Phospholipid-associated annexin A2-S100A10 heterotetramer and its subunits: characterization of the interaction with tissue plasminogen activator, plasminogen, and plasmin.

Annexin A2 (p36) is a highly alpha-helical molecule that consists of two opposing sides, a convex side that contains the phospholipid-binding sites and a concave side, which faces the extracellular milieu and contains multiple ligand-binding sites. The amino-terminal region of annexin A2 extends along the concave side of the protein and contains the binding site for the S100A10 (p11) subunit. The interaction of these subunits results in the formation of the heterotetrameric form of the protein, annexin A2-S100A10 heterotetramer (AIIt). To simulate the orientation of AIIt on the plasma membrane we bound AIIt to a phospholipid bilayer that was immobilized on a BIAcore biosensor chip. Surface plasmon resonance was used to observe in real time the molecular interactions between phospholipid-associated AIIt or its annexin A2 subunit and the ligands, tissue-type plasminogen activator (t-PA), plasminogen, and plasmin. AIIt bound t-PA (Kd = 0.68 microm), plasminogen (Kd = 0.11 microm), and plasmin (Kd = 75 nm) with moderate affinity. Contrary to previous reports, the phospholipid-associated annexin A2 subunit failed to bind t-PA or plasminogen but bound plasmin (Kd = 0.78 microm). The S100A10 subunit bound t-PA (Kd = 0.45 microm), plasminogen (Kd = 1.81 microm), and plasmin (Kd = 0.36 microm). Removal of the carboxyl-terminal lysines from the S100A10 subunit attenuated t-PA and plasminogen binding to AIIt. These results show that the carboxyl-terminal lysines of S100A10 form t-PA and plasminogen-binding sites. In contrast, annexin A2 and S100A10 contain distinct binding sites for plasmin.

The plasminogen activators, tissue plasminogen activator (t-PA) 1 and urokinase-type plasminogen activator activate plasminogen by cleaving the Arg 561 -Val 562 site, resulting in a conformation change and formation of the substrate-binding site. Plasminogen binds with moderate affinity and high capacity to protein and non-protein receptors via its kringle domains. These kringle domains typically interact with lysine residues of the plasminogen receptors (1)(2)(3)(4). Studies from several laboratories establish that although a wide variety of proteins bind plasminogen, only those with carboxyl-terminal lysine residues participate in the conversion of plasminogen to plasmin (1,3). It is now apparent that t-PA and urokinase-type plasminogen activator utilize fundamentally different mechanisms to activate plasminogen. Although t-PA and plasminogen share common cellular binding sites, urokinase-type plasminogen activator binds to a specific receptor, uPAR (5,6).
The annexin A2-S100A10 heterotetramer (AIIt) is a Ca 2ϩbinding protein that is composed of two annexin A2 (p36) and two S100A10 (p11) subunits (for review, see Refs. 7-11). The annexin A2 subunit is a planar, curved molecule with opposing convex and concave sides. The convex side faces the biological membrane and contains the Ca 2ϩ -and phospholipid-binding sites. The concave side faces the extracellular milieu and contains both the amino-terminal and carboxyl-terminal regions of the protein. The carboxyl-terminal region consists of multiple ligand-binding sites including binding sites for F-actin, fibrin, and heparin. A high resolution crystal structure of the AIIt complex has not been determined. However, a low resolution structure of the complex bridging two phospholipid membranes has been determined by cryo-electron microscopy. This structure shows that the S100A10 subunits are positioned in the center, with an annexin A2 molecule on either side, in contact with the membrane surfaces. The high resolution crystal structure of the S100A10 dimer in complex with a peptide to the amino terminus of annexin A2 shows that the annexin A2binding site consists of the amino terminus of one S100 molecule and the carboxyl terminus of the other molecule within the S100A10 dimer. Thus, it is well established that the amino terminus of annexin A2 binds to both amino-and carboxylterminal helices of S100A10 (12)(13)(14)(15)(16). S100A10 has been shown to regulate plasma membrane ion channels (17,18) as well as cytosolic phospholipase A2 (19). Although originally identified as an intracellular protein, a large number of studies demonstrate the presence of both annexin A2 and S100A10 on the extracellular surface of many cells (20 -26). Extracellularly, the S100A10 subunit functions as the plasminogen regulatory subunit of AIIt (27). The penultimate and ultimate carboxyl-terminal lysines of this subunit * This work was supported by a grant from the Canadian Institutes of Health Research and by National Institutes of Health Grant CA78639. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 1 The abbreviations used are: t-PA, tissue-type plasminogen activator; AIIt, annexin A2-S100A10 heterotetramer; S100A10 desKK , S100A10 subunit lacking the carboxyl-terminal lysines Lys-95 and Lys-96; RU, surface plasmon resonance response units; HEK 293 cells, human em-have been shown to participate in the stimulation of t-PA-dependent plasminogen activation by AIIt (28).
We originally reported that AIIt stimulated t-PA-dependent plasminogen activation about 77-fold compared with about 2or 46-fold for recombinant annexin A2 or S100A10 subunit, respectively (25,27). Because the S100A10 subunit possesses two lysines at its carboxyl terminus, we postulated that the carboxyl-terminal lysines of S100A10 played a key role in the stimulatory activity of AIIt (27). To support this hypothesis we have shown that the loss of the two carboxyl-terminal lysines of S100A10 blocks the ability of AIIt to stimulate t-PA-dependent plasminogen activation (28). In the present report we have directly addressed the issue of which subunit of AIIt contains the t-PA-and plasminogen-binding site. The development of a BIAcore biosensor chip with an immobilized phospholipid bilayer has allowed analysis for the first time of the interaction of annexin A2 and AIIt, which are bound to this phospholipid bilayer with components of the plasminogen regulatory system. This methodology also has the advantage that because these interactions are studied by surface plasmon resonance, kinetic parameters are determined rapidly and in real time. Using this novel approach and the traditional method of covalent immobilization of proteins on the biosensor chip, we found that AIIt bound t-PA, plasminogen, and plasmin. Analysis of the activities of the individual subunits has shown that the annexin A2 subunit bound plasmin but not t-PA or plasminogen. In contrast, the S100A10 subunit bound t-PA, plasminogen, and plasmin. Furthermore, the carboxyl-terminal lysines of the S100A10 subunit formed the binding sites for t-PA and plasminogen.

EXPERIMENTAL PROCEDURES
Materials-Primary human umbilical vein endothelial cells were a generous gift from Dr. K. Patel (University of Calgary). HEK 293 cells were obtained from American Type Tissue Culture collection. Gluplasminogen and plasmin were purchased from American Diagnostica. The wild-type annexin A2, wild-type S100A10, and S100A10 desKK deletion mutant (lacking the last two lysine residues of the carboxyl terminus) were expressed in Escherichia coli and purified as described previously (28). This recombinant annexin A2 was used in the binding experiments. AIIt and other annexins were purified from bovine lung as described in Khanna et al. (29). Monoclonal anti-annexin A2 and antiannexin A2 light chain antibodies were purchased from Transduction Laboratories. Anti-mouse horseradish peroxidase-conjugated secondary antibody was purchased from Santa Cruz Biotechnology. Streptavidinagarose was purchased from Sigma. Mutant AIIt composed of wild-type annexin A2 and S100A10 desKK subunits was produced as follows. Both subunits were combined with an equal molar ratio, and the resultant mutant, AIIt, was isolated by gel permeation chromatography and characterized according to Kang et al. (30). Studies comparing the interaction of the wild-type and mutant AIIt to heparin, phospholipid, and F-actin established that the structure of the mutant recombinant AIIt was identical to the wild-type recombinant AIIt. Recombinant yeast annexin A2 was produced in the yeast strain W303-1B with the expression vector pYeAxII, a generous gift from Dr. Jesus Ayala-Sanmartin (Unite de Biologie Cellulaire et Moleculaire de la Secretion, Institut de Biologie Physico-Chimique, Paris, France). Annexin II was subjected to limited chymotrypsin digestion as described previously (27). SDS-PAGE confirmed the loss of the amino-terminal region of the protein. AIIt was digested with carboxypeptidase B, and the loss of the carboxyl-terminal lysines was confirmed as previously described (28).
Transfection of HEK 293 Cells with the Annexin A2 Gene-Human annexin A2 was obtained by PCR using primers containing unique restriction enzyme sites for NotI and BglII. The cDNA was digested, gel-purified, and cloned into a NotI-and BglII-cut retroviral vector, pLin (provided by Dr. L.-J. Chang, Powell Gene Therapy Center, University of Florida, Gainesville, FL). The orientation of the insert was verified by sequencing. The vector harboring annexin A2 was termed pLin-p36. HEK 293 cells were transfected with pLin and pLin-p36 by using LipofectAMINE 2000 Reagent (Invitrogen) as described in Choi et al. (31). The stably transfected cells were selected with 200 g/ml G418, and clonal cell lines were screened by annexin A2 protein expression levels as assessed by Western blot analysis.

Real Time Binding Measured by Surface Plasmon Resonance-The
BIAcore 3000 system and reagents, including sensor chips, flow buffers, the amine coupling kit, and the thiol coupling kit were obtained from BIAcore (BIAcore inc., Piscataway, NJ). The HBS-EP flow buffer consisted of 10 mM Hepes (pH 7.4), 150 mM NaCl, 3 mM EDTA, and 0.005% (v/v) P20 filtered through 0.2-m filter and degassed before use. The HBS-Ca 2ϩ buffer used with the phospholipid surface was 10 mM Hepes (pH 7.4), 150 mM NaCl, and 1 mM CaCl 2 . Annexin II, t-PA, and Gluplasminogen were immobilized on CM5 sensor chips by primary amine coupling according to the manufacturer's instructions. Briefly, the chip surface was activated with a 0.2 M N-ethyl-N- (3-diethylaminopropyl) carbodiimide, 0.05 M N-hydroxysuccinimide solution followed by the injection of ligand at a concentration between 5 and 50 g/ml in acetate buffer (pH 5.0). When the desired level of binding was achieved, unreacted N-hydroxysuccinimide-ester groups were blocked with 1 M ethanolamine hydrochloride. Wild-type and mutant S100A10 were coupled to CM5 chips using the ligand thiol method according to the manufacturer's instructions. The sensor chip was treated with N-ethyl-N-(3diethylaminopropyl carbodiimide/N-hydroxysuccinimide solution as above followed by treatment with 80 mM pyridinyldiethaneamine. Ligand was bound to the desired level, and unreacted sites were blocked with 100 mM cysteine in 1 M NaCl. Sensorgrams (response units versus time) were recorded by the BIAcore 3000 at 25°C with a typical flow rate of 30 l/min and injection times between 60 and 120 s. Controls for the contribution of the change in bulk refractive index were performed in parallel with flow cells derivitized in the absence of ligand (or coated with phospholipid vesicles in the case of sensor chip L1) and subtracted from all binding sensorgrams. Regeneration conditions varied with the interactions studied: 2 mM EGTA for calcium-dependent binding, 10 mM ⑀-aminocaproic acid for lysine-dependent binding, and 10 mM glycine (pH 1.5) for all other interactions. The binding of the various analytes (annexin A2, S100A10, AIIt, plasminogen, plasmin, and t-PA) was examined by multiple injections of the analyte diluted in flow buffer over a concentration range between 0.003 and 2.0 M. Because the mechanism of interaction between t-PA, plasminogen, or plasmin with AIIt is not known, it is not possible to identify an exact binding model. For example, there are likely two binding sites for plasminogen on AIIt (composed of the carboxyl-terminal lysine of each of the two S100A10 subunits) and potentially several lysine binding kringles on plasminogen. Furthermore, plasminogen is heterogeneous and consists of two glycoforms (N-linked plasminogen-1 and O-linked plasminogen-2 glycosylation variants), with each glycoform consisting of six to eight additional isoforms varying in the sialic acid content. It is therefore impossible at this time to identify a binding model to explain these complex interactions. Consequently, we did not attempt to determine the association and dissociation rate constants. We obtained an approximate equilibrium dissociation constant (K d ) by measuring the equilibrium resonance units (R eq ) at several ligand concentrations at equilibrium. Binding data were analyzed by Scatchard analysis using the BIA evaluation software according to the relationship R eq /C ϭ K a R max Ϫ K a R eq , where R max is the resonance signal at saturation, C is the concentration of free analyte, and K a is the equilibrium association constant.
Preparation of Phospholipid Surfaces on Biosensor Chips-In a 12 ϫ 75-mm test tube, 10 l of 20 mg/ml phosphatidylserine, phosphatidylcholine, and cholesterol (all dissolved in chloroform) were shelled by N 2 gas and dried under vacuum for 2 h. The resultant residue of lipids was resuspended in 0.5 ml of HBS-Ca 2ϩ buffer and subjected to 10 freezethaw cycles in liquid nitrogen. Large unilamellar vesicles were prepared by passing the resulting suspension through polycarbonate filters (100 nm pore diameter) in a LiposoFast™ Basic apparatus (Avestin Inc., Ottawa, ON). The large unilamellar vesicle solution was then injected over a BIAcore Pioneer L1 sensor chip at a flow rate of 3 l/min for 15 min. Unbound material was removed with an injection of 50 mM NaOH. Typically, this procedure resulted in the immobilization of about 10,000 -12,000 RU of phospholipid on the chip surface. Bovine serum albumin binds strongly to the dextran matrix of the L1 biosensor chip but only weakly to the phospholipid bilayer. Because we did not observe significant binding of bovine serum albumin to the biosensor chip (data not shown) we concluded that the dextran surface was completely covered by the phospholipid bilayer.

Ca 2ϩ -dependent Binding of AIIt to the Phospholipid Surface
Formed on the Biosensor Chip-The interactions between the hydrophobic chains of phospholipid liposomes and the lipophilic alkyl groups covalently attached to the dextran-coated surface of the L1 biosensor chip results in the fusion and immobilization of the vesicles on the surface of the L1 biosensor chip. Analysis of the structure of the immobilized phospholipid has shown that under these experimental conditions a flexible phospholipid bilayer is formed on the surface of the biosensor chip (32,33).
Phospholipid liposomes composed of phosphatidylserine, phosphatidylcholine, and cholesterol were injected onto the surface of a L1 biosensor chip (Fig. 1A). Typically about 10,000 -12,000 RU of phospholipid was immobilized on the biosensor chip. Because injection of 50 mM NaOH did not change the mass immobilized to the biosensor chip, it was concluded that a stable phospholipid bilayer had formed on the chip. The surface plasmon resonance signal was stable for at least 12 h and was not changed upon washing with any of the incubation buffers (data not shown).
Next we examined the interaction of AIIt with the phospholipid surface. Phospholipid was immobilized on the biosensor chip, and the interaction of AIIt with the phospholipid-coated biosensor chip was measured in real time by recording the changes in surface plasmon resonance upon injection of AIIt. A typical binding curve (sensorgram) is presented in Fig. 1B. We observed that AIIt bound to the phospholipid-coated surface that was generated from liposomes composed of phosphatidylcholine/phosphatidylserine/cholesterol. Although the association of AIIt with this phospholipid surface was rapid, dissociation was extremely slow and could not be measured over the time course of our experiments. Over a time period of 2 h, AIIt formed a stable complex with the phospholipid surface (data not shown). AIIt also bound to this phospholipid surface in a Ca 2ϩ -dependent manner (Fig. 1B). AIIt is known to bind only to anionic phospholipids such as phosphatidylserine and, therefore, does not bind to phosphatidylcholine (34 -36). When we examined the interaction of AIIt with a phosphatidylcholine surface we could measure only a minimal interaction when a high concentration of AIIt was injected. Therefore, the phospholipid specificity and Ca 2ϩ dependence of the AIIt-phospholipid interaction observed in our experiments suggested that the binding of AIIt to the phospholipid surface formed on the L1 biosensor chip resembled the well characterized interaction of AIIt with intact phospholipid liposomes.
Next we quantitated the interaction of annexin A2 and AIIt with the phospholipid surface. For these experiments annexin A2 or AIIt (100 nM) was injected followed by buffer. This procedure allowed for control of the amount of these proteins immobilized on the biosensor chip. Typically we did not exceed 20% of the binding capacity of the biosensor chip. As shown in Fig. 1C, annexin A2 also showed a fast association with the phospholipid surface but remained bound to the phospholipidcoated biosensor chip during the course of our experiments. About 300 RU of annexin A2 and 640 RU of AIIt bound to the phospholipid-coated biosensor chip, which corresponds to about 300 pg (7.9 fmol/mm 2 ) of annexin A2 and 640 pg (7.1 fmol/mm 2 ) of AIIt. This level of immobilization was subsequently utilized in the binding experiments.
Interaction of Plasminogen with AIIt and Its Subunits-To observe the interaction of plasminogen with AIIt, we began by injecting AIIt onto the phospholipid-coated biosensor chip followed by buffer. Once a stable base line was obtained, plasminogen was subsequently injected. We observed that plasminogen bound to the phospholipid-associated AIIt (Fig. 2A). The equilibrium resonance units at several ligand concentrations were determined, and equilibrium binding data were subjected to Scatchard plot analysis (Fig. 2B). We determined a K d of 0.11 M for the interaction of AIIt with plasminogen. The interaction of AIIt with plasminogen was inhibited by 10 mM aminocaproic acid (data not shown), suggesting that this interaction was dependent on the binding of plasminogen to a lysine residue of AIIt.
To establish which subunit of AIIt contained the plasminogen-binding site, we immobilized each subunit on a biosensor chip and examined the plasminogen binding kinetics. S100A10 does not bind phospholipid; therefore, S100A10 was immobilized by thiol-coupling the protein to a CM5 biosensor chip. We then examined the interaction of the thiol-coupled S100A10 with plasminogen. We observed that S100A10 bound plasminogen with moderate affinity (K d of 1.81 M) (Fig. 2, C and D). The binding of plasminogen to the immobilized S100A10 was also blocked by 10 mM aminocaproic acid (data not shown).
Next we injected annexin A2 on the phospholipid surface of the L1 biosensor chip followed by buffer. Once a stable base line was established, plasminogen was then injected. Surprisingly, we found that phospholipid-associated annexin A2 failed to bind plasminogen over a concentration range of 50 nM to 1 M plasminogen. The annexin A2 used in these experiments was human recombinant protein expressed in E. coli. Similarly, the annexin A2 purified from bovine lung or human recombinant annexin A2 (expressed in yeast) also failed to bind plasminogen (data not shown). In contrast, S100A10 bound to the phospholipid-associated annexin A2 (Fig. 2E), suggesting that the annexin A2 was not denatured upon interaction with the phospholipid surface. These results conclusively show that annexin A2 is not a plasminogen-binding protein.
The inability of phospholipid-associated annexin A2 to bind to plasminogen was difficult to reconcile with the published observation that annexin A2, immobilized on the surface of a polystyrene plate, bound plasminogen (37). As a second approach to investigating the interaction of plasminogen with annexin A2, we prepared a plasminogen surface by aminecoupling plasminogen to a CM5 biosensor chip. We then flowed annexin A2 over the immobilized plasminogen surface. We observed that annexin A2 also failed to bind to the immobilized plasminogen (Fig. 2F). AIIt or S100A10 were also injected onto the surface of the chip followed by buffer. As shown in Fig. 2F, AIIt and S100A10 bound to the immobilized plasminogen surface under these conditions. Interestingly, the mutant S100A10, S100A10 desKK , which lacks the carboxyl-terminal lysines, did not bind plasminogen. Thus, these data confirm that the carboxyl-terminal lysines form the plasminogen-binding site of S100A10.
As a third approach to determining if an interaction existed between annexin A2 and plasminogen, we amine-coupled annexin A2 directly to the CM5 biosensor chip. We observed that plasminogen failed to bind to the immobilized annexin A2 (Fig.  2G). In contrast, S100A10 bound to the amine-coupled annexin A2, suggesting that the S100A10-binding site of annexin A2 remained intact in the amine-coupled annexin A2. Similar results were obtained with the thiol-coupled annexin A2 (data not shown). Thus, by three independent approaches we were unable to demonstrate an interaction between annexin A2 and plasminogen.
We also examined the interaction of plasminogen with a mutant AIIt that consisted of a wild-type annexin A2 subunit and the mutant form of S100A10 in which the carboxyl-terminal lysines were deleted (AIIt desKK ). This allowed us to directly evaluate the importance of the carboxyl-terminal lysines of S100A10 within the AIIt heterotetramer. Previous data from our laboratory had shown that the loss of the two carboxyl-FIG. 2. Interaction of plasminogen with annexin A2-S100A10 heterotetramer or its subunits. Panel A, binding of plasminogen to AIIt immobilized on the phospholipid-coated biosensor chip. About 425 RU of AIIt was immobilized, and the plasminogen was applied at a flow rate of 30 l/min. The Scatchard analysis of the binding isotherm is presented in panel B. Panel C, binding of plasminogen to immobilized S100A10. About 1500 RU of S100A10 was immobilized, and the plasminogen was applied at a flow rate of 30 l/min. The Scatchard analysis of the binding isotherm is presented in panel D. Panel E, binding of S100A10 and plasminogen to annexin A2 immobilized on the phospholipid-coated biosensor chip. About 540 RU of annexin A2 was immobilized, and the plasminogen (500 nM) or S100A10 (500 nM) was applied at a flow rate of 30 l/min. Panel F, comparison of the binding of AIIt, S100A10, S100A10 des-KK , and annexin A2 to an immobilized plasminogen surface. About 575 RU of plasminogen was immobilized, and 500 nM AIIt, S100A10, S100A10 desKK , or annexin A2 was applied at a flow rate of 30 l/min. Panel G, interaction of plasminogen with amine-coupled annexin A2. About 240 RU of annexin A2 was amine-coupled to a CM5 biosensor chip, and S100A10 (500 nM) and plasminogen (500 nM) were applied at a flow rate of 30 l/min. Panel H, interaction of plasminogen with phospholipid-immobilized AIIt and AIIt desKK . About 600 RU of wild-type AIIt or mutant AIIt desKK (consisting of wild-type annexin A2 and mutant S100A10 carboxyl-terminal lysine deletion mutant) were immobilized. Plasminogen (500 nM) was applied at a flow rate of 30 l/min. Panel I, S100A10 (2 M) (solid line) or buffer (dotted line) was injected on the immobilized plasminogen surface followed by an annexin A2 injection (1 M).
terminal lysines from the S100A10 resulted in an 80% loss in the ability of AIIt to accelerate t-PA-dependent plasminogen activation (27,28). As shown in Fig. 2H, the loss of the carboxyl-terminal lysines of the S100A10 subunit resulted in a 70 -80% reduction in the binding of plasminogen to the phospholipid-immobilized AIIt. These results show that the carboxylterminal lysines of S100A10 form the plasminogen-binding site within AIIt and that the S100A10 subunit but not the annexin A2 subunit is the key plasminogen binding subunit of AIIt.
The amino terminus of annexin A2 binds to both the aminoand carboxyl-terminal helices of S100A10 (14,15). Our kinetic (27,28) and binding experiments (Fig. 2) show that the carboxyl-terminal lysines of S100A10 (Lys 95 -Lys 96 ) form the plasminogen-binding site and also play a key role in the stimulation of t-PA-dependent plasminogen activation by AIIt. This implies that the annexin A2-binding site residues such as Lys 91 (38) and plasminogen-binding site of S100A10 (Lys 95 -Lys 96 ) are in close proximity. It is unclear if the annexin A2-binding site of S100A10 is accessible to annexin A2 once S100A10 has formed a complex with plasminogen. To address this issue, S100A10 or buffer was injected on the immobilized plasminogen surface, and after a stable base line was obtained, annexin A2 was subsequently injected. We observed that annexin A2 bound to the S100A10-plasminogen complex but not to the isolated plasminogen surface (Fig. 2I). This established that the binding sites for annexin A2 and plasminogen on S100A10 are freely accessible to their respective ligands.
Interaction of t-PA with AIIt and Its Subunits-Next we examined the interaction of t-PA with AIIt. Unfortunately, we observed that t-PA interacted with the phospholipid surface of the biosensor chip, and this high background prevented an accurate measurement of the interaction of t-PA with phospholipid-immobilized annexin A2 or AIIt. We therefore aminecoupled t-PA to a CM5 biosensor chip and examined the interaction of AIIt and its subunits with this immobilized t-PA. We found that AIIt or its S100A10 subunit bound to the immobilized t-PA (Fig. 3, A, B, and D), and this interaction was inhibited by aminocaproic acid (data not shown). Scatchard analysis established K d values of 0.68 M or 0.45 M for the interaction of AIIt or S100A10 with immobilized t-PA (Fig. 3, C  and E). Surprisingly, we were unable to measure any interaction between annexin A2 and the immobilized t-PA (Fig. 3A). We also were unable to observe any interaction between amineimmobilized t-PA and bovine lung annexin A2 or human recombinant annexin A2 (expressed in yeast). Furthermore, amine-immobilized annexin A2 did not bind to t-PA (data not shown).
We also observed that S100A10 desKK did not bind t-PA, therefore, identifying these residues as forming the t-PA-binding site of S100A10 (data not shown). Furthermore, the mutant AIIt, AIIt desKK , did not bind to the immobilized t-PA (Fig. 3A). This result, therefore, shows that the S100A10 subunit of AIIt is the important t-PA binding subunit of AIIt. Furthermore, based on these binding studies we concluded that the carboxylterminal lysines of the S100A10 subunit are a common binding site for both t-PA (Fig. 3A) and plasminogen (Fig. 2, F and H).
Binding of Plasmin to AIIt and Its Subunits-We initially reported that AIIt stimulated the plasmin autoproteolytic re- FIG. 3. Interaction of tissue-type plasminogen activator with annexin A2-S100A10 heterotetramer or its subunits. Panel A, interaction of AIIt and its subunits with t-PA. About 1000 RU of t-PA was amine-coupled on the CM5 biosensor chip. AIIt, AIIt desKK , annexin A2, or S100A10 were injected at a flow rate of 30 l/min and at a concentration of 1 M. Panel B, interaction of AIIt with amine-coupled t-PA. The Scatchard analysis of the binding isotherm is presented in panel C. Panel D, interaction of S100A10 with amine-coupled t-PA. The Scatchard analysis is presented in panel E.
action (39,40). It was unclear from those studies which subunit was involved in this interaction with plasmin. In subsequent studies we showed that both subunits of AIIt possessed plasmin reductase activity and catalyzed the release of kringlecontaining plasminogen fragments from plasmin (41). This suggested that both subunits of AIIt could bind plasmin. As shown in Fig. 4, A and B, phospholipid-associated AIIt binds plasmin with a K d of 75 nM. Phospholipid-immobilized annexin A2 (Fig. 4, C and D), and thiol-immobilized S100A10 (Fig. 4, E and F) also bound plasmin with K d values of 780 and 360 nM, respectively.
Next we attempted to identify the location of the plasmin binding region of AIIt. Removal of the first 23 residues of the amino-terminal region of annexin A2 by limited chymotrypsin digestion did not affect plasmin binding (Fig. 4G). This suggests that the plasmin-binding site of annexin A2 is located in the core domain of the molecule. We also found that removal of the carboxyl-terminal lysines of the S100A10 subunit either by carboxypeptidase B treatment of AIIt (Fig. 4H) or by construction of the AIIt desKK mutant (data not shown) did not affect plasmin binding. Because loss of the carboxyl-terminal lysines of the S100A10 subunit inhibited plasminogen binding, it is therefore likely that S100A10 contains distinct binding sites for plasmin and plasminogen.
Transfection of Cells with Annexin A2 Stimulates the Expression of S100A10 -Our observation that annexin A2 did not bind plasminogen was difficult to reconcile with the report that transfection of 293 cells with annexin A2 resulted in increased extracellular annexin A2 and enhanced plasminogen binding to the cell surface (42). However, it has been reported that annexin A2 regulates the expression of S100A10 protein by a post-translational mechanism (43,44). We therefore examined the possibility that annexin A2 might affect the levels of S100A10. HEK 293 cells were transfected with annexin A2, and stable clones were isolated and propagated. We found that, as had been reported for other cells, the transfection of 293 cells with annexin A2 resulted in the up-regulation of S100A10 (Fig. 5). DISCUSSION Surface plasmon resonance techniques offer the means of presenting membrane-associated proteins in an environment and orientation that closely resembles their in vivo environment. This technology can provide insights into the interactions between membrane-associated proteins and their ligands because it monitors the interactions under conditions that approach the physiological. The physical characteristics of phospholipid surfaces immobilized on the L1 BIAcore biosensor chip have been well characterized and shown to consist of a flexible bilayer (32,33). Because annexin A2 and AIIt bound essentially irreversibly to the phospholipid surface (Fig. 1C), we could examine the binding of various ligands to these proteins. Most FIG. 4. Interaction of plasmin with annexin A2-S100A10 tetramer or its subunits. Panel A, interaction of plasmin with phospholipidimmobilized AIIt or annexin A2. 500 RU of AIIt or annexin A2 were immobilized on the phospholipid-coated biosensor chip. Plasmin was injected at a flow rate of 30 l/min. The Scatchard analysis of the binding isotherm is presented in panel B. Panel C, interaction of plasmin with annexin A2. The Scatchard analysis of the binding isotherm is presented in panel D. Panel E, the interaction of thiol-coupled S100A10 with plasmin. About 1200 units of S100A10 were coupled. The Scatchard analysis of the binding isotherm is presented in panel F. Panel G, interaction of plasmin with phospholipid-associated annexin A2. About 500 RU of annexin A2 or chymotrypsin-treated annexin A2 was coupled. Panel H, interaction of plasmin with phospholipid-associated AIIt. 500 RU of AIIt-or carboxypeptidase-treated AIIt was immobilized.
importantly, the orientation and conformation of annexin A2 and AIIt would be similar to that under the physiological situation. Thus, the convex side of annexin A2 would face and interact with the phospholipid bilayer, whereas the ligand binding concave surface would be exposed to the fluid phase. Using this novel approach we have demonstrated that phospholipid membrane-associated AIIt binds plasminogen and plasmin. In contrast, phospholipid membrane-associated annexin A2 binds plasmin but does not interact with plasminogen. Conversely, S100A10 binds both plasminogen and plasmin. Although both AIIt and the S100A10 bound to aminecoupled t-PA, annexin A2 failed to interact with this protein.
Removal of the carboxyl-terminal lysines of S100A10 blocked t-PA and plasminogen binding but did not affect its interaction with plasmin.
Our results show that annexin A2 does not bind t-PA or plasminogen. This finding was based on three separate approaches. We observed that phospholipid-associated annexin A2 did not bind plasminogen, and amine-coupled annexin A2 did not bind t-PA or plasminogen. Furthermore, we also observed that annexin A2 did not bind to amine-coupled t-PA or plasminogen. These three approaches formed the basis for our conclusion that annexin A2 is not a plasminogen-binding protein. Furthermore, all three forms of annexin A2 (bovine lung or human recombinant annexin A2 expressed in E. coli or yeast) failed to interact with t-PA or plasminogen.
It was unclear if t-PA or plasminogen could bind to annexin A2 when the protein was a subunit of AIIt, i.e. bound to the S100A10 subunit. We chose two approaches to address this issue. First, we thiol-linked the S100A10 subunit to the BIAcore biosensor chip and demonstrated that this immobilized S100A10 could bind t-PA, plasminogen, and plasmin. This result is consistent with our previous data that S100A10 regulates plasmin formation and autoproteolysis (25,27,31,40,41). Second, we showed that incubation of AIIt with carboxypeptidase B, which results in the removal of the carboxyl-terminal lysines of the S100A10 subunit (28) but does not affect the annexin A2 subunit, blocked the binding of t-PA and plasminogen to AIIt or to the isolated S100A10 subunit. We concluded from these approaches that within AIIt, the carboxyl lysines of the S100A10 subunit form the t-PA-and plasminogen-binding site.
Our results contrast to earlier results in which annexin A2 was shown to bind both t-PA and plasminogen (37,42). The annexin A2 used in these early experiments was purified by passive elution of the protein from polyacrylamide gels. In subsequent reports a recombinant annexin A2 expressed in E. coli that contained an extended amino-terminal tag (MAS-MTGGQQMGRDP) and an extended carboxyl-terminal tag (LEHHHHHH) was utilized in binding experiments (45). These experiments and others (46) utilized an approach that involved the drying of annexin A2 on polystyrene plates followed by incubation with t-PA or plasminogen. The conformation or orientation that annexin A2 assumes when dried and rehydrated on polystyrene is unclear. A possible explanation for the differences between these studies and our work is that annexin A2 could have been denatured by the polystyrene surface used for immobilization. In fact, the denaturation of proteins caused by binding to polystyrene is a well established phenomenon (47)(48)(49). For example, it has been shown that when polyclonal or monoclonal antibodies were adsorbed onto polystyrene, they retained less than 20% or 1-2%, respectively, of their functional activity. Furthermore, the adsorption of bovine serum albumin to polystyrene particles causes irreversible changes in the stability and secondary structure of the protein.
It was originally demonstrated with nine different cell types that t-PA could inhibit the binding of plasminogen to cells (50). These authors also showed that under several conditions of cell treatment, t-PA and plasminogen receptor expression was modulated in parallel. Based on these data it was concluded that the cellular binding sites for t-PA were shared by plasminogen. Therefore, the presence of identical sites for t-PA and plasminogen within AIIt is consistent with previous reports. Furthermore, we have used an antisense strategy to lower the extracellular levels of S100A10 on the surface of HT1080 cells. Although the transfection of these cells with S100A10 antisense dramatically reduced the extracellular levels of S100A10, the extracellular levels of annexin A2 were unchanged. Interestingly, we showed that plasmin production by these cells was reduced by 80 -90% (31). These results established that annexin A2 was not involved in plasmin formation by these cells. Similarly, other laboratories report that annexin A2 does not colocalize with plasminogen on the cell surface (51)(52)(53).
Perhaps the most convincing evidence supporting the role of annexin A2 in plasminogen regulation was the demonstration that transfection of 293 cells with annexin A2 resulted in the stimulation of plasmin formation (42). However, these experiments require reinterpretation in the light of new data. Specifically, it was shown that annexin A2 regulates the expression of S100A10 protein by a post-translational mechanism (43,44). These authors demonstrated that transfection of HepG2 cells with annexin A2 resulted in the expression of both annexin A2 and S100A10 protein, suggesting that annexin A2 regulates the expression of S100A10 protein in these cells. We have also shown that transfection of 293 cells with annexin A2 results in an increase in annexin A2 and S100A10 protein (Fig.  5). Therefore, it was likely that the increased plasminogen binding observed with annexin A2-transfected 293 cells resulted from annexin A2-dependent stimulation of S100A10 expression. At the very least, this observation shows that it is very difficult to make conclusions about the function of annexin A2, based on observations made with annexin A2-transfected cells. Considering the demonstration that transfection of cells with annexin A2 results in the translocation of the normally cytosolic S100A10 to the plasma membrane (54 -56), it is likely that annexin A2 functions not as a plasminogen regulatory protein but to localize the S100A10 protein (presumably as AIIt) to the extracellular cell surface.
It was interesting that our binding experiments showed that both annexin A2 and S100A10 bound plasmin. The location of the plasmin-binding site on either of these subunits is not known. Recently, we demonstrated that both subunits of AIIt possessed a plasmin reductase activity (41). This activity was shown to be responsible for the release of anti-angiogenic plasminogen fragment, A 61 , from the plasmin molecule. It is, therefore, very likely that annexin A2 functions in the regulation of aspects of plasmin activity distinct from the regulation of plasmin protease activity.
A previous report suggested that annexin A2 served as a profibrinolytic coreceptor for both plasminogen and t-PA on the surface of endothelial cells and facilitated the generation of plasmin. These investigators reported that t-PA and plasminogen bound to annexin A2 at distinct amino-( 7 LCKLSL 12 ) and carboxyl-terminal (Lys 307 ) sites, respectively. Central to this model was the suggestion that the proteolytic cleavage of annexin A2 at Lys 307 -Arg 308 generated a carboxyl-terminal lysine that participates in plasminogen binding. This model has been difficult to reconcile with the reports that t-PA and plasminogen share a similar extracellular binding site in vivo (50).
Our current data favors a new model for the interaction of AIIt with plasminogen. We propose that the S100A10 subunit of AIIt is a key plasminogen regulatory protein (25,27). To support this hypothesis we have demonstrated that recombinant AIIt stimulates t-PA-catalyzed plasminogen activation about 77-fold compared with about 2-or 46-fold for recombinant annexin A2 or S100A10, respectively. We also have demonstrated that the loss of the carboxyl-terminal lysines of the S100A10 subunit of AIIt blocks AIIt-dependent stimulation of plasminogen activation. Most convincingly, we have shown that the loss of S100A10 from the extracellular surface of HT1080 cells results in inhibition of cellular plasmin production (31). The data presented in the current report demonstrating that membrane-associated annexin A2 does not bind t-PA or plasminogen further supports our hypothesis that the S100A10 subunit is the plasminogen receptor. Furthermore, we now map for the first time that the t-PA and plasminogenbinding site of S100A10 to its carboxyl-terminal lysines and propose distinct plasmin-binding sites on each subunit.