Involvement of Phosphoinositide 3-Kinase γ, Rac, and PAK Signaling in Chemokine-induced Macrophage Migration*

In macrophages, chemotactic stimuli cause the activation of Rac and PAK, but little is known about the signaling pathways involved and their role in chemotactic gradient sensing. Herein, we report that in macrophages, the chemokine RANTES (regulated on activation normal T cell expressed and secreted)/CCL5 activates the small GTPase Rac and its downstream target PAK2 within seconds. This response depends on Gi activation and largely on the subsequent triggering of phosphoinositide 3-kinase γ (PI3Kγ) and Rac. Retroviral transduction of tagged Rac1 and -2 indicates that RANTES/CCL5-mediated activation of PI3Kγ triggers Rac1 but not Rac2. In agreement, silencing of Rac1 by shRNA blocks PAK2 activity and inhibits RANTES/CCL5-induced macrophage polarization and directional migration. On the other hand, the tyrosine kinase receptor agonist CSF-1 activates PAK2 independently of PI3Kγ and Rac. Our results thus demonstrate a chemokine-specific signaling pathway in which Gi and PI3Kγ coordinate to drive Rac1 and PAK2 activation that eventually controls the chemotactic response.

In macrophages, chemotactic stimuli cause the activation of Rac and PAK, but little is known about the signaling pathways involved and their role in chemotactic gradient sensing. Herein, we report that in macrophages, the chemokine RANTES (regulated on activation normal T cell expressed and secreted)/CCL5 activates the small GTPase Rac and its downstream target PAK2 within seconds. This response depends on G i activation and largely on the subsequent triggering of phosphoinositide 3-kinase ␥ (PI3K␥) and Rac. Retroviral transduction of tagged Rac1 and -2 indicates that RAN-TES/CCL5-mediated activation of PI3K␥ triggers Rac1 but not Rac2. In agreement, silencing of Rac1 by shRNA blocks PAK2 activity and inhibits RANTES/CCL5-induced macrophage polarization and directional migration. On the other hand, the tyrosine kinase receptor agonist CSF-1 activates PAK2 independently of PI3K␥ and Rac. Our results thus demonstrate a chemokinespecific signaling pathway in which G i and PI3K␥ coordinate to drive Rac1 and PAK2 activation that eventually controls the chemotactic response.
Rac1, Rac2, and Rac3 constitute a subfamily of the Rho family of monomeric GTPases and cycle between active GTP-bound (Rac GTP ) and inactive GDP-bound (Rac GDP ) states (1)(2)(3). Activation is accomplished by guanine-nucleotide exchange factors (GEFs), 1 which catalyze GDP dissociation (4), and inactivation by GTPase-activating proteins, which increase the intrinsic GTPase activity (5). Rho GTPases integrate signals from cellular receptors and membrane components to regulate the cytoskeleton dynamics required for cell locomotion during chemotaxis, phagocytosis, and many other cellular responses (2,6).
Leukocyte chemotaxis toward sites of inflammation (7) is primarily mediated by chemokine signaling. RANTES CCL5 is a disease-relevant and potent chemoattractant for macrophages. Moreover, it has been shown that RANTES-mediated T-cell activation and chemotaxis requires Rho GTPase activity (8,9). RANTES is a member of the CC-subfamily of chemokines and activates the seven-transmembrane CC-chemokine receptors CCR1, CCR3, CCR4, and CCR5, which are coupled to pertussis toxin sensitive heterotrimeric G i␣ proteins (10). PI3K was identified as a target of these G-protein-coupled chemokine receptors because chemotaxis and polarization of T cells could be inhibited by isoform non-selective class I PI3K inhibitors, such as LY294002 (11). All class I PI3Ks are heterodimers consisting of a 110-kDa (p110) catalytic subunit and an 85-(p85) or 101-kDa (p101) regulatory subunit (12). The two catalytic subunits p110␣ and -␤ are ubiquitously expressed, whereas p110␥ and -␦ expression is largely confined to leukocytes. In addition to their characteristic expression, PI3Ks are differentially regulated by GPCRs and receptor tyrosine kinases (RTKs) (13). RTKs have been shown to activate p110␣, -␤, and -␦ via the p85 regulatory subunit (class I A: PI3K-␣, -␤, and -␦). In contrast, it has been shown that regulation of p110␥ (class I B: PI3K␥) is mediated via the p101 adapter, engaged by G i␤␥ subunits released after activation of GPCRs (14).
Class I B PI3K␥-deficient mice, PI3K␥Ϫ/Ϫ, showed in vitro and in vivo impaired migration of neutrophils and macrophages toward chemoattractants (15)(16)(17). PI3K␥Ϫ/Ϫ neutrophils are unable to produce 3Ј-phosphorylated phosphoinositides (PIP 3 ; e.g. phosphatidylinositol 3,4,5-trisphosphate and phosphatidylinositol 3,4-bisphosphate) when stimulated with GPCR agonists such as fMLP, complement component 5a, or interleukin-8. When activated by cell surface receptors, all isoforms of PI3K generate intracellular membrane-integral 3Јphosphorylated phosphoinositides as second messengers. Proteins can bind via different protein modules to PIP 3 , among them pleckstrin homology domains (18). Like for several PI3Kregulated protein kinases, such as AKT/PKB (19,20), Rho-GEFs contain pleckstrin homology domains but in addition have an adjacent Dbl homology domain (21). It has been proposed that the Dbl homology domain is responsible for the GEF activity and regulated by the pleckstrin homology domain that binds phospholipids such as PIP 3 . In many cell types, PI3K activity and elevated PIP 3 are necessary and sufficient for receptor-driven stimulation of Rac activation (22,23).
To trigger the reorganization of the cytoskeleton during migration and phagocytosis, Rac GTP interacts with several downstream effectors. One of these downstream targets comprises a family of serine/threonine protein kinases termed PAK (24).
The isoforms PAK1 and PAK2 belong to the well-studied PAK-I subfamily.
RhoGTPases and PI3Ks have been shown to be involved in the regulation of macrophage chemotaxis (25). It is noteworthy that PI3K␥-deficient macrophages show only partial loss of migration in vitro and in vivo in response to chemokine stimuli, suggesting that PI3K-independent signaling mechanisms are capable of supporting chemotaxis.
The objective of this study was to investigate possible signaling cross-talk between Rac and PI3K␥ and, furthermore, to elucidate PI3K-independent pathways involved in chemokine signaling in murine macrophages. Our strategy combined genetic inactivation of PI3K␥ and pharmacological inhibition of PI3Ks, tyrosine kinases, and GPCRs in murine bone marrowderived macrophages (BMDM). To gain better insight into signaling specificity addressed by ligands that can induce comparable cellular responses in macrophages (such as cell migration) yet act via GPCR or RTK, we have stimulated BMDM with either RANTES or macrophage colony-stimulating factor (CSF-1), respectively (26). We monitored Rac GTP activity by PAK-p21 Rac-binding domain (PAK-PBD) affinity precipitation (27,28) and the activation status of PAK and AKT with phosphospecific antibodies as well as in-gel kinase assays. Furthermore, retrovirus-transduced primary macrophages with constructs coding for epitope-tagged HA-Rac proteins or shRNA has specified Rac1 as the isoform regulated by PI3K␥ and required for PAK activation and cell migration.
Rac activation assay-Active GTP-bound Rac was precipitated with a PAK-PBD-based assay (Upstate Biotechnology, Lake Placid, NY). Stimulation of BMDM (10 cm dish) was stopped by addition of 0.75 ml of ice-cold 1ϫ MLB (Upstate Biotechnology). Dishes were placed on ice and scraped, and samples were rotated for 10 min and clarified by centrifugation for 1 min. Aliquots were saved for immunoblot analysis. Lysates were rotated with 10 g of PAK-PBD agarose at 4°C for 30 min. The agarose pellet was washed twice and resuspended in 30 l (1:1; v/v) reducing sample buffer. Samples were separated by SDS-PAGE (4 -12% Bis-Tris gels, MES buffer; Invitrogen), transferred to nitrocellulose membrane (Invitrogen), and blotted for Rac.
PAK In-gel Kinase Assay-This assay is based on the ability of renatured PAKs to phosphorylate their specific peptide substrate p47phox (amino acid residues 297 to 331; 0.5 mg/ml) polymerized into the gel (28). The assay was performed according to published methods with minor modifications (32,33).
Positions of PAK were visualized by autoradiography after incubation of the gel with 150 Ci of [␥-32 P]ATP. MnCl 2 (2 mM) was added to the gel equilibration solution, and 0.08% Tween 20 was added to the renaturation buffer. Recombinant active PAK2 (Upstate Biotechnology) was used as a positive control for gel mobility and kinase activity.
Retroviral Transduction-The retroviral packaging cell line Platinum-E (Plat-E) and the vector pMX have been described previously (34). Human cDNA clone of Rac1 was purchased as pUSEamp-Rac1 (Upstate Biotechnology). The sequence of Rac2 (GenBank accession no. NM_009008) was cloned by PCR from a mouse thymus cDNA library into pAcGTC8 vector. Both Rac1 and Rac2 coding sequences were subcloned into the pMX vector in frame with a N-terminal triple-HA epitope-tag. Plat-E cells were cultured in Dulbecco's modified Eagle's medium (4500 mg/ml glucose; Sigma), 10% fetal bovine serum, 100 units/ml penicillin, 100 g/ml streptomycin, 2 mM glutamine, 1 g/ml puromycin (BD Biosciences Clontech) and 10 g/ml blasticidine S (Calbiochem). Before calcium phosphate transfection, cells were cultured without puromycin and blasticidine S. Virus was collected 48 and 72 h later, concentrated, and processed as described previously (35).
Isolated bone marrow cells were cultured in suspension for 1 day in the presence of 10 ng/ml interleukin-3 (Peprotech) to facilitate retroviral transduction. Cells were collected and resuspended in retrovirus-BMDM medium supplemented with 10 ng/ml interleukin-3, 50 ng/ml CSF-1, and 8 g/ml polybrene. 3-4 ϫ 10 6 cells/1.5-ml volume were distributed in each well of a non-adherent 6-well plate (Evergreen Scientific, Los Angeles, CA). For spin-culture-infection, plates were centrifuged for 90 min at 820 ϫ g at 32°C. Thereafter, 1.5 ml of medium supplemented with interleukin-3 and CSF-1 was added. At day 5, transduced suspension cells were harvested and seeded on tissue-culture dishes in BMDM-medium, which was exchanged 1 day later. Experiments were performed at day 8 after preparation with differentiated BMDM characterized as described above. Transduction efficiency was confirmed by GFP control virus and was in the range of 60 -80%.
Real-time PCR-with SYBR Green PCR Master Mix (Applied Biosystems) was performed using Rac1 primers (forward primer, 5Ј-GCA TTT CCT GGA GAG TAC ATC-3Ј; reverse primer, 5Ј-TGT GTC CCA TAG GCC CAG-3Ј) and cDNA of NIH 3T3 cells. The specificity and the optimal primer concentrations were tested. The absence of nonspecific amplification was confirmed by analyzing the PCR products on a 3.5% agarose gel electrophoresis. Potential genomic DNA contamination was excluded by performing PCR reactions with specific intron-glyceraldehyde-3-phosphate dehydrogenase primers. 40 PCR cycles were performed, and the Ct (cycle threshold) values were normalized to the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase.
shRNA Constructs-The retroviral pMX-IRES-GFP (36) construct was converted into a destination vector according to the Gateway technology (Invitrogen) by introducing a 'reading frame cassette B' into the SnaBI site to generate pMX-DEST-IRES-GFP. Five different shRNA constructs targeting Rac1 (shRac1) were cloned by PCR using reverse primers containing a portion that anneals to the 3Ј end of the human U6 promoter and the specific hairpin sequence flanked by attB1. The forward primer contained the attB2 site and a portion that anneals to the 5Ј end of the U6 promoter (5Ј-GGG GAC CAC TTT GTA CAA GAA AGC TGG GT CCC GAG TCC AAC ACC CGT GG-3Ј). A vector containing the U6 promoter was used as a template for PCR amplification. Recombination of the attB2-U6-shRac1-attB1-PCR product with the donor vector pDONR221 containing attP sites created an entry clone. Recombination of the entry clone with the designed destination vector pMX-DEST-IRES-GFP generated pMX-DEST-shRac1 constructs. All pMX-DEST-shRac1 constructs were tested for functionality regarding Rac1 silencing by real-time PCR analysis, and NIH 3T3 total cell lysates were analyzed by immunoblot with Rac antibody (data not shown). The most efficacious suppression was detected with the construct pMX-DEST-shRac1.3 (within text Rac1 shRNA) using the following primer: 5Ј-GGG GAC AAG TTT GTA CAA AAA AGC AGG CTA AAA AAG ACA GAC GTG TTC TTA ATT TGA AGC TTG AAA TTA AGA ACA CGT CTG TCT CGG TGT TTC GTC CTT TCC ACA-3Ј. Note that the selected interfering sequence (underlined) is specific for Rac1 and does not recognize any known related mouse gene. As a control, we used the construct pMX-DEST-mismatch shRac1 using the following primer: 5Ј-GGG GAC AAG TTT GTA C A AAA AAG CAG GCT  AAA AAA ATA CAT CCC CAC TGT ATT TGG AAG CTT GCA AAT  ACA GTG GGG ATG TAT TCG GTG TTT CGT CCT TTC CAC A-3Ј. Constructs were sequenced twice in the two directions for corroborating the sequence. PCR reactions were performed using Pwo DNApolymerase (Roche).
Time-lapse Videomicroscopy-Glass-bottomed plates (6 cm; Milian) were coated overnight with 2 g/ml Vitronectin. Day-5 BMDM were seeded in L929-free 10% fetal bovine serum containing medium; 15 min after adherence, they were stimulated with a point source of chemokine (femtojet). Live cells were imaged every 12 s during stimulation from the micropipette containing 2 mM RANTES (Zeiss microscope). The recorded data were used for analysis of macrophage migration using openlab software.

RANTES-induced Rac Activation Significantly Depends on
PI3K Activity-We investigated whether RANTES is able to activate Rac and, if so, whether Rac is a target of PI3K. To compare the specificity of GPCR-versus RTK-signaling regarding PI3K isoform engagement, we investigated the RTK ligand CSF-1 in parallel.
BMDM were stimulated with RANTES or CSF-1 for the indicated times, and activation of Rac was investigated using the PAK-PBD binding assay to precipitate Rac GTP . Different concentrations and timing for RANTES-induced Rac activation were tested, and submaximal doses with respect to Rac activation and optimal timing were used. Stimulation by RANTES led to a rapid and transient activation of Rac, which was partially inhibited by LY294002 (Fig. 1, A and D, left). In contrast, no activation was detected in response to CSF-1 within the same experiment (see also Fig. 5C for value at 10 s). Activation of Rac peaked at 10 s, followed by a very rapid decrease in levels of active GTPase (see also Fig. 3A). Because semiquantitative PCR analysis showed that BMDM expresses the Rac1 and Rac2 isoforms but not Rac3 (data not shown), activated Rac analyzed by immunoblot with the Rac antibody corresponds to the 92% identical Rac1 and Rac2 isoforms. 2 We conclude that stimulation of serum-starved macrophages by RANTES resulted in a very rapid, transient, and partially PI3K-dependent activation of Rac, whereas incubation with CSF-1 under the same conditions was unable to activate Rac.
Rac Activation Correlates with PAK2  Phosphorylation and PAK2 Activity-Subsequently we examined whether the observed partial PI3K-dependent activation of Rac by RAN-TES is correlated with an activation of PAK. The Ser/Thr kinase PAK is activated in a multistep process, including steps such as binding to Rac GTP and autophosphorylation of threonine (PAK Thr ) and serine (PAK Ser ) residues to establish and maintain the catalytically active state (38,39).
We analyzed PAK Ser phosphorylation by immunoblot with a phospho-PAK Ser antibody. BMDM were stimulated with RAN-TES and CSF-1 for the indicated time points, and total cell lysates were analyzed. RANTES rapidly and transiently triggered phosphorylation of PAK2   (Fig. 1, B and D, middle). As seen for Rac activation, this phosphorylation of PAK2  was partially inhibited by LY294002.
In contrast, CSF-1 was unable to induce Rac and PAK activation (Fig. 1, A and B, right). Nevertheless, phosphorylation of AKT confirmed that CSF-1 signaling to PI3K was intact (Fig.  1C). Within the shown time window, RANTES and CSF-1 both triggered phosphorylation of AKT Ser-473 to a similar extent. This phosphorylation was completely blocked by LY294002 (Fig. 1, C and D, right). In contrast to Rac and PAK activation, LY294002 also reduced basal levels of AKT Ser-473 phosphorylation (data not shown).
To determine the kinase activity of PAK2, we used an in-gel kinase assay together with a PAK-specific peptide substrate derived from p47Phox (28). Recombinant PAK2 (500 ng) confirmed the identification of the PAK2 isoform activated in response to RANTES stimulation based on activity and gel motility (59 and 62 kDa; Fig. 1E, left lane).
We observed a rapid and transient activation of PAK2 in BMDM treated with RANTES (Fig. 1E). PAK2 exhibited maximal activation within 10 s to 1 min of RANTES stimulation (see also Fig. 3E), followed by significant loss of activation at 3 min. In correlation with the inhibition of PAK2  phosphorylation, LY294002 partially inhibited RANTES-induced PAK2 activation. We conclude that RANTES activated PAK2 in a class I PI3K-dependent manner and that this activation correlated with Rac activation.
CSF-1 Promotes PAK2  Phosphorylation but Not Rac Activation-GPCRs often operate instantly upon ligand binding, whereas RTKs require dimerization and subsequent recruitment of effectors (40). To examine whether RANTES and CSF-1 differed only with respect to their kinetics, BMDM were challenged with RANTES or CSF-1 for 5 or 10 min. In addition, prolonged stimulation of macrophages did not result in Rac activation by CSF-1 ( Fig. 2A). However, analysis of total BMDM lysates with phosphospecific antibodies revealed that CSF-1 triggered profound phosphorylation of PAK2  and AKT Ser-473 (Fig. 2, B and C). At these time points, RANTES-mediated responses returned to basal levels within the same experiment (see also Fig.  1, B and C). In contrast to the partial inhibition of RANTESinduced PAK2  phosphorylation, CSF-1 induced phosphorylation of PAK2  was entirely blocked by PI3K inhibition with LY294002. CSF-1 triggered PAK2  phosphorylation resembled AKT Ser-473 phosphorylation with respect to its complete dependence on PI3K. Thus, the rapid response to CSF-1 involves PAK activation through a PI3K-dependent but Rac-independent mechanism.
PI3K␥ Contributes to Rac and PAK2 Activation Triggered by RANTES-Because inhibition of class I PI3K with LY294002 partially impaired RANTES-induced Rac and PAK2 activation, we examined whether the PI3K␥ isoform was required to fully activate Rac and PAK2. For this purpose, bone marrow-derived macrophages from wild-type and PI3K␥-deficient mice were generated.
At first, equal differentiation of mouse bone marrow precursor cells into BMDM was confirmed by fluorescence-activated cell sorting analysis (see "Experimental Procedures") to ensure that differences in signaling described below between differentiated WT and PI3K␥Ϫ/Ϫ BMDM are mainly caused by lack of PI3K␥ expression. To investigate the kinetics of RANTESinduced Rac activation in PI3K␥Ϫ/Ϫ and WT macrophages, BMDM were stimulated with RANTES for various times. As already described for WT BMDM (Fig. 1, A and D, left) RANTES induced an extremely rapid and transient activation of Rac, peaking at 10 s also in PI3K␥Ϫ/Ϫ BMDM (Fig. 3A). However, we repeatedly detected a significant reduction of Rac activation (about 30% in average) in PI3K␥Ϫ/Ϫ compared with WT BMDM (Fig. 3B). To investigate whether this reduction in Rac activation was also reflected in a diminished transduction of the signal to PAK2, we analyzed in parallel PAK2 phosphorylation and kinase activity in PI3K␥ Ϫ/Ϫ macrophages as described above. In addition to the serine residue, PAK Ser-192/197 , we investigated the threonine phosphorylation of PAK, which occurs in the activation  Fig. 5C shows stimulation for 10 s with CSF-1. A, Rac activation analysis. Cleared cell lysates were used for affinity precipitation with 10 g of PAK-PBD for 30 min at 4°C. Active Rac GTP precipitated by PAK-PBD was separated on SDS-PAGE, transferred to nitrocellulose membrane, and blotted for pan-Rac, followed by ECL detection. Left lanes, the total signal detected using cytosol preloaded with loop of the kinase domain and is known to be required for activation (41)(42)(43). The phospho-PAK Thr antibody demonstrated that RANTES stimulation of macrophages for 10 and 30 s triggered phosphorylation of PAK2 Thr-402 , but less so in PI3K␥Ϫ/Ϫ BMDM (Fig. 3C). In addition, RANTES-induced phosphorylation of PAK2  was reduced in PI3K␥Ϫ/Ϫ macrophages (Fig.  3D). To determine whether this diminished PAK2 phosphorylation was indicative of a regulatory role of PI3K␥ in PAK2 function, we assessed the kinase activity by in-gel kinase assay. This experiment confirmed a reduced PAK2 activity in PI3K␥Ϫ/Ϫ BMDM stimulated with RANTES (Fig. 3E). A correlation of onset (10 s) and decline (after 1 min) of PAK2 Ser/Thr phosphorylation (legend to Fig. 3) and PAK2 kinase activity was observed; however, the maximal phosphorylation signal (20 to 40 s; Fig. 3D and data not shown) seemed to precede the maximum of kinase activity (1 min).
In contrast to PAK2, phosphorylation of AKT Ser-473 in response to RANTES occurred in a more sustained manner. It was observed as early as 10 s, reached a maximum after 3-5 min ( Fig. 3F; see also Fig. 1, C and D) and was reduced after 5 min (Fig. 2C). In PI3K␥Ϫ/Ϫ BMDM, AKT Ser-473 phosphorylation did not rise above background levels after 1 min of stimulation (Fig. 3F).
Hence, the data obtained with PI3K␥Ϫ/Ϫ BMDM indicate that the partial contribution of PI3K to RANTES-induced Rac activation is indeed mediated by the PI3K␥ isoform. The observed reduction in Rac activation caused by PI3K␥ deficiency seems to result in an impaired phosphorylation of serine and threonine sites and subsequent deficiency in PAK2 activation.
Activation of Rac1, but Not Rac2, Depends on PI3K␥-Our results demonstrate that PI3K␥ is required for full Rac activation triggered by RANTES. To determine whether RANTES-activated PI3K␥ couples selectively to a specific Rac isoform, we generated retroviral constructs coding for N-terminal HA epitope-tagged Rac1 or Rac2. Transduced WT and PI3K␥Ϫ/Ϫ BMDM were challenged with RANTES, followed by the PAK-   Fig. 1. A, Rac activation assay. PAK2  (B) and AKT Ser-473 (C) phosphorylation analysis of the same experiment. Bottom, reprobing of blots with AKT and PAK antibodies confirmed equal protein expression and loading; however, strong AKT Ser-473 phosphorylation interfered with epitope recognition by the AKT antibody.
PBD affinity precipitation assay. The presence of either active HA-Rac1 GTP or HA-Rac2 GTP was assessed by immunoblot analysis using an HA antibody. We first verified that GTP-dependent Rac binding to PAK-PBD was maintained after the introduction of an N-terminal HA-tag (Fig. 4, A and B, far left). Rac1 was very rapidly activated in response to RANTES stimulation (Fig. 4A,  top) with similar kinetics as observed with the Rac antibody (Fig.  3A). However, PI3K␥ was clearly essential for the activation of Rac1 such that little Rac1 activation was observed in PI3K␥Ϫ/Ϫ BMDM. On the other hand, Rac2 showed only a small response after RANTES stimulation (Fig. 4B, top). This response was not impaired in PI3K␥Ϫ/Ϫ BMDM, indicating that PI3K␥ is responsible for Rac1 but not Rac2 activation. Phosphorylation of PAK2  triggered by RANTES was partially reduced in PI3K␥Ϫ/Ϫ compared with WT BMDM in both HA-Rac1 (Fig. 4C) and HA-Rac2 (Fig. 4D) transduced macrophages. These results parallel our findings in untransduced cells (Fig. 3D and data not  shown). In summary, we show that Rac1, but not Rac2, is the isoform activated in a PI3K␥-dependent manner in RANTESstimulated macrophages.
Within the same experiment (Fig. 5A), CSF-1 treatment did not result in Rac activation in either WT or PI3K␥Ϫ/Ϫ serumstarved BMDM (Fig. 5C), whereas Rac was activated by RANTES (Fig. 5C, right lane). Hence, we confirmed our initial observations that CSF-1 was able to stimulate phosphorylation of PAK2  , potentially independently of Rac.
The recent identification of a novel effector route for Rac activation triggered by the serum-borne GPCR agonist lysophosphatidic acid (44,45) has guided us to assay Rac activation with 5% serum in the medium. Under those conditions, treatment of BMDM with CSF-1 for the indicated time points evoked indeed a transient activation of Rac that declined after 1 min (Fig. 5D). Pretreatment of macrophages with pertussis toxin diminished CSF-1-induced Rac activation, indicating that in the presence of serum, co-activation by a GPCR might be the mechanism for CSF-1-induced Rac activation (data not shown). Values are the mean of seven independent experiments Ϯ S.E. C, phosphorylation of PAK2 Thr-402 analyzed with a phosphospecific antibody that is able to recognize PAK1 Thr-423 and PAK2 Thr-402 . Blot was reprobed with PAK and PAK2 (not shown) antibody. D, phosphorylation of PAK2  . Bottom, the blot was stripped and reprobed with PAK antibody. Four independent experiments under equal conditions were quantified by densitometry. The 20-s value of RANTES-induced Rac activation in WT cells was normalized to an arbitrary value of 100. Values are mean Ϯ S.E. In more concentrated lysates, phosphorylation of PAK2 Ser/Thr was still detectable after 1-min RANTES stimulation (Fig. 4, C and D). E, autoradiograph and quantification of PAK in-gel kinase assay. One representative result of two independent experiments is shown. Positions of PAK2 are designated by arrows. F, phosphorylation of AKT Ser-473 . The 10-s value of RANTES-induced AKT activation in WT cells was normalized to an arbitrary value of 50. Values are mean Ϯ S.E.

PI3K␥-dependent and PI3K-independent G i -mediated Activation of Rac Induced by RANTES-
To better understand which other signals, in addition to PI3K␥, link RANTES CC receptors to partial Rac activation, we pretreated WT or PI3K␥Ϫ/Ϫ macrophages with the tyrosine kinase inhibitor genistein, LY294002, and pertussis toxin before challenge of cells with RANTES. Rac activation in PI3K␥Ϫ/Ϫ macrophages was not inhibited by genistein (Fig. 6A) or LY294002. However, pertussis toxin totally blocked Rac activation in both WT and PI3K␥Ϫ/Ϫ macrophages. Treatment of WT or PI3K␥Ϫ/Ϫ macrophages with pertussis toxin also abolished RANTESinduced phosphorylation of PAK2  (data not shown).
In summary, these results demonstrate that complete activation of Rac relied on two cooperative signals, a PI3K␥ signal dependent on G i -activation and a PI3K-and tyrosine kinase-independent G i -activation signal. Because Rac activation in PI3K␥Ϫ/Ϫ macrophages seemed to be reduced as a result of the lack of PIP 3 production by PI3K␥, an induction (GTP␥S)-preloaded HA-Rac1 and HA-Rac2 strongly associated with the PAK-PBD, whereas no association with the GDP-preloaded HA-Rac was observed. Bottom, in each case, equal transgene expression was confirmed by immunoblot using total cell lysates and HA antibody. The 10-s value of RANTES-induced Rac1 and Rac2 activation in WT cells was normalized to an arbitrary value of 100. Values are mean Ϯ minimal and maximal range (n ϭ 2). Phosphorylation of PAK2  of HA-Rac1-(C) and HA-Rac2-(D) transduced cells analyzed by immunoblot with a phosphospecific antibody. Bottom, blots were stripped and reprobed with PAK antibody to confirm equal protein loading.
of PIP 3 production generated by an alternative class I PI3K agonist should rescue RANTES-mediated Rac activation in PI3K␥Ϫ/Ϫ macrophages.
CSF-1 efficiently triggered PI3K activation, as monitored by AKT Ser-473 phosphorylation (Fig. 6B), but as was shown already (Figs. 1A, 2A, and 5C), CSF-1 alone was ineffective in Cell lysates were subjected to immunoblot analysis with phosphospecific antibodies recognizing phosphorylation of PAK2  (A) and of AKT Ser-473 (B). Bottom, blots were reprobed with PAK and AKT antibodies respectively. C, as a control for Rac activation, cells were stimulated with 0.8 M RANTES. Rac activation was analyzed as in Fig. 1 and was activated by RANTES, but not by CSF-1 stimulation. The 10-s value of RANTES-induced Rac activation in WT cells was normalized to an arbitrary value of 100 and compared with the 30 s value of CSF1. Values are mean Ϯ S.E. D, cells were starved in medium containing 5% fetal bovine serum for 4 h, stimulated with 50 ng/ml CSF-1 for the indicated times, and Rac activation was monitored. Bottom, equal Rac content in lysates was confirmed in each experiment.
Silencing of Rac1 Expression Inhibits RANTES-induced PAK2 Activation and Cell Migration-We have demonstrated that Rac1, and not Rac2, predominantly depends on PI3K␥ activity (see Fig. 3). To investigate whether RANTES-induced Rac1 activation indeed couples to PAK2 activation and to cell migration, we designed retroviral constructs to introduce shRNA specific for Rac1 into primary murine macrophages (Fig. 7A). Silencing efficacy of Rac1 mRNA expression was analyzed by real-time PCR for five different constructs target- ing different portions of sequence within Rac1 (data not shown; see "Experimental Procedures"). Analysis of total lysates showed decent silencing of Rac1 expression in shRac1-transduced cells on the mRNA and the protein levels (Fig. 7B). RANTES-induced PAK2  phosphorylation was markedly reduced in shRac1 expressing BMDM compared with control cells transduced with empty control virus (Fig. 7C). In notable contrast to PAK2  phosphorylation, RANTESinduced phosphorylation of AKT Ser-473 was not affected in shRac1 expressing BMDM (data not shown). These data thus clearly indicate that Rac1 is signaling downstream PI3K and that, in these conditions, Rac does not exert a feedback activation of PI3K (46).
To asses the role of Rac1 in chemotactic gradient sensing, cells were exposed to a RANTES-releasing micropipette and images taken according to the indicated time intervals (Fig.  7D). Cells treated with control virus formed a well-defined pseudopod and crawled toward the source of attractant (see also Fig. 7E, left). In contrast, cells transduced with Rac1 shRNA displayed still ruffle formation but no clear leading edge and showed no significant translocation in the direction of the RANTES-containing micropipette (Fig. 7E, right). Our data thus indicate that the activation of Rac1 and, subsequently, of PAK2 are essential steps for cell polarization and chemotaxis. DISCUSSION We report PI3K␥-dependent and -independent RANTES-induced activation of the small GTPase Rac1 and one of its major targets, PAK2, in primary murine macrophage migration. ShRNA-based silencing of Rac1 blocks PAK2 activation and cellular migration in response to RANTES.
In a similar manner to chemokines, the chemoattractant fMLP signals via GPCRs and has been demonstrated to induce Rac activation in neutrophils (16,27,47,48). Many studies suggest a critical role of PI3K in the regulation of Rac GTPase (49); however, contradictory findings were reported concerning the PI3K dependence of fMLP-stimulated Rac activation (27,47,48,50). Rac activation seemed surprisingly independent of PI3K␥, because neutrophils pooled from PI3K␥Ϫ/Ϫ mice showed no obvious defect of Rac activation after fMLP stimulation for 5 s (16). Our data showed a clear contribution of PI3K␥ in Rac activation after stimulation of macrophages with RANTES. Pharmacological inhibition and lack of PI3K␥ expression resulted in a similar reduction of Rac activation. As expected, AKT Ser-473 phosphorylation was impaired in PI3K␥Ϫ/Ϫ cells (15,17); however, maximal AKT Ser-473 phosphorylation (around 3 min) did not correlate with the peak of Rac activation (10 s). It would seem that Rac-GEFs sensed PIP 3 produced by PI3K␥ at the plasma membrane instantly, and Rac-GTPase-activating proteins rapidly increased intrinsic Rac-GTPase activity to inactivate Rac GTP .
The apparent contradiction of findings on the regulation of Rac by PI3K␥ may have several reasons. The extremely rapid kinetics might hamper analysis, especially because the contribution of PI3K␥ is partial. Furthermore, the regulation seems to vary with cell type; for instance, macrophages rely more on PI3K␥ signaling than do neutrophils (51). In PI3K␥Ϫ/Ϫ mice, migration of neutrophils toward chemokines in vitro and in vivo was merely reduced, whereas recruitment of macrophages in vitro and to the inflamed peritoneal cavity in vivo was more drastically impaired. Recent reports suggest that adhesion of cells critically influences the coupling of Rac to signaling modules (52). Activated Rac in non-adherent cells failed to stimulate the Rac effector PAK. We used differentiated adherent macrophages in our study. There are indeed reports that human and murine neutrophils activated Rac independently of PI3K (48) or PI3K␥ (16) in response to fMLP stimulation, whereas transfected COS-7 cells that express the fMLP receptor showed PI3K␥-dependent activation of Rac (50).
Signaling divergence of GPCRs is another possible reason for the observed variations. fMLP activates the prototype formyl peptide receptor (53), whereas RANTES-mediated signaling in our cellular system (BMDM, day 7) was triggered mainly by CCR5 and CCR1 (Ref. 10; see "Experimental Procedures"). Our results using several pharmacological inhibitors suggest that a PI3K-independent G i␤␥ and a pertussis toxin-sensitive G i␤␥ -PI3K␥-PIP 3 pathway cooperate to trigger complete Rac activation induced by RANTES. A novel Rac-GEF, P-Rex1 (PIP 3 -dependent Rac exchanger1), was recently biochemically purified (54). P-Rex1 integrates G i␤␥ and G i␤␥ -PI3K-PIP 3 signaling and therefore represents, together with other as-yet-unidentified proteins of this emerging subgroup, a likely candidate for a RANTES activated Rac-GEF. It is noteworthy that the RTKagonist CSF-1 alone was unable to activate Rac, even though PIP 3 production was induced, because phosphorylation of AKT-Ser-473 was blocked by PI3K inhibition. The hypothesis that a P-Rex1-like-Rac-GEF is at work in our cellular system was further supported by the following result. Priming with CSF-1 rescued RANTES-mediated Rac activation in PI3K␥Ϫ/Ϫ macrophages, presumably by re-establishing PIP 3 production via an alternative class I PI3K in the time window, when RANTES stimulates G i -protein coupled GPCR activation. However, a tyrosine kinase-based mechanism to activate Rac-GEFs like Vav or Tiam may also be involved in this priming effect (54).
Another explanation for the controversial observations regarding the PI3K dependence of Rac activation may lie in the difficulty in distinguishing between Rac isoforms. Different mouse knockout phenotypes of Rac1 and Rac2, as well as different expression patterns, suggest overlapping but also distinct functions for both isoforms (55)(56)(57)(58). However, it has been proven difficult to study differential activation of the nearly identical Rac-isoforms in primary cells (59,60). To bypass some experimental barriers, we retrovirally transduced mouse bone marrow cells with constructs encoding either HA-Rac1 or HA-Rac2. Our results provide evidence that in chemokine-stimulated macrophages Rac1, not Rac2, depends on PI3K␥ for efficient activation. RANTES has been reported to induce PAK activation in neutrophils by a stimulatory pathway that relies on PI3K activity (28). We show that PI3K␥ and subsequently Rac1 are required for RANTES-induced PAK2 activation. Silencing of Rac1 resulted in a drop of PAK2 activity and inhibition of migration. Macrophages deficient in PI3K␥ were impaired in RANTES-triggered PAK2 activation as a result of reduced Thr-402 and Ser-192/Ser-197 phosphorylation and of partial loss of Rac1 GTP formation. Experiments using pertussis toxin, LY294002, and PI3K␥Ϫ/Ϫ compared with WT BMDM demonstrated the requirement of PI3K␥ and G i␤␥ in a coregulatory mechanism for PAK2 activation. The RTK agonist CSF-1 itself did not activate Rac but did stimulate PAK2 activation via an alternative pathway involving PI3Ks other than PI3K␥, probably PI3K␤ and -␦ (61,62). Kinetics observed for PAK activation showed variations between experiments, but Rac activation within the same experiment was never observed. The precise mechanism of CSF-1-induced PAK2 activation requires further investigation. The alternative signaling pathways triggered by RANTES and CSF-1 (Fig. 8) seemed to converge on the level of PAK2, because both CSF-1 and RANTES stimulation enhanced PAK2 activation. In parallel to the RANTES-PI3K␥-Rac1-PAK2 link described here, other Rho-GTPases might act in a comparable mechanistic manner, and their action might differ with cell type and stimulus. In line with this idea, a recent, elegant report demonstrated a G ␤ ␥-Cdc42 and PAK1 signaling complex in neutrophils that becomes PIP 3 -dependently located at the plasma membrane (63).
We show that PI3K␥ contributes to Rac1 and PAK2 activation in chemokine-stimulated primary murine macrophage migration. We further demonstrate that a substantial amount of Rac activation originates from a pertussis toxin-sensitive pathway that is PI3K-independent as well as tyrosine kinase-independent. In contrast, CSF-1 activation of PAK seems to be independent of Rac1 and PI3K␥ but does require the activity of other class I A PI3K isoforms. We conclude that partial loss of chemokine-induced Rac and Pak activation provides a mechanistic explanation for the reduced chemotaxis of PI3K␥-deficient macrophages (15,64). Furthermore, a PI3K-independent, FIG. 8. Model of signaling pathways leading to PIP 3 generation and Rac1-PAK2 signaling in RANTES-stimulated macrophages. The GEF integrates G i␤␥ and PIP 3 signals to activate Rac1, which couples directly to PAK2 and cell migration. PAKs including PAK2 are believed to become autophosphorylated on Thr and Ser residues upon activation, but an alternative phosphorylation mechanism by PDK1 has also been described (42). CSF-1 activates class I A PI3Ks and PAK2 by a mechanism that might involve Rho-GTPases other than Rac (not shown in the model). RANTES-and CSF-1-signaling converges on the level of PAK2.
pertussis-sensitive Rac pathway may be responsible for the remaining chemotactic activity in the absence of PI3K␥. It will be interesting to clarify dependence and interplay of PI3K-and Rac-isoforms in amplification loops of signal transduction (65) and to study macrophage-specialized functions such as phagocytosis and migration in response to chemokines.