis buffer and once with phosphate-buffered saline to remove nonspecifically bound proteins. The bound proteins were analyzed with immunoblotting.
For immunoblotting, the clarified cell lysates were resolved by SDS-PAGE, transferred to nitrocellulose membrane, blocked with 3% dry milk in TBS-Tween, and exposed to specific primary antibodies as described for each experiment. Antibody binding was detected using horseradish peroxidase-conjugated goat anti-rabbit or anti-mouse secondary antibodies and enhanced chemiluminescence (ECL, Amersham Biosciences). Phosphoblots were re-probed with a second antibody, e.g. anti-ERK1/2 antibody to assure equal loading.
Subcellular Fractionation-For the isolation of membrane fractions, cells were stimulated for the indicated times and lysed in ice-cold lysis buffer (50 mM Tris-HCl, pH 7.4, 50 mM NaCl, 1 mM MgCl 2 , 2 mM EDTA, 1 mM DTT, 2 mM PMSF, 2 g/ml leupeptin, 2 g/ml aprotinin) by sonication. Nuclei and cell debris were removed by centrifugation at 500 ϫ g for 10 min. The supernatants were subjected to ultracentrifugation at 100,000 ϫ g for 40 min and the pellets were washed once. Supernatant fractions (cytoplasm) and pellet fractions (membranes) were separated on SDS polyacrylamide gels.
Raf Kinase Assay-Washed c-Raf immunoprecipitates were suspended in 40 l of kinase buffer (30 mM HEPES, pH7.4, 10 mM MnCl 2 , 5 mM MgCl 2 , 1 mM DTT, 2 mM Na 3 VO 4 , 25 mM ␤-glycerophosphate, 50 M ATP) containing 15 Ci of [␥-32 P]ATP and 0.4 g of purified kinaseinactive MEK1 (Upstate Biotechnologies), and then incubated at 30°C for 30 min. Kinase reactions were stopped by adding Laemmli sample buffer and boiling, followed by separation on SDS polyacrylamide gels, gel transfer, and autoradiography. To assess equal loading, membranes were probed with anti-c-Raf antibody and stained with the ECL kit.
Fluorescence Microscopy-For fluorescence microscopy, HEK 293 cells grown on collagen-coated glass coverslips were transiently cotransfected with different combination of plasmids as described in the figure legends. After stimulation with 100 nM IL-8 or SDF-1 for the times indicated at 37°C, cells were fixed for 25 min with 4% paraformaldehyde in phosphate-buffered saline and mounted with AntiFade (Molecular Probes, Eugene, OR). Z-stack images were taken on a Leica DM Erbe microscope (Leitz, Bannockburne, IL) connected with an ORCA camera (Hamamatsu Photonics, Hamamatsu City, Japan) and deconvoluted using Openlab 3.1.5 software (Improvision, Boston, MA).
To detect apoptosis an annexin V-fluorescein isothiocyanate/pro-2 I. U. Schraufstatter and M. Burger, unpublished observation. pidium iodide staining kit was used (Calbiochem) according to the manufacturer's instructions and green-versus red-stained cells were counted under a fluorescent microscope.
Detection of Intracellular Oxidant Formation-Two methods were used to detect intracellular oxidant formation. First, nitroblue tetrazolium (NBT, Sigma) was added at 1 mg/ml at the time of addition of IL-8, which leads to the formation of blue formazan in the presence of O 2 Ϫ (32), and cells were incubated for 90 min at 37°C. Second, membranes were prepared as described (33), and H 2 O 2 production was detected with the Amplex Red reagent (Molecular Probes). In short, serumstarved ␤-arrestin 1/2 knockout cells were stimulated for 10 min Ϯ 100 nM IL-8 in the presence or absence of DPI, trypsinized, centrifuged through serum-containing medium at 4°C, and resuspended in 400 l of 340 mM sucrose, 10 mM Tris, pH 7.1, 2 mM PMSF, 1 mM EDTA, 2 g/ml leupeptin, and 2 g/ml aprotinin. The samples were sonicated on ice (4 ϫ 15 s), and ultracentrifuged as described above. The pellets were rinsed twice with sucrose buffer, and centrifuged again under the same conditions. The pellets were resuspended in 100 l of sucrose buffer and homogenized with a Teflon pestle. The protein content was determined using the BCA reagent (Pierce), and 10 -15 g of protein were used for each assay. H 2 O 2 formation was detected with the Amplex red reagent according to the manufacturer's protocol in the presence of horseradish peroxidase and in the presence or absence of 1 mM NADPH. Fluorescence intensities at 530 nm excitation, 590 nm emission were measured on 100-l duplicates in a 96-well plate warmed to 37°C. Measurements were made at 10-min intervals for 30 min using a fluorescence plate reader (Fluoroscan, Packard, Meriden, CT). (34), we asked whether activation of CXCR2 caused translocation of arrestin and the components of the MAPK cascade to the plasma membrane. As expected, arrestin, and the kinases c-Raf and MEK, were translocated to the membrane fraction of CXCR2-expressing cells following stimulation with IL-8 (Fig. 1A). This translocation of arrestin was seen in HEK293 cells stably expressing CXCR2-GFP fusion protein (Fig. 1A) as well as in RBL2H3, rat basophilic leukemia cells (Fig. 1B), transfected with wild-type CXCR2. For comparison the same experiment was repeated in CXCR4-expressing cells, which showed only a rather subtle translocation of the same components to the plasma membrane ( Fig. 1C).

IL-8-mediated Translocation of Arrestin and ERK1/2 Cascade Components in Cells Expressing CXCR2-Since arrestin translocation plays an important role in MAPK activation and endocytosis of a number of GPCRs
Activation of MAPK-Arrestin translocation has been shown to play an important role in MAPK activation of many GPCRs, e.g. the ␤ 2 -adrenergic receptor (35), the AT1a-angiotensin receptor (36), and the protease-activated receptor 2 (PAR-2) thrombin receptor (37). Therefore MAPK phosphorylation was assessed in CXCR2-expressing cells stimulated with IL-8. IL-8 was found to induce a robust, but transient ERK1/2 activation in all cells types tested (HEK293, RBL2H3, and SYF cells) ( Fig.  2A), which was maximal around 2 min of stimulation. This activation did not depend on Src activity as it did not appear decreased in SYF cells, which are deficient of Src, Yes, and Fyn. Although activation of CXCR2 can cause transactivation of the EGF receptor (38), this pathway played only a minor role in signaling to ERK1/2, and inhibition of the EGF receptor with AG1478 showed no effect on ERK1/2 phosphorylation (results not shown). In contrast, the ERK1/2 response was almost entirely inhibited by a Raf-1 kinase inhibitor (Fig. 2B). Kinasedeficient Raf (Raf K375M) similarly attenuated ERK1/2 phosphorylation (Fig. 2B). Involvement of c-Raf could further be shown directly in Raf activity assay, in which MEK1 is phosphorylated in the presence of activated c-Raf (Fig. 2C). Since c-Raf has been shown to be part of multicomponent complexes formed between the MAPK cascade, arrestin, and some GPCRs, we next asked whether this was the case for CXCR2.
Coimmunoprecipitation and Colocalization of CXCR2, Arrestin, and Components of the MAPK Cascade-Since activation of the Raf/MAPK cascade has been shown to be mediated through complex formation with arrestin, e.g. in the ␤ 2 -adrenergic receptor system (1), we determined whether IL-8 would induce similar complexes. For this purpose, HEK293 cells expressing CXCR2-GFP were transiently transfected with ␤-arrestin 2-FLAG, Myc-c-Raf, and ERK2-RFP expression plasmids. After stimulation with IL-8, the amount of ERK2 and c-Raf in complex with ␤-arrestin 2 and CXCR2 increased considerably as shown in Fig. 3, A and B, and a high level of phosphorylation of ERK2 was detected in the complexes. This could be shown by immunoprecipitation with either an antibody, which recognizes the FLAG tag fused to arrestin (Fig. 3A) or an antibody that recognizes GFP, which is fused to CXCR2 (Fig. 3B). Similarly, components of the ERK cascade could be coimmunoprecipitated with GFP-or FLAG-arrestin following IL-8 stimulation of the untagged CXCR2, indicating that it was specific for the receptor and not a function of the GFP (Fig. 3C). These results provide strong evidence that CXCR2 stimulation causes recruitment of complexes consisting of ␤-arrestin, c-Raf, and ERK2, in which arrestin appears to function as a scaffold for signal transduction. As shown in Fig. 3D, the absence of ERK2 had no effect on the association of c-Raf with ␤-arrestin 2, whereas c-Raf expression significantly enhanced the precipitation of phospho-ERK2 with ␤-arrestin 2. Raf kinase inhibitor or dominant-negative c-Raf mutant both attenuated the complex formation (Fig. 3D). These results indicated that ERK2 activation and association with ␤-arrestin 2 depended on c-Raf, and that the interaction of c-Raf with ␤-arrestin 2 facilitated ERK2 activation, suggesting an integral role of c-Raf in these complexes, which was compatible with the fact that Raf-1 inhibitor almost totally blocked IL-8 induced ERK1/2 activation (see 2B). Complex formation could also be visualized by fluorescence microscopy of cells transfected with CXCR2-GFP, ␤-arrestin 2, and ERK2-RFP or with CXCR2, ␤-arrestin 2-GFP, and ERK2-RFP. In unstimulated cells CXCR2-GFP was located in the plasma membrane, GFP-fused arrestin homogeneously in the cytoplasm and ERK2-RFP homogeneously in the cytoplasm and to a lesser extent in the nucleus (results not shown). Following stimulation with IL-8, co-localization between CXCR2 and ERK2 or between ␤-arrestin 2 and ERK2 could be detected in intracellular vesicles (Fig. 3E.) These findings together with previous reports showing that GPCR-arrestin complexes are essential for MAPK activation (10), suggested that arrestin played an integral role in CXCR2-mediated MAPK activation. One known function of ␤-arrestin in complex with seven transmembrane receptors and components of the MAPK cascade, is to retain these complexes in the cytoplasm rather than to allow translocation of MAPK to the nucleus, where it induces transcriptional activation (36). However, in CXCR2expressing cells a major fraction of ERK1/2 phosphorylation following stimulation with IL-8 was detected in the nuclear fraction (Fig. 3F). So the function of CXCR2-mediated ERK1/2 nuclear translocation remains to be investigated.
MAPK Activation in ␤-Arrestin 1/2 Knockout Cells-The finding of CXCR2/␤-arrestin/MAPK complexes together with previous reports showing that GPCR-␤-arrestin complexes are essential for MAPK activation suggested that ␤-arrestin play an essential role in CXCR2-mediated MAPK activation. To prove this point ERK1/2 phosphorylation was determined in MEF ␤-arrestin 1/2-double knockout cells (13) stimulated with IL-8. These cells express the mouse CXCR2 constitutively as was verified by reverse transcription-PCR. Although this receptor has a higher K d for human IL-8 than human CXCR2 (4), the concentration of IL-8 used here fully activates the mouse receptor. Surprisingly, when these cells were stimulated with IL-8, a strong and prolonged ERK1/2 phosphorylation was observed (Fig. 4A), which was more pronounced than any response we had seen in any CXCR2-expressing cell type previously, and much higher than in the parent, arrestin-competent MEFs (Fig. 4A). Transient transfection with human CXCR2 further augmented the response despite a low level of transfection efficiency (Fig. 4A). In addition other MAPK pathways, the stress kinases p38 and JNK that are only poorly activated by IL-8 stimulation of CXCR2-transfected cells, were highly phosphorylated in the arrestin knockout cells (Fig. 4, B and C). For comparison phosphorylation of p38 and JNK is shown in the parental MEF, which contains normal ␤-arrestin levels (Fig. 4, B and C). It therefore appears that the complexes formed between CXCR2, arrestin, and the components of the MAPK cascade, attenuate the CXCR2-mediated MAPK response rather than augment it.
Activation of MAPK was associated with membrane translocation of Rac in IL-8 stimulated arrestin knockout cells as shown in Fig. 4D. In contrast, in the parent MEFs no Rac translocation was observed (Fig. 4D). Similarly, in a human microvascular endothelial cells line (HMECs), which constitutively expresses CXCR2 and shows lamellipodia formation in the presence of IL-8 (39), a morphological indicator of Rac activation (40), Rac translocation to the membrane was minimal (Fig. 4D).
Interestingly Rac, which is normally absent from the membrane fraction in unstimulated, arrestin-competent cells, could already be detected to a lesser degree in the membrane fraction in unstimulated ␤-arrestin Ϫ/Ϫ cells.

IL-8 Causes Cell Death in ␤-Arrestin 1/2 Knockout Cells-
The activation of MAP kinases was so excessive in ␤-arrestin 1/2 knockout cells that it appeared likely that it might lead to cell death (41,42) rather than cell activation. Indeed stimulation of these cells with IL-8 caused apoptotic appearance with cells rounding up and de-adhering over a period of several hours (Fig. 6A). This was confirmed by annexin V staining. At 3 h 11% of the cells stained positive for annexin V, an early indicator of apoptosis, whereas only 2.5% of all cells took up propidium iodide, an indicator of late apoptosis or necrosis. By 5 h 60.9% of the annexin V-positive cells were also stained by propidium iodide (results not shown).
Because of the prominent translocation of Rac to the membrane fraction in these cells, we hypothesized that oxidative stress due to activation of the NADPH oxidase system was the pathway that mediated this behavior. When an inhibitor of the oxidative burst, DPI, was added during the incubation with IL-8, cell rounding could be prevented (Fig. 6A). IL-8-mediated cell death was not observed in MEFs, in which one of the two ␤-arrestins had been knocked out (Fig. 6A), indicating that either of the ␤-arrestins could protect from cell death.
Furthermore, DPI attenuated phosphorylation of p38 (Fig.  6B) suggesting that activation of this stress kinase was at least partially mediated by ROS. DPI showed little effect on ERK1/2 and JNK phosphorylation (results not shown), suggesting these kinases were activated directly and not secondary to oxidant production.
Detection of ROS in IL-8-stimulated ␤-Arrestin 1/2 Knockout Cells-To verify an oxidative burst, two methods were used. First, the formation of the blue NBT-diformazan from NBT was followed microscopically. This method is used as a clinical test for NADPH oxidase function in neutrophils, but it is usually not sensitive enough to detect ROS formation in non-phagocytic cells. However, in ␤-arrestin 1/2 knockout cells IL-8 caused intensive blue staining in the presence of NBT (Fig. 6C). This was not observed in any other cell type stimulated with IL-8, including neutrophils, which normally form NBT-diformazan, when stimulated with N-formyl peptide or C5a. The formation of NBT-diformazan in IL-8 stimulated ␤-arrestindeficient cells could be prevented in the presence of DPI (Fig.  6C). Again, either ␤-arrestin 1 or ␤-arrestin 2 could prevent oxidant formation as shown in Fig. 6C. In contrast, stimulation with SDF-1 failed to induce NBT-diformazan formation (Fig.  6C) and did not result in cell death (results not shown).
Second, the formation of red fluorescent resorufin from am-plex red was detected in the membrane fraction of MEF ␤-arrestin 1/2 knockout cells (43). Even cell membrane fractions of unstimulated ␤-arrestin 1/2 knockout cells showed a considerable amount of NADPH-dependent H 2 O 2 formation (Fig. 6D). This was further increased in the presence of IL-8 (Fig. 6D), and attenuated in the presence of DPI (Fig. 6D). The high baseline formation of H 2 O 2 is consistent with the presence of a considerable portion of Rac in the membrane fraction of these cells even in the absence of stimulation. DISCUSSION We show here that CXCR2 forms complexes with ␤-arrestin and components of the MAPK cascade. However, in contrast to the situation for the ␤ 2 -adrenergic receptor or the AT1a-receptors (10, 36), these complexes appear to attenuate MAPK activation rather than promote it. ␤-arrestin-deficient cells showed a markedly increased activation of the various MAP kinases. In were stimulated with 100 nM IL-8 for the indicated times, cell lysates were separated into membrane and cytoplasmic fractions as in Fig. 1, and blotted with anti-Rac antibody. HMECs were chosen, because we had shown previously that IL-8 induces lamellipodia formation in these cells (39), which is Rac1-mediated. E, ␤-arrestin 1/2 Ϫ/Ϫ MEFs were transiently transfected with CXCR2-GFP, stimulated with 100 nM IL-8 for 5 or 10 min, fixed in paraformaldehyde, and green fluorescent images were taken (ϫ100 objective). All experiments were performed at least three times with similar results. particular the stress kinases p38 and JNK, which are barely activated by stimulation of CXCR2 in a variety of cell types, showed a large response in the ␤-arrestin knockout cells. Activation of these stress kinases can lead to cell death (44,45), and even ERK1/2 activation can promote cell death rather than proliferation, when it is stimulated excessively (46). A report, which just appeared in press, indicates that the activation of several GPCRs in ␤-arrestin 1/2 knockout cells results in apoptotic cell death (47). These GPCRs included the formyl peptide receptor, the angiotensin II type 1A receptor and CXCR2, but not the ␤-adrenergic receptor and CXCR4 (47). Our results confirm this report for CXCR2 and CXCR4. While the cited report recognized activation of numerous signaling pathways in stimulated ␤-arrestin-deficient cells including PI3k, Src, MAPK, and cytochrome c leakage from mitochondria (47), it did not investigate the mechanism, which causes cell death. Our results clearly indicate that ␤-arrestin-deficient cells undergo oxidative stress when stimulated with IL-8, and that ROS are responsible for the cell death. This was first suggested by the protective effect of DPI, which inhibits NADPH oxidase, and then confirmed by the ability to detect ROS in IL-8-stimulated ␤-arrestin 1/2 Ϫ/Ϫ MEFs. Furthermore, IL-8 caused translocation of Rac to the membrane fraction, which is an essential component of the NADPH oxidase complex (22,48). Although some oxidant formation was already observed in unstimulated ␤-arrestin knockout cells, in particular if membrane fractions were used, in which cellular antioxidant enzymes would have been removed, this amount appeared below the threshold dose that caused cell death.
Activation of CXCR2 caused translocation of Raf and Rac to the membrane fraction. In the presence of ␤-arrestin this led to a transient activation of ERK1/2 and Rac (39). When ␤-arrestin was absent, the activation was sustained and ultimately caused cell death. Although we did not investigate CXCR4mediated response in much detail, it appears that CXCR4 behaves like the ␤ 2 -adrenergic receptor, where ␤-arrestin is necessary for receptor endocytosis and enhances MAPK activation. SDF-1, which did not cause ROS production, failed to recruit Rac. It is worth noting that only GPCRs, which were endocytosed in an arrestin-independent fashion, like CXCR2, but not like CXCR4, were capable of mediating cell death of arrestin deficient cells (47). Since Rac plays an essential role in endocytosis (49) as well as in the activation of the oxidative burst, these two functions appear intimately linked. One may speculate that GPCRs that cause apoptosis in ␤-arrestin-deficient cells are generally able to recruit Rac.
Furthermore, these results are compatible with a recent report in which it was shown that ␤-arrestin 1 is anti-apoptotic following stimulation of the insulin-like growth factor 1 receptor in a phosphatidylinositol kinase (PI 3-kinase)-mediated mechanism (50). Because PI 3-kinase is necessary for Rac activation (51), it appears that certain receptor kinases may activate the same ␤-arrestin-dependent pathway as the Racactivating GPCRs.
It has been shown that endocytosis of CXCR2 is dependent on dynamin (52). At the time of the publication of this report, dynamin was primarily seen as a component of clathrin-mediated endocytosis. While dynamin certainly plays a role in clathrin-mediated endocytosis, this is not its only function. Both caveolar uptake and type 1 phagocytosis, which is mediated by Rac, are dynamin-dependent (53,54). While we cannot exclude that CXCR2 is endocytosed in an ␤-arrestin/clathrin-dependent fashion in normal cells, there exists an alternate internalization pathway in ␤-arrestin-deficient cells, which presumably is dynamin/Rac dependent (54). In phagocytic cells this pathway is referred to as phagocytosis type 1 (55).
CXCR2, C5a receptor and formyl peptide receptors, are all expressed in neutrophils and can all activate Rac. This causes a respiratory burst for the formyl peptide receptor and C5a-receptors, but not following stimulation of CXCR2 (21). If the oxidative burst, the consequence of Rac translocation, is terminated or inhibited in the presence of ␤-arrestin, the rate of GPCR phosphorylation and of ␤-arrestin recruitment, will determine the extent of the oxidative burst. Phosphorylation of CXCR2 is extremely rapid, -it appears complete within 15 s (30), and maximal complex formation with arrestin was detected by 2 min, which is undoubtedly an overestimate of the time required due to the poor time resolution of preparing cell lysates form adherent cells. It therefore appears likely that the rate of complex formation with arrestin supercedes the rate of assembly of the NADPH FIG. 5. Effect of dominant-negative ␤-arrestin on MAPK activation and receptor internalization in HEK293 cells expressing CXCR2 or CXCR4. A, plasmid encoding ␤-arrestin 319 -418 (5 g of DNA) and CXCR2-GFP (1 g of DNA) was transfected into HEK293 cells as indicated. Cells were then treated with 100 nM IL-8 for the indicated times. Cell lysates were prepared, and ERK1/2 phosphorylation was detected by Western blotting (one experiment representative of four). B, HEK293 cells expressing CXCR4-GFP were transfected with ␤-arrestin 319 -418 as indicated and stimulated with 50 nM SDF-1 for 5 min. ERK1/2 phosphorylation was detected as described above. C, HEK293 cells were transiently transfected with CXCR2-GFP or CXCR4-GFP and dominant-negative ␤-arrestin. Following stimulation with 100 nM IL-8 or 50 nM SDF-1 for 15 min GFP fluorescent images were obtained with a confocal microscope (ϫ63 objective). oxidase in the case of CXCR2. Such a system allows the fine tuning of activation of the NADPH oxidase, which is essential for the protection from bacterial infection, but at the same time deleterious to oxidant producing cells.
Since all cells express ␤-arrestin, it will have to be determined whether there are situations in which ␤-arrestin concentrations are too low to limit the NADPH oxidase activation following stimulation with GPCR ligands. It is intriguing in this respect that cellular arrestin levels can be considerably decreased due to ubiquitin-mediated protein degradation in cells that show excessive activation of receptor-tyrosine kinases (56), which are highly activated in many cancer cells. It will have to be shown, whether these cells undergo increased oxidative stress due to insufficient ␤-arrestin function. Such a process would explain the genetic instability of aggressive cancers. It is known that these cells produce higher than normal concentrations of ROS (57), but it remains to be shown whether insufficient ␤-arrestin plays a role in this process.
It is clear that CXCR2, ␤-arrestin, and components of the MAPK cascade form complexes following the stimulation with IL-8. This poses the question, how similar molecular complexes can fulfill opposite functions, which leads to attenuation of ERK1/2 phosphorylation in the case of CXCR2 and to activation of ERK1/2 in the ␤ 2 -adrenergic system. One may speculate that CXCR2 complexes may contain additional inhibitory factors. Indeed several proteins have been shown to bind to the C-terminal tail of CXCR2 (52,58), which is also the ␤-arrestin binding region. Protein phosphatase 2A core enzyme, which is one of these proteins (52), is interesting in this respect, since it is known to dephosphorylate MAPKs (59,60). Since it also binds to the ␤ 2 -adrenergic receptor (61), its presence does not explain the different behavior of this receptor, however. A second possibility is that heat shock protein 70 (Hsp70), which can be co-immunoprecipitated with CXCR2 (58), prevents activation of the stress kinases p38 and JNK in ␤-arrestin-containing cells. Such an apoptosis inhibitory function for Hsp70 has been described (62). It will have to be shown, whether the binding of Hsp70 depends on the presence of ␤-arrestin.
Although many pathways lead from stimulation of a seven transmembrane receptor to MAPK activation, the majority of the ERK1/2 phosphorylation following IL-8 stimulation of CXCR2 depended on activation of a Ras/Raf pathway independent of the cell type tested. It was associated with translocation of Raf to the membrane fraction, a prerequisite for Raf activation (63). Although IL-8 can transactivate the EGF receptor (38,64), and this transactivation is important for CXCR2-dependent cell migration of endothelial cells, this pathway made only a minor contribution to the overall ERK1/2 phosphorylation in the cell types (HEK293 and SYF) used here. Decreases in ERK1/2 phosphorylation were barely perceptible in AG1478-treated cells. Similarly, inhibition of Src with PP1, or using the Src family-deficient SYF cells, had no effect on ERK1/2 phosphorylation.
FIG. 6. ␤-arrestin 1/2 ؊/؊ MEFs stimulated with 100 nM IL-8 undergo a lethal oxidative burst. A, ␤-arrestin 1/2 Ϫ/Ϫ MEFs as well as ␤-arrestin 1 or ␤-arrestin 2 single knockout MEFs were stimulated with 100 nM IL-8 for 6 h at 37°C as indicated. Some samples were preincubated with 5 M DPI for 30 min prior to the addition of IL-8. After 6 h phase contrast images were taken (ϫ20 objective). B, effect of ROS inhibition on MAPK activation of IL-8-stimulated ␤-arrestin 1/2 Ϫ/Ϫ MEFs. Cells were preincubated with 5 M DPI for 30 min and stimulated with IL-8. Cell lysates were blotted with antiphospho-p38 as labeled. C, ␤-ar-restin1/2 Ϫ/Ϫ cells were treated with DPI and 100 nM IL-8 or 50 nM SDF-1 as above. NBT in phosphate-buffered saline was added, and the cells were incubated for 90 min at 37°C. Images were taken through a ϫ20 objective. D, H 2 O 2 production by plasma membranes derived from ␤-arrestin 1/2 Ϫ/Ϫ MEFs. ␤-Arrestin 1/2 Ϫ/Ϫ MEFs were stimulated with 100 nM IL-8 for 10 min at 37°C in the presence or absence of 5 M DPI. H 2 O 2 production was detected for 30 min on cell membranes (prepared as described under "Experimental Procedures") in the presence or absence of 1 mM NADPH, which is a cofactor of the NADPH oxidase system. One experiment in triplicate is representative of three experiments.