Regulation of Tissue Plasminogen Activator Activity by Cells

A number of cell types have previously been shown to bind tissue plasminogen activator (tPA), which in some cases can remain active on the cell surface resulting in enhanced plasminogen activation kinetics. We have investigated several cultured cell lines, U937, THP1, K562, Molt4, and Nalm6 and shown that they bind both tPA and plasminogen and are able to act as promoters of plasminogen activation in kinetic assays. To understand what structural features of tPA are involved in cell surface interactions, we performed kinetic assays with a range of tPA domain deletion mutants consisting of full-length glycosylated and nonglycosylated tPA (F-G-K1-K2-P), ΔFtPA (G-K1-K2-P), K2-P tPA (BM 06.022 or Reteplase), and protease domain (P). Deletion variants were made in Escherichia coli and were nonglycosylated. Plasminogen activation rates were compared with and without cells, over a range of cell densities at physiological tPA concentrations, and produced maximum levels of stimulation up to 80-fold with full-length, glycosylated tPA. Stimulation for nonglycosylated full-length tPA dropped to 45–60% of this value. Loss of N-terminal domains as in ΔFtPA and K2P resulted in a further loss of stimulation to 15–30% of the full-length glycosylated value. The protease domain alone was stimulated at very low levels of up to 2-fold. Thus, a number of different sites are involved in cell interactions especially within finger and kringle domains, which is similar to the regulation of tPA activity by fibrin. A model was developed to explain the mechanism of stimulation and compared with actual data collected with varying cell, plasminogen, or tPA concentrations and different tPA variants. Experimental data and model predictions were generally in good agreement and suggest that stimulation is well explained by the concentration of reactants by cells.

Tissue plasminogen activator (tPA) 1 is a multidomain serine protease responsible for converting the zymogen plasminogen into the broad specificity serine protease plasmin. Plasmin is involved in a range of biological processes, including fibrinoly-sis, tissue development, and tumor invasion and metastasis (1). Fibrin binding is a critical step in the expression of significant tPA activity because this enzyme is not produced as a zymogen but has only low levels of activity in free solution (2). Although tPA is mainly associated with hemostasis, in contrast to urokinase-type plasminogen activator (uPA), which is considered to be more important in pericellular proteolysis, several groups have identified or isolated a number of different cell surface tPA receptors on endothelial cells and shown that binding of tPA and plasminogen to the same cell can lead to accelerated plasmin generation (3)(4)(5). Published dissociation constants for cell-tPA interactions vary over a wide range, 10 Ϫ11 -10 Ϫ7 M, although it now appears that the highest affinity interactions may be due to cell surface-associated plasminogen activator inhibitor type 1 (6). Binding of tPA to plasminogen activator inhibitor type 1 requires the enzyme active site and results in an inactive complex. The physiological significance of putative tPA receptors of lower affinity is unclear considering the low plasma level of tPA (70 pM). For example, the tPA receptor annexin II, which binds tPA with a K d of 48 nM, is capable of producing modest stimulation of enzyme activity (Ͻ10-fold) in vitro (7). Monocytes and monocytoid cells have been studied and found to bind tPA with a low affinity, but high capacity, and are able to stimulate tPA enzyme activity up to around 20-fold (8,9). Other receptors on liver cells have also been investigated in an attempt to understand the structural features of tPA, both protein and carbohydrate, involved in receptor binding and responsible for its short in vivo half-life and rapid clearance from the circulation when tPA is administered as a thrombolytic agent (10).
It is known that tPA-surface interactions are critical for the regulation of plasminogenolytic activity, and this has been investigated extensively where fibrin is the biological surface (11,12). Detailed structure/function studies on a large number of mutant enzymes indicate that many areas of the protein are involved in modulating the interaction with fibrin and hence regulating tPA activity (13). The purpose of the present study was to look at the regulation of tPA-catalyzed plasminogen activation on cell surfaces using a number of cultured cell lines and to identify structural features involved in tPA-cell interactions leading to stimulation of plasminogen activation. Five cell types have been investigated in conjunction with five tPA variants lacking various domains and/or carbohydrate. To understand the factors regulating the kinetics of plasminogen activation, our studies included ranges of cell (receptor) concentration, as well as tPA and plasminogen concentration. Using this approach, it has been possible to develop a model that explains how cell interactions with tPA and plasminogen regulate the generation of plasmin activity and changes in apparent K m and k cat . * The Acciones Integradas program provided funds to initiate this research. Support was also provided by SCS-Generalitat Catalunya CICYT:SAF96-0376, Marato TV3/Cancer, and Marato TV3/Cardiovascular. 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.
Full-length glycosylated tPA was International Standard melanoma tPA (code 86/670, National Institute for Biological Standards and Control, South Mimms, United Kingdom). Full-length nonglycosylated tPA was a gift from Rhone-Poulenc Rorer (Vitry sur Seine, France) expressed and purified from Escherichia coli (14). Both full-length enzymes had similar specific activities in fibrinolytic assays. ⌬FtPA and protease domain were generated as described below using the same plasmid as for the full-length protein (pXL130), provided by Rhone-Poulenc Rorer and expressed and renatured from E. coli inclusion bodies in the same way as full-length nonglycosylated tPA (14). Expression of tPA was induced by addition of 40 mg/ml naladixic acid (15) to cultures of E. coli strain N99cI ϩ (Pharmacia, Uppsala, Sweden) grown at 37°C with aeration up to an A 600 nm of approximately 1.0. Cultures were grown at 37°C for a further 90 -120 min before harvesting and processing as described previously for the full-length enzyme (14) with the following modifications. After concentrating renatured protein using SP Trisacryl (IBF Biotechnics, Villeneuve-la-Garenne, France). The final lysine-Sepharose column step was replaced by fast protein liquid chromatography using a Mono S column (Pharmacia) run in 25 mM sodium acetate buffer, pH 4.3, containing 0.01% Tween 80, and purified tPA was eluted with a gradient of 0 -1 M NaCl. The serine protease domain purification was further modified by omitting the pH 5.5 wash step on the SP Trisacryl column.
⌬FtPA was produced by deleting residues Ser-1-Lys-49 and serine protease domain by deleting residues Ser-1-Arg-275 using the polymerase chain reaction technique of deletion by overlap extension (16,17) using the plasmid pXL130. Primers A and D were used for both mutant proteins: primer A, 5Ј-ACCTGCAGCCAAGCTT-3Ј, and primer D, ATC-CTGAAATCAGACCAAGTCCTG. For ⌬FtPA, additional primers were B, TCGCTGCAACTCATATGTAAGTAT, and C, ACTTACATATGAGT-TGCAGCGAGC. For serine protease domain generation, primers were B, CCTCCTTTGATCATATGTAAGTAT, and C, ATACTTACATATGA-TCAAAGGAGGGCTC. The final polymerase chain reaction product, ABCD, was digested with AvaI (New England Biolabs, Beverly, MA) for insertion into pXL130. Re-ligated plasmids were checked for correct size on agarose gels, and proteins were examined by SDS-polyacrylamide gel electrophoresis to ensure correct molecular weights. The final step in protein purification was fast protein liquid chromatography and variants were checked for purity and correct molecular weight by SDSpolyacrylamide gel electrophoresis.
K2P (also known as BM 06.022 or Reteplase) was a gift from Roche Molecular Biochemicals. This protein was expressed, renatured and purified in a nonglycosylated form from E. coli (18). All tPA variants were measured in direct assays against chromogenic substrate S-2288 (Chromogenix, Mölndal, Sweden), and activity was quantitated against the 2nd International Standard for tPA (code 86/670, National Institute for Biological Standards and Control, South Mimms, United Kingdom) in a plasminogen activation assay in the absence of stimulator as outlined below.
Cell Culture-THP1, U937, Nalm6, Molt4, and K562 cells were cultured in RPMI 1640 medium with 2-4 mM glutamine, 7-10% fetal bovine calf serum, and 1 mM sodium pyruvate (pyruvate excluded for U937 cells) (Sigma or Life Technologies Ltd.). Cells were grown to a density of 1.5 ϫ 10 6 to 2.5 ϫ 10 7 cells/ml in the presence of 5% CO 2 . Cells were harvested by centrifugation and washed three times in serum-free RPMI 1640 medium at 0 -4°C. The cell pellet was drained and resuspended with gentle vortexing and washed with 0.05 M glycine buffer, pH 3.5, containing 0.1 M NaCl for 2-3 min on ice (19). Cells were washed twice with RPMI medium for final counting, resuspension and dilution in assay buffer (below) containing albumin at 1 mg/ml.
Plasminogen Activation Kinetics-Plasminogen activation reactions were carried out in microtitre plates in reaction volumes of 100 l. In initial studies using full-length tPA enzymes, reaction mixtures were 20 l of tPA (final concentration 5 ng/ml, Ͼ Ͼ70 pM), 40 l of cells (final concentration, between 10 3 and 10 8 cells/ml) and 40 l of substrate mix containing Glu-plasminogen (Enzyme Research Laboratories, Swansea, United Kingdom) at a final concentration of 100 nM and chromogenic substrate S-2251 (Val-Leu-Lys-p-nitroanilide, Chromogenix, Mölndal, Sweden) at a final concentration of 0.15 mM. Reactions were performed in assay buffer consisting of Tris-HCl buffer, pH 7.4, at 37°C and a final ionic strength of 0.12, containing 1 mg/ml human serum albumin. In experiments with higher concentrations of tPA (up to 1.5 nM), reaction volumes remained the same. In experiments with tPA variants having different molecular weights, starting activities of tPA had equivalent activation rates with Glu-plasminogen in the absence of cells and stimulation was calculated as activation rate with cells/activation rate without cells under identical conditions. In studies using a range of plasminogen concentrations reaction mixtures were made up of 20 l of tPA, 20 l of cells (to give a range of cell densities as above), 20 l of plasminogen (to give a range of concentrations up to a final concentration of 520 nM), and 40 l of substrate solution containing S-2251 as above. Cells were incubated for 15 min with tPA to equilibrate to 37°C before substrates were added to begin the reaction. Absorbance was monitored at 405 nm, using a Thermomax thermostatted plate reader (Molecular Devices Corporation, Stanford, CA), producing the expected exponential increase for p-nitroanilide resulting from plasmin production. Rates of plasmin production were calculated from slopes of plots of A versus s 2 generated automatically from Thermomax data by a program specifically written for this purpose (J Waterman-Smith, Molecular Devices, Crawley, United Kingdom). These slopes are proportional to plasmin generation and were calculated using Enzfitter (Elsevier, Cambridge, United Kingdom), as outlined previously (20,21). Plasmin generation could be calculated from rates of A/s 2 by dividing by 22 310 A M Ϫ1 s Ϫ1 from previously determined values for the K m and k cat of plasmin on S-2251 and the extinction coefficient of p-nitroanilide under these conditions (21). Simultaneous kinetic experiments were carried out with and without tPA to control for intrinsic activator (most probably uPA) synthesized by the cells and bound to the surface, which was not removed in the low pH wash. Different cell lines had reproducible levels of associated intrinsic activator giving background rates in the order THP1 Ͼ U937 Ͼ K562 Ͼ Molt4 Ͼ Nalm6 (essentially zero background rate). Where background rates were high, i.e. Ͼ10% of maximum rate achieved with added tPA (found with THP1 and U937 cells), these were subtracted from observed rates with added tPA in order to calculate rates due to exogenous activator only.
Plasminogen activation kinetics in the presence of varying concentrations of 6-aminohexanoic acid (6-AHA) (Sigma) to investigate conformational changes and enzyme resulting from lysine analogue binding were conducted in a similar manner. Concentrations used for these experiments were 50 pM tPA, 1 pM uPA, and 450 nM plasminogen.
To investigate the involvement of oligosaccharide moieties in binding to cells, kinetic experiments were performed in the presence of 20 mM monosaccharides, mannose, fucose, N-acetylglucosamine, N-acetylgalactosamine, N-acetylneuraminic acid, and ovalbumin (10 mg/ml) (all from Sigma).
tPA Binding Studies-Nalm6 and U937 cells were washed as described above and finally resuspended in Hepes-buffered saline, pH 7.4. Ligand binding assays were performed by incubating full-length 125 I-tPA (0.25-8.0 nM in a total reaction volume of 100 l) or K2P (0.5-16 nM in a total reaction volume of 200 l) with the washed cells (10 7 cells/ml) for 2 h at 4°C. Cells were then separated from the whole reaction mixture by centrifugation of aliquots, in triplicate, in 20% sucrose solution. A parallel set of reactions were incubated in the presence of an excess of cold, unlabeled full-length tPA (glycosylated) or K2P to determine nonspecifically bound radioactive counts.
Low affinity binding sites were investigated using higher concentrations of tPA (up to 12 M), and cell-bound tPA was measured by enzyme activity as described previously (22). In these experiments 200 l of 10 7 cells/ml were incubated with a range of tPA concentrations for 30 min at room temperature before bound and free tPA were separated by centrifugation through 20% sucrose, as above. Bound enzyme was determined in plasminogen activation assays, as described above. Parallel experiments were performed without cells to determine the amount of free tPA carryover in kinetic assays. These activities were subtracted from specifically bound rates prior to data analysis. Each tPA concentration was incubated and activity measured in triplicate.
Models-The usual way to analyze tPA data in the presence of a stimulator is to use a simple Michaelis-Menten equation, where [tPA] and [Pgn] are the total added concentrations of enzyme and substrate, respectively. K m Ј and k cat Ј refer to the kinetic parameters of the bound enzyme. The contribution of the solution reaction is ignored. Model 2 is based on the concentration effect of cells on tPA and plasminogen, which interact with cell surface binding sites leading to a localized raised reactant concentration in a cell-associated compartment. The principle of the model is as previously developed by Nesheim et al. (23) to study regulation of the prothrombinase complex on phospholipid vesicles and can be understood in our system with reference to Scheme 1. The abbreviations used in Scheme 1 are listed below, and the prime notation denotes cell-associated protein or parameter. Thus, PgnЈ and tPAЈ are bound plasminogen and tPA, respectively; v c is the cell volume; v ca is the volume of the cell-associated compartment. V c and V ca are the corresponding volumes for the whole reaction mixture. V t is the total reaction volume, and V bulk ϭ V t Ϫ (V c ϩ V ca ). The intrinsic catalytic parameters and binding constants applied in the present study are shown in Table I. Molar concentrations of bound enzyme and substrate, relative to the whole reaction mixture, can be calculated from simple equilibrium considerations knowing K d , the number of binding sites/ cell, and cell concentration (hence receptor concentration). However, rate ca will depend not only on the equilibrium between free and bound reactants but also on the local concentration, which will depend on the size of V ca . Thus it is necessary to calculate [tPAЈ l ]and [PgnЈ l ] (local concentration of tPAЈ and PgnЈ, respectively) in order to calculate the local rate using K m Ј and k cat Ј.
Thus, it can be seen that the concentration factor (R) for reagents in the cell-associated compartment relative to the concentration for the whole reaction mixture is a ratio of 2 volumes and will vary with the cell concentration, or for a fixed reaction volume (e.g. 1 ml), as follows.
Because we are assuming that v ca is a constant, we define as a constant equivalent to 1/v ca in ml. Thus, under these conditions, the following equation is used.
can be used to calculate the volume of v ca for one cell and also h in Scheme 1. To calculate local concentrations, a concentration relative to the whole solution is multiplied by R. An equation for the reaction rate in the cell-associated compartment can be derived using the Michaelis-Menten approach for the reaction A ϩ S N AS 3 A ϩ X, where A is enzyme (tPA) and S is substrate (plasminogen).
The measured rate, relative to the whole reaction mixture, is as follows.
This is a result of the following equation, and therefore the following equations are used.
Applying the enzyme conservation equation, where [B] is the molar concentration of enzyme receptors and K d is the dissociation constant for enzyme and receptor. Rearranging results in the following equation.
Then, dividing Equation 12 by Equation 15 results in the following equation.
This can be rearranged as follows and used as model 2, with respect to tPA and plasminogen.
Bound plasminogen [PgnЈ] is calculated in the usual way, This equation is closely related to the relationship for specific activation of an enzyme-catalyzed reaction (24) but with the modification of K m Ј to K m Ј/R. This model was used to derive values for k cat Ј and (and hence v ca and h) by nonlinear regression analysis using Grafit version 4 (25). Model data were generated using Mathcad, version 7 (Mathsoft Inc., Cambridge, MA).
Model 3 was also considered; it could account for the observed data and is related to the model of Lu and Nelsestuen (26) for prothrombinase activity on vesicles. This model was similar to model 2, being a version of the specific activation equation, although in this case it was assumed that binding (giving tPA b *) modified the intrinsic enzyme parameters K m Ј and k cat Ј directly. To account for the profile of rate ca with cell concentration (the "template effect"), it was necessary to assume that bound enzyme could only react with nonbound plasminogen in free solution (pgn f ). This is shown in Scheme 2.
Model 2 can then be modified to give model 3.

Full-length tPA and Deletion
Variants-All of the cell types tested in these studies, U937, THP1, K562, Molt4, and Nalm6, were able to act as promoters in kinetic assays measuring plasminogen activation by full-length tPA. These cell types gave similar results in terms of level of stimulation and cells/ml required for peak stimulation. Results are presented in detail for U937 cells, a widely used monocytoid line, and Nalm6 cells, a pre-B leukemic line that expresses negligible levels of intrinsic plasminogen activators. Fig. 1 shows the relationship between activation rate and cell density for full-length glycosylated tPA, nonglycosylated tPA, and nonglycosylated deletion variants. To ensure good comparability between variants, all data for one cell type were collected simultaneously on one microtitre plate using one batch of cells. Here, initial enzyme activities in the absence of cells gave similar rates of plasminogen activation and the effectiveness of cells in promoting activation can be seen in changes in rate of plasmin production. Degree of stimulation was also calculated, as rate with cells/ rate without cells. Over a range of cell densities, a peak of stimulation was seen between 10 6 and 10 7 cells/ml for all tPA variants. The bell shaped pattern for stimulation versus log of cell density is in accord with a template mechanism dependent on ternary complex formation of cell-bound activator and substrate. This is supported by radioligand binding studies that demonstrate that both these cell types are able to bind tPA and plasminogen to the cell surface (8). All cell types used in these studies showed this relationship between stimulation and cell density, although peak activity could fall between 10 6 -10 8 cells/ ml, and stimulation could approach 80-fold, with some variation between batches of cells. The ranking order of variants shown in Fig. 1 was constant. For the data presented in Fig. 1, maximum levels of stimulation were observed with full-length glycosylated tPA and were 51-fold with U937 cells and 24-fold with Nalm6 cells. ⌬FtPA and K2P showed a reduced level of stimulation with both cell types examined, 12-and 21-fold, respectively, with U937 cells and 9-and 12-fold, respectively, with Nalm6 cells in the experiments shown. There was a small but measurable stimulation for the protease domain of Ͻ2-fold with Nalm6 cells. This was difficult to measure in the presence of U937 cells due to the higher background rates (from intrinsic activator) associated with these cells. Unexpectedly, K2P stimulation was always greater than ⌬FtPA even though ⌬FtPA lacks only one domain and is structurally much closer to fulllength tPA than K2P. However, it is clear that loss of finger domain in either mutant enzyme results in a marked drop in stimulation by cells, but there was still a significant level of stimulation, apparently due to sites on kringle 2. The lower activity of nonglycosylated versus glycosylated full-length tPA was studied in more detail.
Glycosylated and Nonglycosylated tPA- Fig. 2 shows the degree of stimulation for glycosylated (A and C) and nonglycosylated (B and D) full-length tPA with U937 and Nalm6 cells. Over a range of cell densities, there was a peak of stimulation that was independent of the tPA concentration from 0.075 to 1.5 nM. Over this range of added tPA, there was no evidence of any saturation of binding sites, indicating that binding is not restricted to low numbers of receptors per cell. In the experiments shown in Fig. 2, comparative stimulation for U937 and Nalm6 were 78.9 Ϯ 2.8 (mean stimulation Ϯ S.D.) and 60.0 Ϯ 4.8, respectively, for full-length glycosylated tPA and 34.9 Ϯ 5.8 and 30.5 Ϯ 3.3, respectively, for full-length nonglycosylated tPA. Lack of glycosylation resulted in a drop in peak stimulation of around 50%. Thus, it is possible that glycosylation sites are directly involved in cell interactions, or alternatively, these could be indirect effects such that glycosylation of kringles affects lysine binding. In competition experiments including monosaccharides up to 20 mM or 10 mg/ml ovalbumin in kinetic assays, normal patterns of stimulation were observed with no detectable inhibition over a range of cell densities from 0 to 3 ϫ 10 7 cells/ml. These levels of monosaccharides have previously been shown to inhibit binding of tPA to glycosylation sitespecific cell surface receptors (27). Our results indicate that lack of glycosylation has an indirect effect on stimulation of tPA activity by cells.
Plasminogen Concentration-The experiments shown in Figs. 1 and 2 were conducted at a fixed plasminogen concentration of 100 nM, and further experiments were performed to investigate the effects of varying substrate concentration. If the interaction between tPA and cell receptors is affecting only the K m of the plasminogen activation reaction it should be possible to overcome the lower relative stimulation of the nonglycosylated tPA variants using higher substrate concentrations. Fig.  3A shows similar Michaelis-Menten curves for the tPA variants at the optimal cell density of 10 7 cells/ml. Curve fitting was by nonlinear regression to the standard Michaelis-Menten equa- tion, although there is some suggestion of apparent substrate inhibition with the full-length enzymes (see under "Discussion"). The apparent K m values were similar for the full-length enzymes (20 Ϯ 4 and 12 Ϯ 5 nM for glycosylated and nonglycosylated, respectively) and somewhat higher for the deletion variants (46 Ϯ 11 and 33 Ϯ 8 nM for K2P and ⌬FtPA, respectively). The maximum rate of activation was 0.22 pM/s for full-length tPA, which is lower than for the free solution reaction (compare Fig. 3, A and B). Further reductions in V max were apparent for the deletion variants. Clearly, cell binding affects both apparent K m and V max . The activity of two variants, fulllength tPA and ⌬FtPA, in the absence of cells is shown for comparison, and these variants have similar activity over this range of plasminogen, where the slope of the line can be used to estimate k cat /K m . The literature values for full-length tPA applied in these studies were K m ϭ 65 M and k cat -0.06 s Ϫ1 , k cat /K m ϭ 923 M Ϫ1 s Ϫ1 (11). This is close to the value derived from Fig. 3A of 786 M Ϫ1 s Ϫ1 . Fig. 3B shows the change in plasminogen activation in the absence of cells for full-length tPA up to 16 M plasminogen. Rates did not approach saturation, and hence it was not possible to calculate K m and V max or k cat.
tPA-Cell Binding Studies-To investigate the levels of tPA binding under the conditions used in these kinetic experiments, radiolabeled tPA was incubated with cells over a low concentration range, up to 16 nM. Under these conditions, very little specific binding of 125 I-tPA could be observed, indicating the absence of significant levels of a high affinity binding site with a K d in this range of tPA concentrations (results not shown). In order to investigate lower affinity binding sites, cells were incubated with higher concentrations of tPA and detection was by enzyme activity rather than radioactivity. This approach allows higher concentrations of tPA to be used as no excess of unlabeled tPA is required, and only bound, active enzyme is detected, which is the species of interest in these studies. Results from these studies are shown in Fig. 4 using full-length tPA and K2P. Values for maximum binding and K d were determined from direct fitting to these data using a single site binding isotherm model. The K d values were 487 Ϯ 117 nM for the full-length enzyme and 1122 Ϯ 320 nM for K2P. There was also an approximately 2-fold difference in the maximum level of bound activity of these two enzymes (1659 Ϯ 250 for full-length tPA and 835 Ϯ 166 A/s 2 for K2P). This could be due to a decreased number of binding sites for K2P compared with full-length tPA, or alternatively a reduction in the activity of K2P compared with full-length tPA. In either case, it is possible in theory to use these data to calculate the number of binding sites per cell. However, in practice, this is likely to be inaccurate, as weak interactions such as this will undergo some dissociation during processing and determination of activity (28).
Comparison of Model and Experimental Data-Model 1 was not considered in detail as it cannot account for the rate versus cell concentration profile observed in experimental data ( Figs.  1 and 2). The values used with models 2 and 3 to simulate cell-associated plasminogen activation rates are shown in Table I. The initial assumptions of the model outlined above stated that the effect of cells in this system is to concentrate reactants and intrinsic enzyme parameters, k cat and K m , remain the same whether bound or free. However, it became clear that concentration of reactants alone was insufficient to explain the results as no satisfactory fitting of the data could be achieved. It was necessary to allow for a reduction in k cat Ј for the cell-bound reaction. This requirement is apparent from the equation for model 2 and Fig. 3. Because model 2 is a related to competitive inhibition, V max Ј (cell-associated) should approach V max (in solution) at high [PgnЈ]. This is not the case from the data shown in Fig. 3, A and B, indicating that k cat Ј is lower than k cat .
Using the values for K d , binding sites, and K m shown in Table I, data were fitted by nonlinear regression analysis to determine values for k cat Ј and to a set of data for which the activation rate was measured with varying cell and plasminogen concentrations. Because there are two independent variables in this system (plasminogen and cells), multiple regression analysis was used to derive values for the unknown parameters, using Grafit, version 4.0, software (25). The results from fitting in this way gave the results shown in Table I for k cat Ј and . The k cat Ј of 0.004 s Ϫ1 represents a 15-fold decrease relative to the free solution reaction. The value of is a measure of the concentration of reactants around a cell. This estimate of 3.65 ϫ 10 11 can be thought of as the reciprocal of the reaction volume (in ml) around 1 cell. Model data and experimental data are shown in Fig. 5 for comparison and are in reasonable agreement as can be seen from the pattern of stimulation, the absolute rates, and the optimum cell concentration. Both sets of data have the same trends for increasing apparent K m with increasing cell concentration and increasing optimum cell concentration with increasing plasminogen concentration. The level of stimulation of bound enzyme activation rate was also in good agreement with experimental results over  4. Ligand binding data for the interaction of full-length tPA and K2P with Nalm6 cells. Nalm6 cells (10 7 cells/ml) were incubated with varying concentrations of tPA, either full-length glycosylated (q) or K2P (ࡗ), and bound enzyme was determined, after separation of nonbound, in plasminogen activation assays. Parallel experiments without cells were conducted to determine nonspecific enzyme carryover (open symbols). Nonspecific binding was subtracted from bound activity (ϫ for K2P), and nonlinear regression analysis was used to determine K d and maximum level of activation by fitting to a single binding isotherm using corrected data. this range of cell and plasminogen concentrations.
Model 3 was also applied to the data shown in Fig. 5. Fitting by multiple regression provided estimates for cell-bound enzyme parameters of K m Ј ϭ 7 Ϯ 2 nM and k cat Ј ϭ 0.0039 Ϯ 0.0003 s Ϫ1 . Thus, the same value for k cat Ј as with model 2, but with a decrease of almost 10,000-fold in K m from 65 M to 7 nM on cell binding. The three-dimensional plot was essentially superimposable on the plot shown in Fig. 5 (model), with only minor differences at the extremes. Further exploration of model 2 was performed using constants and variables for full-length tPA and the deletion variant K2P given in Table I, in an attempt to replicate the results shown in Figs. 1 and 3. Model results are shown in Fig. 6 for fixed plasminogen (100 nM) and varying cells (Fig. 6A) and fixed cells (10 7 cells/ml) with varying plasminogen (Fig. 6B). Data were generated applying the two possible alternative explanations for the results in Fig. 4. That is, the reduced maximum binding to cells of K2P relative to full-length tPA was due to a 2-fold reduction in binding sites, or to a 2-fold reduction in the cell-bound k cat Ј. By comparing Fig. 6 with the experimental data shown in Figs. 1 and 3, it is clear that the best explanation is a reduction in k cat Ј, which is 0.004 s Ϫ1 for full-length tPA and 0.002 s Ϫ1 for K2P at the cell surface. This provides levels of stimulation in Fig. 6B of 36.1-and 16.7-fold for full-length tPA and K2P, respectively, in the same range as the data from Fig. 1. Model 2 also explains the approximate 2-fold increase in apparent K m Ј noted above for deletion variants of tPA relative to full-length enzyme (Fig. 3) as being due to the 2-fold increase in K d (from Fig. 4

and Equation 17).
Plasminogen and tPA Conformational Changes-To investigate possible involvement of lysine binding sites on tPA or plasminogen inducing conformational changes leading to stimulation of enzyme activity, kinetic experiments were performed in the presence of a range of 6-AHA concentrations instead of cells. Fig. 7 shows the effect of 0.005-10 mM 6-AHA on activation rate of Glu-plasminogen by full-length glycosylated tPA, tPA protease domain, and an alternative activator, uPA. Clearly, there was little effect of 6-AHA binding to tPA as full-length tPA and protease were identical (overall, a slight inhibition of 14% was observed in both cases). A different mechanism operates in the case of uPA where the conformational change in Glu-plasminogen at around 2 mM 6-AHA, previously found (29), did produce an enhanced activation rate (3.2-fold stimulation in this case), in agreement with earlier findings (30). Activation by tPA was much less sensitive to this conformational change. Similar experiments using Lys-plasminogen showed only inhibition with increasing 6-AHA such that over the same concentration range, activation rates were inhibited by 69.5, 60.9, and 35.1% for tPA, protease domain, and uPA, respectively.
All of the studies shown above using Glu-plasminogen as substrate were repeated using Lys-plasminogen and gave similar results for these cell types and tPA variants, although activation rates were higher with and without cells resulting in lower levels of stimulation (data not shown).

DISCUSSION
From the data presented here, it is clear that the cells used are able to act as efficient promoters of plasminogen activation by tPA. These cells have previously been characterized for binding of plasminogen, uPA and tPA (8,9,31,32). In the present detailed study, the levels of stimulation with fulllength glycosylated tPA were up to 30 -80-fold, which is as good as or better than observed in previous studies with cells or isolated receptors or heparin or fibrinogen fragments (5, 21). Furthermore, the present study was performed using physiological tPA concentrations. The two cell types studied in detail a Taken from previous studies that show that tPA and plasminogen interact with the same cellular binding sites (3,8).
b K d values determined in the present study (Fig. 4). c Cell-bound activity was lower for K2P than full-length tPA (Fig. 4). This could be due to fewer binding sites or lower activity for K2P. Lower activity is more likely from our results (Fig. 6).
d Taken from previous studies (11), k cat /K m agrees with data in Fig. 3. e k cat Ј and are derived from curve fitting to model 2 (Equation 17) by multiple nonlinear regression analysis, where R ϭ /cells/ml.

FIG. 5.
Model results and experimental data showing the dependence of activation rate on cell and plasminogen concentration. Experimental data were collected using Nalm6 cells and Gluplasminogen over the ranges shown in the presence of 70 pM full-length glycosylated tPA. Model data were generated for full-length tPA from model 2 and the parameters in Table I using k cat Ј and from nonlinear regression analysis. Model results using model 3 and values for K m Ј ϭ 7 nM and k cat Ј ϭ 0.004 s Ϫ1 from nonlinear regression analysis are not shown, but were very similar to those presented for model 2.  Table I in the presence of 100 nM plasminogen for full-length tPA (q) and K2P (diamonds) for comparison with data in Fig. 1. B shows the effect of varying plasminogen at 1 ϫ 10 7 cells/ml for the same enzymes and tPA with nl cells (ϫ) for comparison with Fig. 3. Open diamonds show the assumption that K2P binds to Nalm6 cells with a K d ϭ 1122 nM and has half the binding sites/cell as full-length tPA and k cat Јϭ0.004 s Ϫ1 , as for full-length tPA, to account for the data in Fig. 4. Closed diamonds are for K d ϭ 1122 nM with the same number of binding sites per cell as full-length tPA (1.6 ϫ 10 7 /cell), but k cat Ј for K2P is now 0.002 s Ϫ1 .
here were of different origins (U937, a monocytoid cell line, and Nalm6, a pre-B leukemic cell line) but were very similar in their behavior as promoters of tPA activity. In fact, all cells investigated showed similar levels and patterns of stimulation. The pattern of stimulation of the different tPA variants, the ubiquitous binding sites, and the results of binding studies all suggest that relatively weak nonspecific interactions regulate the activity of cell-associated tPA. It is also interesting that stimulation of kinetics by cells or by fibrin are similar in so far as they both depend on multiple sites of interactions. Kringle 2 and finger-growth factor domains have been identified as important in fibrin binding and activity (33,34), but it is now apparent that multiple interactions with fibrin throughout the tPA molecule may play a role in regulating activity (13), and the mechanism of binding may be complex (35). The situation appears to be similar when cells are promoters, although in our studies finger domain is very important, as suggested in previous studies using endothelial cells (36,37).
The effects of domain deletion on stimulation seen here with tPA variants was best explained by lower affinities for binding and a reduction in activity of cell-bound enzyme. It is interesting in this regard that a recent study using vascular smooth muscle cells found a 3-fold lower maximal cell-bound activity for K2P relative to full-length tPA and concluded this was due to the loss of a higher affinity site available only to the fulllength molecule (22). In the present study, data fitting was not improved using a two site model. The observation of Kohnert et al. (18) that fibrin stimulation of K2P was only 35% the stimulation observed for full-length tPA is also very similar to our results with cells. Furthermore, optimal heparin concentrations have been found to stimulate full-length tPA more than K2P, by 22-and 13-fold, respectively (38). All these observations would be expected for a mechanism in which bound K2P has a lower k cat Ј than bound full-length tPA.
Mechanism of Stimulation-From a practical viewpoint, reporting observed k cat and K m values is useful as a measure of degree of stimulation by comparing k cat /K m in the presence and absence of cells. However, caution is necessary not to overinterpret these parameters in proposing direct effects on bound proteins. Furthermore, the apparent K m and k cat values from model 1 are related to total substrate and enzyme added, irrespective of whether it is cell-bound or in solution.
Model 2 has been applied in the present study as an alternative approach. Here, stimulation is primarily a result of the concentration of reactants by the cell. The initial assumptions in model 2 were that K m and k cat are the same whether reactions are in free solution or cell-associated. The experimental data in Fig. 3 showed that this simple assumption could not be applied and there was a need to reduce the k cat Ј for the cellbound enzyme. Curve fitting could then be used to generate values for and k cat Ј. When this was done, model and experimental results were in good agreement, within the variation observed from batch to batch of cells. The reason for the reduction in cell k cat Ј relative to solution reaction is not known but was necessary in all models. The lower activity could result from different conditions close to the cell surface due to different physical conditions (pH, ionic strength, etc.) due to the effect of cell membrane and components (39) or factors relating to replenishment of substrate and removal of product in this compartment such that the maximum k cat Ј is not observed (40). Alternatively, this could be a real regulatory effect on the cell surface. For example, under conditions of high local plasminogen concentration close to the cell, there may be allosteric regulation by substrate, which has been noted in some previous studies (41,42).
Model 2 is also able to account for the template profile seen in Figs. 1 and 2 for varying cell concentration, which is explained by two competing effects. As cell concentration increases, mass action causes the equilibrium to shift toward bound reactants up to a plateau, with concomitant increase in observed rate. Conversely, as cell concentration increases, the volume of V ca also increases, and hence the concentration of reactants in V ca decreases.
The value of derived from multiple regression fitting can be used to calculate the volume of the cell-associated compartment and from this the height of the layer (Scheme 1, h) can be determined. A value of 3.65 Ϯ 0.97 ϫ 10 11 translates into a value of h ϭ 11 nm (range, 8 -15 nm) using a uniform cell diameter for Nalm6 cells of 9 m. This is a significant value from several points of view. The previous study by Nesheim et al. (23), using the same principles to develop their model to study prothrombinase activation in the presence of phospholipid vesicles, used previously published physical studies to estimate the height of their "interface shell." The height of this shell was initially set at 12 nm but was adjusted upwards to 39 nm to fit the data more closely. Furthermore, light scattering studies on plasminogen have provided information on dimensions of the molecule (29). Native Glu-plasminogen is a prolate elipsoid of dimensions 14.7 ϫ 5.7 ϫ 5.7 nm, which extends to a longer open form in the presence of 6-AHA and presumably as Lys-plasminogen and plasmin. The longest dimension of these forms is then increased to 24 nm. From these values, it would appear that the cell-associated compartment is a single layer of molecules around the cell. Interestingly, prothrombin is also a prolate elipsoid of similar dimensions to plasminogen and has been postulated to bind and extend from the vesicle membrane to a distance of 11 nm, the longest dimension (43). The significance of the height of interface shell has also been discussed by Nesheim et al. (23) and clearly does not represent a distinct isolated compartment. What these observations most likely mean is that the probability of plasminogen activation taking place is high at distances very close to, or in contact with, the cell surface. Another observation arising from the model developed by Nesheim et al. (23) for prothrombin activation was apparent substrate inhibition resulting from competition between enzyme and substrate for the same binding sites. The consequences were similar in the tPA and prothrombin systems.
Model 3 has also been included in the analysis and, like model 2, can explain the experimental data well. The crucial difference between models 2 and 3 is that the latter requires a huge change in intrinsic enzyme K m of almost 10,000-fold on cell binding. There is no evidence that this magnitude of change can happen, for example, as a result of conformational changes. Indeed, attempts to induce conformational changes using lysine analogues that are known to act through kringle domains did not produce any significant stimulation of activation rate with tPA, as seen in Fig. 7. Previous studies have also suggested that the protease domain acts independently of the N-terminal, A-chain domains (34,44). It is interesting that the uPA system is stimulated in this way and is more sensitive to the conformation of plasminogen as affected by 6-AHA, fibrinogen fragments, or monoclonal antibodies to the 6-AHA binding site (30,45). Model 3 also assumes that only free plasminogen is available for reaction with cell-bound tPA if the correct "template" profile is to be obtained. Again, there is no evidence to support this. On the contrary, lysine analogues that are able to block fully plasminogen binding, but only partially block tPA binding to cells, are completely effective at inhibiting cellular stimulation of plasminogen activation in our studies (data not shown) and elsewhere (22).
Although there is reasonably good agreement between the data and model 2, there are a number of sources of error in the parameters we have applied. A simplification that was made in multiple regression analysis was to ignore any competition between plasminogen and tPA for the same binding sites although this has been demonstrated. It was not possible to include this feature in the data fitting. However, it could be included in simulated data using model 2 and had a noticeable effect at high plasminogen concentrations (Ͼ500 nM) at which rates began to fall. This effect would lead to an underestimation of the true cell-bound k cat Ј. In vivo, this phenomenon may play a major role in inhibiting excess plasminogen activation as the concentration of plasminogen is around 2 M. In the present study any effects of single chain to two chain tPA conversion or generation of Lys-plasminogen have also been disregarded. This is justified because of the conditions chosen for the study in which reactants were maintained at low concentrations and plasmin generation monitored at early times to gather initial rates of plasmin generation. Furthermore, previous work has shown that single chain and two chain tPA behave similarly in the presence of cells (21). Other inherent problems are the difficulties in accurately measuring dissociation constants for weak complexes, as here for both tPA and plasminogen. The cell lines used also showed some inevitable batch to batch variation, although Nalm6 cells were ideal, being exceptionally homogeneous in terms of size and shape, and were very robust in our assays. tPA is not an easy enzyme to work with due to low solubility and difficulties in quantitation: for example, active site titration and artifacts arising from allosteric interactions with chromogenic substrates (46). However, the approach used in the present study to standardize free solution activities of plasminogen activation of all tPA variants against full-length tPA would minimize these difficulties and afford the most direct way of studying stimulation. Another potential problem with model 2 is that it ignores the contribution of the free solution reaction. Although this is insignificant at low substrate concentrations, it cannot be disregarded as substrate is increased. Model 2 is closely related to the equation for specific activation of an enzyme, which makes the simplifying assumption that the unstimulated enzyme is not active (24). More complex equations have been developed when this simplification does not hold (47). As apparent from the equation for model 2, the value for is closely related to K m Ј. There is no way of measuring K m Ј directly, which adds uncertainty to the value for and h. However, what can be said from the results using model 2 is that large changes in K m on cell binding are not essential to explain stimulation of activity.
Despite these problems, model 2 did provide a good explana-tion for the experimental data obtained. Nevertheless, it is undoubtedly a simplification, and as with the work of Nesheim et al. (23), it is open to criticisms. A number of alternative theoretical treatments dealing with surface activation kinetics have been proposed, some of which are also concerned with coagulation enzymology in the presence of phospholipid vesicles (26, 48 -50) or on platelets (51) or involved other systems (40). These approaches deal with problems related to heterogeneous catalysis in a variety of ways, placing more emphasis on the effect of co-localization of reactants, kinetics of interactions with and on the surface, orientational effects, conformational changes, receptor occupancy, and the effects of transition from three-dimensional to two-dimensional geometry. Model 2 does have the advantage of being intuitively simple and relies primarily on the concentration effect of cells with only minor modifications to cell-associated enzyme behavior, which appear reasonable. The model can simulate observed data over tPA, plasminogen, and cell concentration ranges, with different tPA variants. The data support the idea that low affinity, high capacity interactions can regulate tPA activity by cells, even at physiological tPA concentrations. This approach may be useful in understanding tPA regulation by fibrin and in the regulation of other, surface-bound, enzyme activities.