Signaling and Cross-talk by C5a and UDP in Macrophages Selectively Use PLCβ3 to Regulate Intracellular Free Calcium*

Studies in fibroblasts, neurons, and platelets have demonstrated the integration of signals from different G protein-coupled receptors (GPCRs) in raising intracellular free Ca2+. To study signal integration in macrophages, we screened RAW264.7 cells and bone marrow-derived macrophages (BMDM) for their Ca2+ response to GPCR ligands. We found a synergistic response to complement component 5a (C5a) in combination with uridine 5′-diphosphate (UDP), platelet activating factor (PAF), or lysophosphatidic acid (LPA). The C5a response was Gαi-dependent, whereas the UDP, PAF, and LPA responses were Gαq-dependent. Synergy between C5a and UDP, mediated by the C5a and P2Y6 receptors, required dual receptor occupancy, and affected the initial release of Ca2+ from intracellular stores as well as sustained Ca2+ levels. C5a and UDP synergized in generating inositol 1,4,5-trisphosphate, suggesting synergy in activating phospholipase C (PLC) β. Macrophages expressed transcripts for three PLCβ isoforms (PLCβ2, PLCβ3, and PLCβ4), but GPCR ligands selectively used these isoforms in Ca2+ signaling. C5a predominantly used PLCβ3, whereas UDP used PLCβ3 but also PLCβ4. Neither ligand required PLCβ2. Synergy between C5a and UDP likewise depended primarily on PLCβ3. Importantly, the Ca2+ signaling deficiency observed in PLCβ3-deficient BMDM was reversed by re-constitution with PLCβ3. Neither phosphatidylinositol (PI) 3-kinase nor protein kinase C was required for synergy. In contrast to Ca2+, PI 3-kinase activation by C5a was inhibited by UDP, as was macropinocytosis, which depends on PI 3-kinase. PLCβ3 may thus provide a selective target for inhibiting Ca2+ responses to mediators of inflammation, including C5a, UDP, PAF, and LPA.

Calcium is an important messenger involved in the regulation of multiple cellular processes, and levels of intracellular free calcium ([Ca 2ϩ ] i ) 5 are precisely regulated (1)(2)(3). Increases in intracellular [Ca 2ϩ ] i are initiated by the phospholipase C (PLC) family of enzymes, which hydrolyze membrane-associated phosphatidylinositol 4,5-diphosphate (PIP 2 ) to produce inositol-1,4,5-trisphosphate (IP 3 ) and diacylglycerol (4). IP 3 triggers the release of Ca 2ϩ from stores in the endoplasmic reticulum, whereas diacylglycerol activates members of the protein kinase C (PKC) family. Following activation of stored Ca 2ϩ by IP 3 , influx of extracellular Ca 2ϩ across the plasma membrane may further contribute to an increase in [Ca 2ϩ ] i , which is regulated by several Ca 2ϩ pumps and buffers (1). The net level and duration of these Ca 2ϩ signals regulate cellular responses, including transcription, apoptosis, endocytosis, chemotaxis, and metabolism (3).
Simultaneous stimulation of two GPCRs coupled to different G␣ subunits, often G␣ i or G␣ s in combination with G␣ q , has been shown to yield synergistic Ca 2ϩ responses in several model systems (reviewed in Ref. 5). Limited studies have demonstrated this synergy in primary cells, including neurons and platelets, but the mechanisms of synergy vary and are not well defined (6,7).
Synergistic Ca 2ϩ responses resulting from heterologous GPCR ligation have been little studied in macrophages, where members of the GPCR superfamily can stimulate an increase in [Ca 2ϩ ] i by activating members of the PLC␤ family (4). As part of a systematic screen of RAW264.7 macrophage cells, C5a and UDP demonstrated synergy in producing a rise in [Ca 2ϩ ] i (8)). C5a is an important inflammatory mediator for macrophages and UDP, which is released following cell damage, is also present at sites of injury or infection (9,10). Both ligands signal through GPCRs; C5a signals through C5aR (11), and UDP signals through P2Y6 receptors (12). To examine GPCR cross-talk by these ligands in mouse macrophages we studied both RAW264.7 cells and primary bone marrow-derived macrophages (BMDMs).
Our studies show that signals generated by C5a and UDP, acting through G␣ i -and G␣ q -coupled pathways, respectively, converge at the level of PLC␤, and that these ligands, both individually and in concert, selectively use one PLC␤ isoform, PLC␤3, to activate the production of IP 3 and the consequent release of Ca 2ϩ from intracellular stores.

EXPERIMENTAL PROCEDURES
Reagents-UDP, UTP, LPA, platelet activating factor (PAF), human C5a, and fluorescein isothiocyanate-dextran were from Sigma. Mouse IgG 2a was from BD Pharmaceuticals. F(abЈ) 2 fragment of goat anti-mouse IgG was from Jackson ImmunoResearch Inc. Anti-PLC␤3 was from P. Sternweis, University of Texas Southwestern Medical Center. Anti-P-Akt and anti-P-ERK were from Cell Signaling Technologies. Fura2 was from Molecular Probes. Ionomycin, thapsigargin, pertussis toxin, LY294002, wortmannin, Calphostin C, staurosporine, U-73122, and U-73343 were from Calbiochem. Additional detailed protocols for reagents, procedures, and solutions are available on the internet and are referenced according to protocol number (e.g. PP00000226).
Mice and Culture of BMDM-Mice genetically deficient in G␣ q , G␣ 11 , PLC␤3, PLC␤4, or PLC␤2 were previously described (13)(14)(15)(16)(17)(18)(19). All strains were on the C57BL/6 background except PLC␤2-deficient mice, which were on 129SV. Genetically deficient and corresponding wild-type (WT) strains were bred and housed under approved animal protocols. For BMDM culture femurs and tibias were removed from sex-and agematched mice (4 -20 weeks of age, matched Ϯ 4 weeks) (PP0000017200). Briefly, marrow was flushed from bones, erythrocytes were lysed, and the white cells were seeded in nontissue culture Petri dishes for selection by growth and adhesion. After 6 days, over 99% of the surviving cells were macrophages, and these cells were maintained for up to 35 days in culture. Cells were cultured overnight in tissue culture plates prior to use in assays.
Lentivirus-mediated RNAi-Lentivirus was produced with a combination of three plasmids: (i) pCMV⌬R8.91 packaging plasmid, (ii) pMD.G envelope plasmid (20,21), and (iii) a lentiviral vector plasmid. The packaging and envelope plasmids were generously provided by D. Trono, Geneva. The lentiviral vector plasmids contained shRNA sequences expressed under RNA polymerase III promoters (U6 or H1) upstream of a Ubi-C promoter driving bicistronic expression of either enhanced green fluorescent protein or an hCD4 marker, followed by a resistance gene for either puromycin or hygromycin (22). Transfection of the 3 plasmids into 293T cells utilized Lipofectamine 2000 (Invitrogen) and 20 g of total DNA in a ratio of 4:3:2 for vector, packaging, and envelope plasmids, respectively (PP00000200). Two days post-transfection, lentivirus was concentrated by using Centricon microfiltration tubes (PP00000202). Macrophages were infected at a multiplicity of infection of ϳ10 in the presence of Polybrene at 4 g/ml (PP00000215) (22) . shRNAs targeting murine PLC␤2, PLC␤3,  and PLC␤4 employed the sequences GAA CAG AAG TTA  CGT TGT C; GCA GCG AGA TGA TTT GAT T; and ACG  CGA TTG AGT TTG TAA ATT A, respectively. Retrovirus-mediated Transduction of Macrophages with PLC␤-pFB-neo vectors carrying YFP-tagged murine PLC␤3 or YFP-FLAG epitope were transfected into the PlatE packaging line (23) to produce ecotropic retroviruses for transduction of day 2 cultures of bone marrow cells, which were differentiated into BMDM as described above.
Population Calcium Assays-Ca 2ϩ responses were measured by monitoring the fluorescence of Fura2-loaded cells (PP00000211). Baseline readings were collected for 30 -40 s. Calibration steps included additions of a Ca 2ϩ -minimizing solution (PS00000607) and Fura2 Ca 2ϩ -saturating solution (PS00000608) at the end of each recording, to allow calculation of [Ca 2ϩ ] i values according to the method of Grynkiewicz et al. (24), assuming a cytoplasmic K d of 250 nM for Fura2. Ca 2ϩ signals during the response period were quantified by features as indicated, including the peak offset response (difference between baseline Ca 2ϩ level and the maximal Ca 2ϩ level observed, reported in nanomolar) and an integrated response (integrated Ca 2ϩ level above the average baseline over the indicated time period, reported in nanomolar ϫ seconds).
Single-cell Calcium Assays-BMDM were plated in chambered coverglasses (Nunc, 8 wells/coverglass), cultured overnight, and loaded with Fura2-AM as described above. Video microscopy was performed on a Nikon TE-300 fluorescence microscope equipped with a Photometrics HQ2 camera, 37°C stage incubator (Bionomics), Xe lamp (Sutter), and filter/shutter/dichoric controllers (Sutter and Conix). Simple PCI software was used to control collection parameters and extract fluorescence intensity data for individual cells.
IP 3 Assay-After cell culture overnight, cells were placed in serum-free medium containing, 0.01% bovine serum albumin. After 1 h, ligands were added and, after varied time periods, dishes were transferred to ice, the media aspirated, and the cells washed with cold phosphate-buffered saline. Cells were scraped into 125 l of 5.4% perchloric acid solution and transferred to siliconized microcentrifuge tubes on ice. Samples were centrifuged at 14,000 ϫ g for 15 min at 4°C. 120 l of supernatant was neutralized with 5 N KOH containing 60 mM HEPES, and the samples recentrifuged at 14,000 ϫ g for 15 min at 4°C. The IP 3 content of the final supernatant was assayed with an Amersham Biosciences IP 3 [ 3 H] Biotrak assay kit. Results were reported as picomoles of IP 3 per 10 6 cells.
SDS-PAGE and Western Blot Analysis of Phosphoproteins (Protocols PP00000168 and PP00000181)-Cells were stimulated under the same conditions used for the IP 3 assay. Then buffer was aspirated, the cells were scraped into Laemmli sample buffer, and the samples heated. SDS-PAGE gels were loaded with 20 g of protein per lane, and Western blots were probed with anti-P-Akt, anti-P-ERK, and anti-Rho-GDI. Fluorescent signals, measured for P-Akt and P-ERK by using a phosphorimager, were normalized to Rho-GDI.
Macropinocytosis Assay-Macropinocytosis was assessed by measuring the cellular uptake of fluorescently labeled fluorescein isothiocyanate-dextran. In brief, BMDM cells were cultured overnight in non-tissue culture plates. Medium was replaced with Hanks' balanced salt solution with 1 mg/ml bovine serum albumin, pH 7.4. After 1 h ligands were added together with fluorescein isothiocyanate-dextran (150 kDa, Sigma, 1 mg/ml final concentration in well). Activity was stopped with cold medium. After washing, cells were harvested in phosphate-buffered saline with 5 mM EDTA and 5 mg/ml bovine serum albumin, and analyzed by flow cytometry (FACSCalibur, BD Biosciences) in 0.4% trypan blue solution (Sigma) to quench extracellular fluorescence. Ligand-stimulated activity was expressed as a ratio ("-fold stimulation") to baseline activity.
Statistical Analyses-The error bars in graphs depict the mean Ϯ S.E. The statistical significance of each comparison was evaluated by performing Student's t tests for one-way analysis of variance, followed by Dunnett or individual t tests (with Bonferroni correction), or non-linear mixed effects modeling, as appropriate. The effects of RNAi on Ca 2ϩ responses in RAW264.7 cells were analyzed by non-linear mixed effects modeling because of the non-normal distribution of Ca 2ϩ response features, and because of variation in responses between cell lines and assays. A p value of Ͻ0.05 was considered significant.

UDP and C5a
Interact to Produce a Synergistic Calcium Response in Macrophages-As part of a largescale screen, we observed a synergistic interaction between UDP and C5a for Ca 2ϩ signaling (8). This response showed a faster rise time and an increased peak offset (peak response minus baseline) compared with the predicted additive response by the individual ligands (Fig. 1A). For the peak offset, the observed dual ligand response was increased by 1.5-2-fold over the predicted additive response. Integration of the response over the first 20 s yielded similar values. The ratio of observed/predicted values for such features of the response was referred to as the synergy ratio. BMDM showed a similar synergistic increase in the Ca 2ϩ response, but the synergy ratio was much greater than in RAW264.7 cells (Fig. 1B).
For both cell types the optimal synergistic concentrations of each ligand were at or near the threshold for stimulation ( Fig. 1C for BMDM, RAW264.7 data not shown). Synergy was nonetheless still observed at higher concentrations of both ligands, which fell within the linear portions of their dose-response curves. For RAW 264.7 cells, optimal concentrations for synergy were 0.25-100 nM C5a and 40 -500 nM UDP, whereas for BMDM they were 0.1-3 nM C5a and 150 -500 nM UDP.
Synergy Is Dependent on Signaling through Both G␣ i -and G␣ q -dependent GPCRs-C5a engages C5aR, which signals predominantly through G␣ i -coupled heterotrimers (11,25,26). Thus, in BMDM and RAW264.7, the Ca 2ϩ response to C5a was inhibited following treatment with pertussis toxin (PTX), whereas that of UDP was not ( Fig. 2A and supplemental Fig. S1). PTX-mediated inhibition of signaling by low concentrations of C5a (Ͻ1 nM) was complete, but at high concentrations residual PTX-insensitive calcium signaling was detected both in WT BMDM and in BMDM from mice lacking either G␣ q or G␣ 11 (data not shown). Saturation of PTX intoxication of the C5a Ca 2ϩ response was reached using 5 ng/ml for 18 h (supplemental Fig. S1). Others have demonstrated a role for G␣ 15 in C5a signaling in primary macrophages (27). Although our data indicate that most C5a signaling is PTX-sensitive, they are consistent with some signaling through G␣ 15 when C5a is present in high concentrations. The Ca 2ϩ response to C5a in BMDM from FIGURE 1. UDP and C5a produce a synergistic Ca 2؉ response in macrophages. Intracellular Ca 2ϩ levels were calculated for Fura2-loaded adherent macrophage populations from kinetic assays in 96-well plates. After 40-s baseline readings, ligands were added and responses were monitored for 2.5 min. Each line in the graphs represents the average of three to four individual wells per assay and the error bars (S.E.) are shown for the dual ligand line in each graph. Synergy was evaluated by comparing the experimentally observed dual ligand responses to the predicted additive responses of the individual ligands. The dual ligand response was quantified as the ratio of the observed/predicted additive responses and the term "synergy ratio " was applied to the ratio of the peak offsets (PO ϭ peak height Ϫ baseline). A, RAW264.7 cells were stimulated with UDP (2.5 M), C5a (10 nM), or simultaneous UDP and C5a. Synergy ratio ϭ 1.38. This is a representative experiment of n ϭ 12 with similar results. B, BMDM were stimulated with UDP (500 nM), C5a (0.37 nM), or simultaneous UDP and C5a. Synergy ratio ϭ 2.62. This is a representative experiment of Ͼ25 with similar results. C, dose response pattern for simultaneous UDP-and C5a-stimulated Ca 2ϩ responses quantified by the synergy ratio. Ligand concentrations are expressed as log (nM). The optimal synergy ratio was at ϳ300 nM UDP ϩ 0.3 nM C5a. The surface was interpolated from 84 individual experiments composed of 15 samples each. mice genetically deficient in G␣ i2 was intact (data not shown), suggesting that G␣ i3 is sufficient to support C5a signaling in these cells, because G␣ i1 and G o are not expressed in macrophages (data not shown). Inhibition by PTX was similar to that observed for WT cells (data not shown).
UDP binds to purinergic receptors of the P2Y family, which usually signal through members of the G␣ q family (12,28,29). In accord with this, we found that the Ca 2ϩ response was completely lost in BMDM from G␣ q -deficient mice (Fig. 2B). Surprisingly, although BMDM express other members of the G␣ q family, including G␣ 11 and G␣ 15 (Fig. 2C), these are evidently unable to substitute for G␣ q in response to UDP. Furthermore, the BMDM from G␣ 11 -deficient mice had no reduction in Ca 2ϩ responses for UDP (or for C5a, data not shown) compared with wild type.
UDP only binds with high affinity to the P2Y6 receptor on macrophages. UTP, which has much lower affinity for P2Y6, binds also to P2Y2 and P2Y4 receptors (12,28). To demonstrate that the responses to UDP did not involve contaminating UTP, we separately tested UTP and UDP before and after treatment with hexokinase (12), which catalyzes conversion of UTP to UDP. Hexokinase treatment of UDP had no effect on its capacity to increase [Ca 2ϩ ] i in BMDM (not shown), indicating that contaminating UTP was not responsible for the observed responses in BMDM. The efficacy of hexokinase treatment was confirmed by showing that hexokinase treatment of UTP ablated its capacity to increase [Ca 2ϩ ] i in NIH 3T3 cells, which respond to UTP but not UDP.
The removal of either G␣ q or G␣ i , via genetic deletion or PTX intoxication, respectively, also eliminated any synergistic Ca 2ϩ response to dual ligand stimulation (Fig. 2D). Thus, synergy between C5a and UDP is dependent on the G i -and G qlinked subunit effectors that are activated by C5a and P2Y6 receptors, respectively.
LPA and PAF Also Synergize with C5a for Ca 2ϩ Responses-To determine whether synergy for Ca 2ϩ signaling occurred with other ligand pairs, we also examined responses to C5a or UDP in combination with PAF or LPA, both of which induce a Ca 2ϩ response in macrophages through GPCRs. Pairing of C5a with either PAF or LPA demonstrated a robust synergy in Ca 2ϩ signaling. Little or no synergy was seen with UDP/PAF, UDP/ LPA, or PAF/LPA (data not shown). The levels of synergy observed for C5a paired with LPA or PAF (Fig. 3, A and B) were comparable with those for C5a paired with UDP.
PAF and LPA, like UDP, signaled Ca 2ϩ primarily through G␣ q in BMDM (Fig. 3C), but unlike UDP this was not exclusive; in the G␣ q -deficient cells residual Ca 2ϩ responses for PAF were abrogated by PTX, indicating a minor contribution from G␣ icoupled pathways. PTX did not reduce the Ca 2ϩ response to LPA, but instead surprisingly enhanced it in both G␣ q -deficient and WT BMDM (Fig. 3D). These data suggest for the first time that G␣ i -coupled receptors basally inhibit LPA Ca 2ϩ signaling. As with UDP, synergy by either LPA or PAF with C5a was lost in G␣ q -deficient BMDM (data not shown). Thus, although these receptors can activate some Ca 2ϩ signaling independently of G␣ q , synergy with C5a nonetheless requires G␣ q activation.
Overall, these results indicate that the simultaneous activation of G␣ q and G␣ i heterotrimers results in a synergistic Ca 2ϩ response in macrophages. C5aR was the only endogenous G␣ icoupled GPCR on BMDM that we found to be capable of generating a robust Ca 2ϩ response independently, and it was also the only receptor that synergized with ligands for G␣ q -coupled receptors.
Synergy Requires Dual Receptor Occupancy-We next examined the possibility that one ligand might prime cells for subsequent responses, for example, by increasing the supply of PIP 2 to provide a heightened state of responsiveness to the second stimulus (30). Although synergy was greatest when C5a and UDP were added simultaneously, it was also evident when ligands were added as much as 10 min apart. The sequence of addition was irrelevant. Fig. 4A shows the results for ligands Wild type (WT), G␣ q heterozygote (ϩ/Ϫ), and G␣ q -deficient (Ϫ/Ϫ) BMDM were stimulated with either UDP (2.5 M) or C5a (10 nM). Peak offsets of responses are shown normalized to those of the wild-type cells from each experiment. Values are mean Ϯ S.E. from three experiments. *, p Ͻ 0.001. C, quantitative reverse transcriptase-PCR for G␣ q family isoforms q, 11, and 15 was performed on RAW264.7 cell and BMDM samples to determine relative prevalence. Transcript levels were normalized to those for G␣ q for the same cell type. Data shown are mean Ϯ S.E. from n ϭ 3 samples per cell. D, synergy following dual ligand stimulation requires both G␣ q and G␣ i subunits. WT or G␣ q -deficient (Ϫ/Ϫ) BMDM were stimulated with UDP (500 nM), C5a (0.75 nM), or simultaneous UDP and C5a. WT cells were cultured overnight with or without PTX (100 ng/ml). Data shown are from a representative experiment of three to four with similar results. Each line in the graphs represents the average of three to four individual wells in the assay. separated by 100 s. However, removal of the first ligand during the interim eliminated synergy (Fig. 4B). Thus, if either ligand primes the synergistic response, this effect is rapidly lost. Functionally, synergy requires simultaneous receptor occupancy by both ligands.
Synergy Affects the Initial Release of Ca 2ϩ from Intracellular Stores and IP 3 Production-Synergy between C5a and UDP affected the early rise in [Ca 2ϩ ] i , suggesting an effect on the release of intracellular calcium stores. To test this, the Ca 2ϩ responses to C5a and UDP, either alone or in combination, were measured after acute addition of EGTA to deplete extracellular Ca 2ϩ . Synergy occurred in the presence of EGTA, confirming an effect on the release of intracellular Ca 2ϩ (Fig. 4C). Without EGTA, however, synergy also extended to the sustained phase response, which is dependent on the influx of extracellular Ca 2ϩ . Thus, synergy between C5a and UDP begins with the release of intracellular Ca 2ϩ stores but extends to the influx of extracellular Ca 2ϩ .
The release of Ca 2ϩ from intracellular stores is activated by IP 3 binding to IP 3 receptors on the endoplasmic reticulum to open ER calcium channels (1). Simultaneous stimulation of BMDM with C5a and UDP in amounts that produced a synergistic Ca 2ϩ response also resulted in synergy in the production of IP 3 (Fig. 4D), suggesting synergistic mechanisms are manifest at the level of PLC␤ activation. Levels of IP 3 measured at 30 s and 1 min after ligand additions were increased. Ca 2ϩ levels began to decline while IP 3 was still rising, indicating that levels of [Ca 2ϩ ] i are not solely regulated by levels of IP 3 .
The Synergistic Ca 2ϩ Response Is Independent of Feedback Pathways Involving PI 3-Kinase (PI3K) or PKC-Downstream of GPCR activation, PLC␤ may be regulated by other signaling components, including those generated following the activation of PI3K (by G␤␥ subunits) or PKC (by diacylglycerol) (5). In our studies, however, inhibition of PI3K by LY294002 or PKC by Calphostin C or staurosporine did not significantly affect synergy (supplemental Fig. S2.). The activity of the inhibitors was confirmed by inhibition of Akt or myristoylated alanine-rich c-kinase substrate (MARCKS) phosphorylation (supplemental Fig.  S3). These data are further evidence that an early signaling event is involved in the mechanism of synergy.
C5a and UDP Make Selective Use of PLC␤ Isoforms-To examine the role of PLC␤ in the signaling response to C5a and UDP, we first determined levels of transcripts for PLC␤ isoforms in RAW264.7 cells and BMDMs. By both microarray analysis (data not shown) and reverse transcriptase-PCR (Fig. 5), we found that both cell types express PLC␤2, PLC␤3, and PLC␤4, with little or no PLC␤1. At the transcript level, the proportions of these PLC␤ isoforms, however, differ between RAW264.7 cells and BMDM; normalized to PLC␤3, RAW264.7 cells express similar levels of transcripts for PLC␤2, PLC␤3, and PLC␤4, whereas BMDM express PLC␤2 Ͼ PLC␤3 Ͼ PLC␤4.
To determine whether C5a and/or UDP made selective use of these PLC␤ isoforms, we examined the Ca 2ϩ response in BMDM from mice genetically deficient in PLC␤2, PLC␤3, or PLC␤4. BMDM from mice deficient in PLC␤3 demonstrated a marked loss of signaling in response to all GPCR ligands, including, C5a, UDP, PAF, and LPA ( Fig. 6 and Table 1). Activation of Ca 2ϩ responses, however, was intact in response to ligation of Fc␥RI by cross-linked IgG 2a (Fig. 6), demonstrating that macrophages from PLC␤3-deficient mice are not deficient in the generation of [Ca 2ϩ ] i to a non-GPCR ligand. In BMDM from mice deficient in PLC␤4, the Ca 2ϩ response to UDP was also consistently reduced, whereas the response to C5a was slightly elevated (Table 1 and Fig. 6) and Fc␥RI signaling was normal. No loss of signaling to either UDP or C5a was seen in BMDM from mice deficient in PLC␤2. Thus, in BMDM, signaling by both C5a and UDP is selectively dependent on PLC␤3, but signaling by UDP is also partly dependent on PLC␤4.
To examine the role of PLC␤ in RAW264.7 cells, we used RNAi against the different PLC␤ isoforms. The loss of PLC␤ isoforms in response to RNAi was incomplete (supplemental Fig. S4), but this approach allowed the testing of a uniform cell line, and it avoided possible developmental effects on macrophages due to PLC␤ isoform loss. The depletion of PLC␤3 from RAW264.7 cells by RNAi reduced signaling by C5a, although not to the same extent as in BMDM genetically deficient in PLC␤3 (Table 2). Cells depleted of PLC␤3 by RNAi were not deficient in their response to UDP, but RNAi against PLC␤4 caused a loss of signaling in response to UDP, with a slight elevation in C5a signaling (Table 2). Thus, signaling by C5a depends mostly on PLC␤3 in both BMDM and RAW264.7 cells. Signaling by UDP is partially dependent on PLC␤3 in BMDM, but we could not detect this dependence in RAW264.7 cells by RNAi of PLC␤3. Signaling by UDP is also dependent on PLC␤4 in both BMDM and RAW264.7 cells, whereas deficiency of PLC␤4 augments C5a signaling in both cells.
Ca 2ϩ Responses Are Restored in PLC␤3-deficient BMDM Reconstituted with PLC␤3-Retroviruses were used to transduce wild-type and PLC␤3-deficient BMDM with either YFP-tagged murine PLC␤3 or control YFP-tagged FLAG cDNAs. Single-cell calcium assays were performed, which allowed identification of transduced cells by YFP fluorescence and comparison of responses by transduced and non-transduced cells (Fig. 6, B and C). Reconstitution of PLC␤3-deficient BMDM with PLC␤3 reconstituted the Ca 2ϩ response to both C5a and UDP, alone and in combination, indicating that the loss of Ca 2ϩ response in the PLC␤3-deficient cells is not due to an associated developmental defect.
Synergistic Ca 2ϩ Responses Also Show Isoform Dependence-We next tested the role of the PLC␤ isoforms in synergy between C5a and UDP. In BMDM lacking PLC␤3 or PLC␤4, only those deficient in PLC␤3 were deficient in synergy ( Fig. 7A and Table 3), as reflected by a reduced synergy ratio. Because signaling by individual ligands was lower than wild type in these cells, the predicted additive responses were also lower, but a residual synergistic response was still detected in PLC␤3-deficient cells (Table 3). Thus synergy in Ca 2ϩ signaling, like signaling by individual ligands, is primarily dependent on PLC␤3, but some synergy can be seen without it, Notably, lack of PLC␤4 did not reduce synergy in BMDM but instead enhanced it. We conclude that PLC␤3, but not PLC␤4, plays an important role in synergy between these ligands as well as in their individual responses.  . Selective use of the PLC␤3 and -␤4 isoforms in GPCR signaling in BMDM did not correlate with higher levels of expression. Quantitative reverse transcriptase-PCR for PLC␤ isoforms 1, 2, 3, and 4 was performed on RAW264.7 cell and BMDM samples to determine relative prevalence. Transcript levels were normalized to those for PLC␤3 for the same cell type. Data shown are mean Ϯ S.E. from n ϭ three to four samples per cell. Little to no expression of PLC␤1 mRNA was observed, as shown.
As with Ca 2ϩ signaling, BMDM lacking PLC␤3, but not PLC␤4, failed to demonstrate synergy in the production of IP 3 (Fig. 7B). Thus, studies of both Ca 2ϩ and IP 3 indicate that synergy in signaling by C5a and UDP is the result of enhanced activity of PLC␤3.

Dual-Ligand Effects on PI 3-Kinase Contrast to Those on PLC-
To determine whether the synergistic effects of C5a plus UDP dual ligand stimulation were reflected in signaling events other than PLC activation, we examined activation of PI3K. G␤␥ subunits directly activate PI3K-p110␥ (31) and GPCRs can also activate PI3K-p110␣ and PI3K-p110␤ (32). The G␣ q subunit does not activate PI3K, but instead can interact with and inhibit PI3K-p110␣ (33,34). Thus, PI3K activity reflects important proximal GPCR signals. To assess activation of PI3K, we measured the phosphorylation of Akt, which requires anchoring of its pleckstrin homology domain to PIP 3 produced by PI3K at the cell membrane. In BMDM, C5a rapidly activated PI3K, with peak phosphorylation of Akt at ϳ3 min (data not shown). In contrast, UDP did not activate PI3K, and it inhibited the phospho-Akt response to C5a (Fig. 8). This inhibition of Akt phosphorylation by UDP was at least partially selective, as ERK phosphorylation showed additivity. UDP did not inhibit PI3K activation in response to cross-linking of Fc␥RI (data not shown), demonstrating that signaling by UDP did not globally interfere with all forms of PI3K activation. The observation that UDP inhibits PI3K activation by C5a while promoting Ca 2ϩ signaling suggests that these pathways are differentially regulated.
The Opposing Effects of C5a/UDP Signal Interactions on PLC and PI3K Are Reflected in Macropinocytosis-Macropinocytosis, the endocytic process whereby cells internalize substantial volumes of extracellular fluid and solutes, is dependent on both PLC and PI3K (35,36), and this "sampling " of the environment contributes to macrophage antigen presentation (37,38). We found that C5a activates macropinocytosis by BMDM, whereas UDP does not. Macropinocytosis was inhibited by dual ligand stimulation (Fig.  9), in contrast to synergy for PLC and Ca 2ϩ but in parallel with the inhibition of PI3K.

DISCUSSION
Our studies demonstrate the preferential use of PLC␤ isoforms by GPCRs in eliciting a Ca 2ϩ response in macrophages. Furthermore, they indicate that synergy in signaling by the G␣ i -coupled C5aR, together with the G␣ q -coupled P2Y6 receptor for UDP, depends on a selective use of PLC␤3. Synergy in the Ca 2ϩ response to C5a and UDP correlated with synergy in IP 3 production, suggesting signal convergence at the level of PLC␤ activation. In contrast to Ca 2ϩ activation, synergy between C5a and UDP was not observed in PI3K activation. Instead, the activation of PI3K by C5a was opposed by UDP. A similar effect was seen in the activation of macropinocytosis, which is dependent on both PLC and PI3K. Thus, synergy was selective for IP 3 production and Ca 2ϩ response, consistent with a selective effect on PLC␤.
The preferential use of PLC isoforms by GPCRs in macrophages did not simply reflect differential levels of expression of transcripts for the PLC␤ isoforms. Four isoforms of PLC␤ have been identified (4). We found that both BMDM and RAW264.7 cells expressed transcripts for PLC␤2, ␤3, and ␤4 but not PLC␤1, as determined by gene array analyses on Affymetrix chips and by reverse transcriptase-PCR. We have not been able to develop assays that adequately quantify differences in protein expression of these PLC␤ isoforms, but our results, nonetheless, suggest that the selective use of PLC␤3 in macrophages for Ca 2ϩ signaling and synergy is despite the expression of PLC␤2 and PLC␤4. Thus, in contrast to platelets and neutrophils, PLC␤3 appears to be the major functional isoform in macrophages. While this article was in preparation, Wang et al. (39) also reported reduced Ca 2ϩ responsiveness to C5a by macrophages from PLC␤3-deficient mice, and they linked this to increased apoptosis, and diminished atherosclerosis. Our studies demonstrate that in macrophages UDP can use PLC␤4 as well as PLC␤3, but C5a synergizes with UDP and other activators of G␣ q through signals that converge at the level of PLC␤3, and responsiveness can be restored by transduction of cells with PLC␤3, showing that the defect in signaling does not reflect developmental changes in other pathways.
Synergy in the macrophage Ca 2ϩ response was observed both in the initial, rapid release of Ca 2ϩ from intracellular stores and in the sustained elevation of cytoplasmic Ca 2ϩ levels. with similar results. B, IP 3 production in PLC␤ isoform-deficient BMDM. Cells were stimulated with C5a (10 nM), UDP (2.5 M), or simultaneous C5a and UDP as indicated, and signaling was stopped by cell lysis at 1 min after stimulation. IP 3 was measured using a competitive binding assay for the IP 3 receptor, and results are reported as picomole/10 6 cells. Data represent pooled results from two to four assays with 2 replicate samples per condition per assay. *, p Ͻ 0.005.    (40). Synergy between G␣ i -and G␣ q -coupled receptors has previously been observed in other cell types. In several systems, including smooth muscles, astrocytes, and kidney epithelial cells, G␣ i -coupled GPCRs may alone not trigger a Ca 2ϩ response, but responses may be facilitated in combination with, or after priming by, G␣ q -coupled receptors (41)(42)(43). This synergy is reflected in the generation of IP 3 , as in our current studies of macrophages, implicating PLC in the pathway of synergy.
Our findings narrow the possible mechanisms by which synergy in Ca 2ϩ signaling by macrophages may occur. All PLC␤ isoforms can bind G␣ q subunits, albeit with differing affinities (44 -48), and under certain conditions PLC␤4 demonstrates the highest specific activity for hydrolyzing PIP 2 (49). Consistent with this, the absence of PLC␤4 reduced mobilization of [Ca 2ϩ ] i by all ligands that activate G␣ q , including UDP, LPA, and PAF. The loss of PLC␤4, however, did not impair synergy but instead increased it. Thus, PLC␤4 appears to inhibit rather than promote synergy in macrophages. PLC␤2 and -␤3 are both potently activated by G␤␥ (25,47,50,51), whereas PLC␤4 is not (49). Although most Ca 2ϩ synergy was lost in mice lacking PLC␤3, we could still detect low levels of synergy. We hypothesize that PLC␤2 may be capable of mediating synergy, but in macrophages the contribution of PLC␤2 is small in relation to that of PLC␤3. In all, these results suggest that synergy between C5a and UDP in Ca 2ϩ signaling in macrophages does not require multiple isoforms of PLC␤ but instead involves the convergence of molecular mechanisms that primarily activate PLC␤3, but which may to a lesser extent activate PLC␤2.
Synergy between the G␣ i receptor C5aR and G␣ q receptors does not establish that G␣ q itself participates in the synergy. G␤␥ signaling may differ between C5a and UDP, and synergy could reflect interactions between their unique G␤␥ pathways. Indeed, loss of G␤2 subunits via RNAi disrupts C5a but not UDP Ca 2ϩ responses in RAW264.7 cells (Ref. 52, and data not shown).
In our studies, C5a synergized not only with UDP, but also with PAF and LPA in stimulating a rise in [Ca 2ϩ ] i . Studies of G␣ q -deficient BMDM confirmed that both PAF and LPA utilize G␣ q , but they also revealed important and interesting differences between these ligands and UDP. Unlike UDP, neither PAF nor LPA was fully dependent on G␣ q . The remaining Ca 2ϩ signaling with PAF utilized G␣ i , as interruption of this pathway with PTX in the absence of G␣ q removed all Ca 2ϩ signaling in response to PAF. In contrast, LPA Ca 2ϩ signaling was not reduced by PTX. Instead it was markedly increased. The G proteins used by LPA to elevate [Ca 2ϩ ] i in G␣ q -deficient BMDM are unknown, but it appears that they are normally inhibited by G␣ i .
At high ligand concentrations, C5a also demonstrated some PTX-insensitive activation of Ca 2ϩ signaling. We found that this response was still present in G␣ q -or G␣ 11 -deficient mice, suggesting coupling of C5aR to the more promiscuous G␣ 15 , as has been observed by others (27). However, no Ca 2ϩ synergy was observed with the combination of two G␣ q family-linked ligands. Optimal synergy was observed at low concentrations of C5a, where the C5a-stimulated Ca 2ϩ response was entirely PTX sensitive, so we infer that the synergy is attributable to the G␣ i activation by C5aR. In BMDM, the C5a receptor was the only Ca 2ϩ signaling receptor identified that was primarily dependent on G␣ i . GPCR-mediated PLC␤ activation can be regulated by positive or negative feedback loops. The pleckstrin homology domain of PLC␤ preferentially binds to the phosphatidylinositiol 3-phosphate product of PI3K (4) and thus PI3K has the potential to modulate PLC␤ activity. In our studies, however, inhibition of PI3K by LY294002 did not alter the Ca 2ϩ synergy, indicating that PI3K does not measurably contribute to synergy. PKC may interact with PLC at several levels. It can directly phosphorylate PLC␤, inactivating it (53). It can also phosphorylate and regulate signaling via GPCRs, and can phosphorylate some G protein-coupled receptor kinases (54). In our studies, however, inhibition of PKC with either Calphostin C or staurosporine did not alter Ca 2ϩ synergy. After 1 min of stimulation, the assay was stopped by sample lysis. Western blots were probed with specific antibodies and quantified using a phosphorimager. Phosphoprotein values were normalized to levels of Rho-GDI in the samples and then expressed as a -fold increase above the baseline level, represented by the average of control cell samples not stimulated with specific ligands. Data are shown for P-Akt and P-ERK from nine replicate experiments as mean Ϯ S.E. *, p Ͻ 0.05; **, p Ͻ 0.01. The acute nature of the synergy observed (occurring within seconds of dual ligand addition) and the demonstrated requirement for simultaneous dual receptor occupancy argue against the possibility that one receptor might drive "priming " events affecting responses to the second receptor. Mechanisms for synergy reflecting priming effects have been proposed in a number of other systems (55). The immediate synergy in macrophages precludes changes in receptor or other protein expression levels. Alternatively, a priming event could increase the supply of the PLC substrate PIP 2 to enhance production of IP 3 (56 -58), and in some cases this has been shown to persist for hours after the first ligand stimulation. This synergy mechanism would not require dual receptor occupation during heterologous ligand stimulations of Ca 2ϩ unless increases in the supply of PIP 2 were lost rapidly (we tested 3 min after 1st ligand removal by which time synergy was lost).
The consequences of combined signaling by C5a and UDP in macrophages may be particularly important in areas of inflammation, where C5a is produced, and where UDP may be released from dying cells (59). C5a in particular plays a central role in inflammation, and consequences of Ca 2ϩ signaling would be augmented by UDP, whereas consequences of PI3K activation would be inhibited. The recent report describing a reduction of atherosclerosis in PLC␤3-deficient mice, due to macrophage hypersensitivity to apoptotic induction, links inflammatory outcome to Ca 2ϩ signaling and survival in macrophages (39).