Interaction of the fibrinolytic receptor, annexin II, with the endothelial cell surface. Essential role of endonexin repeat 2.

Endothelial cells express a cell surface co-receptor for plasminogen and tissue plasminogen activator (t-PA) which we recently identified as annexin II (Hajjar, K. A., Jacovina, A. T., and Chacko, J. (1994) J. Biol. Chem. 269, 21191-21197). This protein enhances the catalytic efficiency of t-PA-dependent plasmin generation by 60-fold (Cesarman, G. M., Guevara, C. A., and Hajjar, K. A. (1994) J. Biol. Chem. 269, 21198-21203). Here, we demonstrate that annexin II is constitutively translocated to the endothelial cell surface within 16 h of biosynthesis, and that cell surface annexin II comprises 4.3 +/- 1.0% of the total cellular pool. Exogenous 125I-annexin II bound to EGTA-washed endothelial cells with high affinity (Kd 49 nM) and in a calcium-dependent (I50 = 3 microM), phospholipid-sensitive manner. Peptides KASMKGLGTDED and YDSMKGKGTRDK, mimicking the calcium-binding "endonexin" motif (KGXGT) of annexin II, blocked its interaction with endothelial cells. Recombinant annexin II, bearing the calcium-binding site substitution D161A of core repeat 2, failed to compete with binding of the wild type protein to the cell surface, while E246A and D321A mutants, corresponding to core repeats 3 and 4, behaved as effective competitors. These data suggest that translocated annexin II interacts with cell surface phospholipid via a high affinity calcium-dependent binding site that includes residues 118-122 (KGLGT) and the coordinating Asp161 of core repeat 2. Thus, calcium-regulated expression of annexin II on the endothelial cell surface may play a central role in control of plasmin-mediated processes.

Endothelial cells express abundant binding sites for the fibrinolytic zymogen plasminogen and its serine protease activator, tissue plasminogen activator (t-PA) 1 (1). Binding of plasminogen to endothelial cells enhances the catalytic efficiency of its activation by tissue plasminogen activator by at least 12fold (2). Cell surface plasmin activity may contribute to several endothelial cell functions including thromboresistance, angio-genesis, and activation of growth and differentiation factors (1). Recently, we identified annexin II as an endothelial cell surface co-receptor for plasminogen (PLG) and t-PA (3). When overexpressed in a renal epithelial cell line, this protein was translocated to the cell surface, conferred the capacity to bind both t-PA and PLG, and enhanced cell surface plasmin generation (3). Upon treatment of human endothelial cells with either EGTA, polyclonal anti-annexin II IgG, or antisense oligonucleotides directed against annexin II mRNA, binding of both t-PA and PLG decreased by approximately 50%, suggesting that annexin II accounts for about one-half of available endothelial cell binding sites for these ligands (3). Purified annexin II, moreover, specifically enhanced the catalytic efficiency of t-PA-dependent PLG activation by 60-fold in a cell-free system (4).
The annexins represent a family of some 13 calcium-dependent phospholipid-binding proteins, at least 10 of which occur in mammals (5,6). All annexins possess a variable amino-terminal "tail" domain followed by a conserved "core" region that imparts membrane-binding capability (7). The latter consists of either four or eight homologous "endonexin" motifs which contain the general consensus sequence (L/M)-K-G-X-G-T-(38 residues)-(D/E) (8).
Recent crystallographic studies have served to identify potential calcium-dependent phospholipid-binding sites of annexin V and annexin I (9,10). The pentapeptide sequence K-G-X-G-T within the endonexin motif constitutes a loop connecting the first and second ␣-helices of a given repeat within the core region (11). In addition, an acidic amino acid, located 38 residues downstream and within the junction of the fourth and fifth ␣-helices, coordinates with KGXGT motifs forming a potential calcium-binding site (11). Based on such studies, the major calcium binding domains of annexin V appear to reside within repeats 1, 2, and 4, while those of annexin I are located in repeats 1, 2, and 3 (9,10). Similarly, Gerke and co-workers have used site-directed mutagenesis to locate three potential calcium-binding sites within repeats 2, 3, and 4 of annexin II, as well as a weaker site within repeat 1. Half-maximal saturation of these sites occurs at 5-10 M and 200 -300 M Ca 2ϩ , respectively (12,13).
In the present study, we examined the mechanism by which annexin II associates with the endothelial cell surface. The data indicate that this interaction is an equilibrium-based, calcium-dependent binding event, with an absolute requirement for endonexin repeat 2. These findings suggest that calcium-mediated binding of annexin II to the cell membrane may play a central role in the control of cell surface plasminogen activation.

EXPERIMENTAL PROCEDURES
Materials-Bovine serum albumin, EGTA, CaCl 2 ⅐H 2 O, lactic dehydrogenase assay kit (no. 340-LD), MgCl 2 , phosphatidic acid (P9511), * This work was supported in part by National Institutes of Health Grants HL 42493 and HL 46403, and by an American Heart Established Investigatorship (to K. A. H.). 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 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed: Division of Hematology-Oncology, Dept. of Pediatrics, 1300 York Ave., Box  were purchased from Fisher. -Nitroblue tetrazolium chloride and 5-bromo-4-chloro-3-indolylphosphate -toluidine were obtained from Bio-Rad. Tris base was purchased from PolyScience, Na 125 I from ICN, and D-Val-Leu-Lys-aminofluoromethylcoumarin from Enzyme Systems Products. Phosphatidylinositol-specific phospholipase C was obtained from Boehringer Mannheim. Anti-annexin I, II, IV, VI, and p11 monoclonal IgG were from Zymed.
Purified Proteins-Native annexin II was purified from human placental membranes and characterized as described previously (4). This protein reacted strongly with antibody to annexin II, but not annexins I, IV, or VI. Purified annexin II was specifically found to be free of plasminogen activator activity in a fluorogenic assay as described previously (4). Purified recombinant wild type and mutant annexin II (D161A, E264A, and D321A) were a generous gift from Dr. Volker Gerke (University of Munster, Munster, Germany). These have been extensively characterized in previous studies (12,13).
Recombinant Annexin II-Using the wild type pCMV5-Ann-II plasmid (3) as a template, polymerase chain reactions were carried out using 26-mer oligonucleotide forward and reverse primers and Pfu polymerase. Primers (5Ј-GGTCGGGATCCGTCTACTGTTCACGA-3Ј and 5Ј-AAAAACTCGAGGTCATCTCCACCACA-3Ј) corresponded, respectively, to bases 52-66 and 1052-1066, numbered according to Huang et al. (14), and introduced a 5Ј BamHI restriction site and a 3Ј XhoI site while eliminating the mammalian initiation (ATG) and stop (TGA) codons. The modified insert was purified from agarose gel slices using the QIAEX kit, ligated into pET 21b(ϩ) plasmid (Novagen), and used to transform BL21 Escherichia coli. Bacteria were propagated to OD 0.6 in a 1-liter volume of Luria-Bertani medium containing chloramphenicol and carbenicillin. Protein expression was induced with 1 mM isopropyl-␤-D-thiogalactopyranoside (37°C, 3 h). The cells were pelleted (8000 rpm, JA-14 rotor), resuspended in a buffer containing 50 mM sodium phosphate, 300 mM NaCl, pH 8, and lysed by 1 cycle of freeze-thaw followed by sonication (4°C, two 3-min bursts, 1 min cooling, 250 watts). The supernatant was treated with DNase (5 g/ml) and RNase (10 g/ml) and then centrifuged at 12,000 ϫ g (TLA 100.4 rotor, 15 min, 4°C). The supernatant was loaded on a 4-ml QIAexpress nickel-nitrilo-triacetic acid-agarose column (20 ml/h) preequilibrated with sonication buffer. The loaded column was washed sequentially with sonication buffer (20 ml/h) and with wash buffer (50 mM Na phosphate, 300 mM NaCl, 10% glycerol, pH 6.0) until the effluent reached A 280 Ͻ0.01. The column was eluted with a 40-ml linear gradient (pH 6.0 -4.0) in wash buffer. Two-ml fractions containing monomeric annexin II and dimeric annexin II that eluted at pH 5.2 and 4.8, respectively, were each pooled and dialyzed overnight against 4 liters of PBS. Purified proteins were evaluated by SDS-PAGE on 8 -16% linear gradient gels (Novex) and by immunoblot analysis with a panel of anti-annexin monoclonal IgGs and stored at neutral pH at Ϫ70°C.
Cell Culture-Human umbilical vein endothelial cells (HUVEC, passage 1-4), HepG2 human hepatoma cells, and renal epithelial 293 cells were propagated and characterized as described previously (2,3,15). A human aortic smooth muscle cell line was propagated according to Legrand et al. (16). All cell cultures were supplied with fresh complete medium within 24 h of starting each experiment.
Flow Cytometric Analysis-Confluent HUVEC in 75-cm 2 flasks were washed three times with HBS (Hepes-buffered saline: 11 mM Hepes, 137 mM NaCl, 4 mM KCl, 3 mM CaCl 2 , 1 mM MgCl 2 , 1 mM glucose), and then soaked in HBS or HBS, 10 mM EGTA (HBS/EGTA, 30 min, 37°C). Cells were removed atraumatically from flasks by rinsing with HBS or HBS/EGTA, centrifuged (500 ϫ g, 7 min, 4°C), resuspended in 200 l of ice-cold HBS or HBS/EGTA, and treated with monoclonal anti-annexin II or anti-annexin VI IgG (50 g/ml, 60 min, 4°C). After washing in 10 volumes of either HBS or HBS/EGTA, the cells were resuspended in 200 l of the same buffer and treated with fluorescein isothiocyanateconjugated goat anti-mouse IgG (100 g/ml, 30 min, 4°C). The cells were washed again in 10 volumes of ice-cold HBS or HBS/EGTA, and resuspended in 300 l of the same buffer. Flow cytometric analysis (fluorescein-activated cell sorter) was carried out in a Coulter Epics XL flow cytograph.
Evaluation of Cell Integrity-Lactate dehydrogenase assays were employed as an index of cell lysis according to the method of Wroblewski and LaDue (17). 51 Cr release assays were carried out as described previously (18).
Competitive ELISA-According to the method of Hunter and Bosworth (20), microtiter plate wells (MaxiSorp, Nunc) were coated with 50 l of purified native annexin II (10 g/ml) diluted in 0.1 M NaHCO 3 , pH 9.6 (18 h, 4°C). Known concentrations of purified annexin II (0 -250 nM; 50 l), or unknown samples, were incubated (1 h, 4°C) with polyclonal anti-annexin II IgG (50 l, 1:1000) in a second microtiter plate pretreated with PBS, 0.5% Tween 20. After washing the annexin II-coated plate five times (PBS, 0.5% Tween 20), annexin II-anti-annexin II mixtures were added and incubated at 37°C, 30 min. The wells were washed five times with PBS, 0.5% Tween 20, and incubated with 50 l Possible calcium-binding motifs are shown between ␣-helices A and B in all four repeats, and their potentially coordinating amino acids, 38 residues downstream, between helices D and E. Adapted from Gerke (11).
Radiolabeling of Annexin II-Purified annexin II was radiolabeled according to the lactoperoxidase method of Martin et al. (21). Labeling was carried out for 3 min at 20°C in a total volume of 390 l at pH 7.4. The reaction mixture consisted of 7.5 M purified native annexin II, 1.3 M EDTA, 22 units/ml lactoperoxidase, 70 M H 2 O 2 in PBS. The reaction was stopped with excess KI, and bound 125 I was separated from unbound on a Sephadex G-25 (PD-10) column. The structural and functional integrity of labeled annexin II was verified by SDS-PAGE/ autoradiography and by plasmin generation assay (4).
Phospholipid Vesicles-Phospholipid vesicles were prepared according to the method of Ravanat et al. (22). Briefly, 4.5 mol of phosphatidylcholine plus 4.5 mol of phosphatidic acid, phosphatidylinositol, or phosphatidylserine, or 9.0 mol of phosphatidylcholine alone were dried from stock solutions in chloroform at 37°C under N 2 , and resuspended in 1.0 ml of 50 mM Tris, 0.1 M HCl, pH 7.5 (23). The lipid suspensions were sonicated continuously for 45 min under flowing N 2 at 4°C using a Branson sonifier 250 with a titanium probe tuned to constant duty cycle at 20% of maximum power. The resulting unilamellar vesicle suspension was centrifuged at 25,000 ϫ g (30 min, 4°C) to remove titanium particles, and the supernatant was stored under N 2 at 4°C in an airtight container.
Immunoblotting-Western blotting was carried out exactly as described previously (3).

RESULTS
Expression of Annexin II on the Endothelial Cell Surface-In previous studies, we and others have demonstrated the pres-ence of annexin II on the surface of cultured human umbilical vein endothelial cells using a variety of methods including flow cytometry, immunoprecipitation of surface labeled protein, and radioantibody binding analyses (3,24,25). In the present study, our goal was to characterize the mechanism by which annexin II, a protein lacking a classical transmembrane domain, associates with the endothelial cell surface. To test the potential role of calcium-regulated phospholipid binding domains of annexin II, HUVEC were treated with (Fig. 2, panels B and D) or without (panels A and C) 10 mM EGTA and then incubated at 4°C with anti-annexin II (panels A and B) or anti-annexin VI (panels C and D) monoclonal antibody followed by fluorescein isothiocyanate-conjugated goat anti-mouse IgG at 4°C. While cells treated with anti-annexin II in the absence of EGTA displayed positive immunofluorescence (panel A), cells pretreated with EGTA showed a 7-fold reduction in cell surface immunofluorescence (panel B) (mean relative fluorescence 1.10 and 0.16, respectively). Control HUVEC treated with anti-annexin VI showed low level fluorescence regardless of whether EGTA treatment was included in the protocol (panels C and D) (mean relative fluorescence 0.34 and 0.14, respectively). These data indicate that annexin II is expressed on the endothelial cell surface in a calcium-dependent manner.
To further characterize anti-annexin II cross-reactive material on the surface of endothelial cells, EGTA eluates of HU-VEC monolayers were analyzed by Western blotting (Fig. 3). As an index of cell lysis, EGTA eluates were assayed for lactate dehydrogenase activity. Lactate dehydrogenase activity in cell surface eluates did not differ from that in cell-free controls or in HBS cell surface eluates (9.0 Ϯ 2.4 units/ml, 8.9 Ϯ 2.9 units/ml, and 8.8 Ϯ 2.6 units/ml, respectively, S.D., n ϭ 3). Lactate dehydrogenase values were routinely less than 1% of the maximal release induced by freeze-thaw cell lysis (975 Ϯ 47 units/ ml, S.D., n ϭ 3). Similarly, release of 51 Cr from EGTA-treated endothelial cells did not differ from the spontaneous rate of release (12.4 Ϯ 0.8 versus 9.8 Ϯ 3.4% of maximal, respectively, S.D., n ϭ 3). These data indicated that the presence of annexin II in EGTA eluates was not the result of cell lysis.
EGTA eluates (Fig. 3A, lane 2), but not HBS eluates (lane 1) from HUVEC contained a ϳ36-kDa protein which reacted specifically with a monoclonal antibody directed against annexin II. This band co-migrated with authentic native annexin II (lane 4). A similar band was observed in eluates of fresh, intact umbilical vein (lane 3), suggesting that cultured HUVEC accurately reflect expression of annexin II on the surface of blood vessels in situ. When EGTA eluates of HUVEC were tested for reactivity with monoclonals directed against annexins I, IV, VI, or p11, an annexin II-associating protein, no reactivity was observed (Fig. 3B, lanes 2-5), although the same samples did react with anti-annexin II (Fig. 3B, lane 1). In addition, while EGTA eluates from human smooth muscle cells (Fig. 3C, lane  2), like those of HUVEC (lane 1), showed a 36-kDa crossreactive band, eluates from HepG2 hepatoma cells and renal epithelial 293 cells did not (lanes 3 and 4).
Finally, phosphatidylinositol-specific phospholipase C (26) failed to release annexin II from HUVEC as judged by immunoblot analysis (data not shown). This result contrasted with the release of the urokinase receptor, a known glycosylphosphatidylinositol-linked protein (27) which was easily detected in the phosphatidylinositol-specific phospholipase C supernatant. Together these data suggest that HUVEC, endothelial cells in situ, and smooth muscle cells all express annexin II as a peripheral membrane protein linked to the cell surface through a calcium-dependent mechanism.
To determine whether cell surface annexin II was synthesized by the endothelial cell, metabolic labeling studies were carried out (Fig. 4). HUVEC were treated for 16 h with [ 35 S]methionine and [ 35 S]cysteine and then surface-eluted with HBS with (panel A) or without (panel B) 10 mM EGTA. While whole cell lysates from both groups of cells contained numerous labeled bands upon SDS-gel fluorography (lanes 1 and 5), only the EGTA eluates (lane 2), and not HBS eluates (lane 6), contained labeled bands. When the same eluates were immunoprecipitated with polyclonal anti-annexin II IgG conjugated to Sepharose beads, a ϳ36-kDa band was recovered from EGTA eluates (lane 4), but not from the HBS samples (lane 8). This band was not recovered upon precipitation with preimmune IgG-conjugated beads (lanes 3 and 7). These data demonstrate that cell surface annexin II is synthesized by cultured endothelial cells and translocated to the cell surface within 16 h.
To quantify the relative amounts of annexin II both on the cell surface and within the cell, a competitive ELISA was developed (20). A standard curve was linear at concentrations of purified native annexin II up to ϳ250 nM (r ϭ 0.98) and displayed a lower detection limit of ϳ0.5 ng/well. While cell lysates from ϳ4 ϫ 10 6 HUVEC contained annexin II in a concentration of 2.5-5.3 M, EGTA eluates contained 70 -190 nM annexin II. HBS eluates, prepared in parallel, routinely showed no detectable antigen. EGTA-elutable cell surface an-nexin II represented 4.3 Ϯ 1.0% (mean Ϯ S.E., n ϭ 4) of total cellular annexin II.
Equilibrium Binding of Purified Annexin II to Cultured Endothelial Cells-To further characterize the interaction between annexin II and the endothelial cell surface, direct binding studies using radiolabeled native annexin II were conducted (Fig. 5). In time course studies, binding of a fixed concentration of 125 I-annexin II (8 nM) to EGTA-washed HU-VEC reached a steady state at approximately 60 min. At least 60% of total binding was reversible upon "infinite dilution" of the unbound ligand (Fig. 5A), indicating a dynamic equilibrium between bound and unbound ligand. In addition, EGTA-treated endothelial cells bound native annexin II in a dose-dependent and apparently saturable fashion (Fig. 5B). Binding was halfmaximal at an input dose of ϳ50 nM, and approached a plateau at 75-100 nM. Scatchard analysis (28) (Fig. 5B, inset) suggested a single saturable site with K d 49 Ϯ 9 nM (S.E., n ϭ 3), and B max 1,637,000 Ϯ 403,000 (S.E., n ϭ 3), indicating a high affinity, high capacity interaction.  3 and 7).

FIG. 5. Binding of 125 I-annexin II to HUVEC.
Binding studies were conducted as described under "Experimental Procedures." A, time course and reversibility. Triplicate sets of HUVEC monolayers were incubated with 125 I-labeled native annexin II (329,000 cpm/pmol, 8 nM, 200 l/well) for 0 -120 min (q). In parallel sets of wells (E), unbound ligand was rapidly removed at 60 min and the wells filled with IB(5) to approximate "infinite dilution." At the indicated times, wells were washed rapidly three times, and the bound radioactivity was assayed upon solubilization with 1% SDS, 0.5 M NaOH, 10 mM EDTA (1 h, 58°C). B, binding isotherm. Confluent HUVEC were incubated with a range of concentrations of 125 I-annexin II (0 -100 nM, 116,000 cpm/pmol, 60 min, 4°C), and bound and free ligand estimated. Inset, Scatchard plot. The binding isotherm was linearized according to the "LIGAND" program (28).
Annexin II may be subject to at least four types of posttranslational modification including myristoylation (29), glycosylation (30), phosphorylation (31), and proteolytic cleavage (3). Interestingly, both native and recombinant human annexin II produced in E. coli interacted with HUVEC with comparable affinity (Fig. 6), suggesting that cell surface binding does not require post-translational modification. Both labeled and unlabeled recombinant annexin II competed with labeled native protein for binding to HUVEC (I 50 ϳ10and 12-fold molar excess, respectively) (Fig. 6A). Similarly, both unlabeled native and recombinant annexin II competed with the labeled recombinant protein for binding to HUVEC (I 50 ϳ5and 8-fold molar excess, respectively) (Fig. 6B). In all cases, 60 -80% of total binding was inhibited in the presence of excess competing protein. Interestingly, unlabeled recombinant annexin I also competed with native 125 I-annexin II for cell surface binding (I 50 ϳ7-fold molar excess; data not shown).
Binding of Annexin II to the Cell Surface Is Calcium-dependent-One of the salient properties of annexin II is its capacity for calcium-dependent binding to phospholipid-containing membranes (5,6). This process is enhanced by the annexin II-binding protein, p11, and reduced upon phosphorylation of annexin II by p60 v-src (32). We examined the mechanism of annexin II surface association in more detail. As shown in Fig.   7, binding was calcium-dependent, with maximal specific binding observed at pCa 4.0 -4.5 (ϳ30 -100 M), and a nadir at pCa 6.5-8.0 (10 -300 nM) (Fig. 7A). Half-maximal binding was observed at pCa ϳ5.5 (ϳ3 M), consistent with the involvement of high affinity calcium-binding sites residing within domains 2, 3, and 4 (K d 5-10 M) (13). In the presence of other divalent cations such as Mg 2ϩ , Sr 2ϩ , Zn 2ϩ , and Cu 2ϩ , however, specific binding was no more than 16% of that observed in the presence of Ca 2ϩ (Fig. 7B). These data indicate a specific requirement for calcium in a manner that implicates one or more high affinity endonexin repeats of the core region of annexin II.
Since calcium-binding sites regulate interactions between the annexins and cellular membranes, one would expect binding to be inhibitable by anionic phospholipid. Binding of 125 Iannexin II to HUVEC was blocked in a dose-dependent manner in the presence of vesicles composed of anionic phospholipid in a 1:1 molar ratio with phosphatidylcholine (Table I). Vesicles containing phosphatidylserine or phosphatidic acid blocked approximately 50% of specific binding of native annexin II to HUVEC at concentrations of 9 and 43 M, respectively. At a maximum dose of 125 M, these vesicles blocked between 66 and 94% of specific annexin II binding. Phosphatidylinositol was somewhat less effective, inhibiting little more than 60% of specific binding with an I 50 of 85 M. The uncharged phospholipid, phosphatidylcholine, was the least effective, inhibiting only 26% of annexin II binding at doses of 100 -150 M. These results agree reasonably well with the reported K d (1.8 M) for annexin II and phosphatidylserine-containing vesicles in the presence of micromolar Ca 2ϩ (6) and suggest that annexin II can target cell surface phospholipid with greatest affinity for phosphatidylserine.
In competitive binding experiments, peptide 114 -125 from repeat 2 inhibited binding of 125 I-annexin II to HUVEC in a dose-dependent manner (Fig. 8A). Half-maximal inhibition was observed at a peptide concentration of 0.35 mM. Maximal inhibition of binding occurred in the 0.5-1.0 mM range. A partially scrambled peptide (KASMTGLGKDED) in which the first and fifth pentapeptide residues were interchanged inhibited binding by no more than 20%. In addition, neither the repeat 1 peptide (ETAIKTKGVDEV) nor an unrelated control peptide (DRVYIHPFHL) inhibited binding at all. These data suggested a role for annexin II repeat 2, but not repeat 1, in binding to endothelial cells.
At the same time, peptides mimicking annexin repeats 3 and 4 were studied (Fig. 8B). The peptide YDSMKGKGTRDK from repeat 4 blocked binding of annexin II to HUVEC in a dose-dependent manner. Half-maximal inhibition was observed at 0.2 mM peptide, and maximal inhibition (80%) at doses between 0.5 and 1.0 mM. A partially scrambled control peptide (YDSMT-GKGKRDK), in which the first and fifth pentapeptide residues were interchanged, was much less effective as an inhibitor, blocking no more than 40% of binding at a dose of 1 mM. Neither the peptide representing repeat 3 (DAGVKRKGT-DVP), nor an unrelated control peptide (RNPDADTGPW), blocked binding to a significant degree. These data suggest a possible role for repeat 4, but not repeat 3, in its interaction with endothelial cells. Together, the data from panels A and B of Fig. 8 indicate that the KGXGT motif is necessary for cell surface binding of annexin II.
To further explore the contributions of repeats 2 and 4 in cell surface binding, acidic residues (D or E) located 38 residues downstream of the KGXGT motifs were studied. Three well characterized recombinant annexin II mutants (D161A, E264A, and D321A from repeats 2, 3, and 4, respectively) (12,13) were examined in competitive binding experiments (Fig. 9). Recombinant wild type annexin II, produced by the same vector and bacterial cell host, was used as a positive control. Wild type annexin II, as well as the E246A and D321A mutants, effectively blocked binding of annexin II to HUVEC with 50% inhibition observed at a 2-, 5-, and 4-fold molar excess, respectively. Maximal inhibition of binding was 70 -80% at a 20 -50-fold molar excess. The D161A mutant, on the other hand, blocked only 20 -30% of specific binding even at the highest doses tested (250 nM). These data suggest that the KGLGT motif, in coordination with Asp 161 , is the major calcium-binding site that regulates cell surface association of annexin II.

DISCUSSION
This study demonstrates for the first time that cell surface association of the fibrinolytic receptor, annexin II, is a high affinity binding event that specifically requires endonexin repeat 2. The data indicate that annexin II is synthesized by cultured endothelial cells and translocated within hours to the cell surface in a calcium-dependent interaction. By quantitative ELISA, cell surface annexin II appears to represent approximately 4 -5% of the total endothelial cell pool. Cell surface annexin II was not detected on hepatoma cells or renal epithelial cells, but did appear on the surface of human smooth muscle cells.
As a class, the annexins were originally identified in intracellular locations and shown to be involved in membrane fusion events (5,6). Recently, however, several annexins have been found to be localized on cell surfaces or within extracellular compartments (33). Annexin II, for example, has been identified on the surface of human colon adenocarcinoma cells (34), murine large cell lymphoma cells (35), and endothelial cells (3,24,25). Annexin II, as well as annexin V, is released in microvesicles from chondrocytes during bone mineralization (36 -38), and from skin keratinocytes (39). Annexin I is selectively secreted in high concentrations by the human prostate (40) and by inflammatory cells (41). Annexin V is found in plasma, amniotic fluid, and post-culture medium from endothelial cells and endometrium (42). The specific stimuli for these redistribution events have yet to be determined, and their elucidation may provide insight into the regulation of pericellular protease activity.
In the present study, several lines of evidence support an annexin II-cell surface interaction which is absolutely dependent upon micromolar free Ca 2ϩ . First, surface-associated immunoreactive material was efficiently eluted upon chelation of extracellular Ca 2ϩ with EGTA (Fig. 2). Second, 36-kDa crossreactive material that comigrated with authentic annexin II was recovered in EGTA eluates of endothelial cells both in vitro and in situ (Fig. 3). Third, metabolically labeled annexin II was Endothelial Cell Surface Annexin II eluted from HUVEC cell surface with EGTA, but not HBS (Fig.  4). Fourth, calcium-dependent rebinding of annexin II to EGTA-treated endothelial cells was demonstrated (Fig. 5). The half-saturating calcium concentration (pCa 5.5 or ϳ3 M) agreed closely with that of the high affinity calcium-binding sites within repeats 2, 3, and 4 of the core region of annexin II (K d 5-10 M Ca 2ϩ ) (13), and with the calcium dependence of annexin II binding to phosphatidylserine-containing vesicles (K d 2-5 M Ca 2ϩ ) (32,43). Since plasma free Ca 2ϩ is ϳ1 mM, this interaction would be supported in vivo.
To further define the cell surface binding domain of annexin II, two types of experiments were conducted. In the first, peptides mimicking the four potential endonexin repeats of annexin II were tested for their ability to block binding of the parent molecules to cultured endothelial cells. Of these peptides, two, corresponding to repeats 2 and 4 and sharing the K-G-X-G-T motif (KGLGT and KGKGT), specifically blocked binding (Fig. 8). The essential nature of this pentapeptide sequence was further underscored by the fact that reversal of the initial K and final T residues rendered the peptides almost completely inactive. Given the similarity between these two peptides (KG(L/K)GT), it is not possible to determine whether repeat 2 and 4 peptides were blocking repeat 2, repeat 4, or both.
To further refine the analysis, therefore, recombinant forms of annexin II mutated at the coordinating downstream acidic residue (D or E) of repeat 2, 3, or 4 were examined for their ability to compete with the wild type protein for binding to the cell surface. Because only the mutant corresponding to repeat 2 failed to block binding (Fig. 9), the data suggest that repeat 2 plays a dominant role in mediating the cell surface interaction. This result is also consistent with our previous observation that cell surface annexin II is subject to proteolytic cleavage at Lys 307 -Arg 308 . This modification creates a carboxyl-terminal lysine residue, enabling binding of plasminogen via lysinebinding "kringle" structures (3). Cleavage at Lys 307 -Arg 308 would release a carboxyl-terminal fragment (Arg 308 -Asp 328 ) that contains Asp 321 , thereby inactivating the Ca 2ϩ -binding site of repeat 4. Thus, carboxyl-terminal modification of annexin II is consistent with the hypothesis that repeat 2 is primarily responsible for Ca 2ϩ -mediated cell surface binding of annexin II.
The binding isotherm depicted in Fig. 5 indicates a high affinity interaction between annexin II and the endothelial cell surface (K d 49 nM). The dissociation constant is reasonably close to K d values reported for the interaction of annexin V with platelets (7 nM) (44), ovarian carcinoma cells (9 nM) (45), or endothelial cells (16 nM) (46). The slightly higher affinity for annexin V, compared with annexin II, is consistent with the former's relatively more avid binding to phospholipid (47). The preferential inhibition of binding of annexin II to endothelial cells by anionic phospholipid over neutral phospholipid most likely reflects greater affinity for phosphatidylserine over phosphatidylcholine.
In summary, the present data indicate that expression of annexin II on the endothelial cell surface results from translocation of newly synthesized annexin II, and its calcium-dependent association with outer leaflet phospholipid. This interaction requires the high affinity calcium-binding site of repeat 2. Although the mechanism of annexin II transport by the endothelial cell is unknown, several other proteins including fibroblast growth factor-1, interleukin-1␤, thioredoxin, and lectin L-29 that lack typical signal peptides, are known to undergo leaderless secretion in response to specific stimuli (48 -52). Highly regulated, nonclassical release of these proteins may represent a newly recognized form of host defense (33, 53). Since annexin II represents a major fibrinolytic receptor that FIG. 8. Effect of mimetic peptides on binding of annexin II to HUVEC. HUVEC were incubated with 125 I-labeled native annexin II (40 nM, specific activity 161,000 cpm/pmol) in the presence of 0 -1.0 mM peptide (60 min, 4°C) in IB (5). A, peptides mimicking repeat 1 (ETAIK-TKGVDEV) and repeat 2 (KASMKGLGTDED and KASMTGLGKDED), as well as an unrelated peptide (DRVYIHPFHL) were employed. B, peptides mimicking repeat 3 (DAGVKRKGTDVP) and repeat 4 (YDSMKGKGTRDK and YDSMTGKGKRDK), as well as an unrelated control peptide (RNPDADTGPW) were employed. Sequences related to K-G-X-G-T motifs are underlined.
FIG. 9. Inhibition of annexin II binding to HUVEC by mutant annexin II proteins. HUVEC equilibrated at 4°C were incubated with recombinant 125 I-annexin II (5 nM, specific activity 179,000 cpm/ pmol) in IB (5) in the presence of 250 nM unlabeled recombinant wild type (q), or D161A (f), E246A (ç), or D321A (å) mutants (60 min, 4°C). co-binds plasminogen and tissue plasminogen activator and augments the efficiency of plasmin generation, it is reasonable to hypothesize that regulation of secretion and/or binding of annexin II to cell surfaces may play a central role in the control of plasmin-mediated processes.