Leukocyte-Endothelium Interaction Promotes SDF-1-dependent Polarization of CXCR4* □ S

Chemokine-driven migration is accompanied by polarization of the cell body and of the intracellular signaling machinery. The extent to which chemokine receptors polarize during chemotaxis is currently unclear. To analyze the distribution of the chemokine receptor CXCR4 during SDF-1 (CXCL12)-induced chemotaxis, we retrovirally expressed a CXCR4-GFP fusion protein in the CXCR4-deficient human hematopoietic progenitor cell line KG1a. This KG1a CXCR4-GFP cell line showed full restoration of SDF-1 responsiveness in assays de-tecting activation of ERK1/2 phosphorylation, actin polymerization, adhesion to endothelium under conditions of physiological flow, and (transendothelial) chemotaxis. When adhered to cytokine-activated endothelium in the absence of SDF-1, CXCR4 did not localize to the leading edge of the cell but was uniformly distrib-uted over the plasma membrane. In contrast, when SDF-1 was immobilized on cytokine-activated endothelium, the CXCR4-GFP receptors that were present on the cell surface markedly redistributed to the leading edge of migrating cells. In addition, CXCR4-GFP co-lo-calized and subsequently exposed to 100 ng/ml SDF-1 for the indicated times. Western blot analysis shows phosphorylation of ERK1/2 ( pERK ) at 1–2 min in KG1a CXCR4-GFP cells, whereas the KG1a wt cells did not induce phosphorylation of ERK1/2 upon SDF-1 exposure. Addition of medium alone did not influ-ence the phosphorylation of ERK1/2 in both cell types. Lower blots for total ERK2 confirm equal protein loading. Results are representative for two independent experiments. of ICAM-3 of

Recent studies on interactions between peripheral blood leukocytes and vascular endothelium have provided much insight into the molecular details of chemokine-induced cell trafficking from the circulation to the tissues (1,2). Although much is known about the migration of granulocytes and T cells, the factors that control the transendothelial migration of primary human hematopoietic progenitor cells, i.e. CD34 ϩ cells, are still poorly understood. In the process of homing, transplanted CD34 ϩ cells migrate from the peripheral blood across the endothelium back to their niche in the bone marrow. Stromal cell-derived factor-1␣ (CXCL12) is the major chemokine involved in the homing of primary human CD34 ϩ cells but is also a chemoattractant for other leukocytes, such as monocytes and B and T lymphocytes (3)(4)(5).
SDF-1 is produced by several stromal cell types, including those of the bone marrow (3), and signals through the Gprotein-coupled receptor CXCR4 (6). It has been shown that SDF-1 stimulates integrin-mediated arrest of CD34 ϩ cells and also of T lymphocytes on vascular endothelium under flow, and that SDF-1 is constitutively expressed on human bone marrow endothelium (2,7). Moreover, SDF-1 activates the integrins LFA-1, VLA-4, and VLA-5 on immature CD34 ϩ cells (8). Thus, the interplay between SDF-1, presented on the bone marrow endothelium, and adhesion molecules on the CD34 ϩ cells represents the main driving force for the homing process.
Although directional migration of leukocytes is accompanied by polarization of the cell body and of the actin cytoskeleton, redistribution of chemokine receptors has been the subject of contradictory reports. Studies with C5a receptor-GFP 1 fusion proteins in neutrophils or cAMP receptor-GFP fusion proteins in Dictyostelium cells show uniform distribution of these receptors over the plasma membrane in cells that migrate to chemoattractants (9,10). Recent evidence indicates that differential distribution of G␤␥-subunits leads to amplification of the chemokine receptor signaling (11). Other chemokine receptors, such as the formylmethionylleucylphenylalanine receptor on neutrophils and CCR2 and CCR5 on T lymphocytes, are distributed to the leading edge upon exposure to their corresponding chemokines (12,13). In B lymphocytes, SDF-1 induces polarization of CXCR4 to the leading edge of the cell (14). In addition, it has been reported that SDF-1 can bind to fibronectin and subsequently induces a polarized distribution of CXCR4 on adherent T lymphocytes (15). However, these studies were all performed with artificial substrates, and to our knowledge it is unclear whether leukocytes that adhere to activated endothelial cells redistribute chemokine receptors, such as CXCR4, over their surface to drive directional migration. In addition, there is no information on the distribution and localization of CXCR4 in human hematopoietic progenitor cells.
To study SDF-1-CXCR4 signaling and the distribution of CXCR4 during the interaction of human hematopoietic progenitor cells with human bone marrow endothelium in more detail, we generated a CXCR4-GFP fusion construct and retrovirally expressed the protein into the CXCR4-deficient CD34 ϩ progenitor cell model KG1a. The CXCR4-GFP protein was functionally expressed at the cell surface and mediated SDF-1-dependent actin polymerization, ERK1/2 activation, adhesion to endothelium under flow, and directional migration. By using this cell model, we found that SDF-1, immobilized on activated human bone marrow endothelium to mimic the in vivo situation, induced formation of lamellipodia and membrane ruffles and a marked redistribution of surface-expressed CXCR4 to the leading edge of the cell. In addition, CXCR4 co-localized with lipid rafts at the leading edge, at the sites of contact with the endothelial apical surface. In conclusion, these findings suggest that SDF-1, when presented by the endothelium, induces CXCR4 polarization during directional and transendothelial migration of human hematopoietic progenitor cells.
Isolation of CD34 ϩ Hematopoietic Progenitor Cells-Cord blood was collected after delivery, according to the guidelines of Eurocord (Leiden, The Netherlands). Cord blood CD34 ϩ cells were isolated with the VarioMacs system (Miltenyi Biotec GmbH, Gladbach, Germany) as described (4). At least 95% of the cells from cord blood expressed CD34 as determined by fluorescence-activated cell sorter with CD34 mAb (number 581, Beckman Coulter).
Construction of the CXCR4-GFP-The full-length cDNA encoding CXCR4 was obtained from the National Institutes of Health AIDS Research and Reference Reagent Program, operated by McKesson Bioservices (Rockville, MD). The CXCR4 cDNA was subcloned as an EcoRI-ApaI fragment into the N2pEGFP vector (Clontech, Palo Alto, CA). The insert was then cloned into the EcoRI-NotI sites of the modified vector LZRS-IRES-Zeocin (17). The resulting construct, LZRS-CXCR4-GFP-IRES-Zeocin, was transfected into amphotropic Phoenix packaging cells (18) by means of the calcium-phosphate transfection system (Invitrogen) to produce retroviruses. KG1a cells were transduced with viruscontaining supernatant in the presence of 10 g/ml N-[1-(2,3-dioleoyloxy)propyl]-N,N,N-trimethylammonium methylsulfate (Roche Applied Science) and were centrifuged twice at 2000 rpm for 30 min. After 6 h, the supernatant was replaced with fresh medium, and the cells were allowed to recover overnight. This procedure was repeated twice on 2 consecutive days. Transduced cells were sorted for GFP expression by FACStar (BD Biosciences).
Transmigration Assay-Transmigration assays were performed in 6.5-mm, 5-m pore Transwell plates (Corning Costar, Cambridge, MA). HBMEC were plated at 50,000 cells/Transwell on FN-coated filters. Non-adherent cells were removed after 18 h. HBMEC were cultured to confluency and stimulated overnight with TNF-␣ (10 ng/ml). Freshly isolated CD34 ϩ cells or cell line cells (100,000) were added to the upper compartment, and the assay was performed as described previously (19). After the assays, the filters were fixed and stained with Texas Red-phalloidin to confirm the integrity of the HBMEC monolayer by CLSM.
Immunocytochemistry-Cells were fixed with 2% paraformaldehyde and immunostained with Abs or DiI as indicated in figure legends, followed by staining with fluorescently labeled secondary Abs (10 g/ ml) when necessary. Cells were permeabilized with 0.5% Triton X-100 when indicated. F-actin was visualized with Texas Red-phalloidin (1 unit/ml). Images were recorded with a Zeiss LSM510 confocal laserscanning microscope (CLSM) with appropriate filter settings. Crosstalk between the green and red channel was avoided by the use of sequential scanning. Time-lapse microscopy was performed at 37°C in a temperature-controlled incubation chamber. FN-coated glass coverslips or confluent endothelial monolayers cultured on FN-coated glass coverslips and stimulated overnight with TNF-␣ (10 ng/ml) were used. FN or endothelium was preincubated for 30 min with 100 ng/ml SDF-1 at 37°C, and medium was refreshed at the start of the experiments, as described previously (2,8). SDF-1 presence on endothelium was checked by CLSM with anti-SDF-1 polyclonal Ab (PeproTech). 200,000 KG1a wild type (wt) or KG1a CXCR4-GFP cells were added to the incubation chamber, and cell migration was recorded for at least 30 min. Imaging interval between recordings was ϳ10 s.
Quantification of CXCR4 Polarization-Cells were incubated on activated endothelium, which was coated or not with SDF-1, for 30 min at 37°C. Then cells were fixed with 2% paraformaldehyde and immunostained with ICAM-3-PE mAbs. ICAM-3 was used to determine the back end in polarized cells, and four different microscopic fields per glass coverslip were counted. By determine the back end with ICAM-3 expression and the leading edge with CXCR4-GFP expression, the amount of polarized CXCR4-GFP cells was quantitated.
CXCR4 Internalization-Following the indicated preincubations, the cells were fixed and not permeabilized, subsequently washed and kept in PBS containing 0.25% BSA and 1 mM calcium with CXCR4-PE Ab (1:4) for 30 min at 4°C, and then washed with a 30-fold excess of ice-cold PBS with 0.25% BSA. The relative fluorescence intensity was measured by flow cytometry (FACScan, BD Biosciences).
Adhesion of Cells under Flow-Perfusions under steady flow (0.4 dynes/cm 2 ) were performed in a modified version of a transparent parallel plate perfusion chamber as described previously by Ulfman et al. (20). The endothelial monolayer was cultured to confluency on a gelatin-coated glass coverslip. Monolayers were pretreated overnight with TNF-␣ (10 ng/ml). Additional incubation with SDF-1 (100 ng/ml) was done at 37°C for 30 min. Migration medium with 2 ϫ 10 6 KG1a wt or KG1a CXCR4-GFP cells/ml was perfused. After 5 min, the chamber was flushed with migration medium to wash out all non-adhered cells, and thereafter at least 20 areas of the perfusion surface were analyzed with the image analysis software Optimas 6.1 (Media Cybernetics Systems, Silver Spring, MD). Adhesion of CD-treated cells to TNF-␣treated endothelium was determined under static conditions, after three washing steps with migration medium and by counting the adhered cells per field of view, and 4 fields of view were analyzed per experiment.
Actin Polymerization Assay-Cells were washed three times in assay medium and resuspended in a concentration of 1 ϫ 10 6 cells/ml. Prior to the experiment, the cells and SDF-1 were incubated separately at 37°C for 10 min. SDF-1 (100 ng/ml) was added to the cell suspension, and at indicated time points 100 l of cell solution was transferred to 100 l of fixation solution (Intraprep Permeabilization Reagent, Beckman Coulter). Cells were incubated in the fixation solution for at least 15 min. Thereafter, the cells were centrifuged and resuspended in 100 l of permeabilization reagent (Intraprep Permeabilization Reagent). After 5 min, 1 unit/ml Texas Red-phalloidin was added to visualize the F-actin. After 20 min the cells were centrifuged and resuspended in PBS with 0.25% BSA. Mean fluorescence intensity was measured by FACScan. The distribution of F-actin was analyzed on cytospins of 50,000 cells. Images were analyzed by CLSM.
Statistics-Student's t test for paired samples (two-tailed) was used for statistical analysis. Student's t test for independent samples was used where indicated.

RESULTS
To gain insight in the dynamics and distribution of the receptor for SDF-1, CXCR4, during hematopoietic progenitor cell-endothelium interactions, we generated a cell line that stably expressed a CXCR4-GFP fusion protein. For this approach we used the human hematopoietic progenitor cell line KG1a, which lacks endogenous expression of CXCR4 but does express the progenitor cell marker CD34 (21). Flow cytometric analysis of cells transduced with the CXCR4-GFP encoding retrovirus confirmed stable expression of the CXCR4-GFP protein in the KG1a cells (Fig. 1A). CLSM revealed that the CXCR4-GFP was localized both at the plasma membrane and in sub-membranous vesicles (Fig. 1B). Flow cytometric analysis showed that the CXCR4-GFP fusion protein was recognized by monoclonal antibodies against CXCR4. Moreover, SDF-1 induced rapid internalization of CXCR4-GFP (Fig. 1C).
Chemokine-induced intracellular signaling via G-proteincoupled receptors, including CXCR4, has been studied extensively (9,22). To confirm that the transduced CXCR4-GFP fusion protein was functional in mediating intracellular signaling, we analyzed the phosphorylation of ERK1/2 following SDF-1 stimulation (23). We observed rapid maximal phosphorylation of ERK1/2 in the KG1a CXCR4-GFP cells after 1 min of SDF-1 treatment (Fig. 2). Medium changes did not induce increased phosphorylation of ERK1/2 in either cell type, whereas fetal calf serum, used as a positive control, induced ERK1/2 phosphorylation in the KG1a wt cells as well as in the KG1a CXCR4-GFP-transduced cells (data not shown). p38 mitogen-activated protein kinase phosphorylation was observed in KG1a CXCR4-GFP cells after 15-30 min of SDF-1 stimulation, whereas the wt cells did not show p38 phosphorylation upon SDF-1 stimulation (data not shown). These findings indicate that the transduced fusion protein is capable of inducing intracellular signaling toward ERK1/2 and p38.
Chemokine-induced directional migration is accompanied by localized actin polymerization, mainly at the leading edge of the cell, to induce lamellipodial protrusions and to drive directional migration (24,25). The KG1a CXCR4-GFP cells showed a rapid and significant actin polymerization upon exposure to SDF-1, whereas the KG1a wt cells did not (Fig. 3A). The kinetics and extent of SDF-1-induced actin polymerization in the KG1a CXCR4-GFP cells were similar as in primary human CD34 ϩ cells (24). CLSM analysis confirmed these results (Fig.  3B) and underscored the observation that only the CXCR4-GFP-positive cells responded to SDF-1 by polymerizing and polarizing their actin cytoskeleton. The above data show that transduction of CXCR4 is sufficient to restore SDF-1-responsiveness in these cells.
Several groups have reported that SDF-1 can be efficiently presented by fibronectin to migrating T lymphocytes and induces redistribution of CXCR4 to the leading edge of the cell, suggesting that immobilized SDF-1 promotes directional cell migration and receptor polarization (5,7,15). Analysis of CXCR4-GFP redistribution on immobilized SDF-1 on fibronectin-coated surfaces showed polarization of CXCR4-GFP (Fig.  4A). Labeling of the cells with the membrane probe DiI showed that the polarization of CXCR4-GFP is not simply due to accumulation of plasma membrane, because large parts of the cell that stained strongly positive for DiI were negative for CXCR4-GFP (Fig. 4B). Thus, SDF-1, immobilized on fibronectin, induces redistribution of CXCR4 at the plasma membrane.
The expression levels of adhesion molecules such as VLA-4, VLA-5, ␤ 2 integrins, and selectins, known to be involved in the process of homing of primary CD34 ϩ cells, were not affected by CXCR4-GFP cells were serum-starved overnight and subsequently exposed to 100 ng/ml SDF-1 for the indicated times. Western blot analysis shows phosphorylation of ERK1/2 (pERK) at 1-2 min in KG1a CXCR4-GFP cells, whereas the KG1a wt cells did not induce phosphorylation of ERK1/2 upon SDF-1 exposure. Addition of medium alone did not influence the phosphorylation of ERK1/2 in both cell types. Lower blots for total ERK2 confirm equal protein loading. Results are representative for two independent experiments. the CXCR4-GFP transduction into the KG1a cells (data not shown) and were comparable with those on primary CD34 ϩ cells (data not shown and see Refs. 26 and 27). Peled and colleagues (7) reported that SDF-1 is constitutively expressed on human bone marrow endothelium and induces firm adhesion of primary human CD34 ϩ cells to the endothelium under shear stress. Moreover, on primary human CD34 ϩ cells, SDF-1 activates the integrins LFA-1, VLA-4, and VLA-5 (8). To test whether the SDF-1-induced adhesion of CD34 ϩ cells to the endothelium is specifically mediated by CXCR4, the KG1a CXCR4-GFP cells and the KG1a wt cells were exposed to TNF-␣-stimulated and SDF-1-incubated endothelium under physiological shear stress. Adhesion of KG1a CXCR4-GFP cells to activated endothelium was significantly enhanced in the presence of SDF-1 (Fig. 5). KG1a wt cells did not show a significant increase in adhesion, indicating that the SDF-1-mediated increase in adhesion to endothelium was mainly dependent on CXCR4.
The final step in the homing of CD34 ϩ cells is the migration to the bone marrow stroma. In Transwell-based in vitro migration assays, the KG1a wt cells showed no increase in transmigration across FN-coated filters to SDF-1, whereas KG1a CXCR4-GFP cells showed a significant increase. Primary human CD34 ϩ cells migrated for almost 50% across FN-coated filters to SDF-1 (Fig. 6A). Migration experiments with CXCR4deficient KG1a cells and CXCR4-expressing and SDF-1-responsive HL60 cells, which both were transduced with GFP-actin, showed that the transduction procedure itself did not alter the migratory capacities or SDF-1 responsiveness of these cells.
Dose-response experiments showed optimal migration of the KG1a CXCR4-GFP cells across fibronectin at 70 ng/ml SDF-1 (Fig. 6B). A decline in migration was observed at high SDF-1 concentrations (1000 ng/ml, n ϭ 1; data not shown). Migration of the KG1a CXCR4-GFP cells across TNF-␣-stimulated HB-MEC revealed an optimal concentration for SDF-1 of 30 ng/ml and a bell-shaped dose-response curve (Fig. 6C). Although the absolute migration of the cells across the endothelium was lower than across FN, the SDF-1 concentration, required for maximal migration, was reduced. This suggests that the endothelium presents the chemokine more efficiently than the FN or that the adhesion to the endothelium promotes SDF-1 responsiveness, compared with the adhesion to FN. Previous experiments performed with primary human CD34 ϩ cells showed similar results in migration across endothelium versus FN (26). Thus, migration of KG1a-CXCR4-GFP cells to SDF-1 across FN and HBMEC was comparable with that of primary human CD34 ϩ cells and was dependent on CXCR4.
Under physiological flow, the role of soluble chemokines to create a gradient will be limited. It has been published that chemokines can bind to proteoglycans and glycosaminoglycans on the endothelium (28,29). Moreover, Middleton and colleagues (30) have reported that intradermally injected chemokines are presented on the apical side of the endothelium in vivo. Based on the results that showed immobilized SDF-1 to induce firm adhesion to activated endothelium (Fig. 5), we analyzed the distribution of CXCR4-GFP on untreated or SDF- 1-treated TNF-␣-activated endothelium in real time (Fig. 7). These experiments showed that in the absence of SDF-1, KG1a CXCR-4-GFP cells adhered and migrated on TNF-␣-stimulated endothelial monolayers. Moreover, upon adhesion, the cells formed membrane ruffles and lamellipodia. However, no prominent CXCR4-GFP localization was observed in these ruffles, but CXCR4-GFP seemed to accumulate in the back of the cell (Fig. 7A, see Supplemental Material for 4 Quick-Time movies). However, not all CXCR4-GFP-expressing cells showed such CXCR4-GFP localization in the back of the cell but a more uniform distribution over the plasma membrane. These differences might be due to the different expression levels of CXCR4-GFP per cell. Addition of soluble SDF-1 to the medium did not induce a redistribution of CXCR4-GFP in already adherent cells (data not shown). In contrast, immobilization of SDF-1 onto TNF-␣-stimulated endothelium induced increased adhesion and lamellipodia formation by the migrating KG1a CXCR4-GFP cells. Moreover, under these conditions CXCR4-GFP was redistributed to the leading edge of the cell, in particular to the newly formed lamellipodia (Fig. 7B, see Supplemental Material). These results clearly show that immobilized SDF-1 on activated endothelium induces a redistribution of CXCR4 toward the leading edge of the migrating cell.
To determine whether the CXCR4-GFP was localized intracellularly or on the cell surface, the KG1a CXCR4-GFP cells were fixed but not permeabilized and stained with a CXCR4 antibody, following their incubation on TNF-␣-stimulated endothelium, with or without immobilized SDF-1. Confocal microscopy of non-permeabilized cells confirmed that a fraction of the CXCR4-GFP was localized to the cell surface at the leading edge, whereas the CXCR4-GFP fraction at the back of the cell was not stained by the antibody (Fig. 8A). By using the PEconjugated CXCR4 antibody, also primary human CD34 ϩ cells showed a polarized CXCR4 localization after exposure to immobilized SDF-1 (Fig. 8A). Real time live cell analysis with the PE-conjugated antibody to CXCR4 revealed that SDF-1, immobilized on the endothelium, induced CXCR4 distribution to the front of the migrating KG1a CXCR4-GFP cell (Fig. 8B). CXCR4-GFP that did not co-localize with the antibody staining was present in the back of the cell, similar to fixed cells as described above. Moreover, the change of cell shape and the directionality of migration were paralleled by a corresponding CXCR4 redistribution over the cell surface to a newly formed lamellipodium, whereas the KG1a wt cell shape did not change (Fig. 8B). Although the antibody used (clone 12G5) is described as a blocking antibody, our results indicate that responses induced by immobilized SDF-1 were not inhibited by this antibody. Reports by others (31,32) already showed that this antibody only partially blocks SDF-1 responses. However, on cells that are treated in suspension, the 12G5 antibody is still capable of blocking SDF-1 responses. 2 Together these results showed that, upon exposure to immobilized SDF-1 on endothelium, CXCR4-GFP is present on the cell surface at the front of the cell, whereas CXCR4-GFP at the back of the cell is localized intracellularly. To confirm that the antibody recognized the CXCR4-GFP fusion protein, SDF-1-stimulated KG1a CXCR4-GFP cells were fixed, permeabilized, and stained for CXCR4 with the CXCR4 antibody. Confocal analysis showed a 100% co-localization of staining by the antibody with the CXCR4-GFP fusion protein (Fig. 8C).
To quantify SDF-1-induced CXCR4-GFP polarization at the leading edge of KG1a CXCR4-GFP cells on activated endothelium, the cells were incubated on SDF-1-coated and TNF-␣activated endothelium and stained for ICAM-3, which localizes to the back of a polarized cell (Fig. 9A) (14). The other side of the cell therefore represents the leading edge. After exposure to immobilized SDF-1, detailed examination of ICAM-3 and of CXCR4-GFP distribution showed a significant increase in cells (73%) that expressed CXCR4-GFP at the leading edge, whereas in the absence of SDF-1 only 34% of the cells showed a redistribution of CXCR4-GFP to the leading edge (Fig. 9B). Thus, in order to amplify signaling, induced by immobilized SDF-1 on TNF-␣-stimulated endothelium, migrating cells redistribute CXCR4 to the leading edge.
To organize chemokine-induced signaling, migrating and polarized cells associate glycosphingolipids with cholesterol to cluster membrane microdomains or so-called lipid rafts. Several reports have shown that depletion of cholesterol from the plasma membrane, which disturbs lipid raft integrity, results in an impaired responsiveness to extracellular signals (33,34). Visualizing the lipid rafts with CTB showed that in resting KG1a CXCR4-GFP cells CXCR4 co-localized with these rafts (Fig. 10A). Stimulation with soluble SDF-1 resulted in a rapid clustering of lipid rafts and polarization of CXCR4-GFP. Besides the fact that CXCR4-GFP still co-localized with the clustered lipid rafts on the cell surface, we also observed a rapid CXCR4-GFP internalization after SDF-1 exposure. Depletion of cholesterol from the membrane by CD significantly prevented SDF-1-induced CXCR4 internalization, as was measured by flow cytometry (Fig. 10B). In addition, SDF-1-driven transmigration of the KG1a CXCR4-GFP cells across FNcoated filters was also significantly inhibited upon depletion of cholesterol by CD (Fig. 10C). Moreover, pretreatment of the KG1a CXCR4-GFP cells with CD significantly decreased firm adhesion to activated endothelium (Fig. 10D). These results indicate that the complete inhibition of SDF-1-induced migration of CD-treated KG1a CXCR4-GFP cells is partially due to less adhesion to the endothelium but also due to deficient CXCR4 signaling.
SDF-1, immobilized on activated endothelium, induced lipid raft redistribution to the leading edge of migrating cells. Detailed three-dimensional analysis revealed that the lipid rafts co-localized with CXCR4-GFP at the leading edge of the cell, at sites of contact with the apical surface of the endothelium. The back of the cell did not show co-localization between lipid rafts and CXCR4-GFP (Fig. 10E). In order to determine whether lipid rafts are involved in SDF-1-induced CXCR4 polarization, the cells were pretreated with CD and subsequently incubated on activated endothelium, on which SDF-1 was immobilized. These results showed that lipid rafts support SDF-1-induced polarization of CXCR4 (Fig. 10F). In conclusion, proper lipid raft organization is required for efficient SDF-1-mediated responses.  8. CXCR4 is expressed at the outside of the plasma membrane in a polarized fashion. A, antibody labeling showed marked redistribution of CXCR4 to the leading edge of the cells after exposure to immobilized SDF-1 on TNF-␣-stimulated endothelium. After a 30min exposure of the KG1a CXCR4-GFP cells to activated endothelium, without (control, a-c) or with immobilized SDF-1 (e-g), the cells were fixed but not permeabilized and stained with CXCR4 mAb in red (b and f). Green shows localization of CXCR4-GFP (a and e), and yellow represents co-localization (c and g). Asterisk shows CXCR4 polarization at the surface; arrowhead shows intracellular localized CXCR4. Right panel shows CXCR4-PE staining of fixed primary human CD34 ϩ cells incubated on control (d) or SDF-1-coated endothelium (h). Bar, 5 m. B, immobilized SDF-1 on TNF-␣-stimulated endothelium redistributes CXCR4 on the cell surface to a newly formed lamellipodium on the migrating cell, indicated by the asterisk. Images were taken with a short time interval, as indicated in seconds in the upper left corner. Cells were labeled with PE-labeled CXCR4 Ab (b and e). Green shows distribution of CXCR4-GFP (a and d). Co-localization between CXCR4-GFP and CXCR4-PE is shown in yellow (merge, c and f). Phase contrast panels on the right show four cells, one of which is CXCR4-GFP-positive and hence spreads upon exposure to immobilized SDF-1 (d and g). Bar, 10 m. C, the antibody 12G5 to CXCR4 recognizes all the CXCR4-GFP fusion proteins. Cells were exposed to immobilized SDF-1 on FN (5 min), fixed, permeabilized, and stained with CXCR4 Ab and secondary Alexa-568 Ab. Green shows CXCR4-GFP (a),and red represents staining with CXCR4 Ab (b). Co-localization appears in yellow (merge, c). Bar, 20 m.

DISCUSSION
Chemokine-induced polarization of the cell body and of the intracellular signaling machinery is well established (35). In contrast, polarization of the relevant chemokine receptors is induced by chemokine exposure according to some authors (13) but not to others (36), possibly due to cell type-specific differences. The present study was designed to test the hypothesis that the context in which the chemokine is presented plays a crucial role in the modulation of receptor (re)distribution. To this end, we expressed a CXCR4-GFP fusion protein in the CXCR4-deficient hematopoietic progenitor cell line KG1a. The expression of the fusion protein restored the ability of the cell to respond to SDF-1 in migration, internalization, and actin polymerization assays. Studies with primary human hematopoietic progenitor cells, i.e. CD34 ϩ cells derived from cord blood, showed similar kinetics in these assays as the KG1a CXCR4-GFP cells (7,26), which also shows that the KG1a CXCR4-GFP cell line is a valid model to study human hematopoietic progenitor cell responses to SDF-1 in detail.
By using this newly transduced cell line, we were able to document the redistribution of CXCR4 in response to SDF-1. Several studies already analyzed the behavior and redistribution of chemokine receptors. From these studies, it was clear that not all chemokine receptors behave similarly in response to their cognate ligands. Studies with GFP-tagged fusion proteins, such as the C5a receptor and the cAMP receptor, have shown that these receptors did not polarize upon activation (10,36). However, Meili et al. (37) showed that the pleckstrin homology domain of protein kinase B/Akt fused to GFP polarizes upon exposure to cAMP, indicating that polarized signaling from G-protein-coupled receptors is involved in sensing and responding to chemokine gradients. With fluorescent antibodies, it was shown that the chemokine receptors CCR5, CCR2, and CXCR4 polarize upon chemokine activation under static conditions (13)(14)(15). These conflicting results were obtained either with Dictyostelium as a model or with mammalian cells adhered to an artificial matrix. Moreover, chemokines were added in solution or via a micropipette to study the mechanism that controls cell polarization.
In vivo, human leukocytes crawl over the endothelium of the vessel wall in response to signals from the underlying tissue. Peled et al. (7) have shown that the chemokine SDF-1 is expressed on human bone marrow endothelium. In addition, Middleton et al. (30) reported that chemokines, such as interleukin-8 and regulated on activation normal T cell expressed and secreted (RANTES), could be presented at the apical surface of endothelial cells. Moreover, because SDF-1 is not only produced by bone marrow stroma cells, but also by other tissues, and because also other human leukocytes, such as monocytes and T and B lymphocytes express CXCR4, SDF-1 presentation on the endothelium might represent a general mechanism to promote CXCR4-mediated chemotaxis. Here we show that CXCR4 rapidly redistributes to the leading edge of the cell in response to immobilized SDF-1, both on FN (this paper and Ref. 15) and on a physiological relevant substrate such as activated human endothelium. Complementary ICAM-3 staining, to visualize the back of the cell, revealed that CXCR4 was mainly polarized to the leading edge. Moreover, we observed that the shape of the KG1a wt cells did not change on activated endothelium, in the presence or absence of immobilized SDF-1, whereas the KG1a CXCR4-GFP cells did spread and migrate over activated endothelium, even in the absence of SDF-1. This observation might indicate that not only the underlying tissue but also the human bone marrow endothelium itself is able to produce and subsequently present chemokines, in this case SDF-1, in order to mediate chemotaxis, although this amount of chemokine, present at the surface of activated endothelium, is not sufficient to induce clear polarization of CXCR4 to the leading edge of the migrating cell.
It has been suggested that cells cluster signaling receptors into so-called lipid rafts to respond rapidly to environmental changes. Previous reports (34,38) suggest that lipid rafts localize to the leading edge of a migrating cell, although this topic is the subject of conflicting reports. Here we show that SDF-1, immobilized on activated endothelium, induces clustering of lipid rafts. Three-dimensional analysis suggests that CXCR4 co-localizes with the lipid rafts at the basolateral side of the cell. Recently, Shamri et al. (39) showed that G-protein-coupled receptor-mediated activation of ␣ 4 ␤ 1 integrin depends on the formation of cholesterol-raft platforms. The specific localization of CXCR4 in the lipid rafts at the basolateral side of the cell suggests that in lipid rafts, integrins may cluster with G-protein-coupled receptors on leukocytes, which would likely favor chemokine-induced integrinmediated activation.
Finally, chemokine receptor distribution can be studied by means of GFP fusion proteins or specific antibodies. Here we show that CXCR4-GFP redistributes to the cell surface at the leading edge of the cell, and we underscore these findings by using specific antibodies, directed to the receptor. However, CXCR4-GFP was also detected at the rear of the cell, where it was not recognized by these antibodies on non-permeabilized cells, indicating that this fraction of the receptor pool was localized intracellularly. This observation shows that GFP fusion protein distribution does not necessarily represent cell surface localization.
In conclusion, we have generated and characterized a suita-FIG. 10. SDF-1-induced CXCR4 polarization and signaling depends on lipid raft formation. A, cells were incubated on poly-L-lysine-and SDF-1-coated glass coverslips at 37°C. Time is indicated in seconds in the upper left corner. Cells were stained with cholera toxin B (CTB), fixed, and analyzed with CLSM as described under "Materials and Methods." Lipid rafts are in red (a, d, and g); green indicates CXCR4-GFP (b, e, and h), and yellow represents co-localization (merge, c, f, and i). Lipid rafts co-localize with CXCR4-GFP and cluster upon exposure to SDF-1. Bar, 5 m. B, depletion of cholesterol from the plasma membrane inhibits SDF-1-induced internalization of CXCR4. Cells were treated with 10 mM methyl-␤-cyclodextrin (CD) (open bars) or with PBS alone (closed bars) for 30 min at 37°C and exposed to SDF-1 for the times indicated on the abscissa. CXCR4 surface expression was measured using CXCR4-PE mAb on fixed but not permeabilized cells. Data are mean Ϯ S.D. of three independent experiments. *, p Ͻ 0.05. C, lipid raft integrity is required migration. Cells were treated with 10 mM CD or were left untreated for 30 min at 37°C and were subsequently allowed to migrate to 30 ng/ml SDF-1 in a Transwell assay as described under "Materials and Methods." Data are mean Ϯ S.D. of three independent experiments. *, p Ͻ 0.001. D, SDF-1-induced firm adhesion of KG1a CXCR4-GFP cells to endothelium is blocked by CD. Cells were treated with 10 mM CD or were left untreated for 30 min at 37°C and were subsequently allowed to adhere to activated endothelium, on which SDF-1 was immobilized. Percentage on the y axis represents the number of cells per field of view, analyzed by CLSM. Data are mean Ϯ S.D. of three independent experiments. *, p Ͻ 0.01. E, CXCR4 and lipid rafts co-localize to the leading edge. Endothelial cells were cultured on FN-coated glass coverslips and stimulated with TNF-␣. 30 min prior to the addition of the cells, the endothelial cells were coated with 100 ng/ml SDF-1. Cells were allowed to adhere to the endothelium for 30 min and were then fixed and analyzed by CLSM. CTB is visualized in red (a) and CXCR4-GFP in green (b), and co-localization appears in yellow (merge, c). Detailed X-Z images revealed that the lipid rafts localize to the baso-lateral side of the cell, together with CXCR4-GFP, as is indicated by the white arrows (d). Red color layer in X-Z image represents CTB staining on the endothelium. Bar, 5 m. F, methyl-␤-cyclodextrin blocks SDF-1-induced polarization of CXCR4. Cells were treated as described under E. CXCR4 polarization was determined as described under "Materials and Methods." Data are mean Ϯ S.D. of three independent experiments. **, p Ͻ 0.05. ble model cell line for human hematopoietic progenitor cells to be used in studies on SDF-1-induced chemotaxis and stem cell homing. Moreover, we have shown that the CXCR4 receptor displays substrate-and chemokine-dependent polarization to the leading edge and that the cells redistribute the receptor to lipid rafts. These data provide new insights in the mechanisms that govern cell polarization and migration and underscore the crucial role of cellular context for the responses to chemokines.