Identification of a Negatively Charged Peptide Motif within the Catalytic Domain of Progelatinases That Mediates Binding to Leukocyte β2 Integrins*

The αMβ2 integrin of leukocytes can bind a variety of ligands. We screened phage display libraries to isolate peptides that bind to the αM I domain, the principal ligand binding site of the integrin. Only one peptide motif, (D/E)(D/E)(G/L)W, was obtained with this approach despite the known ligand binding promiscuity of the I domain. Interestingly, such negatively charged sequences are present in many known β2 integrin ligands and also in the catalytic domain of matrix metalloproteinases (MMPs). We show that purified β2 integrins bind to pro-MMP-2 and pro-MMP-9 gelatinases and that that the negatively charged sequence of the MMP catalytic domain is an active β2 integrin-binding site. Furthermore, a synthetic DDGW-containing phage display peptide inhibited the ability of β2 integrin to bind progelatinases but did not inhibit the binding of cell adhesion-mediating substrates such as intercellular adhesion molecule-1, fibrinogen, or an LLG-containing peptide. Immunoprecipitation and cell surface labeling demonstrated complexes of pro-MMP-9 with both the αMβ2 and αLβ2 integrins in leukocytes, and pro-MMP-9 colocalized with αMβ2 in cell surface protrusions. The DDGW peptide and the gelatinase-specific inhibitor peptide CTTHWGFTLC blocked β2 integrin-dependent leukocyte migration in a transwell assay. These results suggest that leukocytes may move in a progelatinase-β2 integrin complex-dependent manner.

Phage Display-Phage display selections were made using a pool of random peptides CX 7-10 C and X 9 -10 , where C is a cysteine and X is any amino acid (14,25). Briefly, the ␣ M I domain-GST or the GST fusion protein was immobilized on microtiter wells at 20 g/ml concentrations, and the wells were blocked with BSA. The phage library pool was first subtracted on wells coated with GST, and then unbound phage was transferred to ␣ M I domain-GST-coated wells in 50 mM Hepes, 5 mM CaCl 2 , 1 M ZnCl 2 , 150 mM NaCl, 2% BSA, pH 7.5. After three rounds of subtraction and selection, individual phage clones were tested for binding specificity, and the sequences of the phage that specifically bound to the I domain were determined (14).
Peptide Biosynthesis and Chemical Synthesis-The phage peptides were initially prepared biosynthetically as intein fusions. The DNA sequences encoding the peptides were PCR-cloned from 1-l aliquots of the phage-containing bacterial colonies that were stored at Ϫ20°C. The forward primer was 5Ј-CCTTTCTGCTCTTCCAACGCCGACGGGGCT-3Ј, and the reverse primer was 5Ј-ACTTTCAACCTGCAGTTACCCAG-CGGCCCC-3Ј. The PCR conditions included initial denaturation at 94°C for 2 min followed by 30 cycles of 94°C for 30 s, 55°C for 30 s, and 72°C for 30 s. The PCR products were purified using the Qiagen nucleotide removal kit. They were then digested with SapI and PstI restriction enzymes and ligated to a similarly digested and phosphatase-treated pTWIN vector (New England Biolabs). Correct insertions were verified by DNA sequencing. Intein fusion proteins were produced in Escherichia coli strain ER2566 and affinity-purified on a chitin column essentially as described (26). The peptide was cleaved on the column, eluted, and finally purified by high pressure liquid chromatography. Chemical peptide synthesis was done using Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry as described, and the sequences were verified by mass spectroscopy (26).
Phage Binding Assay-Phage (10 8 infective particles/well) in 50 mM Hepes, 5 mM CaCl 2 , 1 M ZnCl 2 , 0.5% BSA, pH 7.5, were added to microtiter wells coated with I domain-GST fusion or GST (20 ng/well). The phages were allowed to bind in the absence or presence of a competitor peptide (15 M) for 1 h followed by washings with PBS containing 0.05% Tween 20. The bound phage was detected using 1:3000 dilution of a peroxidase-labeled monoclonal anti-phage antibody (Amersham Biosciences) and o-phenylenediamine dihydrochloride as a substrate. The reactions were stopped by addition of 10% H 2 SO 4 , and the absorbance was read at 492 nm using a microplate reader.
Pepspot-The peptides were synthesized on cellulose membranes as described (27). The membrane was blocked with 3% BSA in TBS containing 0.05% Tween 20, and incubated with 0.5-5 g/ml ␣ M I domain for 2 h at room temperature. The DDGW peptide was used as a competitor at a 50 M concentration. Bound ␣ M I domain was detected using the monoclonal antibody LM2/1 (1 g/ml) or MEM170 (5 g/ml) and peroxidase-conjugated rabbit anti-mouse antibody (1:5000 dilution) followed by chemiluminescence detection.
Expression and Purification of GST Fusion Proteins-The ␣ L , ␣ M , and ␣ X I domains were produced as GST fusion proteins in E. coli strains BL 21 or JM109 and purified by affinity chromatography on glutathione-coupled beads (30,31). GST containing CTT in the C terminus was constructed using the protocols described for LLG-C4-GST (14), and glutathione-coupled beads were employed for purification. The purity of the GST fusion proteins was confirmed by SDS-PAGE with Coomassie Blue staining and Western blot analysis. For pepspot analysis, GST was cleaved from the ␣ M I domain with thrombin.
Coprecipitation of ␤ 2 Integrin and Progelatinases-Serum-free conditioned medium containing pro-MMP-2 and pro-MMP-9 was collected from human HT-1080 fibrosarcoma cells grown in the presence of 100 nM phorbol ester 4␤-phorbol 12,13-dibutyrate (PDBu) (Sigma) overnight at ϩ37°C. A 500-l volume of the supernatant was incubated with 100 ng of GST-␣ M , GST-␣ L , or GST-␣ X I domain or ␣ M ␤ 2 integrin for 3 h at 25°C. GST and GST-LLG-C4 were used to determine nonspecific binding. CTT, STT, LLG-C4, and ICAM-1 were used as competitors at a 200 g/ml concentration and the antibodies LM2/1 and TL3 at 40 g/ml. After a 1-h incubation at ϩ4°C, complexes of I domain and gelatinases were pelleted with glutathione-Sepharose. Integrin complexes were captured by incubating first with the OKM10 antibody for 3 h at ϩ4°C and then with protein G-Sepharose for 1 h. After centrifugation and washing, samples were analyzed by gelatin zymography on 8% SDS-polyacrylamide gels containing 0.2% gelatin (32).
Effect of Peptides on Pro-MMP-9 Release from Cells-THP-1 cells (40,000/100 l) were incubated in serum-free RPMI medium for 48 h in the absence or presence of 200 M peptide as described in the text. Aliquots of the conditioned media were analyzed by gelatin zymography.
Interaction between CTT and Pro-MMP-9 -CTT-GST and GST control (5 g/well) were coated overnight on 96-well microtiter plates in 50 l of TBS followed by blocking of the wells by BSA. Pro-MMP-9 or APMA-activated form (80 ng/well) was incubated in the absence or presence of competitors for 2 h in 50 M Hepes buffer containing 1% BSA, 5 mM CaCl 2 , and 1 M ZnCl 2 , pH 7.5. After washing, bound MMP-9 was determined with anti-MMP-9 and horseradish peroxidaseconjugated anti-mouse IgG as described above. To examine complexing of CTT with pro-MMP-9 in cell culture, THP-1 cells were activated with PDBu for 30 min and then incubated with CTT, W3 A CTT, or Inh1 (each 200 M) at ϩ37°C in serum-free medium. Samples were taken from the media at 0 -5-h time points and analyzed by zymography and Western blotting with polyclonal anti-MMP-9 antibodies. Experiments with HT-1080 cells were performed similarly except that the medium samples were collected after 6 h.
Cell Surface Labeling, Immunoprecipitation, and Immunoblotting-Non-activated or PDBu-activated THP-1 cells (1 ϫ 10 7 ) were subjected to surface labeling using periodate tritiated sodium borohydride (33). The 3 H-labeled cells were lysed with 1% (v/v) Triton X-100 in PBS, clarified by centrifugation, and precleared with protein G-Sepharose. The lysate was immunoprecipitated with polyclonal anti-MMP-9, ␣ M (OKM-10), or ␤ 2 (7E4) antibodies. After a 1-h incubation at ϩ4°C together with protein G-Sepharose, immunocomplexes were pelleted, washed, and resolved on 8 -16% SDS-PAGE gels (Bio-Rad). The gels were treated with an enhancer (Amplify, Amersham Biosciences), dried, and exposed. Non-labeled THP-1 cells (1 ϫ 10 7 ) were similarly lysed and immunoprecipitated as above. The samples were resolved on 4 -15% SDS-PAGE gels and transferred to nitrocellulose membranes. Immunodetection was performed with ␣ M (MEM170) antibody (10 g/ ml) followed by peroxidase-conjugated anti-mouse IgG and chemiluminescence detection (Amersham Biosciences). The membranes were Immunofluorescence-Immunofluorescence was performed on resting cells, or the cells were activated with PDBu for 30 min. A portion of the cells was treated with ICAM-1 or CTT to block ␤ 2 integrins or gelatinases, respectively. Cells were bound to poly-L-lysine-coated coverslips, fixed with methanol for 10 min at Ϫ20°C or with 4% paraformaldehyde for 15 min at ϩ4°C, and permeabilized with 0.1% Triton X-100 in PBS at room temperature for 10 min followed by several washings. The coverslips were incubated with rabbit anti-MMP-9 polyclonal and mouse anti-␣ M (OKM-10) antibodies diluted 1:500. After washing with PBS, the secondary antibodies, rhodamine (TRITC)-conjugated porcine anti-rabbit or fluorescein isothiocyanate-conjugated goat anti-mouse (FabЈ) 2 (Dakopatts A/S, Copenhagen, Denmark) were incubated at a 1:1000 dilution for 30 min at room temperature. The samples were mounted with Mowiol, incubated in the dark for 2 days, and examined by a confocal microscope (Leica multiband confocal imagine spectrophotometer) at a ϫ400 magnification or a fluorescence microscope (Olympus Provis 70) at a ϫ60 magnification.
Cell Adhesion and Migration-Fibrinogen and ICAM-1-Fc were coated at 40 g/ml in TBS at ϩ4°C. Peptides (2 g/well) were coated in TBS containing 0.25% glutaraldehyde at ϩ37°C. The wells were blocked with 1% BSA in PBS. THP-1 cells (50,000/well) with or without PDBu activation were added in 0.1% BSA/RPMI medium in the presence or absence of 200 M peptides or monoclonal antibodies at 50 g/ml. After 30 -35 min the wells were washed with PBS to remove non-adherent cells, and the adhesive cells were quantitated by a phosphatase assay. The cell migration assay was conducted using transwell migration chambers (8 m pore size, Costar) in serum-containing medium as described (14). Briefly, the membranes were coated on the upper and lower surface with 40 g/ml GST, LLG-C4-GST, or left uncoated. The wells were blocked with 10% serum-containing medium for 2 h. THP-1 cells (50,000/100 l) or HT1080 (20,000/100 l) were preincubated with the peptides for 1 h before transfer to the upper chamber. The lower chamber contained 500 l of the medium without the peptides. The cells were allowed to migrate to the lower surface of the membrane for 16 h and then stained with crystal violet and counted.

Identification of the ␣ M I Domain-binding Peptide Motif
D/E/D/E-G/L-W-By using phage peptide display libraries, we selected peptides that interact with the ␣ M I domain. GSTbinding phage were first eliminated on GST-coated wells, and the unbound phage preparations were incubated on ␣ M I domain GST fusion protein-coated wells. The ␣ M I domain-binding phage were enriched by three rounds of panning, and the peptide sequences were determined. With the exception of one linear peptide, the peptides were derived from the cyclic CX 7 C and CX 8 C libraries. The I domain-binding sequences showed only one conserved motif, a somewhat unexpected finding in terms of the known ligand binding promiscuity of the I domain. The bound peptides contained two consecutive negatively charged amino acids, i.e. glutamic and/or aspartic acids, followed by glycine and tryptophan residues (Fig. 1A). The consensus (D/E)(D/E)(G/L)W determined by this approach was clearly different from LLG-C4 and other ␤ 2 integrin-binding peptides reported so far.
We first prepared the phage display peptides as intein fusion proteins, from which the peptides were cleaved. This allowed us to test rapidly the peptide solubility and the binding specificity before large scale chemical peptide synthesis. The peptides were cloned using oligonucleotide primers that amplify the peptide library insert from the phage vector. Consequently, all the peptides prepared contain the vector-derived sequences ADGA and GAAG in the N and C termini, respectively. Phage binding experiments using soluble peptides as competitors in-dicated that the peptides bearing the two adjacent negative charges bound to a common site (not shown). We chose the peptide ADGA-CILWMDDGWC-GAAG (DDGW) for further experiments as this peptide showed strong binding and was highly soluble in aqueous buffers (soluble in 50 mM NaOH at Ͼ10 mM concentrations). The peptide was also prepared by chemical synthesis. The phage bearing the DDGW sequence avidly bound to the ␣ M I domain, and this was readily inhibited by low concentrations of the DDGW peptide, but only marginally affected by the LLG-C4 peptide, indicating different binding sites for DDGW and LLG-C4 (Fig. 1B). Control phage bearing other peptide sequences did not bind. The DDGWbearing phage also showed also specific binding to the ␣ L I domain that was inhibitable by DDGW, but the interaction was weaker than with the ␣ M I domain ( Fig. 1C and data not shown). No binding was observed with the ␣ X I domain or GST used as a control (Fig. 1C).
Characterization of DELW Sequence on the Catalytic Domain of Gelatinases That Mediates Interaction with the ␤ 2 Integrin I Domains-We searched protein databases for matches to the novel (D/E)(D/E)(G/L)W motif. One of the phage library-derived peptides, CPEELWWLC, was highly similar to the DELW(S/ T)LG sequence present on the catalytic domain of MMP-2 and MMP-9 gelatinases (Fig. 1A). DELW-like sequences with double negative charges are also present in other secreted MMPs but not in the membrane-type MMPs such as MMP-14.
No MMP has been reported to bind to the leukocyte ␤ 2 integrins. We therefore set out to study whether MMP-9 in particular could be a ligand of the ␤ 2 integrins as MMP-9 gelatinase is the major leukocyte MMP and is induced during ␤ 2 integrin activation. As a first step, we synthesized the whole pro-MMP-9 sequence as overlapping 20-mer peptides on a pepspot membrane. Binding assays with the ␣ M I domain revealed a single active peptide that located to the MMP-9 catalytic domain (Fig. 1D). No binding was observed when the I domain was omitted, and the membrane was probed with antibodies only. The sequence of the I domain-binding peptide was QG-DAHFDDDELWSLGKGVVV, and it contained the binding motif identified by phage display (Fig. 1D).
The active MMP-9 peptide contained four consecutive amino acids with negative charges, DDDE. To study the importance of these residues, the aspartic and glutamic acid residues that were closest to the tryptophan were replaced by alanines. At the same time the peptide length was shortened to 15-mer. The alanine mutagenesis significantly abrogated I domain binding on the pepspot filter; the OD value dropped from 2010 to 476 (Table I). To study whether the negatively charged peptide from other MMPs is also active, we synthesized the corresponding 15-mers and the double alanine mutations. Sequences from MMP-1-3, -7-9, and -13, but not the membrane-anchored MMP-14 (MT1-MMP), could bind the ␣ M I domain. Alanine mutations always decreased the binding.
We did similar pepspot analysis for some of the known I domain ligands, which contain (D/E)(D/E)(G/L)W-like sequences. Peptides derived from myeloperoxidase, catalase, thrombospondin-1, and complement protein iC3b strongly bound the I domain in this assay, and the double alanine mutation caused a loss of binding (see Table I). Of the three iC3b peptide permutations tested, ARSNLDEDIIAEENI was the active one. The acidic residues were followed by a hydrophobic isoleucine cluster in this peptide. The soluble DDGW pro-MMP-9. E, alanine-mutated and truncated peptides were synthesized on a pepspot filter and probed with the recombinant ␣ M I domain (5 g/ml). Bound I domain was measured using monoclonal antibody MEM170 (5 g/ml) followed by horseradish peroxidase-conjugated anti-mouse secondary antibody and ECL detection. The binding was quantified by densitometric scanning. The bars show ␣ M I domain binding to single peptide spots as arbitrary optical density units/mm 2 . Similar results were obtained in three independent experiments. peptide efficiently inhibited the binding of this peptide to the I domain. Weaker I domain binding was observed with one complement factor H-derived peptide and one fibronectin-derived peptide. Peptides derived from ICAMs-1-3, neutrophil inhibitory factor, Cyr61, fibrinogen, GP1b, factor X, or E-selectin lacked activity.
Alanine-scanning mutagenesis of the DDGW peptide with the pepspot system similarly indicated the importance of the aspartic acid residues for I domain binding (Fig. 1E). Alanine mutations of the glycine or either one of the tryptophan residues also inactivated the peptide. Mutations of the isoleucine, leucine, or methionine residues were tolerated. Deletion of the ADGA sequence from the N terminus had no effect on I domain binding, but removal of the C-terminal GAAG sequence abolished the binding. As the peptides were immobilized via the C terminus on the filter, a sufficient linker sequence such as GAAG seemed important. We also tested a series of truncated cyclic peptides to identify the shortest active sequence. This analysis showed that ADGA-CEDGWC-GAAG but not ADGA-CDDGWC-GAAG was the minimal peptide that supported ␣ M I domain binding. The longer side chain of glutamate compared with aspartate is probably required to bring the negatively charged carboxyl group in the correct position for I domain binding.
Progelatinases Bind to Purified ␣ M ␤ 2 and ␣ L ␤ 2 Integrins and Their I Domains-We next used a microtiter well-based sandwich assay to study gelatinase binding to purified integrins. Progelatinases bound in a concentration-dependent manner to the coated ␣ M ␤ 2 integrin ( Fig. 2A). Curiously, MMP-2 and MMP-9 lost the integrin binding ability after activation by trypsin or APMA. The binding of pro-MMP-2 and pro-MMP-9 was observed with both ␣ M ␤ 2 and ␣ L ␤ 2 integrins and their corresponding I domains (Fig. 2, B and C). No binding was detected on the ␣ X I domain or the ␣ 1 ␤ 1 and ␣ 3 ␤ 1 integrins.
The DDGW peptide was an efficient inhibitor, and it inhibited pro-MMP-9 binding to the ␣ M I domain with an IC 50 value of 20 M (Fig. 3, A and B). To demonstrate that the negative charges of aspartic acids are essential for the peptide activity, the peptide ADGACILWMKKGWCGAAG (KKGW) containing lysines in place of aspartic acids was prepared. As expected, the KKWG peptide was inactive and did not compete with pro-MMP-9 binding. We were also interested in testing lovastatin, as its binding site in the ␣ L I domain is known (34,35). Lovastatin was not able to compete with pro-MMP-9 even at a high concentration.
Pro-MMP-9 bound like a true integrin ligand, as the cation chelator EDTA (5 mM) nearly completely prevented the binding (Fig. 3, C and D). For background measurement in the sandwich assay, we used antibodies alone (control) or coating with ICAM-1 or wild type GST. The gelatinase-binding peptide CTT (200 M) inhibited pro-MMP-9-integrin interaction with the same efficiency as EDTA did. The control peptides STTHWG-FTLS (STT) and CTTHAGFTLC (W3 A CTT), which lack gelatinase inhibitory activity (26), were without effect. A nonpeptide chemical MMP inhibitor (Inh1) also prevented pro-MMP-9 binding. As EDTA inhibits both the gelatinase and the integrin, we used integrin-blocking antibodies and ligand peptides to demonstrate the specific binding activity of ␤ 2 integrin.  The known ligand-binding blocking antibodies MEM170, MEM83, and LM2/1 inhibited pro-MMP-9 binding. A control antibody TL3 had no effect. The I domain binding peptide LLG-C4 showed a partial inhibitory effect. RGD-4C, a ligand of ␣ V integrins, served as control peptide and had no effect on pro-MMP-9 binding. The purity of the integrins was typically more than 90% and that of I domains 95%, making it unlikely that progelatinases would bind to impurities in the preparations. Progelatinase-integrin complexes were also obtained by coprecipitation experiments using HT1080 conditioned medium as a source of pro-MMP-9 and pro-MMP-2, which were analyzed by zymography. The progelatinases coprecipitated with ␣ M ␤ 2 integrin or ␣ M I domain GST fusion protein when these were used as a bait. The integrin added to the medium was immunoprecipitated with the ␣ M antibody OKM10 (Fig. 4A). The ␣ M I domain GST protein was pulled down with glutathione beads (Fig. 4B). CTT but not STT had an inhibitory effect. Inhibition of the I domain by LM2/1, ICAM-1, or LLG-C4 also affected the pull-down of progelatinases. GST control did not coprecipitate the gelatinases. No active forms of gelatinases were found to coprecipitate with ␣ M ␤ 2 or the I domain, when APMA-treated HT-1080 medium or APMA-activated MMP-9 was used (not shown).
As the gelatinase inhibitors CTT and Inh1 prevented the binding of pro-MMP-9 to the integrin, it can be anticipated that CTT and Inh1 avidly bind to pro-MMP-9. To gain more insight into this, we examined binding of pro-MMP-9 to immobilized CTT peptide. Pro-MMP-9 specifically bound to the CTT-GST fusion protein (Fig. 5A) but not to LLG-C4-GST. CTT and Inh1 at 100 M concentrations effectively competed in binding, but W3 A CTT did not. The pro-MMP-9 preparation did not contain detectable amounts of active MMP-9 on zymography analysis, and after pro-MMP-9 activation with APMA, the CTT-GST binding increased. CTT and Inh1 could also bind to pro-MMP-9 secreted into the medium of PDBu-activated THP-1 leukemic cells (Fig. 5B) or HT1080 fibrosarcoma cells (not shown). A time-dependent reduction in the gelatinolytic activity of pro-MMP-9 was observed with CTT (1st panel) and Inh1 (3rd panel) but not with the W3 A CTT peptide (4th panel). Western blot analysis indicated that CTT does not decrease the secretion of pro-MMP-9 by the cells (2nd panel). Furthermore, the CTT complex was reversible and disappeared after repeated freezing and thawing of the samples.
Demonstration of a Cell-surface Complex between Progelatinases and ␤ 2 Integrins-To study whether the progelatinases occur in a complex with the ␤ 2 integrins on the leukocyte surface, we performed immunoprecipitation and colocalization studies. First, we examined THP-1 monocytic leukemia cells in the resting state and after induction by PDBu, which mimics leukocyte activation in vivo. THP-1 cell stimulation with PDBu led to up-regulation of MMP-9 (data not shown). The cell surface glycoproteins of THP-1 cells were labeled with tritium [ 3 H] followed by immunoprecipitation with ␤ 2 integrin and MMP-9 antibodies. In the PDBu-activated cells, the ␣ M chain antibody OKM10 and ␤ 2 chain antibody 7E4 immunoprecipitated two 3 H-labeled proteins corresponding to the integrin ␣ M chain (165 kDa) and ␤ 2 chain (95 kDa) (Fig. 6A, lanes 9 and 10). Importantly, polyclonal MMP-9 antibodies immunoprecipitated the same two integrin chains (lane 7). In non-activated cells, essentially no coprecipitation of ␣ M and ␤ 2 was observed with MMP-9 antibodies, although the ␣ M and ␤ 2 chains were present. The coprecipitation of the integrin chains by MMP-9 antibodies was prevented by the CTT peptide (lane 8). The control antibody (TL3) did not precipitate any proteins.
With the 3 H-labeled cells, we did not observe any band corresponding to pro-MMP-9, perhaps because the carbohydrates of pro-MMP-9 are poorly labeled. We therefore analyzed the PDBu-activated THP-1 cells by Western blotting (Fig. 6B). Pro-MMP-9 was readily immunoprecipitated with antibodies against MMP-9, ␣ M , or ␣ L but not by the control antibody. MMP-9 antibodies in turn were able to immunoprecipitate the ␣ M but not the ␣ L chain. MMP-2 antibodies similarly coprecipitated ␣ M but not ␣ L . When the cell lysate was precleared with the ␣ M antibody OKM-10, the amount of immunoprecipitated ␣ M and pro-MMP-9 clearly decreased (lane 6). Preclearing with the ␣ L antibody TS2/4 did not significantly remove ␣ M or pro-MMP-9 but abolished the ␣ L precipitation (lane 7).
As THP-1 cells do not express high amounts of the ␣ L chain (11), we examined the Jurkat T cell line, which expresses more ␣ L than ␣ M (28). We observed a significant immunoprecipitation of ␣ L by MMP-9 and MMP-2 antibodies after PDBu activation (Fig. 6C). Furthermore, the ␣ L antibody coprecipitated more pro-MMP-9 in comparison to the ␣ M antibody. No pro-MMP-9 coprecipitated with MMP-2 antibodies in Jurkat or THP-1 cells.
Pro-MMP-9 and ␣ M ␤ 2 were found to colocalize on the cell surface following PDBu activation of THP-1 cells as studied by fluorescence and confocal microscopy (Fig. 7, A and B, respectively). By using a higher magnification, colocalization was

FIG. 2. Binding of progelatinases to purified integrins and I domains.
A, pro-MMP-9, pro-MMP-2, or their trypsin-activated forms, at the concentrations indicated, were allowed to bind to ␣ M ␤ 2 integrincoated wells. Appropriate MMP antibodies were used to determine binding. The results in this and other figures are represented as the means Ϯ S.D. from triplicate wells. B, binding of pro-MMP-9 (80 ng/ well) was examined on microtiter wells coated with an integrin (␣ L ␤ 2 , ␣ M ␤ 2 , ␣ 1 ␤ 1 , and ␣ 3 ␤ 1 ) or an I domain (␣ L , ␣ M , ␣ X ). The binding was determined using anti-MMP-9 antibody. C, pro-MMP-2 (80 ng/well) was allowed to bind to ␣ L ␤ 2 , ␣ M ␤ 2 , ␣ 1 ␤ 1 , or the I domains ␣ L , ␣ M , or ␣ X . The binding was determined using an anti-MMP-2 antibody. primarily seen in cell surface clusters (Fig. 7B) and to a lesser extent on areas where cells contacted each other (not shown). We believe that the MMP-9 colocalizing with ␣ M ␤ 2 is the pro-MMP-9, as the activated MMP-9 did not bind to ␣ M ␤ 2 . Without PDBu activation, there was hardly any colocalization of pro-MMP-9 and ␣ M ␤ 2 . The secondary antibodies did not stain the cells when the primary antibodies were omitted (data not shown). When the cells were preincubated with the CTT peptide or recombinant soluble ICAM-1 to block pro-MMP-9 or ␣ M ␤ 2 , the cell surface clusters did not form, and the pro-MMP-9-␣ M ␤ 2 colocalization was not observed (not shown). ProMMP-9-␣ M ␤ 2 colocalization was also observed on Jurkat cells following phorbol ester stimulation (not shown).
Blocking the Progelatinase-␤ 2 Integrin Complex with DDGW Releases Cell-bound Pro-MMP-9 and Inhibits Cell Migration but Not Adhesion-As the DDGW peptide is an integrin ligand, one of the questions was whether it can support adhesion of leukocytes. We studied adhesion of human myelomonocytic THP-1 cells on immobilized glutaraldehyde-polymerized peptide. Phorbol ester-activated cells efficiently bound to the DDGW peptide, whereas there was no binding in the absence of cell activation (Fig. 8A). As a positive control, the recombinant intein-produced ADGA-CPCFLLGCC-GAAG peptide supported adhesion, but unlike the DDGW peptide, it also supported adhesion in the absence of integrin activation. The acute myeloid leukemic cell line OCI/AML-3 also avidly adhered to DDGW, whereas human fibrosarcoma HT1080 cells which lack ␤ 2 integrins did not (not shown). As THP-1 cells were able to adhere on DDGW, we next studied the effect of the peptide on ␤ 2 integrin-dependent adhesion to fibrinogen and ICAM-1. Interestingly, DDGW did not block cell adhesion to fibrinogen, whereas the LLG-C4 peptide blocked the adhesion as reported previously (Fig. 8B). Similarly, DDGW did not block the bind-ing of recombinant ␣ M I domain to immobilized fibrinogen (not shown). DDGW did not either block cell adhesion on ICAM-1-Fc fusion protein. As a control, the blocking antibody 7E4 against ␤ 2 integrins prevented the ICAM-1 binding, indicating that the THP-1 cells bound in a ␤ 2 integrin-dependent manner (Fig.  8C). We also found no blocking effect of DDGW on THP-1 adhesion to LLG-C4-GST fusion protein (not shown).
The second question raised by these studies was whether the DDGW peptide can release cell-bound pro-MMP-9. When THP-1 cells were cultured for 48 h in the presence of DDGW, an increase of pro-MMP-9 level was observed in the conditioned medium as studied by gelatin zymography (Fig. 8D). The peptide increased both monomeric and dimeric pro-MMP-9 in the culture medium. In contrast, CTT slightly decreased or inhibited active pro-MMP-9. KKGW and W3 A CTT had no effect.
Finally, we studied the role of the progelatinase-␤ 2 integrin complex in leukocyte migration using a transwell assay in which leukocyte migration can be adjusted by the choice of coated matrix or ligand protein. We tested that the CTT, LLG, and DDGW peptides are not toxic to the THP-1 cells in a 48-h time frame at Ͼ200 M concentrations using a 3-(4,5dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay. By using transwells coated with 10% serum in cell culture medium, we first studied the effect of peptides on the basal migration of THP-1 cells in the absence of any stimulus by phorbol ester or an adhesive matrix. Under such conditions, CTT, LLG-C4, or the gelatinase inhibitor Inh1 at a 200 M concentration had no effect on THP-1 migration indicating no active involvement of gelatinases or ␤ 2 integrins (Fig.  9A). We have shown previously that when the transwells are coated with LLG-C4-GST fusion protein, THP-1 cells adhere and migrate in a ␤ 2 integrin-dependent manner (14). Thus, FIG. 3. Inhibitors of the pro-MMP-9/␤ 2 integrin complex. A, ␣ M and ␣ L I domain-GST fusions were immobilized on microtiter wells. Pro-MMP-9 (100 ng/ well) was added in the presence or absence of various peptides (200 M) or lovastatin (100 M) in a buffer containing 0.5% BSA. Pro-MMP-9 binding was detected using a monoclonal antibody against MMP-9. The results are shown as percent binding compared with binding in the absence of inhibitors (100%), and no pro-MMP-9 was added (0%). B, DDGW peptide blocks pro-MMP-9 binding to ␣ M in a dose-dependent manner. The assay was done similarly as in A, except various concentrations of peptides were added to compete for binding. All samples were assayed as triplicates, and results shown are means Ϯ S.D. from a representative experiment. C, pro-MMP-9 binding to ␣ M ␤ 2 and ␣ L ␤ 2 was examined in the absence and presence of EDTA (5 mM), MMP inhibitor-1 (100 M), CTT (200 M), STT (200 M), MEM170 (40 g/ml), or control TL3 antibody (40 g/ml). The background of primary and secondary antibodies was measured by omitting pro-MMP-9 from the wells or by coating with ICAM-1. D, pro-MMP-9 binding to purified ␣ M and ␣ L I domain GST fusion proteins or wild type GST was studied in the absence or presence of competitors as indicated. Control shows the background when pro-MMP-9 was omitted. transwells were coated with LLG-C4-GST fusion protein or GST alone. Both the DDGW and CTT peptide, but not KKGW, inhibited the migration of THP-1 cells on the LLG-C4-GST substratum (Fig. 9B). The soluble LLG-C4 peptide also blocked the migration. In the presence of GST coating, cell migration was negligible. To verify that the effect of DDGW peptide was ␤ 2 integrin-dependent, HT1080 fibrosarcoma cells lacking these integrins were allowed to migrate in the presence of CTT, DDGW, KKGW, or LLG-C4. Of these peptides, only CTT was capable of inhibiting cell migration (Fig. 9C).

DISCUSSION
Analysis of ␣ M I domain-binding peptides led to the finding that MMPs, particularly the MMP-9 and MMP-2 progelatinases, are potent ␤ 2 integrin ligands. Our studies show that pro-MMP-9, the major MMP of activated leukocytes, is colocalized with the ␤ 2 integrin on the cell surface. Cell surface labeling and coimmunoprecipitation further demonstrate the occurrence of the complex in leukemic cell lines. Finally, we have found evidence that this proteinase-integrin complex plays a role in migration of the leukemic cells.
Although phage display has been extensively used with whole integrins (14, 36 -44), to the best of our knowledge, this is the first successful phage display selection on an isolated integrin I domain. We could enrich only one binding motif even though the ␣ M I domain can bind a variety of ligands (1-3). The The immunoprecipitated samples were resolved on an 8 -16% polyacrylamide gel, and the film was exposed for 3 days. Lanes 1-4 are from non-activated cells and lanes 6 -10 from PDBu-activated cells. Lane 5 shows molecular weight markers. B, lysates from PDBu-activated THP-1 cells were immunoprecipitated with integrin or MMP antibodies followed by Western blotting with ␣ M (OKM10), ␣ L (TS2/4)n or MMP-9 antibodies. Preclearings of the cell lysates were done using ␣ M (lane 6) and ␣ L (lane 7) antibodies. C, lysates from PDBu-activated Jurkat cells were subjected to immunoprecipitation followed by blotting with the ␣ L (MEM83) and MMP-9 antibodies. peptide motif we isolated could not compete with the ICAM-1, fibrinogen, or LLG-C4 ligands. The success of phage display depends on the libraries used and the biopanning conditions. Our method favored "high affinity" interactions with the cyclic peptides yielding the (D/E)(D/E)(G/L)W motif. Interestingly, this motif shows a high degree of similarity to the CWDD(G/ L)WLC peptide isolated by phage display as an RGD sequencebinding peptide (40). By recognizing the RGD ligand sequence, CWDDGWLC structurally and functionally behaves like a minimal integrin. Here we have identified the DDGW peptide in a reverse situation, as a ligand to integrin. However, the RGD sequence does not compete with the ␣ M I domain as the  GRGDSP peptide at a 1 mM concentration was unable to inhibit pro-MMP-9 binding to the I domain. 2 It will be interesting to see whether the DDGW peptide recognizes a positively and negatively charged sequence in the I domain.
Some hints for the binding site of DDGW come from the interaction of iC3b with ␣ M ␤ 2 integrin. Our pepspot analysis showed that the iC3b peptide ARSNLDEDIIAEENI, but not the control peptide ARSNLDAAIIAEENI, bound the ␣ M I domain, and the DDGW peptide blocked this binding. The DEDI-IAEENI sequence with multiple adjacent negative charges is required for efficient binding of iC3b to ␣ M ␤ 2 integrin (45). The binding site of complement protein iC3b in the I domain has been mapped indicating a role for the positively charged amino acid residue Lys 245 for iC3b binding (46). Mutation of this residue does not affect the binding of the fibrinogen recognition peptide (47). This may account for the inability of the DDGW peptide to inhibit ICAM-1 and fibrinogen-mediated cell adhesion. These findings suggest that the Lys 245 residue is the positively charged contact site for the DDGW peptide and the (D/E)(D/E)(G/L)W motif as well.
The pepspot analysis indicates that a class of ␤ 2 integrin ligands contains an active (D/E)(D/E)(G/L)W motif. These include the previously identified ␣ M ␤ 2 ligands iC3b, thrombospondin-1, and the enzymes myeloperoxidase and catalase (3)(4)(5)48). In our experiments, the peptides derived from several secreted MMPs, but not membrane-bound MT1-MMP, were also active. It is notable that the (D/E)(D/E)(G/ L)W motif is relatively conserved in the secreted members of the MMP family. Whether other MMPs in addition to the progelatinases can make ␤ 2 integrin complexes remains to be determined.
Finding of a dominant integrin-binding site in the catalytic domain of pro-MMP-9 was unexpected, because previous studies suggested an essential role for another MMP domain, the hemopexin domain, in integrin binding. The hemopexin domains mediated MMP-2 binding to the ␣ V ␤ 3 integrin (49 -51) and MMP-1 binding to the ␣ 2 ␤ 1 integrin (52,53). The cleaved hemopexin domain of MMP-2 has also been shown to occur in vivo and to inhibit angiogenesis (50). Understandably, phage peptide display and pepspot techniques have limitations, and only linear peptide sequences can be analyzed, not protein conformations. Thus, in the present study we cannot make conclusions of the function of separate MMP-9 domains in integrin binding, and it remains to be seen whether cleavage products of MMP-9, if present in vivo, can act as ␤ 2 integrin ligands. Our studies suggest that the peptide sequence from the catalytic domain is essential for the binding of full-length pro-MMP-9 to the ␤ 2 integrin, as the synthetic DDGW peptide could completely inhibit the integrin binding. It was important to use natural pro-MMP-9 because ␣ M ␤ 2 is known to bind to denatured proteins, and this may sometimes be the case for bacterially expressed proteins (4). Furthermore, we did not observe binding of an active MMP-2 or MMP-9 to the integrin, although the DELW(T/S)LG sequence should remain unchanged in the active enzyme. We rather found that AMPA and trypsin, activators of pro-MMP-9, released MMP-9 from THP-1 cells, apparently affecting the integrin complex. 3 These results suggest that the proenzyme presents the integrin-binding site more efficiently than the active enzyme, and the ␤ 2 integrin may even control the activation of the proenzyme.
In the three-dimensional structures of pro-MMP-2 and -9 (54, 55), the I domain-binding site is located in the vicinity of the zinc-binding catalytic sequence HEFGHALGLDH between the catalytic domain and the fibronectin type II repeats. This location suggests a mechanism for evading pro-MMP-9 inhibition by tissue inhibitors of MMPs or ␣ 2-macroglobulin. In the absence of inhibitors, the cell surface-localized pro-MMP-9 would be readily susceptible for activation and substrate hydrolysis, which may also occur in the presence of intact propeptide (20). On the other hand, because the binding site of the I domain is located in the vicinity of the catalytic groove, it also suggests an explanation for the blocking of MMP-9/␤ 2 integrin interaction by the small molecule MMP inhibitors such as CTT and Inh1.
The activity of the DDGW peptide in the THP-1 cell migration assay suggests an important function for the integrinprogelatinase complex in leukocyte migration. Obviously, we cannot exclude the possibility that the DDGW peptide blocks binding of other ligands than gelatinases and in this way inhibits the leukocyte migration. However, as the specific gelatinase inhibitor CTT also blocks the THP-1 cell migration, these results strongly suggest that the pro-MMP-9-␤ 2 integrin complex is the main target for DDGW. Interestingly, the DDGW peptide blocked THP-1 cell migration although it increased the level of pro-MMP-9 in the medium, which suggests that cell surface-bound rather than total MMP-9 level is a critical factor in cell migration.