A Novel Method for Measuring Cell Surface-bound Thrombin DETECTION OF IODINATION-INDUCED CHANGES IN THROMBIN-BINDING AFFINITY*

A method is presented, which we call the hirudin assay, for measuring picogram amounts of thrombin bound to the cell surface of normal mouse fibroblasts. After incubation of cells with hirudin, virtually all cell-bound thrombin is released into the medium. Impor- tantly, over 90% of cell-dissociated thrombin is in complex with hirudin as shown by native gel electropho- resis. Therefore, by incubating cells with i261-hirudin and precipitating '2SI-hirudin-thrombin complexes with an antibody to thrombin, we can quantitatively meas- ure the amount of cell-bound thrombin. Thrombin linked to protease-nexin is tightly bound to the cell surface and only a small amount is released into the medium in the presence of hirudin. Thus, the hirudin assay primarily measures free, unlinked thrombin bound to cells. Comparison of the binding of "%thrombin and unlabeled thrombin to mouse cells using the hirudin assay shows that i261-thrombin detects only about one-half of the binding sites that thrombin does. Additionally, by incubating a constant amount of 'asI-thrombin or thrombin with increasing amounts of membrane preparations, most thrombin molecules, but only 30% of IzaI- thrombin molecules, bind to cellular receptors. The discrepancies between lZ6I-thrombin and thrombin binding to mouse cells are caused, at least in part, by the of a large fraction of '261-thrombin con- taining diiodotyrosine were added to the protein precipitates and samples were hydrolyzed in sealed glass ampules for 3-6 h at 110 "C. After neutralization with sulfuric acid, extracts of the barium sulfate precipitate (in acetone/acetic acid, 32, v/v) were applied to Silica Gel 60 aluminum sheets using a I-butanol/acetic acid/water (lO:l:l, v/v) developing solvent (25). Standards were identified with ninhydrin and samples were identified by autoradiography. 10% or less of input radioactivity was recovered. These results indicate that at least 90% of the acid-precipitable radioactivity in Iz5I-hirudin preparations was not hirudin. This occurred because the hirudin that was iodinated contained substantial amounts of non-hirudin protein as demonstrated by its low specific activity. However, the radioactive non-hirudin proteins in l2'1-hirudin preparations did not interfere with the hirudin assay and background radioactivity in tubes without thrombin was less than 1% of input radioactivity. Additionally, using native gel electrophoresis, we find that after addition of 1Z51-hirUdin to cells containing cell-bound thrombin, only '2sII-hirudin-human thrombin complexes are observed (not shown). These results indicate that labeled non-hirudin proteins in Iz5I-hirudin preparations do not bind tightly to cellular components and are not measured using the hirudin assay.

that each possesses high affinity binding sites for thrombin. By analogy with the action of hormones such as epidermal growth factor (5), these thrombin receptors have been implicated in the action of thrombin on platelets (6, 7 ) , chick embryo fibroblasts (8, 9), and mouse embryo fibroblasts (10).
To study the interaction of thrombin with these binding sites or receptors, thrombin has been iodinated using chloramine-T ( l l ) , lactoperoxidase coupled to Sepharose (7), and soluble lactoperoxidase (6). At low levels of iodination, lz5Ithrombin appears to retain most of its proteolytic activity as evidenced by its ability to cleave fibrinogen. However, Iz5Ithrombin loses stability, especially at high specific activities, and must be used soon after labeling (12). Because of this instability, long term cell-binding experiments cannot be done, which precludes determination of receptor occupancy by thrombin over the 2-day course of a mitogenic experiment.
A number of reports have shown that iodination of polypeptide hormones can both alter their affinities for cell surface receptors (13) and their biological activities (13,14). Although '251-thrombin retains its ability to cleave fibronogen (6), it is not known whether the fibrinogen binding site of thrombin is identical with its cellular binding site. Because of these problems and uncertainties when using Iz5I-thrombin for thrombinbinding studies, we sought an alternative means of detecting cell-bound thrombin. In this paper, we present a cellular binding assay which will quantitatively detect unlabeled thrombin bound to mouse cells. Results obtained using this assay for unlabeled thrombin differ significantly from results obtained using 'z51-thrombin. We show that these differences are caused by iodination-induced changes in the affinity of thrombin for its receptor. solution, glutamine, and antibiotics were from Gibco, and calf serum was from Irvine Scientific (Santa Ana, CA). Na-Iz5I was obtained from Amersham and 1,3,4,6-tetrachloro-3a,6ol-diphenylglycoud (IODO-GEN) was from Pierce.
Iodination and Binding-Hirudin (2 pg) was radioiodinated using IODO-GEN (16) and Na-Iz5I (1 mCi). The specific activity of lZ5Ihirudin was determined by spotting dilutions of the postiodination solution and postdialysis solution, respectively, on Whatman 3MM filter paper discs. After drying, discs were added to a beaker containing 500 ml of ice-cold 10% trichloroacetic acid containing 50 m~ KI, rinsed in this trichloroacetic acid solution, and the radioactivity remaining on each disc was determined using a Beckman gamma counter. Specific activities for '251-hirudin ranged from 2 X 10' to 3 x 10' cpm/pg. Iz5I-Hirudin was frozen at -80 "C in small diquots and was stable for at least 3 weeks.
Thrombin was radioiodinated using the lactoperoxidase procedure described by Martin et al. (6). The specific activity of '251-thrombin was determined as described above for I2'I-hirudin except that samples were first precipitated in trichloroacetic acid and then washed on Whatman GF/C glass filter discs. The radioiodination protocol described above yielded '2s1-thrombin with 0.3 to 0.4 mol of iodide/mol of thrombin and a specific activity of about 5 X IO6 cpm/pg. These preparations retained 100% of their proteolytic activity as determined using a fibrinogen clotting assay (17).
I-thrombin was prepared in parallel with Iz5I-thrombin except that Na-lZ5I was replaced with the same amount of NaI. We assumed that the levels of iodination of I-thrombin and 'Z51-thrombin preparations, prepared at the same time, were equal. I-thrombin containing 1.5 mol of iodide/mol of thrombin was prepared by adding 3-fold more NaI to iodination mixtures than described above for Iz5I-thrombin. These Ithrombin preparations retained about 30% of their fibrinogen clotting activity.
HF fibroblast-like cells and secondary mouse embryo fibroblastlike cells were cultured as described (2). Upon reaching confluent levels, cultures were rinsed and incubated in serum-free DV medium for 48 h. After one additional rinse in DV medium, cultures were incubated with 1251-thrombin in 2 ml of binding medium for the indicated times. The presence of intracellular '"I-thrombin-PN complexes was determined by addition of trypsin (0.125% in DV medium) for 5 min at 37 "C as described (18). To terminate binding reactions, unbound 1251-thrombin was removed by quickly rinsing culture dishes 6 times with cold D-PBS on ice. The total radioactivity remaining on each dish was determined by solubilization of cells in 1 ml of 1 N NaOH for 3 h at 37 "C and quantitation of radioactivity in a gamma counter. Alternatively, cells were solubilized in 200 p1 of sample buffer (20 mM bis-tris, pH 7.5, 1% SDS, 2 mM EDTA, 7 m dithiothreitol, 10% glycerol) and heated at 100 "C for 3 min before analysis by SDS-PAGE, described below.
Hirudin Assay-Increasing amounts of thrombin, diluted in D-PBS containing l% BSA, were added to Microfuge tubes. 0.75 ml of D-PBS containing BSA (1%) and '251-hirudin was then added, followed by addition of 5 pl of rabbit anti-thrombin antiserum prepared as described (19). After incubation for 60 min at 37 "C with rocking, protein A Sepharose (50 pl of a 1:IO dilution in D-PBS) was added. After incubation for 30 min at 37 "C with rocking, the Sepharose beads were carefully washed twice by centrifugation for 2 min at 10, OOO X g in 0.75 ml of D-PBS and once in a detergent solution containing 0.05% Nonidet P-40, 0.1% SDS, 0.3 M NaCl, 10 mM Tris-HC1, pH 8.6 (20). After a final wash in D-PBS, the radioactivity remaining in tubes was determined using a gamma counter.
Measurement of Thrombin Binding Using the Hirudin Assay-To measure the amount of thrombin bound to mouse cells, thrombin was incubated with cultures in binding medium as described above for 1Z51-thrombin binding. To terminate the binding reaction, cells were rinsed 4 times in cold D-PBS and once in D-PBS containing 0.1 mM PhAsO. After a further rinse in D-PBS, 0.75 mI of D-PBS containing BSA (1%) and 35 ng of 1251-hirudin (specific activity of about 6 X lo7 cpm/pg) was added and cultures were incubated for 30 min at 4 "C. D-PBS containing BSA and Iz5I-hirudin were also added to dishes (without cells) containing increasing amounts of thrombin (up to 10 ng) to produce a standard curve. After an additional incubation for 3 h at 37 "C, supernatant solutions were transferred to 1.5-ml Microfuge tubes and processed as described above for the hirudin assay. The amount of radioactivity in each tube was converted to the amount of cell-bound thrombin by reference to the standard curve.
Membrane Binding Assay-Mouse cell membranes were prepared by harvesting secondary mouse cells from dishes in cold 10 mM sodium phosphate buffer (pH 7.0) containing aprotinin (100 Kallikrein inactivating units/ml). Cells were lysed by forcing them through a 27gauge needle. The cell homogenate was centrifuged for 5 min at 1,500 x g to remove nuclei and the supernatant solution was further centrifuged for 30 min at 10, OOO X g to obtain membrane pellets. Membranes were washed twice in D-PBS and resuspended in binding medium before use in the membrane binding assay described in Fig.  5. The amount of thrombin bound to membranes was determined by addition of 0.75 rnl of D-PBS containing BSA (1%) and Iz5I-hirudin (10 ng) to each tube. Membranes were resuspended by agitation and tubes were rocked for 3 h at 37 "C. After centrifugation for 5 min at 12,000 x g to remove membranes, the amount of membrane-dissociated human thrombin in supernatant solutions was determined using the hirudin assay.
Polyacrylamide Gel Electrophoresis and Iodotyrosine Analysis-Medium containing 'z51-thrombin and 1z51-thrombin-hirudin complexes was analyzed using a bis-tris native gel system, pH 7.5 (System 1935 described by Chrambach et al.) (21) except that the stacking buffer pH at 22 "C was 6.8. For SDS-PAGE using this bis-tris buffer system, the upper reservoir buffer contained 0.1% SDS. SDS-PAGE using a Tris buffer system was performed as described by Laemmli (22) using a 5% acrylamide stacking gel and a 10% acrylamide separating gel. Quantitation of radioactivity in gel slices (2 mm each) was carried out as described (23).
Gel slices containing '251-thrombin and Iz5I-thrombin-PN were eluted by electrophoresis through a 5% acrylamide stacking gel (22) into dialysis bags. After dialysis against water, samples were dried using a Speed Vac Concentrator (Savant) and SDS was removed by addition of 200 p1 of acetone/triethylamine/acetic acid/water, 8 5 5 : 5 5, v/v, for 10 h at 0 "C (24). Following addition of 16 nmol of MIT and DIT, 300 pl of 0.38 N barium hydroxide were added to the protein precipitates and samples were hydrolyzed in sealed glass ampules for 3-6 h at 110 "C. After neutralization with sulfuric acid, extracts of the barium sulfate precipitate (in acetone/acetic acid, 3 2 , v/v) were applied to Silica Gel 60 aluminum sheets using a I-butanol/acetic acid/water (lO:l:l, v/v) developing solvent (25). Standards were identified with ninhydrin and samples were identified by autoradiography.

Assay for the Detection of Picogram Amounts of Thrombin
We developed a highly sensitive assay for detecting thrombin by using the leech salivary protein hirudin. Hirudin (Mr -8OOO) has a high affinity for thrombin and its interaction with thrombin is highly specific (26,27). Hirudin was iodinated to a high specific activity (about 2.5 x lo5 cpm/ng) and incubated with increasing amounts of thrombin. These mixtures, containing '251-hirudin-thrombin complexes, were incubated with rabbit anti-thrombin antiserum and protein A Sepharose to specifically adsorb 1251-hirudin-thrombin complexes. As shown in Fig. 1, the maximum sensitivity of the assay was about 50 cpm/pg of thrombin? The linear range of Based on the data shown in Fig. 1 and the specific activities of the thrombin and hirudin used in these experiments, about 35 ng of hirudin (3.5 X lo-* units) effectively bound to 10 ng of thrombin (3.1 X IO-' units). Thus, it appears that iodination did not damage the hirudin in these preparations, since by definition 1 unit of hirudin will inactivate 1 unit of thrombin. However, using the specific activity of '"I-hirudin preparations in similar calculations yields a molar ratio of hirudin/thrombin at saturation of about 0.1. Additionally we found that in assay tubes containing saturating amounts of thrombin, 10% or less of input radioactivity was recovered. These results indicate that at least 90% of the acid-precipitable radioactivity in Iz5I-hirudin preparations was not hirudin. This occurred because the hirudin that was iodinated contained substantial amounts of non-hirudin protein as demonstrated by its low specific activity. However, the radioactive non-hirudin proteins in l2'1-hirudin preparations did not interfere with the hirudin assay and background radioactivity in tubes without thrombin was less than 1% of input radioactivity. Additionally, using native gel electrophoresis, we find that after addition of 1Z51-hirUdin to cells containing cell-bound thrombin, only '2sII-hirudin-human thrombin complexes are observed (not shown). These results indicate that labeled non-hirudin proteins in Iz5I-hirudin preparations do not bind tightly to cellular components and are not measured using the hirudin assay.

Measurement of Cell
Surface-bound Thrombin 150 -) " " -" " " " Measurement of thrombin and DIP-thrombin using the hirudin assay. After addition of increasing amounts of thrombin and DIP-thrombin to Microfuge tubes, either low specific activity (1.3 X lo' cpm/pg; 35 ng) or high specific activity (2.6 X 10' cpm/pg; 15 ng) '251-hirudin was added. Immediately following 12'II-hirudin addition, rabbit anti-thrombin antiserum was added (5 pl/tube). Each tube contained a total volume of 0.75 ml and a final BSA concentration of 1%. After 30 min at 22 "C, protein A Sepharose (50 pl) was added, the tubes were washed, and the radioactivity in each tube was determined as described under "Experimental Procedures." A, measurement of thrombin using high specific activity 1251-hirudin; 0, measurement of thrombin low specific activity '251-hirudin; 0, measurement of DIP-thrombin using low specific activity '2'II-hirudin. the assay was extended by addition of unlabeled hirudin to 'z51-hirudin preparations before incubation with thrombin ( Fig. 1). The rabbit anti-thrombin antiserum used was raised against DIP-thrombin since human thrombin is poorly antigenic (19). Interestingly, DIP-thrombin was detected very poorly by the hirudin assay ( Fig. 1).

Assay for the Detection of Thrombin Bound to the Cell Surface of Fibroblasts Release of 1251-Thr~mbin from the Surface of ME Cells-
To measure thrombin bound to the cell surface of fibroblasts, we first determined the conditions which would allow complete dissociation of thrombin from the cell surface. To do this, '251-thrombin was incubated with mouse cells for 15 min. Cells were rinsed thoroughly and incubated in BSA-containing buffer with or without the addition of hirudin or PhAsO. After 2 h, both cell and medium samples were analyzed by SDS-PAGE. Table I shows that after a 2-h incubation in buffer alone, about 75% of the cell-bound '251-thrombin was lost. However, only a fraction of this "dissociated" '251-thrombin was recovered in the medium, indicating that some '251-thrombin was degraded by the cells. To confirm this hypothesis, cells were treated with PhAsO just prior to the 2-h incubation. We have shown previously that PhAsO does not affect binding of 'Z51-thrombin to human fibroblasts but completely blocks its internalization (18). As shown in Table I, pretreatment with PhAsO did not significantly change the amount of lz5Ithrombin bound to cells after the 2-h incubation. However, all of the Iz5I-thrombin lost from cells during the 2-h incubation was recovered as intact '251-thrombin in the medium. These results indicated that cells internalized and degraded a portion of the IZ5I-thrombin during the 2-h incubation in the absence of PhAsO.
Addition of hirudin during the 2-h incubation resulted in a 95% reduction in cell-bound '251-thrombin. All of this dissociated '251-thrombin was recovered in the medium (Table I). Importantly, this occurred in the absence of PhAsO, indicating that hirudin blocked internalization of '251-thrombin by the mouse cells. Thus, after incubating cells for 2 h in the presence of hirudin, 95% of cell-bound '251-thrombin was lost from the cell surface and could be recovered in the medium.  ~~ only thrombin-hirudin complexes were detected by the assay, it was important to show that thrombin which dissociated from cells in the presence of hirudin was also in complex with hirudin. After addition of '251-thrombin to mouse cells cultures, cells were incubated in the presence or absence of hirudin for 2 h. Media samples were then analyzed by native PAGE. In this native gel system, free '251-thrombin barely entered the separating gel, but '251-thrombin-hirudin complexes migrated well into the separating gel. As shown in Fig.  2 A , Iz51-thrombin which dissociated from cells in the absence of hirudin migrated to the stacking gel/separating gel interface. However, at least 90% of '251-thrombin which dissociated from cells in the presence of hirudin was also in complex with hirudin as evidenced by its migration into the separating gel. Importantly, little if any '251-thrombin which bound to cells nonspecifically was found in complex after the 2-h incubation with hirudin (Fig. 2B). Nonspecifically bound '251-thrombin is defined as '251-thrombin which binds to cell cultures in the presence of a large excess of thrombin. A comparison of Fig.  2A and Fig. 2 3 showed that the amount of '251-thrombin which was not in complex with hirudin was about equal to the amount of nonspecifically bound '251-thrombin. Therefore, in Fig. 2A the fraction of '251-thrombin not in complex with hirudin appeared to be nonspecifically bound '251-thrombin. Taken together, these results showed that after incubation of cells in the presence of hirudin for 2 h, almost all specifically cell-bound '251-thrombin was cell-dissociated and in complex with hirudin.

Measurement of '251-thrombin-PN by the Hirudin Assay-
Over 90% of Iz5I-thrombin specifically bound to mouse cells is bound as free '251-thrombin. The remaining specifically bound 'Z51-thrombin is found in complex with a protein of M , -38,000 that we have named PN (23, 28). PN probably would not play a significant role in thrombin-binding studies using mouse cells, since it accounts for less than 10% of the total specifically bound 1251-thrombin.3 However, about 70% of ' ' ' Ithrombin specifically bound to human foreskin fibroblast-like cells is in complex with PN (28). Thus, it was important to determine if, under the conditions described above, thrombin-PN complexes could dissociate from cells and might thus be available for detection by the hirudin assay. '251-Thrombin was incubated with human cells for 2 min to avoid internalization of thrombin which occurs rapidly in these cells (18). After treatment with PhAsO to prevent possible subsequent internalization, cells were incubated in the presence of hirudin for 3 h. Cells were trypsin treated before and after the 3-h incubation to determine levels of intracellular 1251-thrombin-PN. Both cell and media samples were analyzed by SDS-PAGE. Table I1 shows that cell-bound 1251-thrombin-PN remained at the cell surface throughout the incubation as evidenced by its complete sensitivity to tryspin. After 3 h, only about 16% of total cell bound 1251-thrombin-PN was found in the medium. Thus, '251-thrombin-PN complexes dissociated very slowly from human cells with the consequence that only a small amount of PN would be available for detection using the hirudin assay.
Comparison of the Cellular Binding of Thrombin and lZ51-Thrombin Scatchard Analysis-To measure thrombin binding to the cell surface, increasing amounts of thrombin were added to mouse cell cultures. Incubations were for 30 min, by which time steady state binding levels had been reached (not shown). After rinsing, cultures were incubated with 1251-hirudin for 3 h. As shown above, under these conditions almost all cell surface-bound thrombin had dissociated and formed complexes with hirudin (Table I, Fig. 2). 1251-Hirudin-thrombin complexes were then measured using the hirudin assay. As shown in Fig. 3A, binding of thrombin to mouse cells was saturable. The results, transformed by the method of Scatchard (29), are shown in Fig. 323 and cumulative results of six separate determinations are presented in Table 111. Thrombin appeared to bind to a single class of high affinity ( K d = 2.4 X M) receptors at 37 "C as evidenced by a high linear correlation coefficient. At saturating levels, thrombin detected about 140,000 receptor sites/mouse cell. However, interpretation of binding data obtained at 37 "C can be complicated by ligand internalization (30); therefore, we also performed binding studies at 0 "C to block endocytosis. Our results showed that the apparent affinity of thrombin for its receptor almost doubled, but the apparent number of receptor sites detected remained the same when compared to binding studies performed at 37 "C (not shown).
Because previous studies have used '251-thrombin for the detection of thrombin-binding (6, 7 , 9-12), we compared results obtained using '251-thrombin to those obtained using the hirudin assay. Thrombin was iodinated by a soluble lactoperoxidase technique (6) and retained 100% of its fibrinogen clotting activity at up to 0.4 mol of I/mol of thrombin. Fig. 3A shows that the total binding of 1Z51-thrombin to mouse cultures was nonsaturable, in contrast to the total binding of thrombin to mouse cultures determined using the hirudin assay. However, subtraction of the nonspecifically bound lZ5I-thrombin from total 1251-thrombin at each '251-thrombin concentration yielded a saturable binding curve (Fig. 3A). As shown in Fig.  3B and Table 111, '251-thrombin detected one class of thrombin-binding sites. As seen for unlabeled thrombin, the binding affinity of lZ5I-thrombin at 0 "C increased by about 2-fold over the apparent affinity obtained at 37 "C, but an equal number of Iz5I-thrombin-binding sites was detected at 0 and 37 "C (not shown). However, only about one-half of the thrombinbinding sites detected using the hirudin assay were detected using '251-thrombin. Additionally, the apparent affinity of 1251thrombin was over 2-fold higher than the apparent affinity of uniodinated thrombin for its receptor (Table 111).
The discrepancies between binding results obtained using 1251-thrombin and thrombin could have been caused by radioiodination-induced alterations in thrombin-binding. Alternatively, the different methods of measurement of cellular binding of '251-thrombin and thrombin could have caused

Dissociation of '251-"thrombin-PN from the surface of human cells in the presence of hirudin
Binding medium containing '251-thrombin (100 ng/ml) with or without the addition of thrombin (10 pg/ml) was preheated to 37 "C and incubated with human cells (2 X lo5 cells/dish) for 2 min at 37 "C. (Under these conditions, about 80% of specifically-bound Iz5Ithrombin was in 'Z51-thrombin-PN complexes.) Cell cultures were immediately washed (4X) in D-PBS at 0 "C and treated with PhAsO a~ described in Table I. DV medium (1 ml) containing &pes (20 mM) ovalbumin (20 pg), hirudin (50 ng), and PhAsO (0.1 mM) was then added and cultures were incubated at 37 "C. At the indicated times, the amount of cell surface-bound '251-thrombin-PN was determined by trypsin addition and cell and/or medium samples were analyzed by SDS-PAGE (bis-tris system, pH 7.5) as described under "Experimental Procedures." Specifically bound radioactivity in gel slices containing '251-thrombin-PN was determined as described under "Experimental Procedures." of Cell Surface-bound Thrombin thrombin binding affinity (Fig. 3C). The results presented in Figs. 3B and 3C also showed that iodination of thrombin did not alter determination of the total number of thrombinbinding sites on mouse cells. These results suggested that the reduced number of thrombin binding sites determined using I-thrombin (Fig. 3B) was not solely the result of iodinationinduced changes in thrombin binding.
Specific Activity Effects-Additional data supporting the conclusion that low levels of iodination of thrombin increased its apparent binding affinity are shown in Fig. 4. '251-Thrombin was diluted with thrombin to one-fiih the original specific activity. Both diluted and undiluted '251-thrombin preparations were incubated with mouse cells a t 37 "C and the specific activity of each preparation was used to calculate the amount of thrombin bound to cells. The results in Fig. 4 showed that cellular binding of these '251-thrombin preparations was similar at low thrombin concentrations. However, at higher thrombin concentrations, it appeared that more lZ5I-thrombin was cell-bound using the lower specific activity preparation. These results, in agreement with results presented in Fig. 3, indicated that iodination increased the apparent affinity of thrombin for cellular binding sites.
Membrane Binding Assay-To determine what fraction of the molecules in '251-thrombin and thrombin preparations could bind to cell surface-binding sites, we performed a membrane binding assay (31). Increasing amounts of mouse cell membranes were incubated with a constant amount of lZ5Ithrombin or unlabeled thrombin. The results presented in differences in the results obtained. To determine if either or both of these possibilities was occurring, we iodinated thrombin with unlabeled iodide. Cellular binding of two I-thrombin preparations was measured using the hirudin assay. One thrombin preparation contained a low molar ratio of iodide to thrombin (0.3 mol of I/mol of thrombin) and retained all of its fibrinogen clotting activity. The other thrombin preparation contained a high molar ratio of iodide to thrombin (1.5 mol of I/ mol of thrombin) and retained only 30% of its fibrinogen clotting activity. Our results showed that low levels of iodination increased the apparent affinity of thrombin for mouse cells (Fig. 3B) whereas high iodination levels decreased to mouse cells Binding parameters were determined as described in Fig. 3 based on the assumption that both '251-thrombin and thrombin bind to a single class of binding sites on mouse cells.  . 4. Effect of alteration of the specific activity of "'Ithrombin on the apparent amount of lZ6I-thrombin bound to mouse cells. The specific activity of a preparation of '251-thrombin (6.5 X lo6 cpm/pg, 0.36 mol of I/mol of thrombin) was reduced 5-fold by addition of thrombin. Both lZ5I-thrombin preparations were incubated with mouse cells (8 X lo5 cells/dish) in the presence and in the absence of thrombin (10 pg/ml) for 30 min at 37 "C. After rinsing in D-PBS, the specifically bound radioactivity remaining on dishes was determined as described in Fig. 3 and under "Experimental Procedures." The radioactivity which was specifically bound to each dish was divided by the corresponding specific activity of 1251-thrombin to obtain the amount of cell-bound thrombin. 0, '251-thrombin, 6.5 X lo6 cpm/pg; 0, Iz5I-thrombin, 1.3 X lo6 cpm/pg.  5 showed that at least 85% of thrombin in unlabeled preparations could bind to mouse cell receptors when measured using the hirudin assay. This is a minimal estimate since it appeared that some thrombin was sequestered by the membranes as evidenced by the reduction in detected thrombin, and thus membrane-released thrombin, at the largest membrane addition (Fig. 5). In contrast, results obtained using "' Ithrombin showed that only about 30% of the thrombin in these preparations appeared to bind to mouse cell receptors. Thus, iodination of thrombin at levels which did not appear to affect the ability of thrombin to cleave fibrinogen prevented the binding of a large portion of thrombin molecules to the mouse cell surface.

Measurement of Cell Surface-bound Thrombin
Effects of Tyrosine Iodination on Cellular Binding of Iz5I-Thrombin-As previously noted, the reduction in the number of thrombin-binding sites measured when using IZ5I-thrombin versus thrombin was not due solely to iodination-caused changes in thrombin-binding affinity. In addition, if '"Ithrombin preparations were not uniformly labeled, and if only certain labeled thrombin molecules could bind to cells, this would result in an incorrect determination of the specific activity of bindable '251-thrombin and an erroneous estimation of the number of thrombin-binding sites. We tested the possibility that cell-bound 1251-thrombin contained different 1251-Tyr residues from unbound '251-thrombin by subjecting them to partial enzymatic digestion with Staphylococcus V8 protease and chymotrypsin (32). We found that these proteases yielded the same Iz5I-Tyr-containing pol?rpeptide fragments for both "'I-thrombin and cell-bound '2sI-thrombin, indicating that the '251-thrombin preparations contained similar 1251-Tyr residues? However, our interpretation of these results would be compromised if neither V8 protease nor chymotrypsin could cleave between the different iodinated tyrosines postulated.
Since it appeared that similar tyrosines were labeled in cellbound and unbound Iz5I-thrombin preparations, we determined whether there were differences in labeling at a single Tyr. '251-Thrombin was incubated with mouse cells in the presence or absence of unlabeled thrombin. After 30 min, solubilized cellular proteins were analyzed by SDS-PAGE and Maximal binding of 'z61-thrombin and thrombin to mouse cell membranes. Binding buffer containing 1251-thrombin or thrombin (1.0 ng) was incubated with increasing amounts of mouse cell membranes in Microfuge tubes. Thrombin (5 pg) was added to one-half of the tubes containing 1Z51-thrombin and the total volume of each tube was adjusted to 0.5 ml with binding buffer. After incubation for 30 min at 37 "C with gentle rocking, tubes were centrifuged for 3 min at 15,000 X g and membrane pellets were rinsed once in D-PBS. Centrifugation and rinsing were done at 0 "C. The amount of specifically bound '251-thrombin was determined as described under "Experimental Procedures" and Fig. 3. At the largest membrane addition (760pg/tube), about 80% of 'Z51-thrombin was specifically bound. The amount of membrane-bound thrombin was determined using the hirudin assay as described under "Experimental Procedures." 0, membrane-bound thrombin; 0, membrane-bound '251-thrombin. gel slices containing Iz5I-thrombin were pooled and eluted by electrophoresis, and iodotyrosine standards were added. After hydrolysis for 6 h, samples were analyzed by thin layer chromatography. As shown in Fig. 6a, about 70% of the radioactivity was in MIT and about 30% was in DIT. However, mouse cell-bound IZ5I-thrombin contained 94% of the radioactivity in MIT and only 6% in DIT (Fig. 6b). This small amount of ' "1thrombin containing DIT could have been derived from nonspecifically bound thrombin since about 20% of the '251-thrombin in these experiments was nonspecifically bound. As shown in Fig. 6d, about 22% of the radioactivity in nonspecifically bound I*'I-thrombin was DIT. Therefore, the small amount of '251-thrombin containing DIT which bound to mouse cells appeared to be bound nonspecifically. These results showed that '251-thrombin preparations contained a significant amount of DIT and that this diiodinated thrombin could not specifically bind to mouse cells. We showed previously that PN mediates most of the specific cellular binding of thrombin to human cells (28). Therefore, we determined if '251-thrombin containing DIT could bind to PN. '"I-Thrombin was incubated with human cells and solubilized cellular proteins were analyzed by SDS-PAGE. Gel fractions containing '251-thrombin-PN complexes were pooled and analyzed as described above. Results in Fig. 6c show that '251-thrombin containing DIT linked to PN. Since the fraction of DIT found in complex with PN was similar to that found in the 1251-thrombin preparation, we conclude that 1251-thrombin containing either MIT or DIT is equally able to form a linkage with PN.
The results presented in Fig. 6a indicated that about onethird of the total radioactivity in 1251-thrombin was in DIT. However, this appeared to be a low estimate since experiments in which 1251-thrombin was directly hydrolyzed for only 3 h and analyzed by thin layer chromatography showed that about one-half of the iodotyrosine radioactivity was in DIT (Fig. 7). Again, we found that cell-bound '251-thrombin contained little, if any, DIT. Importantly, a significant amount of Iwas present. Since about 90% of this '251-thrombin preparation was precipitated with trichloroacetic acid, it appears that free Iz5I-was released from iodotyrosine. Release of '251also occurred in the experiment presented in Fig. 6, although a comparison of levels of released 1251-could not be made due to the different assay conditions employed in the experiments presented in Figs. 6 and 7. Detection of iodotyrosine standards added before hydrolysis of samples indicated that DIT was less stable than MIT. This would be expected if the rate of loss of Ifrom MIT and DIT was equal since DIT + MIT 4 Tyr. Therefore, it appears that at least 50% of the radioactivity in '251-thrombin was in DIT.
Together, these results showed that a large fraction of radioactivity in '251-thrombin preparations was present in DIT ( Figs. 6a and 7). '251-Thrombin containing DIT was apparently unimpaired in its ability to cleave fibrinogen and form a linkage with PN (Fig. 6c), but was unable to bind to mouse cells (Figs. 6 and 7). Additionally, thrombin containing MIT appeared to have a higher affinity for mouse cells then unla- pg/ml) was added to mouse cells for 30 min at 37 "C. Cells were then rinsed in D-PBS and treated with PhAsO as described in Fig. 2. Cellbound 1251-thrombin was dissociated by addition of DV medium (0.75 ml) containing Hepes (20 mM), ovalbumin (2 pg/ml), and hirudin (35 ng) for 3 h at 37 "C. Both this culture fluid, which contained celldissociated '251-thrombin, and the initial '251-thrombin preparation (which was adjusted to contain the same amount of ovalbumin and hirudin as cell-dissociated 1251-thrombin) were hydrolyzed in barium hydroxide as described under "Experimental Procedures." '"I-Tyr was extracted (in 2 N ammonium hydroxide/ethanol, 1:1, v/v) and analyzed by thin layer chromatography as described under "Experimental Procedures." The development time was 2.5 h. Arrows indicate the position of standards. beled thrombin (Table 111, Figs. 3 and 4). These iodinationinduced alterations in the interaction of thrombin with mouse cells caused errors in the determination of both the binding affinity and the total number of binding sites for thrombin on mouse cells.

DISCUSSION
Hirudin Assay-The present method for detecting picogram amounts of unlabeled thrombin bound to the surface of mouse cells depends upon quantitative release of thrombin from the cell surface and formation of thrombin-hirudin complexes after incubation with hirudin.* Cell surface-bound but not internalized thrombin is measured by the assay since only thrombin released from cells and in complex with '251-hirudin is detected. We showed that by incubating mouse cells in the presence of hirudin for 2 h, at least 95% of cell-bound 1251thrombin was released into the medium (Table I). However, it appears from results with human cells ( Table 11) that 1251thrombin-PN complexes are tightly bound to the cell surface and only a small fraction of 1251-thrombin-PN is released into the medium in the presence of hirudin. This confirms results we have previously obtained showing that 1251-thrombin-PN binds to human cells with an extremely high affinity (18). Because only about 10% of specifically-bound L251-thrombin of mouse cells is in complex with PN, we did not determine whether '251-thrombin-PN complexes were also tightly bound to mouse cells. Such a determination would be difficult since to avoid rapid internalization of 1251-thrombin-PN, very short times of incubation with 1251-thrombin must be used (18). This severely limits the amount of 1251-thrombin-PN that can be formed at the surface of cells. However, since human cellderived PN can bind to mouse cells,5 it seems likely that 1251thrombin-PN complexes also dissociate very slowly from mouse cells.
In addition to the requirement that thrombin must be dissociated from the cell surface, thrombin must also be in complex with hirudin to be measured by the hirudin assay. We showed in Fig. 2A that almost all of the cell-dissociated '251-thrombin migrated as 1251-thrombin-hirudin complexes. Only a small amount of nonspecifically bound 1251-thrombin appeared to form a complex with hirudin (Fig. 2B). One explanation for this is that most of the nonspecifically bound I-thrombin might be damaged thrombin, either present in the unlabeled thrombin preparation or induced by iodination. Alternatively, interaction of thrombin with nonspecific sites on the cells or dishes could alter thrombin. In either case, our finding that nonspecifically bound '251-thrombin was measured poorly by the hirudin assay was corroborated in cell binding studies. When using 1251-thrombin, the fraction of 9thrombin bound nonspecifically increased as a linear function of the concentration of '"I-thrombin. Thus, at high lZ5I-thrombin concentrations, a large fraction of total cell-bound 1251thrombin was nonspecifically bound (Fig. 3A). However, when measuring binding of unlabeled thrombin using the hirudin assay, as high thrombin concentrations were reached, the amount of cell-bound thrombin approached saturating levels (Fig. 3A). Together, these results showed that almost all specifically bound thrombin released from mouse cells in the presence of hirudin was also in complex with hirudin.
Although we have used the hirudin assay primarily to 125 determine levels of cell-bound thrombin, it is possible that other applications may be found. The hirudin assay is more sensitive than the radioimmunoassay developed for thrombin (19). As shown in Fig. 1, DIP-thrombin is poorly detected using the hirudin assay. This is surprising because DIP-thrombin was used as an immunogen to raise antibodies to thrombin (19) and DIP-thrombin binds hirudin (26). DIP-thrombinhirudin complexes thus appear to have a different conformation than thrombin-hirudin complexes, preventing antithrombin antibody from interacting with thrombin antigenic determinants. Binding of hirudin near the active site of thrombin (33) may play a role in inducing this postulated conformational change.
Comparison of '251-Thrombin and Thrombin Binding to ME CeZZ~--'~~I-Thrombin has been used in previous studies to measure thrombin binding to a variety of cell types (6, [9][10][11][12]. Our results showed that the binding properties of lZ5Ithrombin differ from thrombin when assayed using mouse cells (Figs. 3-6, Table 111). We found two reasons for these binding differences: the binding affinity of '251-thrombin containing MIT appeared to be higher than the binding affinity of thrombin (Figs. 3,4, and 6), and the binding affinity of lZ5Ithrombin containing DIT was greatly reduced or negligible (Figs. 6 and 7 ) .
In six separate determinations, we found that the apparent binding affiiity of '251-thrombin was 2-to 3-fold higher than that of thrombin (Table 111). Our determination of the binding affinity of '251-thrombin for mouse cells is in close agreement with the binding affinity that we reported previously (34). Additional data which suggested that '251-thrombin had a higher affinity for cellular binding sites than thrombin was shown in Fig. 4. In this experiment, if the binding affinities of Iz5I-thrombin and thrombin were equal, then reduction of the specific activity of Iz5I-thrombin preparations with thrombin would not alter measurement of the amount of cell-bound thrombin. We found that more '251-thrombin was apparently bound to cells using the lower specific activity preparation. Thus, it appeared that thrombin could not compete equally with '251-thrombin for cellular binding sites.
'251-Thrombin preparations contained 50% or more of their radioactivity in DIT (Fig. 7 ) . Thus, at least one-third of the '251-thrombin molecules in these preparations were unable to bind to mouse cells, with the effect of lowering the apparent affinity of '251-thrombin. It seems likely, therefore, that the actual binding affinity of '251-thrombin containing MIT is at least 1.5-fold higher than the apparent binding affinity we estimated from Scatchard analysis (Fig. 3, Table 111). We also showed that the binding affinity of thrombin that contained high levels of iodide was reduced by about 2-fold compared to thrombin (Fig. 3C). This may have resulted either from an increase in levels of I-thrombin containing DIT or iodination of another Tyr residue critical for cellular binding, which would have the effect of lowering the apparent binding affinity.
Although the binding affinity of thrombin containing high levels of iodide was about one-half the value determined for thrombin, the same total number of thrombin binding sites was detected (Fig. 3C). In contrast, '251-thrombin detected only about one-half of the binding sites detected by thrombin (Fig. 33, Table 111). The reason for this difference in the detection of binding sites by 1251-thrombin and I-thrombin appears to be that the specific activity determined by lZ5Ithrombin was erroneously high. This occurred because lZ5Ithrombin preparations contained a mixture of MIT and DIT, and only the MIT derivative bound to mouse cells (Figs. 6  and 7 ) . We showed previously that DIP-thrombin bound to mouse cells (34) but did not form a linkage with PN (28). These results indicated that the active site serine of thrombin was involved in the linkage of thrombin to PN but was not required for thrombin-binding to mouse cells. The results presented in this paper also indicated that the thrombin site which binds to mouse cells and the site which links PN are not identical. We found that although lZ51-thrombin containing DIT did not bind to mouse cells, its ability to link to PN was unimpaired (Fig. 612). These results corroborate those which showed that under optimal conditions with human cells almost all Iz5Ithrombin was found in complex with PN6 and that iodination of thrombin did not affect its ability to link to anti-thrombin I11 (35). Through use of techniques presented here and elsewhere (18,28,36), it is now possible to accurately determine the amount of thrombi bound to the mouse cell surface, both free and in complex with PN. We showed in Table I1 that thrombin-PN complexes dissociated very slowly from the cell surface. Therefore, only a small amount of thrombin-PN would be available for interaction with '251-hirudin in the hirudin assay. It also seems unlikely that the hirudin could bind to thrombin-PN since hirudin blocks formation of thrombin-PN (18). By using the hirudin assay for detection of non-PN bound thrombin, the errors inherent in using '251-thrombin, discussed above, can be avoided. Since iodination of thrombin does not impair its ability to link to PN (Fig. 6c), I-thrombin can be used to measure thrombin in complex with PN.
Finally, through use of the hirudin assay, it will now be possible to accurately measure thrombin-binding during the course of an experiment in which cells are mitogenically stimulated by thrombin. These studies, now in progress, should allow us to evaluate whether binding of thrombin to its receptor is a necessary event in thrombin-induced stimulation of cell division.