Amyloid-β Induces Chemotaxis and Oxidant Stress by Acting at Formylpeptide Receptor 2, a G Protein-coupled Receptor Expressed in Phagocytes and Brain*

Amyloid-β, the pathologic protein in Alzheimer's disease, induces chemotaxis and production of reactive oxygen species in phagocytic cells, but mechanisms have not been fully defined. Here we provide three lines of evidence that the phagocyte G protein-coupled receptor (N-formylpeptide receptor 2 (FPR2)) mediates these amyloid-β-dependent functions in phagocytic cells. First, transfection of FPR2, but not related receptors, including the other known N-formylpeptide receptor FPR, reconstituted amyloid-β-dependent chemotaxis and calcium flux in HEK 293 cells. Second, amyloid-β induced both calcium flux and chemotaxis in mouse neutrophils (which express endogenous FPR2) with similar potency as in FPR2-transfected HEK 293 cells. This activity could be specifically desensitized in both cell types by preincubation with a specific FPR2 agonist, which desensitizes the receptor, or with pertussis toxin, which uncouples it from Gi-dependent signaling. Third, specific and reciprocal desensitization of superoxide production was observed whenN-formylpeptides and amyloid-β were used to sequentially stimulate neutrophils from FPR −/− mice, which express FPR2 normally. Potential biological relevance of these results to the neuroinflammation associated with Alzheimer's disease was suggested by two additional findings: first, FPR2 mRNA could be detected by PCR in mouse brain; second, induction of FPR2 expression correlated with induction of calcium flux and chemotaxis by amyloid-β in the mouse microglial cell line N9. Further, in sequential stimulation experiments with N9 cells, N-formylpeptides and amyloid-β were able to reciprocally cross-desensitize each other. Amyloid-β was also a specific agonist at the human counterpart of FPR2, the FPR-like 1 receptor. These results suggest a unified signaling mechanism for linking amyloid-β to phagocyte chemotaxis and oxidant stress in the brain.

In Alzheimer's disease, progressive dementia and neurodegeneration are associated with a complex pathologic lesion made up of neurofibrillary tangles and aggregated extracellular protein deposits, known as senile plaques, which together are surrounded and infiltrated by activated microglial cells (1). Amyloid-␤ (A␤), 1 a heterogeneous 39 -43-amino acid, self-ag-gregating peptide produced by sequential cleavage of amyloid precursor protein by the enzymes ␤-secretase and ␥-secretase, is central to the pathogenesis of this disease (2,3). The main component of senile plaque (4), A␤ is also biologically active and has been proposed to promote neurodegeneration by both direct and indirect mechanisms. It is directly toxic to cultured neurons in vitro (5) and is able to regulate production of the protein tau (6), which accumulates in neurofibrillary tangles. It may also induce neurodegeneration indirectly through its proinflammatory activity (7)(8)(9)(10), which includes the ability to directly induce chemotaxis of mononuclear phagocytes (11,12) as well as production of cytokines and reactive oxygen species (13)(14)(15)(16)(17)(18)(19) by microglial cells, monocytes, and neutrophils. A␤ may also induce phagocyte accumulation and activation indirectly, by inducing C5a production through activation of complement (20) or by inducing macrophage colony-stimulating factor release from neurons (21). Consistent with a proinflammatory role, intravascular injection of A␤ causes endothelial cell leakage and leukocyte adhesion and migration in vivo (17). The notion that inflammation is important in the pathogenesis of Alzheimer's disease is consistent with clinical reports linking nonsteroidal anti-inflammatory drug administration to reduced incidence of disease and milder clinical course in affected patients (22).
The mechanism of A␤ action on cells has not been fully defined yet. A␤ has been reported to bind to several otherwise unrelated receptors, including the receptor for advanced glycation end products (RAGE; Ref. 23), the class A scavenger receptor (19), the p75 neurotrophin receptor (24), glypican (25), neuronal integrins (26), and the N-methyl-D-aspartate receptor (27).
The role of glypican, N-methyl-D-aspartate receptors, integrins, and p75 neurotrophin receptor in mediating A␤ action is not defined. RAGE has been implicated in mediating A␤-induced oxidant stress in endothelial cells and cortical neurons, NF-B activation in endothelial cells, and induction of tumor necrosis factor-␣ production, chemotaxis, and haptotaxis of the mouse microglial cell line BV-2 (19); conflicting results have been reported with regard to the role of RAGE in A␤-induced neurotoxicity (28). Scavenger receptors have been reported to mediate adhesion of rodent microglial cells and human mono-cytes to A␤ fibril-coated surfaces, leading to secretion of reactive oxygen species and cell immobilization (18), and to mediate internalization of aggregated A␤ protein (29); however, these receptors do not appear to mediate A␤ stimulation of peripheral blood monocyte-dependent neurotoxicity (30). A␤ has also been reported to have direct toxic effects on membranes independent of receptors (31). Despite these advances, the precise mechanisms by which A␤ induces chemotaxis and oxidant production in primary phagocytic cells remain undefined.
Most known phagocyte chemotactic receptors are members of the G i class of G protein-coupled receptors (GPCRs), which signal through pertussis toxin-sensitive pathways (32). Recently, pertussis toxin was reported to block A␤ induction of interleukin-1␤ release from the human monocytic cell line THP-1 (14) as well as A␤ induction of calcium flux in HL-60 cells (33). This, together with the fact that calcium flux is strongly associated with GPCR activation by chemoattractants, suggested to us that A␤ may act via a GPCR. Since ligand promiscuity is a common property of chemoattractant receptors, we tested this hypothesis by examining the ability of cloned phagocyte chemoattractant receptors to reconstitute A␤ signaling in a transfected cell line. We also investigated receptors mediating A␤ signaling on mouse phagocytes (reported here) and human phagocytes (reported separately).
Preparation of Mouse Neutrophils-Neutrophils were obtained from the peritoneal cavity of wild type and gene knockout litter mates of F1 and F6 backcrosses of 129/sV FPRϪ/Ϫ mice with C57Bl/6 mice 3-4 h after intraperitoneal injection of a 3% thioglycollate solution, as previously described (41). The cell population was consistently composed of Ͼ90% neutrophils, as determined by light microscopy of DiffQuickstained cytospins. Thus, hereafter we will refer to this cell preparation as neutrophils.
Calcium Flux Analysis-To monitor intracellular Ca 2ϩ concentration, adherent cells were harvested by incubation in phosphate-buffered saline at 37°C for 15 min and then incubated in phosphate-buffered saline containing 2.5 M Fura-2/AM at 37°C for 45 min. Cells were washed twice with HBSS (Life Technologies) and suspended in HBSS at 1-2 ϫ 10 6 /ml. One ml of cells was added to 1 ml of HBSS and stimulated with ligand in a continuously stirred cuvette at 37°C in a fluorimeter (model MS-III; Photon Technology Inc., South Brunswick, NJ). Data were recorded every 200 ms as the relative ratio of fluorescence emitted at 510 nm following sequential excitation at 340 and 380 nm. The following ligands were evaluated: A␤ (nonfibrillated, human residues 1-42; California Peptide Research; Napa, CA), fMet-Leu-Phe (fMLF; Sigma), ATP (Life Technologies), and the chemokines RANTES, SDF-1, I-309, fractalkine, MIP-1␣, and KC (Peprotech, Rocky Hill, NJ). The particular chemokines tested were chosen because of their specificity for phagocyte targets. All chemokines were human with the exception of KC, which is mouse. The receptor targets for these chemokines are as follows: RANTES, CCR1, CCR3 and CCR5; SDF-1, CXCR4; I-309, CCR8; fractalkine, CX3CR1; MIP-1␣, CCR1, and CCR5; CXCR2. A␤, chemokines and ATP were dissolved in water and stored at Ϫ20°C; fMLF was dissolved in Me 2 SO and stored at Ϫ20°C. In some experiments, the cells were incubated in 250 ng/ml pertussis toxin (PTX; Calbiochem) for 4 h at 37°C in medium, harvested, and loaded with Fura-2/AM as described above. Immediately after harvesting, murine neutrophils were incubated in 1-2 ϫ 10 6 /ml of phosphate-buffered saline containing 2.5 M Fura-2/AM for 45 min at 37°C. Neutrophils were washed twice in HBSS and suspended to 1-2 ϫ 10 6 /ml for analysis. Calcium flux was performed with N9 cells preincubated in the presence or absence of 300 ng/ml lipopolysaccharide (LPS) (37°C, 24 h) using similar procedures.
Chemotaxis-HEK 293 cells were harvested from tissue culture flasks by incubation in trypsin (0.05%)/EDTA (0.1%) (Quality Biologicals, Inc., Gaithersburg, MD) for 5 min at 37°C. Cells were suspended evenly by vigorous pipetting, and excess Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum was then added to block trypsin. Cells were washed twice in Dulbecco's modified Eagle's medium and suspended to a concentration of 4 ϫ 10 6 cells/ml in chemotaxis medium (RPMI 1640; 20 mM HEPES (Life Technologies) and 1% bovine serum albumin (ICN Biomedicals Inc., Aurora, OH)). Chemoattractants, diluted in chemotaxis medium, were added to the bottom wells of a 96-well chemotaxis plate (Neuro Probe, Inc., Gaithersburg, MD). A 12-m pore size membrane was placed on top, and 25 l of cell suspension containing ϳ100,000 cells was placed in the upper chamber. Cells were incubated for 5 h at 37°C, 100% humidity, 5% CO 2 . The membrane was carefully removed, and cells in the bottom well were counted using a hemacytometer. Methods for murine neutrophils were the same except that ϳ200,000 cells were added to the top of a 5-m pore size membrane. Chemotaxis assays for N9 cells incubated with or without LPS (300 ng/ml, at 37°C for 24 h) were performed with 48-well chemotaxis chambers (Neuro Probe). Polycarbonate filters with 8-m pore size and 90-min incubation at 37°C were used for measurement of microglial cell migration.
Superoxide Production-Mouse neutrophils were suspended in HBSS containing Ca 2ϩ and Mg 2ϩ at 10 6 cells/ml. 50 l (5 ϫ 10 4 cells) were distributed into wells of a 96-well microtiter chemiluminescence plate and incubated at 37°C for 5 min. Then a mixture of the superoxide-specific chemiluminescence indicator reagent Diogenes (National Diagnostics, Atlanta, GA) was added to the cells (50% of total reaction volume) with appropriate stimuli or vehicle control, and superoxide dismutase-inhibitable chemiluminescence was measured in a luminometer (Labsystems Luminoskan; Helsinki, Finland). Data are expressed as integrated luminescence (relative light units) observed during 0.5-s readings obtained at 12-s intervals over a time course of 10 min. For sequential stimulation experiments, 5 ϫ 10 4 FPR Ϫ/Ϫ neutrophils were distributed into microcentrifuge tubes, and test substances were added. The mixture was then immediately transferred to a chemiluminescence plate. After incubation at 37°C for 8 min (when fMLF was the first stimulus) or 9 min (when A␤ was the first stimulus), Diogenes reagent plus the final stimulus was added, and the activity was monitored for 10 min. To control for desensitization of NADPH oxidase by the first stimulus, cells were stimulated with PMA (100 ng/ml) after the second stimulation, and superoxide was measured for 10 min. To control for scavenging of superoxide by fMLF or A␤, neutrophils were stimulated simultaneously with (i) PMA (100 ng/ml) and Me 2 SO (vehicle for fMLF; 0.2% of the volume in which cells were stimulated); (ii) PMA (100 ng/ml) and fMLF (50 M); (iii) PMA (100 ng/ml); or (iv) PMA (100 ng/ml) plus A␤ (10 M). Each condition was tested in triplicate, and the mean of the mean number (grand mean) of superoxide dismutase-inhibitable relative light units throughout the duration of the assay and the corresponding standard errors of grand means were calculated. Differences between conditions were tested for significance by two-tailed paired t tests or unequal variance tests (Mann-Whitney rank sum) where appropriate. A value of p Ͻ 0.05 indicated significant differences.
RNA Analysis by PCR-Wild type littermates of an F7 backcross of 129/sV FPR Ϫ/Ϫ mice with C57Bl/6 mice were euthanized by cervical dislocation, and brains from three mice were removed, pooled, and washed in phosphate-buffered saline for 15 min at room temperature. Brain tissue was sliced in a Petri dish on ice using a clean razor blade and homogenized in an ice-cold Teflon homogenizer. RNA was extracted using the RNA STAT-60 kit (Tel-Test, Inc., Friendswood, TX) according to the manufacturer's instructions. RNA was reverse-transcribed using the cDNA Cycle Kit (Invitrogen, San Diego, CA) following the manufacturer's instructions. Gene-specific primers were used for PCR amplification of the cDNA using the GeneAmp PCR System 9700 (PerkinElmer Life Sciences). For mouse FPR2, the 5Ј primer 5Ј-TCTAC-CATCTCCAGAGTTCTGTGG and 3Ј primer 5Ј-TTACATCTACCA-CAATGTGAACTA were used to generate a 268-base pair product. The PCR conditions for amplification were 3 min at 95°C for the initial melting followed by 30 cycles of 1 min of melting at 95°C, 1 min of annealing at 55°C, 2 min of synthesis at 72°C, with a final extension of 10 min at 72°C and cooling to 4°C. PCR products were analyzed by gel electrophoresis using a 1% agarose gel in TBE containing 10 g of ethidium bromide/100 ml. Data were recorded on a UVP Gel Imaging System (Appropriate Technical Resource, Laurel, MD). For analysis of the N9 microglial cell line, RT-PCR was performed with 0.5 g of total RNA extracted from cells treated with 300 ng/ml LPS for different time periods (High Fidelity ProSTARTM HF System, Stratagene, Kingsport, TN). The procedure consisted of a 15-min reverse transcription at 37°C, 1-min inactivation of Moloney murine leukemia virus reverse transcriptase at 95°C, and 40 cycles of denaturing at 95°C (30 s), annealing at 55°C (30 s), and extension at 72°C (1 min), with a final extension for 10 min at 72°C. Primers for murine ␤-actin gene were used as controls (Stratagene). The RT-PCR products at different dilutions were electrophoresed on 1% agarose gel and visualized with ethidium bromide staining.

RESULTS
Mouse FPR2 and Its Human Counterpart FPRL1R Are Receptors for Amyloid ␤-Using induction of calcium flux as a highly sensitive and specific real time assay of receptor activation, we screened a panel of stable cell lines transfected with plasmids encoding the known phagocyte formylpeptide receptors (human and mouse FPR, human FPRL1R, and mouse FPR2), four chemokine receptors (human CCR1, CCR5, CCR8, and CX3CR1), the mouse lipoxin A4 receptor (encoded by mouse Fpr-rs1), and an orphan receptor highly related in sequence to formylpeptide receptors (Fpr-rs3), as well as untransfected control cells, for responsiveness to 10 M A␤ (Fig. 1). This concentration was chosen based on A␤ dose-response studies published previously for human neutrophils and monocytes and rat microglial cells (13). The lipoxin A4 receptor and Fpr-rs3 were included because of their high sequence similarity to the formylpeptide receptors (37).
A␤ induced a response in HEK 293 cells expressing FPRL1R and FPR2, which are human and mouse low affinity formylpeptide receptors, respectively. Activation of each receptor produced a robust transient that was similar in magnitude and duration to the response induced by the prototypical N-formylpeptide fMLF in the same cells (Fig. 2, A and B) and was similar kinetically to the transients induced by other classic chemoattractants and chemokines (Fig. 1). A␤ was specific for these receptors, since none of the other cell lines tested responded. The CCR1, CCR5, CCR8, and CX3CR1 and the human and mouse FPR (high affinity formylpeptide receptor) cell lines did respond to appropriate known agonists as previously described (34 -36). The Fpr-rs1 and Fpr-rs3 cell lines were unresponsive to fMLF but did respond to ATP through an endogenous signaling pathway. Although RNA for Fpr-rs1 and Fpr-rs3 is present in these two cell lines, we have not yet obtained direct evidence of receptor protein expression.
A␤ signaling could be completely blocked by pretreatment of the cells with pertussis toxin (Fig. 1, column 1, tracing labeled FPR2 ϩ PTX), which inactivates G i type G proteins. Pertussis toxin also blocks signaling by other FPR2 agonists (34,42,43). When FPR2-and FPRL1R-expressing cells were sequentially stimulated with 10 M A␤, they responded to the first but not the second stimulus (Fig. 1, column 1, tracing labeled FPR2, and data not shown) indicating homologous desensitization of the signal transduction pathway, which is characteristic of G protein-coupled receptors (44). Moreover, A␤ and fMLF reciprocally interfered with each other's signaling at FPR2 (Fig. 2, A  and B) in a concentration-dependent manner, providing further evidence that both agonists act at the same receptor. This was specific, since A␤ did not affect signaling by agonists acting at any of the other receptors considered ( Fig. 1 and data not  shown).
A␤ induced calcium flux in both FPR2-and FPRL1R-transfected HEK 293 cells in a graded concentration-dependent manner, with an EC 50 of 5 M (Fig. 3A). In contrast, HEK 293 cells expressing either mouse or human FPR did not respond to A␤ from 0.5 to 20 M (Fig. 3A). However, all four cell lines responded to fMLF in a concentration-dependent manner, with EC 50 consistent with those previously reported (data not shown; Ref. 34).
To test whether native FPR2 also functions as an A␤ receptor, we first focused on primary mouse neutrophils, which, as we have previously shown, express FPR2 endogenously (34) and which can be analyzed in an FPR-deficient background due to the availability of FPR knockout mice (41). A␤ induced calcium flux in FPR Ϫ/Ϫ neutrophils with an EC 50 of 1 M, similar to the value for FPR2-transfected HEK 293 cells (Fig. 3,  A and B). FPR Ϫ/Ϫ neutrophils also mimicked FPR2-transfected HEK 293 cells in sequential stimulation experiments; fMLF and A␤ were able to reciprocally cross-desensitize each other (Fig. 4, A and B). Specificity was again confirmed by the lack of cross-desensitization in this assay between A␤ and either SDF-1, MIP-1␣, or KC in mouse neutrophils (Fig. 4C). It is important to note that FPR and FPR2 both mediate fMLF signaling in mouse neutrophils (34,41). However, the desensitization experiments were carried out using neutrophils from FPR knockout mice, which rules out cross-desensitization of A␤ action by fMLF signaling through FPR and strongly implicates A␤ usage of endogenous neutrophil FPR2, the only other known neutrophil fMLF receptor. As with FPR2-transfected HEK 293 cells, A␤ induction of calcium flux in mouse neutrophils was completely blocked by pretreatment of the cells with pertussis toxin, indicating a G i -dependent signaling pathway (Fig. 4C). A␤ potency was indistinguishable in neutrophils from FPR Ϫ/Ϫ and ϩ/ϩ mice (Fig. 3B). Although there was a trend toward lower efficacy (maximal response) in cells from FPR Ϫ/Ϫ mice, this difference was not statistically significant (Fig. 3B).
A␤ Is a Chemotactic Agonist at FPR2-To assess the potential biological significance of A␤-FPR2 signaling, we used in vitro chemotaxis assays as a model of cell migration. Consistent with the calcium flux results, A␤ induced chemotaxis of FPR2transfected HEK 293 cells but not mouse FPR-transfected HEK 293 cells; likewise, A␤ induced migration of mouse neutrophils (Fig. 5). In each case, the peak responses occurred at ϳ10 M, and the EC 50 values were consistent with the values for induction of calcium flux in these cells, 5 M (Fig. 5, B and C). We have previously shown that the fMLF dose-response curve for chemotaxis in neutrophils from wild type mice has two peaks, one with an optimum at ϳ500 nM and the other with an optimum at 10 M. The 500 nM optimum is due to FPR activity, since it is absent in cells from FPR Ϫ/Ϫ mice (34). The second peak is consistent with FPR2 pharmacology in transfected HEK 293 cells. Since the dose-response curve for A␤ chemotaxis is the same in neutrophils from FPR Ϫ/Ϫ and ϩ/ϩ mice, A␤ chemoattraction of mouse neutrophils is not mediated by FPR. Since in FPR Ϫ/Ϫ neutrophils the A␤ and fMLF chemotactic and calcium flux optima are similar and match the A␤ optimum in FPR2-transfected HEK 293 cells, A␤ chemoattraction of these cells is most likely mediated by FPR2. Since application of A␤ on both sides of the chemotaxis filter gave net results equivalent to the background control, we conclude that A␤-induced cell migration was due to chemotaxis, not chemokinesis (Fig. 5A).
Evidence That FPR2 Mediates Induction of Superoxide Generation by Amyloid ␤-To test whether FPR2 can also mediate production of reactive oxygen species by A␤, we examined whether A␤ could induce superoxide production in mouse neutrophils and, if so, whether this activity could be desensitized by prestimulation with fMLF. Again, FPR Ϫ/Ϫ neutrophils were used to eliminate the possibility of cross-desensitization of A␤ activity by fMLF signaling through FPR. As shown in Fig.  6A, A␤ at 10 M, a concentration that saturated the chemotactic and calcium flux response in mouse neutrophils and FPR2/ HEK 293 cell transfectants, induced superoxide production with similar efficacy in FPR Ϫ/Ϫ and FPR ϩ/ϩ neutrophils. This is consistent with the calcium flux and chemotaxis results. Additional experiments (n ϭ 2) showed a similar graded A␤ dose-response relationship and equivalent potency for FPR Ϫ/Ϫ versus FPR ϩ/ϩ neutrophils (data not shown). This is consistent with the chemotaxis and calcium flux results and indicates that A␤ induction of superoxide generation is not mediated by FPR. fMLF also induced superoxide generation in both FPR ϩ/ϩ and Ϫ/Ϫ neutrophils; however, the EC 50 was 10-fold lower at FPR Ϫ/Ϫ neutrophils, which is consistent with our previous report of weaker potency of fMLF at FPR2 versus FPR for induction of both calcium flux and chemotaxis in both neutrophils and receptor-transfected cells (34). 2 The superoxide response of FPR Ϫ/Ϫ neutrophils to 10 M A␤ was markedly attenuated when the cells were pretreated with 5 M fMLF compared with pretreatment with vehicle alone (Fig. 6B). Likewise, the response to 5 M fMLF was markedly attenuated when the cells were pretreated with 10 M A␤ (Fig.  6C). The reduced response is not due to depletion or inactivation of NADPH oxidase by the first stimulation, because PMA could induce large amounts of superoxide production in cells when added after completion of the response to the second stimulus (data not shown). Moreover, costimulation experiments in which PMA was added simultaneously with fMLF or A␤ ruled out scavenging as the mechanism by which each agent reduced superoxide production by the other (data not shown).
FPR2 Is Expressed in Mouse Brain-Previously, we reported that by Northern blot analysis FPR2 mRNA was detectable in mouse spleen, lung, and liver but not brain (34). Because of the importance of A␤ to the pathogenesis of Alzheimer's disease and our finding that it is an agonist at FPR2, we reexamined brain expression of FPR2 by RT-PCR (Fig. 7) and were able to detect a relatively weak band of the appropriate size, 268 base pairs.
FPR2 Expression in a Mouse Microglial Cell Line-We next tested whether microglial cells, the major phagocytic cells of the central nervous system, expressed FPR2. For this purpose, we used the murine microglial cell line N9, which expresses typical markers of resting mouse microglia and has been extensively used as a representative of primary mouse microglial cells (40). Low levels of FPR2 mRNA could be detected in this cell line under resting conditions using RT-PCR; however, the cells did not respond to A␤ either in calcium flux or chemotaxis assays (Fig. 8). Cell activation with LPS induced FPR2 mRNA expression in a time-dependent fashion (Fig. 8A) and rendered the cells responsive to A␤ in a concentration-dependent manner in both calcium flux and chemotaxis assays (Fig. 8, B and C). The potency of A␤ was consistent for both functions and was consistent with the values obtained in studies of mouse neutrophils and FPR2-HEK 293 cells. As we observed with mouse neutrophils and FPR2-transfected HEK 293 cells, fMLF and A␤ were able to reciprocally crossdesensitize each other in sequential stimulation experiments using calcium flux as the functional readout (Fig. 8D). Finally, chemotaxis of LPS-activated N9 cells to A␤ was completely blocked by pretreatment of the cells with pertussis toxin (Fig. 8E), demonstrating a G i -dependent signaling pathway. This is consistent with the results obtained using the calcium flux assay in neutrophils and FPR2-transfected HEK 293 cells (Figs. 1 and 4). DISCUSSION These data directly demonstrate that A␤, the major component of senile plaque in Alzheimer's disease, can function as a specific chemotactic agonist at the mouse N-formylpeptide receptor subtype FPR2 expressed in HEK 293 cells. This receptor is constitutively expressed in neutrophils, lung, liver, spleen, and brain and at low levels in macrophages and the mouse microglial cell line N9; FPR2 expression in N9 cells can be induced by LPS. Furthermore, we provide a series of observations that together strongly suggest that A␤ activates mouse phagocytic cells via endogenous FPR2. First, the prototypical FPR2 agonist fMLF can desensitize A␤ action (calcium flux, chemotaxis, and superoxide production) in mouse neutrophils, including neutrophils from mice lacking the other known fMLF receptor subtype FPR. Second, LPS induces responsiveness of the mouse microglial cell line N9 to A␤ in both calcium flux and chemotaxis assays, and fMLF and A␤ can reciprocally crossdesensitize each other in activation of calcium flux in these cells. LPS induction of this activity correlates with the induction of FPR2 mRNA expression. Third, A␤ activates mouse phagocytes and FPR2-transfected HEK 293 cells with similar potency. We have also demonstrated that A␤ can activate the human counterpart of FPR2, FPRL1R, in transfected HEK 293 cells. This receptor is expressed constitutively on human neutrophils and monocytes. Recently, we have reported that human monocyte and FPRL1R-HEK 293 cell chemotactic responses induced by fMLF and A␤ can be reciprocally crossdesensitized, consistent with our observations in mouse cells (45).
Our results suggest a unified molecular mechanism involving FPR2 for A␤ activation of chemotaxis and oxidative stress in phagocytic cells. This signaling system could support beneficial functions such as host defense and tissue repair in settings where A␤ is produced homeostatically (46). In Alzheimer's disease, where A␤ accumulates pathologically, our results suggest a novel mechanism to explain why activated microglial cells accumulate at senile plaques. Our gene expression data suggest that involvement of the receptor in this process in Alzheimer's disease would require cell activation, since FPR2 mRNA is expressed at very low levels in brain and N9 microglial cells. Since LPS can induce FPR2 gene expression and responsiveness to A␤ by this cell line, then local production of LPS-sensitive cytokines such as tumor necrosis factor and interleukin-1␤ in the brains of patients with Alzheimer's disease could provide such a stimulus in vivo for receptor expression. In this regard, it is important to note that A␤ has been shown to induce interleukin-1␤ production in THP-1 cells (14).
FPR2 was originally identified as the product of a mouse gene that cross-hybridized with FPR1, a human gene for the high affinity N-formylpeptide receptor FPR (37,47). FPR1 cross-hybridizes with two other human genes: FPRL1, which encodes the low affinity N-formylpeptide receptor FPRL1R, and FPRL2, which encodes an orphan receptor (48 -50). These genes, which are located in a cluster on human chromosome 19q13.3 (51), are expressed in phagocytes (52), and their protein products have 56 -69% amino acid identity. FPRL1R has multiple and structurally diverse agonists, which, in addition to A␤ and fMLF (50,53), include human immunodeficiency virus-1 envelope-derived peptides (54) and two endogenous mediators, serum amyloid A (55) and the eicosanoid lipoxin A4 (56). This receptor has also been called LXA4R, for lipoxin A4 receptor (56). In mouse, six FPR-like genes have been cloned (37), and agonists have been identified for three of them: Fpr1, which encodes the mouse FPR orthologue (53), and Fpr-rs1 and Fpr-rs2, both of which encode proteins with ϳ65% amino acid identity to FPRL1R (34,37,57). To date, Fpr-rs1 has been reported to encode a receptor named LXA4R which is specific for lipoxin A4 (57), whereas Fpr-rs2 encodes FPR2, which is specific for the other FPRL1R agonists including A␤ (34,42,43). The specificity of FPR2 for lipoxin A4 has not been clearly resolved. Thus, FPRL1R functions may be split between two mouse receptors, FPR2 and LXA4R.
Although several other A␤ receptors have been reported previously, FPR2 and FPRL1R are the only ones that are GPCRs. Consistent with this, most known leukocyte chemoattractants act via GPCRs (32) and, like FPR2, activate calcium flux. Moreover, FPR2 was already known to function as a chemotactic receptor for serum amyloid A, human immunodeficiency virus-derived peptide T21, and fMLF (34,42,43). A␤ had previously been shown to induce calcium flux and inter-leukin-1␤ production in a pertussis toxin-sensitive manner (14,33), suggesting involvement of a G i -coupled receptor, and FPR2 is such a receptor. RAGE, the receptor for advanced glycation end products and a member of the immunoglobulin gene superfamily, is also an A␤ receptor and has been reported to mediate A␤-induced oxidant stress in endothelial cells and cortical neurons, NF-B activation in endothelial cells, and induction of tumor necrosis factor-␣ production, chemotaxis, and haptotaxis of the mouse microglial cell line BV-2 (21,23). Although RAGE is expressed by primary microglial cells and is found at higher levels in brains from patients with Alzheimer's disease versus unaffected controls (21), its role in mediating A␤ functional responses in primary microglial cells and other mononuclear phagocytes has not been defined. A␤ has been reported to induce chemotaxis of rodent microglial cells with an EC 50 in the low nanomolar range, which is consistent with the reported binding affinity of A␤ for RAGE (21). Yet A␤ functions reported by most other studies have occurred with half-maxi- mal potency in the low micromolar range. These include A␤ action on mouse neutrophils and N9 microglial cells (present study), A␤ binding to human macrophages (58), A␤-induced production of tumor necrosis factor by THP-1 (59) and N9 (10) cells, and A␤ induction of reactive oxygen species by human neutrophils (13), human macrophages (13), rat macrophages (60), and rat microglial cells (13). The explanation for this discrepancy is not clear, but it does not appear to simply involve differences in potency of A␤ for activating different cell types or for inducing different cell functions. One possibility is that in primary cells A␤ acts via two or more receptors that differ in affinity for ligand, a paradigm that has many examples, including the fMLF receptors FPR and FPR2 (34).
In summary, we have identified a novel A␤ receptor, FPR2 in mice and FPRL1R in humans, and have presented evidence that FPR2 mediates A␤ action on mouse neutrophils and the LPS-activated mouse microglial cell line N9. In resting N9 cells and in mouse brain, FPR2 mRNA is expressed at low levels. Our combined genetic and pharmacologic analysis of mouse neutrophils strongly suggests that A␤ can induce both chemotaxis and oxidant production via the same receptor, FPR2, which suggests a unified molecular basis for microglial cell recruitment to senile plaques and induction of oxidant stress in microglial cells in Alzheimer's disease. Consistent with this, we have found that FPRL1R RNA is expressed by inflammatory cells infiltrating senile plaques in brain tissues from patients with Alzheimer's disease (45). The role of this receptor in Alzheimer's disease may now be tested further by developing an FPR2 knockout mouse backcrossed onto the amyloid precursor protein transgenic mouse model of Alzheimer's disease, which exhibits the inflammatory aspect of the human disease (61,62).