Peptides Derived from the Complementarity-determining Regions of Anti-Mac-1 Antibodies Block Intercellular Adhesion Molecule-1 Interaction with Mac-1*

Peptides or small molecules that can block the interaction of the integrin Mac-1 with its receptor, intercellular adhesion molecule-1 (ICAM-1), have not previously been developed. We studied this interaction by measuring the adherence of ICAM-1-expressing Chinese hamster ovary (CHO) cells to immobilized, purified Mac-1. Nucleotide sequence information was obtained for the complementarity determining regions (CDRs) of three antibodies (44aacb, MY904, and 118.1) shown to block Mac-1-mediated cell adherence. Peptides were synthesized based on the predicted amino acid sequences of the CDRs and tested for the ability to block cell adhesion to Mac-1. Peptides derived from CDR1 of 44aacb, CDR2 of 118.1, and CDRs 1 and 3 of MY904 heavy chains were found to possess blocking activity at 10–100 μm. This may indicate that one or two CDRs contribute disproportionately to the antibody binding affinity. The binding of ligands to Mac-1 has been shown to require a region of the α-chain known as the I- or A-domain. We have recombinantly produced Mac-1 I-domain, and show that it is also capable of supporting the adherence of ICAM-1-expressing CHO cells. The adherence of ICAM-1-CHO cells to the I-domain is inhibited by 44aacb and 118.1 and by the CDR peptides from 44aacb and 118.1. By using phage display of peptide libraries based on the 118.1 CDR peptide with five residues randomized, we were able to identify a novel peptide inhibitor of Mac-1 with substitutions at all five positions. These peptides provide lead structures for development of Mac-1 antagonists.

The extravasation of white blood cells to sites of inflammation and the phagocytosis of opsinized microorganisms by these cells is clearly crucial to host defense. However, the mounting of an inappropriately large mobilization of phagocytic cells is thought to contribute to organ damage in sepsis, in adult respiratory distress syndrome, and following reperfusion of ischemic tissue (1,2). Activated neutrophils are recruited into tissues or are sequestered in the microcirculation of the lung and liver; in either case tissue is damaged upon their degranulation and release of enzymes and activated oxygen species. Interruption of neutrophil extravasation or oxidative burst may be an effective means of damage control in these situations.
Mac-1 is a cell surface glycoprotein contributing to several myeloid cell functions including adherence to and transmigration across the endothelium, binding and phagocytosis of opsinized particles, and the oxidative burst (3)(4)(5). It is a heterodimer of two transmembrane proteins, CD11b (␣ M ) and CD18 (␤ 2 ), the latter also being part of the related integrins LFA-1, p150/95, and ␣ D ␤ 2 . The major adhesion partner for Mac-1 is ICAM-1 1 (6), which is a member of the Ig supergene family and contains five extracellular Ig domains of the C2 type, characteristic of Fc receptors and proteins involved in cell adhesion (7). However, the complexity of the functions involving Mac-1 results in part from the fact that an array of ligands besides ICAM-1 is also recognized by this molecule, including iC3b, fibrinogen, and factor X (3). The precise residues of Mac-1 and ICAM-1 mediating their interaction are not known, although the domains responsible have been elucidated. Ig domain 3 of ICAM-1 has clearly been implicated in binding to Mac-1, whereas domain 1 mediates binding to the related adhesion molecule, LFA-1 (8). CD11b contains a 200-amino acid "inserted domain" or "I domain," so called due to its presence only in the other ␤ 2 integrins and in the VLA ␣1 and ␣2 subunits and its absence in most other integrins. Antibodies to this domain can block ICAM-1 binding, as well as that of iC3b and fibrinogen (3). Mutations within the I-domain of Mac-1 have been shown to prevent binding of ICAM-1 and iC3b (9 -11). The Mac-1 I-domain has been expressed recombinantly, and there are some data suggesting that it interacts with fibrinogen, iC3b, and soluble ICAM-1 (4,12). Whether the I-domain can support the adherence of ICAM-1-expressing cells independent of other domains of Mac-1 has not been previously demonstrated.
The lack of information on the precise sequences within ICAM-1 and Mac-1 that interact has precluded modeling of small molecule inhibitors. One approach to overcoming this problem that has proven successful in other cases has been to use CDR sequences from antibodies directed at the active site as lead structures. Antibodies useful in this regard have been either developed as anti-idiotypic to the ligand (13) or simply chosen by virtue of their ligand blocking activity (14). As well as providing lead inhibitors, in some cases CDRs have shared sequence similarity with a portion of the known ligand, implicating that sequence in receptor binding (14 -16).
We have produced the Mac-1 I-domain recombinantly and show that it supports the adherence of ICAM-1-expressing cells. Several antibodies that block the adherence of ICAM-1 to both Mac-1 and to the I-domain have been sequenced to allow determination of CDR structures. Peptides based on these CDR structures are shown to block the adherence of ICAM-1 to both Mac-1 and the I-domain.

MATERIALS AND METHODS
Purification of Mac-1-Peripheral blood leukocytes were purified from "Buffy coats" (Stanford Medical School Blood Center, Stanford, CA). Buffy coats were diluted 1:1 with HBSS and layered on Histopaque (Sigma) gradients as described (17). Both mononuclear and polymorphonuclear cell layers were collected, and red blood cells were lysed one or two times as necessary. Mac-1 was purified from lysates of the remaining leukocytes by immunoaffinity chromatography essentially as described by Diamond et al. (6) and purity assessed using SDSpolyacrylamide gel electrophoresis (see Fig. 1A).
Construction of Mac-1 I-domain-The construction of the I-domain was based on a plasmid construct reported by Michishita et al. (11). The I-domain of human Mac-1 from the glycine residue at position 111 to the alanine at position 318 was generated using synthetic oligonucleotides and PCR. Eight overlapping oligonucleotides were synthesized and combined in a stepwise PCR procedure to generate the final 603-base pair fragment. The 5Ј-most oligo included a BamHI site present naturally in the Mac-1 gene sequence, whereas the 3Ј-most oligo included an added EcoRI site. The internal EcoRI site in this I-domain region was eliminated via a single base change from A to T at the third position in the glutamate codon at position 179 (a silent mutation). Each pair of oligonucleotides sharing partial complementarity was annealed and subjected to PCR (3Ј at 94°C, followed by 10 cycles of 1Ј at 94°C, 2Ј at 55°C, and 3Ј at 72°C) using 250 M dNTPs and VENT polymerase (New England Biolabs, Beverly, MA). The PCR products were then mixed, melted for 3Ј at 94°C, and subjected to PCR as described above. After assembly of all of the oligos, the resulting BamHI-EcoRI fragment was cloned into the pGEX-2T (Pharmacia Biotech Inc.) vector at the BamHI and EcoRI sites, resulting in an in-frame fusion with an Nterminal domain encoding glutathione S-transferase.
Purification of Recombinant I-domain-The glutathione S-transferase-I-domain fusion protein was expressed in Escherichia coli cells (strain JM101). Overnight cultures of E. coli JM101 were diluted 1:10 with L broth medium containing ampicillin and grown for 1 h at 37°C. Isopropyl-␤-D-thiogalactoside (1 mM) was added to induce expression of the fusion protein, and after 3 h of growth, bacteria were pelleted and frozen at Ϫ80°C. Pellets derived from a 1-liter culture were then thawed and resuspended in 18 ml of cold PBS, pH 7.4, phenylmethylsulfonyl fluoride was added at 1 mM, and the samples were disrupted on ice at 10,000 p.s.i. in a French Press (SLM Instruments, Inc., Rochester, NY). 20% Triton X-100 was then added to a final concentration of 1%, and the lysate was incubated on ice for 30 min with occasional rocking. After centrifugation at 12,000 ϫ g for 10 min, the supernatant was incubated with glutathione-Sepharose 4B (1.5 ml, Pharmacia) for 1 h at room temperature. The beads were then washed once in PBS plus 0.35 M NaCl and washed four times with PBS and were resuspended in 5 ml of PBS. 200 units of human thrombin (Enzyme Research Laboratories, South Bend, IN) was added to cleave the I-domain from the fusion protein, and the mixture was incubated for 2 h at room temperature. NaCl and MgCl 2 were added to the cleaved soluble recombinant I-domain at final concentrations of 0.35 M and 1 mM, respectively, and the samples were then passed through a 700-l benzamidine-Sepharose 6B (Pharmacia) column to remove the thrombin. The flow-through was assayed for thrombin activity using Chromozym TH (Boehringer Mannheim, Indianapolis, IN) as a substrate, and the clearance was greater than 150-fold. Protein concentration was determined using the Bio-Rad protein assay (Bio-Rad). Typical yields of the recombinant I-domain were 3-5 mg/liter bacterial culture.
Mass Spectrometric Analysis of I-domain-Electrospray ionization was performed on a Finnigan SSQ 7000 mass spectrometer (San Jose, CA) in the positive ion mode. Liquid chromatography/mass spectroscopy was performed using a capillary reversed phase column with a flow rate into the mass spectrometer of 5 l/min. To map tryptic peptides, electrophoresis of I-domain protein was carried out in 12% polyacrylamide followed by transfer to Immobilon (70 V, 60 min; Ref. 18). In situ tryptic digestion was carried out according to the method of Wong et al. (19). Capillary HPLC was performed using a Valco tee to split the 200 l/min flow from an HP 1090 HPLC PV5, driving the capillary system (Vydac C18 0.32 ϫ 250-mm column, maintained at 40°C, Microtech Inc., Sunnyvale, CA) at 5 l/min. The gradient was as follows: 0 min, 100% A (0.1% trifluoroacetic acid in H 2 O); 40 min, 30% B (0.09% trifluoroacetic acid in acetonitrile); 50 min, 60% B; 51 min, 0%B; 72 min, 0% B.
Adherence Assays-The adherence of human neutrophils to wells coated with human serum was carried out exactly as described previously (17). CHO cells were stably transfected with an expression vector encoding human ICAM-1 (20), and their adherence was assessed as follows. ICAM-1-CHO cells were grown in RPMI 1640 with 10% fetal bovine serum and used at 80% confluence. Cells were loaded with calcein-acetoxymethyl ester (Molecular Probes, Eugene, OR) at 5 M in HBSSϩϩ for 30 min at 37°C. The cells were then detached from the flask with PBS containing 5 mM EDTA (15 min, 37°C), washed in HBSSϩϩ containing 0.1% HSA (Sigma), and resuspended in HBSSϩϩ containing 0.5% HSA at 2 ϫ 10 6 /ml. 96-well plates were coated with purified Mac-1 or I-domain in HBSSϩϩ (50 l/well) for 2 h at 37°C. In the case of Mac-1, ␤-octylglucoside at a final concentration of 0.15% was present in the coating solution. For each lot of Mac-1, the optimal protein concentration for coating the wells (5-13 g/ml) was determined as that which best supported the adherence of ICAM-1-CHO cells but not of vector-CHO cells. Wells were washed twice with HBSSϩϩ containing 0.1% HSA and blocked with HBSSϩϩ supplemented with 0.5% HSA for 30 min at 37°C. Test compounds were preincubated in blocked wells for 15 min at 37°C in 50 l, and then 50 l of cells were added for an additional 60-min incubation. Nonadherent cells were removed by gently inverting plates and blotting on paper towels. Wells were washed twice with HBSSϩϩ containing 0.1% HSA. Adherent cells were quantitated in a 96-well fluorescence plate reader (IDEXX Labs, Westbrook, ME). HSA lots were tested in this assay to find those that gave a minimal background adherence and a maximal adherence to Mac-1. All incubations in HBSSϩϩ were carried out without CO 2 .
Monoclonal Antibodies-Hybridoma cells producing the anti-Mac-1 monoclonal antibodies LM2/1 (murine IgG1), 44aacb (murine IgG2a ) and MY904 (murine IgG1 ) were from ATCC (21,22). The hybridoma cell line secreting mAb 118.1 (murine IgG1) was generated from a Balb/C mouse immunized according to Diamond et al. (3). Spleen cells were mixed in a 5:1 ratio with FOX-NY murine myeloma cells (ATCC 1732 CRL) and fused by slow addition of polyethylene glycol 1500 (0.5 ml/10 8 cells, Boehringer Mannheim). After 1 min at 37°C, the cell suspension was slowly diluted in RPMI, centrifuged at 200 ϫ g for 7 min, and resuspended in selection medium (RPMI 1640, 20% fetal bovine serum, Pen/Strep, AAT media supplement (Sigma), and STM Mitogen (RIBI ImmunoChem Research Inc.)). After 9 days of growth in 96-well culture plates, the hybridoma supernatants were screened for antibodies binding to purified Mac-1 in an enzyme-linked immunosorbent assay adapted from Diamond and Springer (23). Serum-free conditioned media from the hybridoma lines were harvested from confluent roller bottles, and antibodies were purified by protein A chromatography using PROSEP-A affinity resin (BioProcessing, Ltd.) following the manufacturer's suggested protocol.
Cloning and Sequencing of the Variable Regions of 44aacb, MY904, and 118.1-mRNA was isolated from 44aacb, MY904, and 118.1 hybridoma cells (10 8 cells each, grown in RPMI 1640, 10% fetal bovine serum, 10 mM HEPES) using the Mini RiboSep mRNA Isolation Kit (Becton Dickinson, Bedford, MA). Following first strand cDNA synthesis using a cDNA synthesis kit (Amersham Corp.), V H and V L regions were amplified (Vent DNA polymerase, New England Biolabs, Beverly, MA) using the following primers: V H regions, upstream primer 5Ј-GCA-GAATTCSARGTSCARTTRCARCA and downstream primer 5Ј-GCA-GAATTCGGGGCCAGTGGATAGAC; V L regions, downstream primer 5Ј-GCAGAATTCGGTGGGAAGATGGATACAGTT coupled to upstream primers 5Ј-GCAGAATTCACMCARTCHCCAGTNAT (44aacb), 5Ј-GCA-GAATTCGAYATYGTBCTGACNCA (MY904), or 5Ј-GCAGAATTCGAY-GTBGTKATGACMCA (118.1) (S ϭ C or G; R ϭ A or G; M ϭ A or C; Y ϭ C or T; K ϭ G or T; H ϭ A, C, or T; B ϭ C, G, or T; and N ϭ A, C, T, or G). The sequences of the downstream primers were based on the conserved nucleotide sequences of the 5Ј-ends of the constant regions of the heavy and light chains. The upstream primers were designed from either the experimentally determined N-terminal amino acid sequences of the three antibodies or deduced N-terminal "consensus" amino acid sequences of murine immunoglobulin variable regions from GenBank. The mouse codon usage was also considered when degenerate primers were designed. An EcoRI site was introduced at the 5Ј-end of all the primers to facilitate cloning. PCR was performed for 30 cycles of 1 min at 94°C, 1 min at 55°C, and 1 min at 72°C, and bands were resolved by electrophoresis in a 1% agarose gel.
The correctly sized PCR products were digested with EcoRI, gelpurified, and ligated into EcoRI-digested pUC9 vector DNA. The ligated DNA was then transformed into E. coli JM101 by electroporation, and clones were selected on LB plates containing ampicillin. For each V H or V L region, three or four positive clones were picked, DNA was prepared using the Wizard miniprep system (Promega, Madison, WI), and the PCR product carried in the plasmid DNA was sequenced.
Construction of Phage Libraries-Two peptide libraries were constructed in which the library inserts were directly linked to the N terminus of the C-terminal domain of M13 gene III as described (24). The GYXDXYXGXIXYN and GXIXPXYXGXTYN libraries were prepared using synthetic oligonucleotides containing core sequences GGA TAT NNS GAT NNS TAC NNS GGT NNS ATT NNS TAC AAC and GGA NNS ATT NNS CCT NNS TAT NNS GGT NNS ACC TAC AAC, respectively. The nucleotides were made double-stranded by extension with Klenow polymerase (New England Biolabs) from a 5Ј primer. The resulting products were gel purified using MERmaid spin kit (Bio101 Inc., Vista, CA) and inserted into the BstEII and BamHI sites of the phagemid vector pAL53. 2 The library-containing vectors were then transfected into E. coli Top10FЈ cells using 10 -12 electroporations. The yields of the two primary libraries were 0.5-1 ϫ 10 7 .
Panning of Phage Libraries-Libraries were packaged into phage particles by infection of the transfected E. coli Top10FЈ library with M13KO7 helper phage, resulting in expression of less than one copy of the peptide-gene III protein per phage (24). Phage expressing Mac-1 binding peptides were selected by panning on 96-well enzyme-linked immunosorbent assay plates (Corning) coated with Mac-1 (0.5 g/well in HBSSϩϩ containing 0.15% ␤-octylglucoside for 2 h at 37°C). After the wells were washed and blocked with HBSSϩϩ supplemented with 0.5% HSA, 10 l of partially purified phage (ϳ5 ϫ 10 11 cfu) was added to the wells in a total volume of 100 l of the same buffer and incubated at 37°C for 2 h while shaking. The unbound phage particles were then removed by washing with 100 l of HBSSϩϩ/0.05% Tween six times quickly on ice. Bound phage were eluted with 200 l of 100 mM citrate buffer, pH 3.0/0.1% Tween for 30 min at room temperature with shaking and neutralized with 25 l of 1 M Tris base. The wells were then washed three times with 100 l of PBS, and eluates and washes were combined. The eluted phage were titered and amplified and carried through three more rounds of enrichment by panning on Mac-1. The stringency of the PBS wash conditions was increased in successive rounds as follows: nine quick washes at room temperature in round 2, followed by two additional washes for 10 min each with shaking in round 3 and three washes for 10 min each with shaking in round 4. Phage were randomly selected for sequencing after round 4.

Production of Recombinant Mac-1 I-domain-Purified
Mac-1 has been shown to support the adherence of ICAM-1-expressing L-cells (6). To determine whether the I-domain of Mac-1 could similarly support the adherence of cells expressing ICAM-1, this domain was expressed recombinantly in E. coli. Analysis of the purified protein by SDS-polyacrylamide gel electrophoresis showed a major band migrating as expected and in some cases a minor band with slightly slower migration (Fig. 1B). Analysis of the purified protein by mass spectrometry showed a major peak of 24,103 Da, in agreement with the molecular mass predicted for this construct. Also revealed was a minor peak of 25,102 Da, present at approximately 10 -20% the level of the major peak in different preparations. To determine the source of the additional 1000 Da, the proteins in the I-domain preparation were separated by SDS-polyacrylamide gel electrophoresis and subjected to both N-terminal sequence analysis and tryptic peptide mapping. Both the 24-and 25-kDa proteins had the expected N-terminal sequence of GSNLRQQP. The 24-and 25-kDa bands showed a similar tryptic map pattern with the exception of the C-terminal peptide (IFANSS; mass, 638 Da), which disappeared from the digest of the 25-kDa band. A new peptide (mass, 1634 Da) was evident in the map of the 25-kDa protein. The mass of the new peptide was 997 Da larger than the predicted C-terminal peptide, similar to the mass difference observed in the intact molecules, and thus we conclude that the added mass is contained at the C terminus. The 1634-Da peptide could be accounted for by translation of the usual stop codon in the pGEX-2T expression vector as a tryptophan residue, translation of 10 additional vector residues prior to the next stop codon, and proteolytic trimming of three C-terminal vector residues, resulting in the tryptic peptide IFANSSWLTDDLPR.
Adherence Assays Based on Purified Mac-1 or the Mac-1 I-domain-Purified recombinant I-domain was tested for the ability to support the adherence of ICAM-1-CHO cells. As shown in Fig. 2A CHO cells. The adherence of ICAM-1-CHO cells was blocked by the anti-Mac-1 antibody 44aacb. However, at high levels of I-domain, vector-transfected CHO cells were also adherent, and this adherence was blocked by 44aacb. These results indicate that the I-domain of Mac-1 interacts not only with ICAM-1 but also with an unknown receptor on CHO cells. This interaction with a CHO cell receptor is not unique to the recombinant I-domain, because vector-CHO cells also adhered to high levels of purified Mac-1, and this adherence was similarly blocked by 44aacb (data not shown).
To verify that under typical assay conditions with low substrate level, the observed adherence of ICAM-1-CHO cells was due to ICAM-1 rather than endogenous CHO receptors for Mac-1, a blocking ICAM-1 antibody was preincubated with cells prior to addition to the plate. This antibody completely blocked the adherence of ICAM-1-CHO cells to both the I-domain and to Mac-1 (Fig. 2B). The number of fluorescently labeled ICAM-1-CHO cells adhering to the I-domain under optimal conditions was typically 3-4-fold less than that adhering to Mac-1.
Characterization of Anti-Mac-1 Antibodies-The anti-Mac-1 antibodies 44aacb and MY904 block the Mac-1-mediated adherence of neutrophils (17,22). The antibody 118.1 was generated as described under "Materials and Methods" and also shown to block Mac-1-mediated neutrophil adherence (data not shown). 44aacb, MY904, and 118.1 were found to bind to Mac-1 in an enzyme-linked immunosorbent assay format with EC 50 values of 12.0, 9.3, and 10.7 nM, respectively. Preincubation of these antibodies with Mac-1 (Fig. 3A) resulted in inhibition of the adherence of ICAM-1-CHO cells with IC 50 values of 50 -500 pM in four assays. The three antibodies also blocked I-domainmediated adherence (Fig. 3B) with IC 50 values of 1-3 nM in four assays. LM 2/1, which does not block neutrophil adherence, does not block ICAM-1-CHO adherence to Mac-1 (Fig. 3A).
Characterization of CDRs of Anti-Mac-1 Antibodies-mRNA isolated from hybridoma cells producing 44aacb, MY904, and 118.1 was used to determine the nucleotide sequence encoding the CDRs of each of the antibodies. The deduced amino acid sequences are shown in Table I. Peptides were synthesized corresponding to the underlined portions of the sequences in Table I. These peptides were tested for inhibition of ICAM-1-CHO cell adherence to Mac-1 (Table I). One or two HC CDRs from each antibody blocked Mac-1-mediated adherence with an IC 50 at or below 106 M. Of the light chain CDR peptides made for antibodies 44aacb and MY904, none possessed blocking activity. Two of the active HC CDR peptides were also tested for blocking activity in the adherence assay using the I-domain as a substrate. HC CDR1 from 44aacb and HC CDR2 from 118.1 were found to possess comparable blocking activity in this assay (data not shown).
Identification of Additional Blocking Peptides through Phage Display-Variants of the 118.1 HC CDR2 blocking peptide were identified by displaying libraries based on this peptide on phage and selecting for those phage that bound to Mac-1. To select residues for randomization, the structure of 118.1 was modeled on the three-dimensional structure of the antibody D11.15, an anti-lysozyme antibody (Brookhaven Protein Data Bank 1JHL). D11.15 was chosen for this purpose because of significant homology in the residues immediately flanking the hypervariable domains of 118.1 and D11.15. The antigen binding loop of 118.1 HC CDR2 was predicted to consist of GYID-PYYGGITYN, with PYYG making the ␤-turn. Two libraries, each with five randomized residues in this region were expressed as N-terminal fusions with M13 gene III. To select for high affinity peptides, a monovalent phage display system was utilized in which one copy or less of the gene III protein is expressed as a fusion peptide on each phage (24). Phage expressing these libraries were panned on Mac-1 for four rounds with increasingly stringent wash conditions. Sequencing of phage eluted after four rounds revealed a consensus from one of the libraries, in which all five randomized residues differ from   The phage-derived consensus sequence (plus five C-terminal residues from the 118.1 sequence) was synthesized and tested for inhibitory activity. The consensus peptide was equivalent to the corresponding 118.1 peptide in blocking the adherence of ICAM-1-CHO cells to Mac-1 (Fig. 4A) and to I-domain (Fig. 4B). Peptides made from the consensus sequence either with flanking residues from gene III or lacking the five added flanking residues from the 118.1 sequence were inactive (data not shown).

DISCUSSION
Peptides that block the binding of ICAM-1 to Mac-1 have been derived from the CDRs of anti-Mac-1 antibodies. Antibodies were chosen for this purpose based on their ability to block Mac-1-mediated adherence of ICAM-1-expressing CHO cells. The specificity of these antibodies was further verified by showing that they also block ICAM-1-CHO cell adherence to the recombinant I-domain of Mac-1. The amino acid sequences of the three HC CDRs and three light chain CDRs of each of the antibodies were deduced from the nucleotide sequences encoding the antibodies, and peptides were made from 15 of the 18 total CDR sequences. The adherence of ICAM-1-CHO cells to Mac-1 was blocked by four of these peptides, two from MY906 and one each from 44aacb and 118.1. This may result from one or two of the CDRs contributing most of the binding affinity of the antibody, or it may reflect a subpopulation of peptides taking on an active conformation out of their native context.
Because these peptides compete with ICAM-1 for binding to Mac-1, they were analyzed for homology with ICAM-1 (Ref. 25; Lasergene, DNAstar, Madison, WI). Some similarity was found between HC CDR2 from 118.1 and residues Val 238 to Ala 249 of ICAM-1 (Table II). These residues in ICAM-1 constitute the loop between ␤-strands D and E in immunoglobulin domain 3. Mutations in loops between strands C and D and strands E and F of domain 3 have been shown to inhibit adhesion to Mac-1 (8).
A variant of HC CDR2 from 118.1 was identified using a monovalent display system on the phage M13 and selecting for Mac-1 binding. Interestingly, although all five randomized residues differed from the original 118.1 sequence, this peptide was found to possess similar Mac-1 blocking activity to the parental 118.1 peptide, indicating that there is considerable room for variation, and presumably improvement, within this CDR sequence. The selection of this CDR variant through binding of phage to Mac-1 demonstrates that this peptide does bind to Mac-1 and provides evidence that its blocking activity can be attributed to a specific interaction with Mac-1.
Evidence has suggested that protein-protein interaction involves multiple contact sites, and until recently little progress had been made in disrupting such interactions with small peptides. However, several examples now exist of peptides that can block protein-protein interaction; for example, an 8-amino acid peptide has been identified that can can block the activation of FGFR1 by basic fibroblast growth factor (26). Further, the interaction of the two cell surface proteins, CD4 and MHC class II, can be blocked by a heptapeptide based on a protruding loop from one of the immunoglobulin-like domains of CD4. Thus even in this latter example, where many contact sites from two Ig domains of CD4 are thought to interact with MHC class II, binding appears to depend on a single ␤-turn-containing loop (27). In the case of antibody-antigen interaction, varying numbers of hypervariable regions may act together to provide the net affinity. In each of the three antibodies we have examined, we have identified one or two CDRs as candidates for important binding sites.
Peptides that block the interaction of ICAM-1 with Mac-1 have not previously been reported. Peptides derived from factor X and related peptides derived from filamentous hemagglutinin have been shown to prevent factor X binding to cells, presumably via Mac-1 (28,29). However, the I-domain is not the primary binding site for factor X (12), and these peptides have not been shown to affect ICAM-1 binding. There is precedent, however, for peptide inhibition of ICAM-1-␤ 2 -integrin interaction. A peptide from ICAM-2 has been shown to bind to LFA-1 and inhibit endothelial cell adhesion (at 100 g/ml), presumably via ICAM-1 (30). Peptides from ICAM-1, domains 1 and 2, have also been shown to inhibit ICAM-1-mediated adhesion to unknown ligands (at 100 M) (31).
In conclusion, we have utilized sequence information from anti-Mac-1 antibodies to derive the first peptide Mac-1 antagonists. These findings support a growing body of evidence that protein-protein interactions may depend disproportionately on one site or pocket.