Structural and functional analysis of the plasminogen activator inhibitor-1 binding motif in the somatomedin B domain of vitronectin.

Plasminogen activator inhibitor 1 (PAI-1) binds to the somatomedin B (SMB) domain of vitronectin (VN), a domain present in at least seven other proteins. In this study, we investigate the PAI-1 binding activity of these SMB homologs and attempt to more specifically localize the PAI-1 binding site within this domain. SMBVN and several of its homologs were expressed in Escherichia coli, purified, and tested for PAI-1 binding activity in a competitive ligand binding assay. Although recombinant SMBVN was fully active in this assay, none of the homologs bound to PAI-1 or competed with VN for PAI-1 binding. These inactive homologs are structurally related to SMBVN, having 33-45% sequence identity and containing all 8 cysteines at conserved positions. Thus, homolog-scanning experiments were conducted by exchanging progressively larger portions of the NH2- or COOH-terminal regions of active SMBVN with the corresponding regions of the inactive homologs. These experiments revealed that the minimum PAI-1-binding sequence was present in the central region (residues 12-30) of SMBVN. Alanine scanning mutagenesis further demonstrated that each of the 8 cysteines as well as Gly12, Asp22, Leu24, Try27, Tyr28, and Asp34 were critical for PAI-1 binding and were required to stabilize PAI-1 activity. These results indicate that the PAI-1 binding motif is localized to residues 12-30 of SMBVN and suggest that this motif is anchored in the active conformation by disulfide bonds.

Urokinase-type and tissue-type (t-PA) 1 plasminogen activators catalyze the conversion of the zymogen plasminogen into its active form plasmin (1,2). Plasmin is the primary enzyme of the fibrinolytic system and may also function in ovulation, embryonic development, inflammation, wound healing, angiogenesis, and neoplasia. The activities of t-PA and urokinasetype plasminogen activator in vivo are regulated by plasminogen activator inhibitor-1 (PAI-1) (3), a member of the serine protease inhibitor (Serpin) family (4). Although PAI-1 exists in both an active and a latent conformation (5)(6)(7)(8), it appears to be synthesized in the active form but is conformationally unstable in solution and rapidly decays into the more stable but inactive latent form (3). Recent studies suggest that the active to latent transition is caused by the insertion of the reactive center loop into ␤-sheet A (9). For example, variants of PAI-1 constructed to reduce the rate of exchange between the reactive center loop and ␤-strands 3C/4C have a significantly lower rate of transition into the latent form (i.e., they are more stable; Ref. 10). Furthermore, Berkenpas et al. (11) used random mutagenesis to demonstrate that a combination of mutations greatly prolongs the half-life of active PAI-1. Besides possibly reducing strand 4A insertion, these mutations result in markedly increased thermal stability of the resulting active recombinant inhibitor relative to that of active native PAI-1.
Interestingly, the binding of PAI-1 to vitronectin (VN) significantly stabilizes this labile inhibitor in its active conformation (12)(13)(14)(15) and may alter its specificity (16). Although VN binds to active PAI-1 with high affinity (12,13,16), little specific binding of the inactive latent form of PAI-1 to VN could be demonstrated (13,18,19). VN itself is a major component of plasma and also is present in many tissues (reviewed in Ref. 20). It appears to regulate a variety of processes including cell adhesion, complement activation, and fibrinolysis. This 75-kDa adhesive glycoprotein is composed of an NH 2 -terminal somatomedin B domain (SMB), a region containing a number of hemopexin-like repeats, and the heparin binding domain. Residues 1-40 of the SMB domain have been shown to contain the primary high affinity binding site in VN for active PAI-1 (14,21). The amino acid sequence of SMB predicts an intricate compact structure because it contains 8 cysteines that probably anchor the PAI-1 binding motif in its active configuration. Like intact VN, the isolated SMB domain has been reported to stabilize PAI-1 (14).
Several other proteins were recently identified that also contain SMB-like domains (22,23). Interestingly, megakaryocytestimulating factor (MSF) (22), the plasma cell membrane glycoprotein, PC-1 (24), autotaxin (ATX) (25), and the tumor cell surface antigen, gp130   (26), all have two tandem copies of a SMB-like domain at their NH 2 termini. SMB homologs also were identified in placental-specific protein 11 (PP11) (27), the T cell-specific protein, Tcl-30 (28), and the Drosophila scavenger receptor, dSR-CI (29). There is no information in the literature to indicate whether these SMB homologs also bind to PAI-1 with high affinity. Moreover, the structural requirements for the interaction between the SMB domain of VN (SMB VN ) and PAI-1 are not known, and the actual PAI-1 binding motif within SMB VN has not been defined.
In the present study, we express and purify the SMB domains of VN and several of its homologs and examine their PAI-1 binding activities. We show that the recombinant SMB domain of MSF does not bind to PAI-1. Based on this observation, we developed a domain swapping (i.e., homolog-scanning) approach and used it to localize the PAI-1 binding site to the central region of SMB VN . The essential residues for PAI-1 binding in SMB VN were further identified by site-directed mutagenesis, and a consensus PAI-1 binding motif was derived. These findings provide insights into the unique structure of SMB and suggest that this domain may mediate the binding of other proteins to cell surfaces and extracellular matrices.

EXPERIMENTAL PROCEDURES
Materials-Bovine serum albumin, casein, Triton X-100, Tween 80, Tris base, phenylmethylsulfonyl fluoride, dithiothreitol, and ampicillin were obtained from Sigma. Isopropyl-␤-D-thiogalactopyranoside was from Calbiochem (La Jolla, CA). Reagents for SDS-polyacrylamide gel electrophoresis (PAGE) were from Bio-Rad. XAR-5 x-ray film was from Eastman Kodak. Oligonucleotides and Taq-DNA polymerase were from Life Technologies Inc. Restriction enzymes and bacteriophage T4 ligase were from Boehringer Mannheim. PAI-1 was purified from the conditioned medium of a transformed human lung fibroblast cell line (SV40 WI38 VA13 2RA) using standard chromatographic techniques in the absence of SDS (6). Analysis of the final preparation by SDS-PAGE and staining with silver nitrate revealed a single protein with a molecular mass of 50,000 kDa. Activation of purified PAI-1 with 4 M guanidine hydrochloride was performed as described previously (5).
DNA Cloning and Mutagenesis-For protein expression, DNA was cloned into the pET22b vector (Novagen, Madison, WI) using the NcoI and XhoI sites to eliminate vector encoded residues. In this way, the expressed protein contained only the signal peptide and 6 histidines. The SMB VN domain was amplified from VN cDNA (30) by using the 5Ј oligonucleotide primer, 5Ј-GGCGATGGCCATGGACCAAGAGTCATG-CAAGGGC-3Ј, and the 3Ј oligonucleotide primer, 5Ј-GGCGCGCACTC-GAGCATGGGCTTGCACTCAGCCGTATAG-3Ј. PCR amplified products were digested with NcoI and XhoI and ligated into pET22b previously digested with the same restriction enzymes. The DNA for the SMB domain of MSF exon 3 (SMB MSF ) was synthesized using the DNA sequence inferred from the published amino acid sequence (22) and corrected for Escherichia coli preferred codon usage. The coding strand was 5Ј-GCCATGGAACTCAGTTGTAAAGGGCGGTGTTTTGTGAAAG-TTTTGAACGGGGGCGGGAATGTGATTGCGACGCGC-3Ј, and the complementary strand was 5Ј-CACTCGAGCATCTCTCGACAGAAAC-TTTACTAATCCGGACAACAACATTTATCATATTTTTTACA-TTGCGCGTCGCAATC-3Ј. These two oligonucleotides have overlapping 3Ј ends and were extended by PCR to yield the full-length gene for SMB MSF . The PCR product was then cloned using the TA cloning kit (Invitrogen, San Diego, CA). For protein expression, the cloned SMB MSF gene was released from the pCRII vector by NcoI and XhoI digestion and cloned into the pET22b expression vector. The DNA for the SMB homolog of PC-1 (SMB PC-1 ) and other homologs were also cloned into pET22b using the published sequence and expressed as above. Mutagenesis was performed either by PCR with mutations incorporated into PCR primers or by transformer site-directed mutagenesis (Clontech, Palo Alto, CA).
Protein Expression and Purification-E. coli BL21(DE3) cells were transformed with the various expression plasmids and grown at 37°C in 2 ϫ YT medium (15 g/liter Bacto-tryptone, 10 g/liter yeast extract, 5 g/liter NaCl) containing ampicillin (100 g/ml) to a density of approximately 2 ϫ 10 8 cells/ml. The cells were transferred to 30°C, induced by the addition of isopropyl-␤-D-thiogalactopyranoside (1 mM) for 8 h, and collected by centrifugation. The cells were resuspended in one-fourth volume of 30 mM Tris-HCl (pH 8.0) containing 20% (w/v) sucrose and incubated at room temperature for 10 min. Cells were then pelleted by centrifugation at 4,000 ϫ g for 10 min at 4°C, resuspended in one-tenth of the original volume of ice-cold 5 mM MgSO 4 , and stirred for 10 min in an ice bath. The supernatant after centrifugation (10,000 ϫ g for 20 min at 4°C) was the cold osmotic shock fluid from which the target peptide was purified with the His⅐Bind Resin and Buffer kit (Novagen). The eluted peptide was further dialyzed against phosphate-buffered saline and concentrated (approximately 10-fold) with Centriprep-3 (Amicon, Beverly, MA). The concentration of the purified protein was determined using the BCA protein assay reagent (Pierce).
Determination of PAI-1 Binding Activity-The PAI-1 binding activities of VN and recombinant SMB peptides were measured using a competitive binding assay as described (14). Briefly, increasing amounts of VN or related peptides were added to VN-coated wells in the presence of active PAI-1 (0.6 nM active PAI-1). After incubation overnight at 4°C, the wells were washed with phosphate-buffered saline, and bound PAI-1 was detected using rabbit anti-PAI-1 IgG followed by biotin-conjugated goat anti-rabbit IgG and avidin-phosphatase (Zymed, South San Francisco, CA). The reaction was developed using the PNPP substrate (Zymed), and the change in color was determined at 405 nm. The OD reading from duplicate wells was corrected by subtracting values for the binding of the antibodies to VN-coated wells incubated in the absence of PAI-1.
Stabilization of PAI-1 by Recombinant SMB Mutants-Active PAI-1 (50 ng/ml) was incubated in phosphate-buffered saline/bovine serum albumin/Tween for increasing lengths of time at 37°C in the presence of an excess of SMB or mutant peptides (100 g/ml). At the end of the incubation periods, the samples were analyzed for remaining PAI-1 activity by using the t-PA binding assay (31). Briefly, the samples were added to t-PA-coated wells (1 g/ml), and after incubation for 1 h at 37°C, the amount of PAI-1 bound to the well was measured by enzymelinked immunosorbent assay. Only active PAI-1 binds to t-PA. Thus, the extent of binding represents the percentage of PAI-1 activity remaining in the sample at the end of each incubation period.
Miscellaneous-Double-stranded plasmid DNA sequencing was performed by using the Sequenase version 2.0 DNA sequencing kit from U. S. Biochemical Corp. SDS-PAGE was performed in slab gels according to Laemmli (32). The upper stacking gel contained 4% acrylamide, whereas the lower resolving gel contained 15% acrylamide. Chemical reduction of samples was accomplished by adding dithiothreitol to the sample to a final concentration of 50 mM and boiling it for 3 min. The Rainbow-colored protein (low range) from Amersham Corp. was used as a molecular standard for SDS-PAGE. Reverse phase HPLC was performed using a C-18 column on a Waters 501 Unit. The active fraction was directly used for mass spectrometry analysis. In these experiments, the peptides were recovered in approximately 24% acetonitrile and 0.1% trifluoroacetic acid. Mass spectrometry was performed by the Mass Spectrometry Core Facility (The Scripps Research Institute, La Jolla, CA).

Expression and Analysis of Recombinant SMB-
The SMB VN domain (residues 1-41) and the domain from one of its homologs (the SMB MSF domain; residues 42-83; see Table I) were cloned into the pET22b vector by fusing their NH 2 termini to the pelB signal peptide and their COOH termini to a stretch of 6 histidines. The expressed products contained the pelB signal peptide and were thus directed to the periplasm of E. coli, facilitating proper folding of these cysteine-rich peptides. The recombinant peptides were purified from the periplasm of E. coli on a nickel-charged column. The majority of purified SMB VN peptide migrated as a single band when analyzed by reducing SDS-PAGE with an apparent molecular mass of approximately 8 kDa (data not shown). However, varying amounts of SMB VN dimers and trimers were also apparent. The molecular mass of SMB VN was also determined by mass spectrometry after further purification by HPLC, and the measured molecular mass was identical to the calculated molecular mass of 6 kDa (i.e., 5180 Da for SMB plus 822 Da for the 6 histidines). These results indicate that the signal peptide was cleaved correctly. The purified peptide was recognized by monoclonal antibodies raised against human VN purified from plasma. 2 These results suggest that the recombinant peptide was properly folded. Under identical conditions, SDS-PAGE revealed that the majority of the SMB MSF peptide migrated in bands of approximate molecular masses 13 and 20 kDa (data not shown), corresponding to the molecular mass of dimerized and trimerized SMB MSF . Comparison of the sequences of SMB VN and SMB MSF (Table I) predicted that they would have similar structures because the 8 cysteines in each peptide are identically positioned, they share greater than 40% homology, and most of the charged residues are conserved ( Table I).
The PAI-1 binding activities of the recombinant peptides were tested in the competitive ligand binding assay (14). In this assay, a constant amount of PAI-1 was added to VN-coated microtiter wells in the presence of increasing amounts of the recombinant SMB domains. After incubation, the amount of PAI-1 bound to the immobilized VN was determined by enzyme-linked immunosorbent assay. Fig. 1 shows that the purified recombinant SMB VN peptide competed with VN for PAI-1 binding, with half-maximal inhibition at approximately 0.04 g/ml or 6.6 nM. This concentration compares favorably with the concentration of plasma-derived VN or SMB required to inhibit PAI-1 binding to VN (i.e., 2-3 nM for 50% inhibition) (14,33). In contrast to these results with SMB VN , the related peptide (i.e., SMB MSF ) had no detectable PAI-1 binding activity in the competitive ligand binding assay under the conditions employed. Because dimeric and trimeric SMB VN continue to bind PAI-1, 2 these differences in activity do not appear to result from the different forms of the domain present in the two preparations. Thus, although structurally related to SMB VN , the SMB MSF homolog lacks the specific residues necessary to bind to PAI-1 with high affinity.
The PAI-1 Binding Site in SMB VN -To more precisely identify the PAI-1 binding site in SMB VN , a domain swapping strategy was adopted (Fig. 2). In these experiments, portions of the inactive SMB MSF domain were exchanged for the corresponding regions of active SMB VN . The sequence homology and conserved cysteines in these two peptides (Table I) suggest that domain swapping in this way will not lead to gross structural alterations in the resulting peptides. A two-step PCR method (34,35) was used to create the fusion genes in these domain swapping experiments, and each of the resulting constructs was sequenced prior to use to verify that it contained the correct sequence. Again, all peptides were purified from the periplasmic fluid using a nickel column, and then each was tested for PAI-1 binding activity in the competitive ligand binding assay (Fig. 3). When residues 1-20 of SMB VN were replaced with the corresponding residues of SMB MSF , the resulting fusion peptide (i.e., SMB MSF 1-20 VN ) no longer bound to PAI-1 (Fig. 3A). Negative results also were obtained when residues 21-41 of SMB VN were exchanged with the corresponding SMB MSF sequences to create SMB VN 1-20 MSF (Fig. 3B). The lack of activity in both constructs indicates that neither the NH 2 -terminal nor the COOH-terminal half of the SMB VN molecule is sufficient for PAI-1 binding activity. This domain swapping approach was therefore expanded by systematically exchanging progressively larger fragments of the amino-terminal (Fig. 3A) or carboxyl-terminal (Fig. 3B) regions of SMB VN for SMB MSF and monitoring changes in activity. For example, replacing residues 1-9 of SMB VN with equivalent residues of SMB MSF (i.e., SMB MSF 1-9 VN ) did not destroy PAI-1 binding activity (Fig. 3A). Although this activity remained when residues 1-11 were exchanged, the hybrid peptide containing residues 1-12 of SMB MSF (SMB MSF 1-12 VN ) showed no PAI-1 binding activity (Fig. 3A). Because the only difference between SMB MSF 1-11 VN and SMB MSF 1-12 VN is the substitution of Gly 12 with a serine, this glycine must be critical for PAI-1 binding. Hybrid molecules containing residues 1-14, 1-16, and 1-18 of SMB MSF also were inactive. Thus, these initial studies establish that residues 1-11 at the NH 2 terminus of SMB VN are not required for PAI-1 binding when replaced with the corresponding sequence from another SMB homolog. Residue 12 (Gly) appears to be essential for the PAI-1 binding activity of SMB VN .

TABLE I Comparison of the SMB sequence in various VNs and homologs
The plus and minus signs across the top of the table refer to positively and negatively charged amino acids, respectively. The cysteines are numbered according to human VN, and the shaded areas in the table indicate consensus amino acids required for PAI-1 binding (see "Discussion"). These consensus amino acids are also indicated on the Consensus line. The numbers in parentheses refer to the references used to obtain each sequence. The asterisk indicates that pig VN was obtained from a database search (GenBank/EMBL) accession number D63145. Similar experiments were performed to identify the COOHterminal sequences necessary for PAI-1 binding. In these experiments, hybrid peptides were constructed by exchanging residues in the COOH terminus of SMB VN with corresponding regions of SMB MSF . Fig. 3B shows that swapping SMB VN residues 31-41 (i.e., SMB VN 1-30 MSF ) or residues 29 -41 (i.e., SMB VN 1-28 MSF ) with SMB MSF sequences created hybrid peptides that retained PAI-1 binding activity, although the latter peptide had somewhat lower affinity for PAI-1. In contrast, replacing residues 26 -41 created a hybrid peptide (SMB VN 1-25 MSF ) that completely lacked PAI-1 binding activity. Thus, residues 31-40 at the COOH terminus of SMB VN are not required for PAI-1 binding when replaced with the corresponding region of another SMB homolog. Residue 29 is necessary for full PAI-1 binding activity. The importance of residue 28 cannot be established from these experiments because it is conserved in both molecules (see below).
Although the above results suggest that the PAI-1 binding site is located in the central region of SMB VN , it is also possible that the loss of activity in the inactive hybrid molecules reflects conformational changes in the molecule and not the loss of specific sequences. To distinguish between these possibilities, gain of function experiments were performed. In these experiments, the central region of SMB VN was inserted into the non-PAI-1 binding homolog SMB MSF in an attempt to convert it into a PAI-1 binder. When residues 12-30 of SMB VN were exchanged with the corresponding sequences of intact but inactive SMB MSF , the resulting peptide (i.e., SMB MSF 1-11 VN 12-30 MSF ) was indeed converted into an active PAI-1-binding protein (Fig. 4). To further confirm these results, SMB VN 12-30 was introduced into the first SMB domain of PC-1, another inactive SMB homolog, creating the hybrid protein SMB PC-1 1-11 VN 12-30 PC-1 . Again the resulting peptide demonstrated full PAI-1 binding activity (Fig. 4). These results indicate that the 19 residues in the central region of active SMB VN contain the sequences necessary to convert the non-PAI-1-binding homologs into fully active PAI-1-binding proteins.
The Cysteines in SMB VN Are Essential for PAI-1 Binding-Approximately 20% of the residues in SMB VN are cysteines. Because free cysteines were not detected in recombinant SMB VN domain using Ellman's method (36), 2 and because 10 mM N-ethylmaleimide did not interfere with the binding of SMB VN to PAI-1 (data not shown), the 8 cysteines appear to be arranged in disulfide bonds. Complete chemical reduction of VN 2 or of SMB itself (33) abolished PAI-1 binding activity. Thus, the cysteines in SMB VN appear to be essential for PAI-1 binding. To determine whether all 8 cysteines were required for PAI-1 binding, we systematically changed each cysteine to alanine and then determined the PAI-1 binding activity of the resultant mutant peptides. Recombinant peptides with Cys 9 , Cys 19 , Cys 21 , Cys 31 , Cys 32 , or Cys 39 replaced by alanine were completely inactive (Fig. 5A). However, the mutants with Cys 5 or Cys 25 replaced by alanine exhibited detectable PAI-1 binding activities, although their activities were significantly decreased compared to the control. These latter results raise the possibility that Cys 5 and Cys 25 may not be essential for PAI-1 binding activity and that the observed loss of activity results from the creation of free, reactive cysteines. Moreover, the fact that out of all the cysteine mutants only these two retained detectable PAI-1 binding activity suggests that they form a disulfide pair. This possibility seems unlikely because conversion of both Cys 5 and Cys 25 into alanines completely inactivated the molecule. 2 In any case, these results demonstrate that all of the cysteines are critical for PAI-1 binding activity, most likely because they are required to maintain the unique, tightly folded structure of the molecule.
Other Residues in SMB VN Required for PAI-1 Binding-A site-directed mutagenesis approach was employed to identify other residues in SMB VN essential for PAI-1 binding. Initially, we converted clusters of charged amino acids into alanines and then tested the resulting peptides for PAI-1 binding activity. For example, when E3SCKGRCTE (Table I) was changed to A3SCAGACTA, the resulting mutant peptide still bound to PAI-1 (data not shown). Similarly, when D16KK was converted to A16AA, the mutant displayed full PAI-1 binding activity (Fig. 5B). These results indicate that the charged residues in the NH 2 -terminal half of the molecule are not essential for PAI-1 binding and suggest that these 7 charged residues either may be buried inside the molecule or located at a site distal to the PAI-1 binding site. Finally, changing E38CK to A38CA at the COOH-terminal of SMB VN had no effect on the PAI-1 binding activity of the mutant peptide. The remaining three negatively charged residues in SMB VN (i.e., Asp 22 , Glu 23 , and Asp 34 ) were then individually mutated to alanines. Interestingly, the Asp 22 to alanine mutant (D22A) was completely inactive, and the D34A mutant had only partial activity (Fig.  5B). Mutation of Glu 23 to Ala 23 had no effect on PAI-1 binding. These results indicate that among the 12 charged residues in SMB VN , the two negatively charged aspartic acids appear to be directly involved in PAI-1 binding. Because these amino acids are conserved in SMB VN and SMB MSF , they were not revealed in the domain swapping experiments.
Besides the cysteines and the 2 aspartic acids, a number of other residues are conserved in SMB VN and SMB MSF . Thus the question of whether these residues are important in PAI-1 binding could not be addressed in the domain swapping experiments. These residues, along with a number of others in the central region, were therefore converted individually into alanine to examine their role in PAI-1 binding. Among these residues, Gly 12 , Leu 24 , and Tyr 28 were essential because converting any of them into alanine resulted in the total loss of PAI-1 binding activity (Fig. 6). Another tyrosine, Tyr 27 , also appears to play a role in PAI-1 binding because the alanine mutant has lower affinity for PAI-1. Finally, a number of residues were identified (Table II) that did not appear to be essential for PAI-1 binding because each could be converted into alanine without affecting the PAI-1 binding activity of the resulting peptide. Table II lists all of the above mutants and summarizes their PAI-1 binding activities. In brief, a number of individual residues appear to be necessary for PAI-1 binding. These include the 8 cysteines, the polar (Tyr 27 and Tyr 28 ) and hydrophobic (Gly 12 and Leu 24 ) residues, and the two aspartic acids (Asp 22 and Asp 34 ). Fig. 7 shows that biologically active and presumably properly folded SMB VN migrates as a monomer of 8 kDa when analyzed on reducing SDS-PAGE. However, the inactive, G12A mutant migrated as a ladder of progressively larger bands, with very little material migrating at 8 kDa. The mutants L24A and Y28A both migrated in a pattern similar to that of wild type SMB VN . These results indicate that substituting Gly 12 with alanine results in the formation of inactive, reduc- tion-resistant aggregates. Thus, Gly 12 appears to be necessary for the proper folding of the molecule, suggesting that the loss of activity in this mutant may result from gross structural changes in the molecule. The fact that substituting Leu 24 and Tyr 28 with alanine did not alter the electrophoretic profile of the mutants suggests that the loss of activities in these mutants is due to the loss of PAI-1 binding sites rather than from misfolding.
It is notable that in the NH 2 -terminal half of the SMB VN molecule, only Gly 12 and the cysteines are essential for PAI-1 binding. The fact that Gly 12 and the cysteines are conserved in the second SMB homolog from MSF (i.e., in exon 2; SMB MSF2 ; Table I) predicts that swapping only residues 20 -30 of SMB VN into SMB MSF2 would result in an active PAI-1 binding peptide. Fig. 8 indicates that this conclusion is correct. The fusion peptide (SMB MSF2 1-19  ) is fully active in PAI-1 binding.
Stabilization of PAI-1 Activity by the SMB VN Mutants-Active PAI-1 is unstable at 37°C, decaying into the latent form with a half-life of 1-2 h. Intact VN as well as purified SMB VN appear to stabilize PAI-1 (i.e., partially block this conversion) (17,18,37). The ability of the various SMB VN mutants to stabilize PAI-1 was therefore tested. In these experiments, recombinant, active SMB VN or various SMB VN mutants were incubated with active PAI-1 at 37°C for increasing lengths of time. At the end of the incubation period, the amount of PAI-1 activity remaining in the sample was determined by using the t-PA binding assay (31). Because only active PAI-1 can bind to t-PA, this assay provides a simple approach for determining the half-life of PAI-1 under various conditions. After incubation, the bound PAI-1 is quantitated by enzyme-linked immunosorbent assay using anti-PAI-1 antibodies. These experiments demonstrate that the ability of the various SMB VN mutants to stabilize PAI-1 correlates with their ability to bind PAI-1 (Fig. 9). Under conditions of excess peptide, active SMB MSF 1-11 VN and SMB MSF 1-11 VN 12-30 MSF stabilized PAI-1 as well as SMB VN , whereas inactive SMB MSF 1-12 VN and SMB VN 1-25 MSF had no stabilizing activity (Fig. 9). The hybrid peptide constructed between the SMB domains of VN and PC-1 (i.e., SMB PC-1 1-11 VN 12-30 PC-1 ) also exhibited full activity in stabilizing PAI-1. These results indicate that the structural requirements for stabilizing the biological activity of PAI-1 are the same as those required for PAI-1 binding.

DISCUSSION
The interaction between PAI-1 and vitronectin is of potential physiological significance because it not only stabilizes the relatively labile inhibitor, it also appears to localize and concentrate this potent antifibrinolytic molecule in tissues during a variety of pathologic processes (reviewed in Ref. 23). The demonstration that PAI-1 and VN frequently co-localize in the atherosclerotic vessel wall (20,38) is consistent with this idea. In addition to these changes, recent data suggest that the binding of PAI-1 to VN may also alter the specificity of the inhibitor by significantly increasing its affinity for thrombin (15,39). These observations raise the possibility that VN is a cofactor for PAI-1 and suggest that efforts to delineate the residues in each molecule responsible for their interaction will provide insights into the role of these molecules in a variety of biological processes. In spite of this, relatively little is known about the structure and function of the individual domains that govern the PAI-1-VN interaction. In this regard, preliminary mutagenesis studies suggest that polar and/or hydrophobic residues at the amino-terminal half of PAI-1 are involved (39,40). Although the carboxyl-terminal half of the inhibitor contains its reactive center, it does not appear to be involved in VN binding. We (14,17) and others (33) have mapped the high affinity binding site for active PAI-1 in VN to the amino-   terminal region of the molecule, the so called somatomedin B domain (i.e., residues 1-51). Although additional PAI-1 binding sites in VN have been detected (41,42), they are not specific for active PAI-1 and appear to be of relatively low affinity (23).
The SMB domain contains 8 cysteine residues, suggesting that it exists in a tightly folded, compact structure. We previously employed deletional analysis to further define the residues within SMB VN that were important for PAI-1 binding (14). Although deletion of residues 41-52 did not decrease the PAI-1 binding activity of the resulting peptide (i.e., consisting of residues 1-40), further truncation at this end of the molecule or removal of the first 5 residues at the NH 2 terminus created recombinant molecules that no longer bound PAI-1. The data presented in the current study suggest that this loss of activity results from the deletion of single cysteines at either end of the molecule (i.e., at residue 5 or 39). This conclusion is based on the observation that replacing any single cysteine with alanine severely decreased the PAI-1 binding activity of the resulting peptides (Fig. 5). Each of these deletions would create free reactive cysteines that in turn would be expected to disrupt the overall structure of the molecule.
The critical role of the cysteines in maintaining the structure of SMB indicates that any mutational analysis must be performed under conditions that preserve these residues. Thus, in the current study, we have employed domain swapping using homologous proteins with conserved cysteines, as well as alanine scanning of charged residues between the cysteines, to further define the PAI-1 binding motif in SMB VN . The rationale for the domain swapping experiments was based on the observation that a number of proteins were recently identified that contain the SMB domain (Table I). These domains are structurally related to SMB VN because each contains 8 cysteines at conserved positions, and there is up to 45% overall sequence identity between SMB VN and the homologs. The functions of many of these proteins are unknown, and they have not been characterized with respect to PAI-1 binding. Thus, in our initial experiments, we prepared a number of recombinant homologs and directly compared their PAI-1 binding activity with that of recombinant SMB VN (Fig. 1). Although SMB VN bound to PAI-1 and competed with VN for PAI-1 binding, the homologs did not (Figs. 1, 3, and 4). Thus, although there is striking structural and sequence homology between these proteins, their ability to bind PAI-1 is not shared. The structural similarity of these molecules suggested that we might be able to systematically exchange portions of active SMB VN with corresponding regions of the inactive homologs without disrupting their cysteine pairing or tertiary structure. Analysis of the activity of the resulting hybrid peptides should therefore provide initial insights into the regions of SMB VN required for PAI-1 binding. This strategy is outlined in Fig. 2, and Figs. 3, 4, and 8 indicate that it was successful. For example, exchanging the NH 2 -or COOH 2 -terminal halves of each molecule created hybrid molecules that lacked PAI-1 binding activity (Fig. 3). Thus, the PAI-1 binding motif does not appear to be included entirely in either half of the molecule. Exchanging progressively larger regions from both ends of the molecule revealed that the PAI-1 binding site was in fact localized to the central region of the molecule (i.e., to residues 12-30) and thus spanned both halves (Figs. 3, 4, and 8). The fact that residues 12-30 alone convert non-PAI-1-binding proteins into binders (Figs. 4, 8), demonstrates that a specific PAI-1 binding motif is contained within the central region of the molecule. Moreover, these gain of function experiments argue against the possibility that the absence of activity in many of the mutants is due to conformational changes in the molecule.
Critical residues for PAI-1 binding in the central region were further delineated by alanine scanning (Fig. 5; Table II) and include the nonpolar residue, Gly 12 , the charged residues, Asp 22 and Asp 34 , a hydrophobic residue, Leu 24 , and the polar residues, Tyr 27 and Tyr 28 . Comparison of the SMB domains present in human (30), mouse (43), rabbit (44), and porcine ( Table I) VN reveals that these residues are conserved in all of these molecules (Table I). The human and mouse (43), as well as rabbit 3 VNs are all PAI-1-binding proteins. Although there is no published data to indicate that porcine VN also binds PAI-1, it seems likely because it contains the critical residues required for PAI-1 binding (see below). Examination of Table I also reveals that a number of residues are not conserved in these VNs. For example, in the NH 2 -terminal half of the molecule, residues Gly 7 , Glu 11 , and N14VDK of human SMB VN are not conserved in the other VNs (Table I). Similarly, in the COOH-terminal region, Ser 26 , Thr 33 , and T36AE are not conserved. Because each of these VNs bind PAI-1 with high affinity, the nonconserved residues are not essential for PAI-1 binding. This conclusion is consistent with the results from our mutagenesis studies ( Fig. 5; Table II). In general, the residues identified in our mutagenesis studies as being critical for PAI-1 binding are conserved in the various VNs, whereas those that were not essential were not conserved. In any case, these results suggest the following consensus sequence for the PAI-1 binding motif in SMB VN : ----C 5 ---C 9 --G 12 ------C 19 -C 21 D 22 -L 24 C 25 -Y 27 Y 28 --C 31 C 32 -D 34 ----C 39 --(see Table I). The fact that all of the VNs contain this sequence and bind to PAI-1 and that the homologs lack this sequence and do not bind to PAI-1 supports the conclusion that these residues are essential for PAI-1 binding. It should be noted that we also determined the PAI-1 binding activity of a number of the other homologs, including TCl-30, PP11, PC-1 (94 -134), and ATX1. These SMB homologs also lacked PAI-1 binding activity (data not shown). 3 D. Seiffert, unpublished observation.

FIG. 9. Stabilization of active PAI-1 by hybrid mutants of SMB.
Active PAI-1 (1 nM) was incubated (37°C) in the presence of SMB or various hybrid mutants of SMB at a concentration of 100 g/ml for varied lengths of time in phosphate-buffered saline/bovine serum albumin/Tween. The PAI-1 activity remaining at the end of the incubation period was measured by the t-PA binding assay. The results are expressed as the percentages of active PAI-1 remaining in the nonincubated controls.
Although VN and SMB VN stabilize PAI-1 in its active form (14,18,37), the minimum structural requirement in SMB VN for this PAI-1 stabilizing activity has not been defined. Data from the current study show that the ability of SMB VN mutants to stabilize PAI-1 correlates with their ability to bind PAI-1 (Fig.  9). The mutant with the highest affinity for PAI-1 (i.e., SMB MSF 1-11 VN ) stabilized PAI-1 nearly as well as SMB VN , whereas SMB MSF 1-12 VN , which had little affinity for PAI-1, showed no stabilization. These results provide evidence for the hypothesis that SMB VN stabilizes PAI-1 by directly binding to it and preserving it in the active conformation.
In summary, our data suggest that the actual PAI-1 binding site in SMB VN is located in the central region of the molecule and is composed of several residues presented on the surface of this compact structure. Although the positively charged residues do not appear to be essential for PAI-1 binding, the negatively charged residues are critical for the high affinity PAI-1-VN interaction. The finding that the polar residues, Tyr 27 and Tyr 28 , and hydrophobic residue, Leu 24 , were important for PAI-1 binding is consistent with the suggestion from studies of PAI-1 mutants (39,40) that such molecules are involved. In this model, the two negatively charged aspartic acids may provide for the initial protein docking in the interaction or may be necessary for the proper folding of SMB VN .