Identification of a High Affinity FcγRIIA-binding Peptide That Distinguishes FcγRIIA from FcγRIIB and Exploits FcγRIIA-mediated Phagocytosis and Degradation*

FcγRIIA is a key activating receptor linking immune complex formation with cellular effector functions. FcγRIIA has 93% identity with an inhibitory receptor, FcγRIIB, which negatively regulates FcγRIIA. FcγRIIA is important in the therapeutic action of several monoclonal antibodies. Binding molecules that discriminate FcγRIIA from FcγRIIB may optimize receptor activity and serve as a lead for development of therapeutics with FcγRIIA as a key target. Here we report the use of phage display libraries to select short peptides with distinct FcγRIIA binding properties. An 11-mer peptide (WAWVWLTETAV) was characterized that bound FcγRIIA with a Kd of 500 nm. It mediated cell internalization and degradation of a model antigen. The peptide-binding site on FcγRIIA was shown to involve Phe163 and the IgG binding amino acids Trp90 and Trp113. It is thus overlapping but not identical to that of IgG. Neither activating receptors FcγRI and FcγRIII, nor FcγRIIB, all of which lack Phe163, bound the peptide.

Leukocyte IgG receptors (Fc␥Rs) play a crucial role in immune protection by providing a link between antibody-antigen complexes and cellular effector functions. Two general classes of Fc␥Rs are recognized: in humans they are the activating Fc␥Rs I, IIA, and IIIA and the inhibitory Fc␥R IIB. The activating Fc␥Rs are characterized by an intracellular tyrosinebased activation motif (ITAM), 4 which triggers an activating signaling cascade leading to phagocytosis, endocytosis, antibody-dependent cell cytotoxicity, and release of inflammatory mediators (1)(2)(3). By contrast, the inhibitory Fc␥RIIB contains an intracellular tyrosine-based inhibitory motif. Co-ligation of Fc␥RIIB with ITAM-containing receptors results in inhibition of ITAM-mediated functions (4 -6). The balance of activation and inhibition through Fc␥Rs is important for the regulation of immune function, setting thresholds for and ultimately determining the magnitude of the response.
Fc␥RIIA is the most widely expressed Fc␥R that is present on subgroups of leukocytes such as neutrophiles and mononuclear phagocytes, where the receptor exists in two common allelic forms at amino acid 134. The Fc␥RIIA-H134 allelotype (histidine) has higher binding efficiency for human IgG2 and IgG3 antibodies when compared with Fc␥RIIA-R131 (arginine) (7). The inhibitory Fc␥RIIB is expressed on the same cell types as Fc␥RIIA. In addition, Fc␥RIIB is expressed by B-cells, where this receptor is the only Fc receptor expressed.
Several studies have recognized that Fc␥RIIA is of particular importance in the anti-tumor activity of therapeutic monoclonal antibodies. Fc␥RIIA on leukocytes from patients undergoing granulocyte colony-stimulating factor treatment known to up-regulate Fc␥RIIA expression was shown to be the major trigger molecule for antibody-dependent cell cytotoxicity induced by an anti-HER-2/neu-specific IgG when various breast cancer cell lines were target cells in vitro (8). Furthermore, a study by Weng and Levy (9) showed a positive association of response rates in rituximab-treated non-Hodgkin's lymphoma patients with Fc␥RIIA-H134. Similarly, Zhang et al. (10) observed that metastatic colorectal cancer patients treated with an anti-epidermal growth factor receptor monoclonal antibody showed an Fc␥RIIA-R134-dependent progression-free survival. Taken together, these studies clearly demonstrate that Fc␥RIIA-dependent anti-tumor effects mediated by mononuclear phagocytes and/or neutrophils have considerable impact. Furthermore, Fc␥RIIA has a predominant role in infectious diseases (11,12), and enhanced phagocytosis by this receptor may have a broad application in antibacterial therapy.
Although all Fc␥Rs bind IgG immune complexes, individual Fc␥Rs bind with affinities that vary depending on the IgG subclass involved (13). A high activating/inhibitory Fc␥R binding ratio was found to correlate with biological activity in models of tumor clearance and platelet depletion (14). In accordance with this, blocking Fc␥RIIB in mice resulted in enhanced tumor immunity (15). Whereas Fc␥RIIIA-expressing NK cells implicated in antitumor antibody-dependent cell cytotoxicity (16) have been shown to infiltrate solid tumors poorly (17), the Fc␥RIIA/IIB-expressing cells (macrophages, DCs, neutrophils) do this very efficiently (18). Therefore, in such cases, selective engagement of Fc␥RIIA might be very efficacious, also because of the essential role of Fc␥RIIA in promoting uptake and presentation of antigens to both CD4 ϩ (19) and CD8 ϩ T-cells (20).
Because the extracellular domains of Fc␥RIIA and Fc␥RIIB are closely related in structure, having 93% amino acid identity in their extracellular domains, it has been challenging to design modified IgGs with mutations in the Fc region that distinguish between the two receptors (21). Strategies have been utilized to identify smaller molecules that target Fc␥ receptors, including design of IgG-derived peptide sequences that mimic the part of IgG-Fc that binds Fc␥ receptors (22)(23)(24)(25), but still neither of these discriminates between the two receptors.
In the present study we performed selection from randomized phage display libraries to identify short novel peptides with distinct Fc␥RIIA binding properties. We identified, compared, and characterized several binders, and a phage displaying an 11-mer peptide (NNK11-C1) was characterized as the best binder. Binding of the NNK11-C1 phage was competed with free synthetic peptide with the same amino acid sequence, demonstrating that the peptide alone was sufficient for binding. The free peptide bound similarly to both allelic variants (Arg 134 and His 134 ) with a K d of 500 nM, whereas only very weak binding to Fc␥RIIB and no binding to Fc␥RI and Fc␥RIIIB was observed. The peptide was further shown to interfere with IgG binding. Comparing a panel of Fc␥RIIA and Fc␥RIIB mutants for their ability to bind the peptide, this was found to depend on the presence of Trp 90 and Trp 113 in addition to the proximal Phe 163 in Fc␥RIIA. Notably, whereas Fc␥RI, Fc␥RIIB, and Fc␥RIII harbor both Trp 90 and Trp 113 , amino acid 163 is a valine in Fc␥RIII and a tyrosine in Fc␥RI and Fc␥RIIB. Thus, the presence of an aromatic, hydrophobic amino acid in position 163 was found to be the key to the actual binding specificity. Biotinylated peptide complexed on streptavidin-bound Fc␥RIIA on PMN and monocytes and mediated internalization and degradation of streptavidin coupled to 1 M magnetic beads. This peptide may be an interesting candidate for development of therapeutics for optimal engagement of the immune system.

EXPERIMENTAL PROCEDURES
Cells and Antibodies-K562 (CCL-243) and 293E (CRL-10852) cells were purchased from the American Type Culture Collection (LGC Promochem, UK) and cultured as described (26). 293F cells were obtained from Invitrogen and cultured as described by the manufacturer. Leukocytes were drawn from peripheral blood from normal human volunteers as detailed elsewhere (27). Chimeric human IgG3 (chIgG3) was isolated from J558L cell lines as previously described (28). Biotinylated human IgG (hIgG-biot) was obtained by biotin labeling of normal pooled IgG (Tetagam, Aventis Behring, PA). Heat aggregation of IgG was done by incubation at 63°C for 10 min.  The extracellular domains of Fc␥RIIA-R134,  Fc␥RIIA-H134, and Fc␥RIIB as well as the mutants IIA W90A,  IIA W113A, IIA F163Y, IIB K130Q, IIB S135L, IIB N138T, IIB  Y163F, and IIB K130Q/S135L/N138T/Y163F (IIB 4mut) were cloned and expressed as soluble fusions to GST in 293E cells as described in the supplemental text. The IIA mutants were all in Arg 134 . In addition, Fc␥RIIA-R134 and Fc␥RIIB were cloned and expressed as fusions to a His 6 tag (Supplementary methods).
Phage Display Peptide Libraries and Phage Clones-The vector fUSE5 (30), which supports phage fd protein III (pIII) peptide expression, was used for library constructions. Two such libraries, Cys6 and Cys9, that represent 2.2 ϫ 10 7 and 5 ϫ 10 7 different cysteine-flanked peptides of six or nine random amino acids, respectively, have been described (31). A library of 11 random amino acids (NNK11) and three motif libraries, Evo1, Evo2, and Evo3, were similarly constructed using the primer NNK 11 for the NNK11 library and the primers Evo1, Evo2, and Evo3 (supplemental Table S1) (all from DNA Technology, Risskov, Denmark) for the three evolution libraries. The number of different clones in each library was estimated based on the number of primary transformants with insert. Control phages were randomly chosen from the Cys6 and Cys9 libraries. The Cys6 phage clone C6-1, with peptide insert CLRSGLGC, has previously been described and is selected for binding to Fc␥RI (26).
Selection of Fc␥RIIA-binding Peptides-Three consecutive rounds of selection from the Cys6, NNK11, Cys9, as well the three evolution libraries were performed. For the first round, wells precoated with 100 l of goat anti-GST polyclonal antibody (Amersham Biosciences) were blocked for 1 h at room temperature with 1% PBS/skm. Aliquots of 100 l of Fc␥RIIA-R134-GST (10 g/ml) were added and incubated for 1 h at room temperature, followed by three washes with PBS/T. A total of 2 ϫ 10 11 Escherichia coli K91K transducing units (TUs) from the libraries diluted in 400 l of PBS/skm were added to a total of four wells. After incubation for 1 h at room temperature followed by 10 washes with PBS/T, bound phages were eluted with 400 l of 0.1 M HCl-glycine (pH 2.2), and eluates were neutralized with 28 l of 1.5 M Tris (pH 8.8). Eluted phages were rescued as E. coli K91K transfectants, titrated, and amplified as described (30). Phage supernatant of the amplified eluate (E1A) was prepared essentially as described (30). For the second round of selection, two wells of MaxiSorp strips (Nunc) were coated overnight at 4°C with 100 l of Fc␥RIIA-R134-GST (10 g/ml), blocked for 1 h at room temperature with 1% PBS/skm, and washed three times with PBS/T. A portion of E1A corresponding to 2.5 ϫ 10 9 TU in PBS/skm was added to the wells.
After incubation for 1 h at room temperature followed by 10 washes with PBS/T, bound phages were eluted, neutralized, rescued, and amplified as before. A third round of selection was performed with an input of 5 ϫ 10 8 TU of the amplified second eluate (E2A) followed by the same procedure as described for the second round. Phage supernatants from single TUs were prepared as described (30).
Characterization of Phages-Aliquots of 5 or 50 l of phage supernatants from amplified eluates or individual clones in a total of 100 or 200 l 1% PBS/skm were allowed to react with Fc␥RII-GST fusion proteins, GST or PBS/skm coated in microtiter wells for 1 h at room temperature. The phage recovery was determined as described above. The amino acid sequences of peptides displayed by selected phages were determined by DNA sequencing (GATC Biotech, Konstanz, Germany) of PCR products as previously described (31).
Synthetic Peptides-Synthetic peptides were in the form ADGAX n GAAK-Bio (Alta Bioscience), where ADGA and GAA are flanking amino acids as found in the fUSE5 phage, K-Bio represents a biotinylated lysine residue, and X n represents selected peptides CPWFQWPC (C6-D), WAWVWLTETAV (NNK11-C1), and CTLRLGVGVRC (C9-E11) as well as a control peptide CWTSGARWRLC (RB-14). Peptides flanked by cysteines (C6-D, C9-E11, and RB-14) were produced as cyclic peptides. The RB-14 sequence has previously been selected for binding to poly(Ig) receptor (32) and was used as negative control. The synthetic peptides were purified by high pressure liquid chromatography to more than 80% purity (Alta Bioscience). The peptides were dissolved to 10 mM in Me 2 SO (Sigma) and kept at Ϫ20°C.
Competition Assays with a Free Synthetic NNK11-C1 Peptide-Portions of 40 M or 400 M synthetic NNK11-C1 peptide were added to aliquots of 10 l of supernatant of the NNK11-C1, C6-D phage, or a control phage displaying the peptide CDIF-GRDC and incubated in wells coated with 5 g/ml Fc␥RIIA-R134-GST at 1 h at room temperature. The phage recovery was determined as described above.
ELISA Assays with Biotinylated Peptides-ELISA assays with biotinylated peptides were performed essentially as the ELISA for detection of bound hIgG to Fc␥RIIs as described (26). Briefly, synthetic biotinylated peptides in PBS were added in wells coated with Fc␥RII-GSTs and bound peptide detected by streptavidin conjugated to alkaline phosphatase.
Surface Plasmon Resonance Analyses-The instrument used for the SPR analysis was BIAcore 3000 (BIAcore AB, Uppsala, Sweden). The running buffer for all of the experiments was BIA-certified HBS buffer (10 mM HEPES, pH 7.4, 150 mM NaCl, 3.4 mM EDTA, 0.005% Surfactant P20). chIgG3 was covalently immobilized to ϳ1000 RU on a CM5 chip (BIAcore AB) with the amine coupling procedure. Serial dilutions (0.078 -4 M) of Fc␥RIIA-R134-His and Fc␥RIIB-His were injected over the sensor chip at a flow rate of 10 l/min, and the binding reactions were allowed to reach (near) equilibrium. K D was derived by nonlinear regression analysis of plots of R eq (the equilibrium binding response) versus the analyte (chIgG3) concentration. The NNK11-C1-biotin peptide was captured on an SA chip (BIAcore AB) as described by the manufacturer, to ϳ50 -100 RU followed by injection of recombinant human Fc␥RI (2 M), Fc␥RII Sequence Analysis-ClustalW was used for sequence alignment of the extracellular domains of human Fc␥RIIA and Fc␥RIIB. The NCBI nucleotide accession numbers are NM_021642 (human Fc␥RIIA) and NM_004001 (human Fc␥RIIB). The stereo ribbon representation of the Fc␥RIIA structure was designed using MOLMOL with the crystallographic data of the Fc␥RIIA (33) available on the Protein Data Bank site. The structure of Fc␥RIIA (33) was superimposed onto the structure of Fc␥RIIB (34).
Circular Dichroism-The CD sample was dissolved in water. Trifluoroethanol was titrated to the water solution of the peptide at concentrations from 0 to 50%, and the CD data were recorded. The concentration of peptide was determined by absorption measurements at 280 nm using a Shimadzu UVinstrument (Shimadzu Corperation, Kyoto, Japan) before and after CD measurements. CD spectra were recorded using a Jasco J-810 spectropolarimeter (Jasco International Co., Ltd., Tokyo, Japan) calibrated with D-camphor-10-sulfonate (Icatayama Chemical, Tokyo, Japan). All of the measurements were taken up using a quartz cuvette (Starna, Essex, UK) with a path length of 0.1 cm. The samples were scanned five times at 50 nm/min with a bandwidth of 0.5 nm and a response time of 1 s, over the wavelength range 190 -260 nm. The data were averaged, and the spectrum of a sample-free control was subtracted. The ␣-helical content of the peptide was determined by application of the single point method using the mean residual ellipticity at 222 nm ([] 222 ) and the equation: f H ϭ 100% ϫ [] 222 /(ϫ40,000(1-2.5n)), where f H is the ␣-helical content in %, and n is the number of residues (35). Two repetitions of each measurement were done.
Flow Cytometry Analysis-The PMN and monocyte populations of prepared leukocytes were identified in the scatter diagram in the flow cytometer and also verified by staining with mouse IgM anti-Fc␥RIIIB-FITC (Immunotech, Quebec, Canada) as a specific marker for human PMN, and mouse IgM anti-CD14-FITC conjugated to FITC (Coulter, Fullerton, CA) as a specific marker for human monocytes. Mouse IgM-FITC (Coulter) was used as negative isotype control. The binding of synthetic NNK11-C1-biotin (final concentration, 4 M) was either detected by preformed complexes with streptavidin-FITC (Dako, Denmark) or with streptavidin-R-phycoerythrin (Dako). Fluorescence and scatter properties of cells were analyzed in Partec Cyto-Flow ML (Munster, Germany) flow cytometer.
Receptor-mediated Internalization and Degradation-Peptide-mediated internalization and degradation of streptavidin coupled to 1 M magnetic beads (Invitrogen) were investigated essentially as described (26). Briefly, 125 I-labeled streptavidincoupled magnetic beads were charged with the NNK11-C1 peptide and incubated with K562 cells, in the presence or absence of 0.6 M concanamycin A (Sigma) or a combination of 1.6 M leupeptin (Roche Applied Science) and 80 M E64d (Calbiochem, Germany). Samples of supernatants were collected after 0, 3, and 6 h of incubation. Following precipitation with 10% trichloroacetic acid, acid-insoluble material, which represents cell-degraded 125 I-labeled streptavidin, was counted.

Production of Recombinant Fc␥RII-GST Fusion Proteins-
The extracellular domains of Fc␥RIIA-R134, Fc␥RIIA-H134, and Fc␥RIIB were produced as recombinant GST fusion proteins as described in the supplemental text. All were purified on GST column and analyzed by SDS-PAGE electrophoresis where they appeared as bands of expected size (55 kDa) (not shown). Furthermore, all three variants bound heat-aggregated hIgG, the Fc␥RIIA variants slightly better than Fc␥RIIB (26), which is in agreement with data for Fc␥Rs produced by others (36,37).
Affinity Selection of Fc␥RIIA-binding Peptides from a Cys6 Library-Initially, a phage library of six random amino acids flanked by cysteines (the Cys6 library) was searched for binders to Fc␥RIIA-R134-GST as described under "Experimental Procedures." After three rounds of selection resulting in increasing output titer, individual clones were sequenced. Among 29 selected clones, seven different amino acid sequences were found (Table 1) five of which shared the consensus motif XW(F/Y/W)(Q/E)(W/F)P. They revealed up to 10 times increased recovery from wells coated with Fc␥RIIA-R134-GST compared with wells coated with skm or GST only (results not shown), and clone C6-D (CPWFQWPC) showed the highest recovery (Fig. 1a). Neither an Fc␥RI-binding phage clone, C6-1 (Fig. 1a) nor unselected phage clones (not shown) bound the target. Furthermore, the C6-D phage clone bound better to wells coated with Fc␥RIIA-R134-GST than to wells coated with Fc␥RIIB-GST (Fig. 1b).
In Vitro Evolution and Selections of New Fc␥RIIA Peptide Binders-To improve affinity by increasing the putative contact surface area between peptide and target, three evolution libraries based on the C6-D sequence were prepared. In the Evo1 library, the C6-D sequence was extended by three random amino acids (X) on each side (XXXCPWFQWPCXXX). In the Evo2 and Evo3 libraries, one proline and one of the flanking cysteines were excluded from the C6-D motif, and six random amino acids added to either side (XXXXXXWFQWPC and CPWFQWXXXXXX). In addition, a phage library of 11 random amino acids (NNK11) was prepared as described under "Experimental Procedures." The sizes of the evolution libraries and the NNK11 library were estimated to be 1-3 ϫ 10 7 and 9 ϫ 10 8 different clones, respectively. DNA sequencing revealed the presence of diverse inserts of expected length (not shown). Three rounds of selection were performed as described under "Experimental Procedures," and selection from the three evolution libraries was compared with selection from the NNK11 library and also a Cys9 library of nine random amino acids flanked by cysteines, previously described (31) (supplemental Table S2).
Characterization of Individual Phage Clones-To characterize individual phage clones, supernatants from 10 to 30 individual isolates after three rounds of selection from each library were sequenced. The sequences are presented in Table 2, and an analysis of the sequences is included in the supplemental text. The clones were then screened for binding to Fc␥RIIA-R134-GST, and the clones with the best binding capacity from each library were then compared for binding to Fc␥RIIA-R134-GST, Fc␥R-H134-GST, Fc␥RIIB-GST, GST, and 1% skm. The results are summarized in Fig. 2. The NNK11-C1 clone showed thousand times increased recovery from wells coated with Fc␥RIIA-R134-GST as compared with the C6-D clone. In contrast, recovery from Fc␥RIIB-GST was 10-fold lower than that of the C6-D clone, as was background binding to skm. The  clones selected from the evolution libraries (Evo1-G2, Evo2-B2 and Evo3-G10) showed up to 10-fold increased binding to both Fc␥RIIA-R134-GST and Fc␥RIIB-GST compared with the C6-D clone. However, these clones also revealed increased background binding to wells coated with 1% skm. None of the clones selected from the 9-mer library reached the binding capacity toward Fc␥RIIA-R134-GST as the C6-D clone (not shown). All of the clones tested showed equal binding to Fc␥RIIA-R134-GST and Fc␥RIIA-H134-GST (not shown).
Inhibition of the NNK11-C1 and C6-D Phage Binding to Fc␥RIIA by Free Peptide-Competition assays with a free synthetic peptide with NNK11-C1 sequence including the flanking residues ADGA and GAA from the phage format and a C-ter-minal Lys for conjugation of a biotin molecule (ADGAWAWVWLTETAVGAAK-Bio) showed clear inhibition of corresponding phage binding to Fc␥RIIA-R134-GST in ELISA (Fig. 3). Thus, the binding of the NNK11-C1 peptide to Fc␥RIIA was sequence-specific and not dependent on fusion to the phage. Furthermore, the peptide inhibited C6-D phage binding, although to a lesser extent. This suggests proximity of the binding sites for the two sequences on the receptor target. No inhibition of a control phage was observed.

SPR Analysis of NNK11-C1 Peptide Binding to Fc␥Rs and NNK11-C1 Peptide-mediated Inhibition of IgG Binding to
Fc␥RIIA-The kinetics of the interaction between the synthetic NNK11-C1 peptide and Fc␥RIIs were studied using SPR. To circumvent potential aggregation of the GST-tagged receptor proteins (38,39), Fc␥RIIA-R134-His and Fc␥RIIB-His were expressed as described in the supplemental text, and monomeric fractions of the receptors were isolated by size exclusion chromatography (supplemental Fig. S1). Both Fc␥RIIA-R134-His and Fc␥RIIB-His bound chIgG3 immobilized on a CM5 chip, and the steady state levels of the SPR responses (supplemental Fig. S2) were used to calculate the equilibrium affinity constants to be 1 and 2 M (supplemental Table S3), respectively, which agrees with previous kinetic studies (37).

Identification of a High Affinity Fc␥RIIA-binding Peptide
The synthetic biotinylated peptide was then immobilized on a SA chip, and samples of Fc␥RI (2 M), Fc␥RIIA-R134-His (1 M), Fc␥RIIB-His (1 and 2 M), and Fc␥RIIIB (2 M) were injected. The steady state level of the SPR response curve for the binding of Fc␥RIIA was set to 1. The data obtained showed increased binding of Fc␥RIIA compared with Fc␥RIIB, whereas binding of neither Fc␥RI nor Fc␥RIIIB was detected (Fig. 5a). Then serial injections of increasing concentrations (0.078 -4 M) of Fc␥RIIA-R134-His (supplemental Fig. S3) and Fc␥RIIB-His (data not shown) were injected. The kinetic rate constants (supplemental Table S4) were obtained using a heterogeneous ligand model, which gave the best global fit using the BIAevaluation 4.1 software. The model assumes that there are two independent parallel reactions with the immobilized NNK11-C1 peptide, which was determined to be at 0.5 M (KD1) and 0.2 M (KD2) at both 25°C as well as 37°C. This fitted well with the equilibriumderived affinity constant of 0.5 M (supplemental Table S4). The binding responses for the interaction between Fc␥RIIB-His and the NNK11-C1 peptide was too low for an affinity to be determined.
Injections of Fc␥RIIA-R134-His preincubated with an excess amount of chIgG3 over immobilized peptide resulted in decreased binding responses (Fig. 5b). The same was observed when Fc␥RIIA-His was preincubated with an excess amount of  free peptide prior to injection over immobilized hIgG3 (Fig. 5c). Thus, the NNK11-C1 peptide interferes with hIgG3 binding to Fc␥RIIA.
Characterization of the Peptide-binding Site-To characterize the binding specificity of the peptide at the amino acid level, a number of Fc␥RIIA and Fc␥RIIB mutants were generated, coated in wells, and tested for peptide binding as before. First, the lack of NNK11-C1 binding to two Fc␥RIIA mutants, namely IIA W90A and IIA W113A, clearly demonstrated that both of these residues are critical for binding (Fig. 6a). Subsequently, the amino acid sequences of Fc␥RIIA and Fc␥RIIB were aligned (supplemental Fig. S4), and 10 residues found to differ between their extracellular parts, namely P15Q, Q30R, A32T, R33H, M107V, Q130K, L135S, T138N, F163Y, and V174A (Fig. 6b). Six are found in the cleft between the two Ig domains opposite the IgG interaction region involving Trp 90 and Trp 113 . Four (Gln 130 , Leu 135 , Thr 138 , and Phe 163 in Fc␥RIIA) are, however, close to the IgG interaction surface in the folded molecule. Each and all four (IIB 4mut) were made Fc␥RIIA like on an Fc␥RIIB background, making a total of five Fc␥RIIB mutants. All were coated as before and compared with wild type Fc␥RIIA and Fc␥RIIB for peptide binding. Interestingly, of the four amino acids tested, only one had a clear effect, namely IIB Y163F. Both IIB Y163F and IIB 4mut showed increased peptide binding compared with Fc␥RIIB wild type and the other three mutants. An Fc␥RIIA mutant where F163 was made IIB-like (IIA F163Y) lost peptide binding ability (Fig. 6a). Thus, a single hydroxyl group is at the core of the IIA/IIB discrimination shown by the selected peptide.
CD Analysis of the NNK11-C1 Peptide-The NNK11-C1 peptide precipitated fast from water as well as methanol and ethanol solutions. In spite of this, we were able to obtain CD spectra of the peptide in water. In addition, the CD spectra in water:trifluoroethanol was obtained (supplemental Fig. S5). NNK11-C1 did not have a well defined structure in water but formed an ␣-helix in mixtures of water and trifluoroethanol, with an ␣-helical content of 54% in 50% trifluoroethanol solution.
Flow Cytometry Analysis of NNK11-C1 Peptide Binding to PMN and Monocytes-Biotinylated peptide was complexed to streptavidin-FITC as described under "Experimental Procedures" and incubated with peripheral blood leukocytes before analyzes by flow cytometry. The scatter diagrams were gated for PMN, monocytes, and lymphocytes. Both PMN and monocytes bound the NNK11-C1 peptide, whereas no binding to the irrelevant control peptide RB14 was seen (Fig. 7). The phenotype of the gated PMN and monocyte preparations was verified by staining with anti-Fc␥RIIIB-FITC (PMN) and anti-CD14-FITC (monocytes), respectively (data not shown). In contrast, when gating on the lymphocyte fraction, no binding of NNK11-C1 was observed (Fig. 7). Again, this demonstrates the Fc␥RIIA specificity of the peptide, as normal B-cells express Fc␥RIIB because their only Fc receptor and the B-cells in the lymphocyte population were negative. Aggregated hIgG showed a dose-dependent inhibition of binding of NNK11-C1 peptide complexed to streptavidin (supplemental Fig. S6), which again suggested overlapping binding sites.
The NNK11-C1 Peptide Induces Receptor-mediated Internalization-The functional properties of the NNK11-C1 peptide as regarding induction of Fc␥RIIA-mediated internalization and degradation was demonstrated as follows. Aggregates of the NNK11-C1 peptide were prepared by incubation of bio- tinylated NNK11-C1 peptide with iodinated ( 125 I) streptavidin on 1-m magnetic beads. The beads were then incubated with K562 cells at 37°C for 3 or 6 h, and the degradation of streptavidin was estimated as previously described (26). The cells that were stimulated with beads loaded with the NNK11-C1 peptide degraded significantly more streptavidin than cells incubated with beads loaded with the irrelevant control peptide (RB14), at both time points (Fig. 8). Both concanamycin A (inhibitor of the vacuolar proton pump) and a combination of E64d (inhibitor of lysosomal thiol proteases) and leupeptin (inhibitor of lysosomal serine and cysteine proteases) prevented the formation of acidsoluble cpm in the medium, suggesting that degradation of streptavidin took place in late acidic endocytic compartments (late endosomes/lysosomes).

DISCUSSION
Through a comprehensive selection strategy, we identified a peptide that bound specifically and with high affinity to the human activating Fc␥RIIA. The peptide, derived from an 11-mer library and denoted NNK11-C1, bound with similar affinity to both allelic variants of the receptor, Fc␥RIIA-H134 and Fc␥RIIA-R134, showed very low binding to the inhibitory Fc␥RIIB, and showed no binding to two other activating Fc␥Rs, namely Fc␥RI and Fc␥RIIIB. Peptide binding to Fc␥RIIA competed with binding of the natural ligand, hIgG. Polymerized peptide mediated targeting of streptavidin to cells expressing Fc␥RIIA and promoted internalization and degradation of this model antigen in acid vesicles.
We used two different phage display based strategies to select specific peptide binders to Fc␥RIIA. In an "evolution approach" a core consensus motif of six amino acids was isolated, and the best binder identified (C6-D) was further extended with four or six flanking amino acids. In the second approach, two libraries of either 11 random amino acids or nine random amino acids constrained by cysteines were searched for binders. The C6-D phage clone distinguished somewhat between Fc␥RIIA and Fc␥RIIB. After sequence extension and additional rounds of selection, binding to Fc␥RIIA increased. However, so did binding to Fc␥RIIB and blocking reagent. In contrast to this, a single phage clone from the 11-mer library, NNK11-C1, had the desired binding characteristics with three logs better recovery from Fc␥RIIA than C6-D and four logs better recovery on Fc␥RIIA than on Fc␥RIIB. Also, free synthetic peptide bound Fc␥RIIA in ELISA and discriminated exceptionally well between Fc␥RIIA and Fc␥RIIB. Furthermore, the same free synthetic peptide preparation inhibited binding of phages displaying either NNK11-1C or C6-D, and both strategies thus selected binders to overlapping  sites on the receptor. The fact that NNK11-1C was inhibited to a greater extent than C6-D and that NNK11-1C discriminates better than C6-D between Fc␥RIIA and Fc␥RIIB argues against the binding sites being identical. That the NNK11-C1 peptide binds at or close to the IgG-binding site was demonstrated in two independent SPR-based assays where Fc␥RIIA was preincubated with IgG and tested for binding to immobilized peptide, or vice versa, preincubated with peptide and tested for binding to immobilized IgG. In either case, less complexed than free receptor was bound. Also, the interference with the IgGbinding site was verified by flow cytometry analysis of NNK11-C1 binding to PMN and monocytes.
The specificity of the peptide was further confirmed by SPR measurements, which revealed binding to Fc␥RIIA at an affinity of 0.5 M, greatly reduced binding to Fc␥RIIB and almost no binding to Fc␥RI and Fc␥RIIIB. The affinity toward Fc␥RIIA was slightly increased compared with hIgG, which is important for any future clinical application of the NNK11-C1 peptide.
The selected peptides all contain several aromatic amino acids and in particular tryptophane. It should be noted that a high number of aromatic amino acids is commonly found among phage display selected peptides (31, 40 -42). Also in native proteins, aromatic amino acids play an important role in folding and recognition and are often found as part of binding sites for small ligands and enzyme substrates (43,44). Other sequences with specificity for Fc␥RI that were previously selected from the same Cys6 library that was used in the present study did not contain tryptophanes (26). Thus, the selection of tryptophane depends on the nature of the target binding site and not the library per se. Interestingly, peptide Fc␥RIIA binders recently identified (45) also contain two tryptophane residues. The best binder, CWPGWDLNC (C7C1), selected from a 7-mer library, has striking similarities with C6-D described here. Mutagenesis within the C7C1 peptide showed that W2A and P3A lost affinity for Fc␥RIIA, and these amino acids are also present in C6-D. Binding of C7C1 was inhibited by IgG. Because NMR structure analyses of C7C1 revealed a type II ␤-turn between the two tryptophanes and solvent-exposed proline, a binding mode was suggested where the proline is inserted between Trp 90 and Trp 113 on Fc␥RIIA, mimicking the interaction between Fc␥RIIA and IgG (46,47). The argument may hold for C6-D as well. Because Trp 90 and Trp 113 are found in both Fc␥RIIA and Fc␥RIIB, this would explain why the short C6-D peptide and the sequences derived from C6-D distinguish poorly between the two receptors. In contrast, the NNK11-C1 peptide showed a great increase in binding to Fc␥RIIA compared with Fc␥RIIB. Both W90A and W113A IIA mutants lost affinity for the peptide, demonstrating that binding is indeed dependent on Trp 90 and Trp 113 , found on both receptors, and part of the IgG-binding site. Notably, the affinity of NNK11-C1 for Fc␥RIIA was measured to be 200 times better than the 100 M found for C7C1. Because the NNK11-C1 peptide is longer than C7C1, it may offer the possibility for a greater interaction surface between the peptide and Fc␥RIIA. Mutational analysis revealed that Phe 163 in Fc␥RIIA is part of this interaction surface. Phe 163 is in very close proximity to the two tryptophanes in the folded molecule, and the peptide may bind the extended site whether it forms a ␤-sheet or has an ␣-helical conformation.
Residue 163 is a tyrosine in Fc␥RI and Fc␥RIIB and a valine in Fc␥RIIIA and Fc␥RIIIB, which indicates that the peptide has a requirement for an aromatic, hydrophobic amino acid in this position that allows for hydrophobic stacking. This particular amino acid is not a part of the IgG interaction site with Fc␥RIII (46,47).
Both the Fc␥RIIB Y163F mutant and the Fc␥RIIB mutant where all amino acids at the IgG binding surface were IIA-like showed increased peptide binding. The level did, however, not quite reach that of Fc␥RIIA. Amino acids in the cleft between the two Ig domains may contribute. An overlay of Fc␥RIIA and Fc␥RIIB structures (Fig. 9) shows the two tryptophanes central to the IgG-binding site, as well as amino acid residue 163. Notably, the 163 Y/F transition alters neither the main chain nor the side chain position. The peptide thus seems to discriminate between the absence and presence of a single hydroxyl group.
Important features for usefulness in clinical settings of targeting peptides are high affinity to ligand, low immunogenicity, and absence of unwanted side effects. We believe that the ability of the short NNK11-C1 peptide sequence to exploit Fc␥RIIAmediated uptake and degradation without interfering with the inhibitory Fc␥RIIB makes this ligand an interesting candidate for further studies. Furthermore, the ability of the NNK11-C1 peptide to bind to Fc␥RIIA in an allele-independent fashion suggests potential applications of this molecule in patients who carry the low binding allele.