Tight regulation of serine proteases is essential for their physiological function, and unbalanced states of protease activity have been implicated in a variety of human diseases. One key example is the presence of uPA (urokinase-type plasminogen activator) in different human cancer types, with high levels correlating with a poor prognosis. This observation has stimulated efforts into finding new principles for intervening with uPA's activity. In the present study we characterize the so-called autolysis loop in the catalytic domain of uPA as a potential inhibitory target. This loop was found to harbour the epitopes for three conformation-specific monoclonal antibodies, two with a preference for the zymogen form pro-uPA, and one with a preference for active uPA. All three antibodies were shown to have overlapping epitopes, with three common residues being crucial for all three antibodies, demonstrating a direct link between conformational changes of the autolysis loop and the creation of a catalytically mature active site. All three antibodies are potent inhibitors of uPA activity, the two pro-uPA-specific ones by inhibiting conversion of pro-uPA to active uPA and the active uPA-specific antibody by shielding the access of plasminogen to the active site. Furthermore, using immunofluorescence, the conformation-specific antibodies mAb-112 and mAb-12E6B10 enabled us to selectively stain pro-uPA or active uPA on the surface of cultured cells. Moreover, in various independent model systems, the antibodies inhibited tumour cell invasion and dissemination, providing evidence for the feasibility of pharmaceutical intervention with serine protease activity by targeting surface loops that undergo conformational changes during zymogen activation.
- urokinase-type plasminogen activator (uPA)
Many serine proteases with a trypsin-like fold have important pathophysiological functions. Development of new therapeutics for intervention with these are therefore of great interest. The widely used strategy of developing small molecule inhibitors targeting the catalytic site has proved a daunting task since the catalytic site topology of different proteases are often very similar, making it difficult to obtain sufficient specificity. One strategy to overcome this difficulty is to target other steps in the natural regulation of serine protease activity.
In nature, a key mechanism for the regulation of serine proteases is proteolytic activation of the initially secreted zymogens or proenzymes. Zymogen activation allows for rapid amplification of the activation signal and generally occurs by cleavage of the bond between amino acid residues 15 and 16 (using the chymotrypsin template numbering). The new N-terminus inserts into a hydrophobic binding cleft forming, in addition to hydrophobic interactions, a salt bridge to the side chain of Asp194, which stabilizes the substrate-binding pocket and oxyanion hole in a catalytically productive conformation. X-ray crystal structure analyses of trypsinogen and trypsin as well as chymotrypsinogen and chymotrypsin showed that conformational changes after cleavage involve four loop regions collectively called the activation domain, including the activation loop (residues 16–21), the autolysis loop (residues 142–152), the oxyanion-stabilizing loop (residues 184–194) and the S1 entrance frame (residues 216–223). The catalytic activity of a zymogen relative to the mature protease is in general the result of an equilibrium between active and inactive conformational states of the protease domain involving these four surface loops (for reviews, see [1,2]).
The termination of serine protease activity is likewise a key physiological regulatory event, with inhibition mainly occurring by other proteins with surface-exposed loops that can bind covalently or non-covalently to the active site of the proteases. Inhibitors of the serpin family are an important example of such regulatory proteins. Of crucial importance for the inhibitory mechanism of serpins is the surface-exposed RCL (reactive centre loop), tethered between β-strands 1C and 5A. The active site of the protease binds to the P1–P1′-bond of the RCL to form a non-covalent Michaelis complex and attacks it as a substrate, but at the enzyme-acyl intermediate stage, the N-terminal part of the RCL inserts as β-strand 4, thereby pulling the protease to the opposite pole of the serpin and distorting its active site so that it is unable to complete the catalytic cycle (for reviews, see [3–5]).
A serine protease of particular relevance is uPA (urokinase-type plasminogen activator), which catalyses the conversion of plasminogen to the active protease plasmin that in turn directly catalyses the degradation of extracellular matrix proteins. Abnormal expression of uPA is implicated in tissue remodelling in several pathological conditions, and in particular, uPA is central to the invasive capacity of malignant tumours (for reviews, see [6–8]). As with all trypsin-like proteases, uPA has a catalytic serine protease domain, with surface-exposed loops around residues 37, 60, 97, 110, 170 and 185. Besides the catalytic domain, uPA has an N-terminal extension consisting of a kringle domain and an EGF (epidermal growth factor) domain. The latter domain functions in binding to the cell-surface-anchored uPAR (uPA receptor) (reviewed in ). Several proteases, including plasmin (for a review, see ), glandular kallikrein , matriptase  and hepsin , can catalyse the activation of the zymogen pro-uPA. The primary inhibitor of uPA is the serpin PAI-1 (plasminogen activator inhibitor-1). Whereas several three-dimensional structures of the catalytic domain of uPA in the active conformation have been determined by X-ray crystal structure analysis, the structure of pro-uPA remains to be reported and details of zymogen activation are therefore still unclear.
Conformational changes in proteases induced by events such as zymogen activation and cofactor binding have previously been studied using conformation-sensitive antibodies as probes. In a study of Factor VIIa, it was found that binding of a monoclonal antibody was affected by conformational transitions of the protease domain , and conformational changes of a specific surface residue bound by the antibody were found to be linked to changes in the active site . Furthermore, conformational antibodies have been shown to possess interesting inhibitory mechanisms alternative to standard competitive inhibitors binding at the active site. Using X-ray crystallography, it was found that the antibody Fab40 against the trypsin-like serine protease hepatocyte growth factor activator allosterically inhibits the protease activity by binding to a surface loop, keeping the active site in a conformation incompatible with substrate binding .
We have previously described a conformational monoclonal antibody with a novel mechanism of interfering with the activity of uPA, i.e. by inhibiting conversion of pro-uPA to active uPA . On the basis of this finding and with the dual purpose of elucidating mechanisms of zymogen activation and finding new principles for therapeutic intervention with serine protease activity, in the present study we performed a functional characterization of three inhibitory monoclonal antibodies with strongly differential affinity to pro-uPA or active uPA. Mapping of the epitopes was used as a means of localizing regions changing conformation upon pro-uPA activation as well as by deciphering interaction areas on uPA when reacting with its macromolecular substrate, plasminogen, and inhibitor, PAI-1. The autolysis loop of uPA was found to be important for binding of all three conformation-specific antibodies. Consequently, this loop seems to be strongly implicated in conformational changes during the activation of pro-uPA as well as in uPA's interaction with plasminogen and PAI-1. Furthermore, the efficacy of regulating uPA protease activity by the different mechanisms employed by the monoclonal antibodies was evaluated in cancer model systems in vitro and in vivo. Targeting surface loops outside the conserved catalytic site was found to be a feasible approach for the inhibition of cancer cell invasion and dissemination. Additionally, the differential affinity of the antibodies to pro-uPA or uPA allowed us to detect pro-uPA or uPA on the surface of human tumour cells by immunofluorescence. Therefore these antibodies potentially represent new tools for separate analysis of pro-uPA and active uPA levels as relevant biomarkers.
MATERIALS AND METHODS
Unless otherwise indicated, all in vitro reactions were carried out in a buffer containing 30 mM Hepes, pH 7.4, 135 mM NaCl, 1 mM EDTA and 0.1% BSA (HBS-B) or 0.1% polyethylene glycol 8000 (HBS-P).
Human two-chain uPA was purchased from Wakamoto. Recombinant human pro-uPA was a gift from Abbott Laboratories. The uPA–PAI-1 complex was prepared as described previously . The pro-uPA–PAI-1 complex was prepared by incubating pro-uPA and PAI-1 (1:2 molar ratio) in the presence of 2 μg/ml aprotinin and 10 mM isoleucine-valine for 1 h at 37 °C . The complexes were analysed by SDS/PAGE under reducing conditions, confirming that the uPA in the complex was either in the single-chain pro-uPA form or in the two-chain active uPA form, as intended.
Recombinant wt (wild-type) and mutant human pro-uPA and active uPA variants were expressed in HEK-293T (human embryonic kidney-293 cells expressing the large T-antigen of simian virus 40) cells . When expressed under standard conditions, at least 50% of the uPA in the conditioned medium was active uPA, as evaluated by immunoblotting analysis under reducing conditions. When cultured in the presence of 5 μg/ml aprotinin, no conversion of pro-uPA to active uPA was observed. When active uPA was needed, HEK-293T cells were grown in the absence of aprotinin and the pro-uPA in the conditioned medium was converted to uPA by adding plasmin. The concentration of uPA variants in the conditioned medium was determined using SPR (surface plasmon resonance) analysis (see below).
The serine protease domain of recombinant human matriptase (residues 596–855) was purchased from R&D Systems. Glu-plasminogen purified from human plasma was a gift from Lars Sottrup-Jensen (Department of Molecular Biology, Aarhus University, Aarhus, Denmark).
All procedures involving the use of mice were approved by the Danish Experimental Animal Inspectorate. Monoclonal antibodies against pro-uPA were generated by intraperitoneal immunizations of Balb/c mice with recombinant human single-chain pro-uPA yielding 22 clones including mAb-101 and mAb-112 , or with recombinant human two-chain uPA yielding mAb-12E6B10 . Antibodies were purified from hybridoma-conditioned medium using Protein G–Sepharose 4FF . A 1 litre volume of cell culture supernatant yielded between 15 and 50 mg of purified antibody. mAb-PUK was purchased from Technoclone.
The following antibodies were also used in this study: mouse anti-uPA mAb-6 ; rabbit polyclonal anti-uPA antibody F1609 ; mouse anti-PAI-1 mAb-2 ; mouse anti-PAI-1 mAb-7 ; and mouse anti-CD44 mAb 29-7 .
SPR analyses were performed on a BIACORE T100 instrument, using CM5 sensor chips, flow rates of 30 μl/min and HBS-B with 0.05% Tween 20. Concentrations of uPA variants in conditioned medium from HEK-293T cells were determined by measuring the initial rate of binding to a chip with 200 RU (response units) of anti-uPA mAb-6 with an epitope in the kringle domain , using a standard curve of purified pro-uPA or uPA. Affinities of the antibodies for wt or various forms of mutant pro-uPA and uPA were determined by injecting the proteases (0.5–150 nM), either purified or in conditioned medium, over a chip with 200 RU of immobilized mAb-112, mAb-101, mAb-PUK or mAb-12E6B10. The kon, koff and Kd values were determined by global fitting to a 1:1 binding model with the BIACORE evaluation program.
Binding of the antibodies to a complex between uPA and PAI-1 or between pro-uPA and PAI-1 was assessed by capturing 500 RU of preformed complexes on a chip with anti-PAI-1 mAb-7 immobilized to a final response level of 2000 RU. Binding was then evaluated by injecting the antibodies in concentrations of up to 200 nM.
Binding of the antibodies to pro-uPA or uPA prebound to uPAR was analysed by capturing 300 RU of pro-uPA or uPA on a chip with human soluble uPAR immobilized to a final response of 1700 RU. Binding was evaluated by injecting the antibodies in concentrations of up to 1 μM.
The ability of the various antibodies to compete with each other for the binding to either uPA or pro-uPA was evaluated by immobilizing a chip with 200 RU of one antibody and injecting uPA or pro-uPA alone or pre-incubated with an excess of the other antibodies.
In vitro plasminogen activation assay
Pro-uPA or uPA (0.25 nM) was pre-incubated with various concentrations of the antibodies (5–160 nM) at room temperature (23–25 °C) for 30 min in HBS-B. Addition of 0.5 μM Glu-plasminogen and 0.5 mM of the plasmin substrate S-2251 (H-D-Val-Leu-Lys-p-nitroanilide) initiated the reaction at 37 °C. S-2251 hydrolysis was monitored as the increase in absorbance at 405 nm.
Pro-uPA activation assay
Pro-uPA (10 nM) was incubated in the presence or absence of 100 nM of antibodies at room temperature for 30 min in HBS-B. At time zero, 0.5 nM plasmin or 5 nM matriptase was added. Then, the mixtures were incubated at room temperature for the indicated time periods, after which the activity of plasmin and matriptase was quenched by addition of 1 μM aprotinin. The uPA substrate S-2444 (H-D-Glu-Gly-Arg-p-nitroanilide) was added to a final concentration of 0.5 mM and the amount of active uPA generated during the incubation with plasmin and matriptase was estimated by the rate of S-2444 hydrolysis and comparison with a standard curve of active two-chain uPA.
Analysis of pro-uPA cleavage by SDS/PAGE and immunoblotting analysis
Samples of pro-uPA (200 nM) were incubated in the presence or absence of 300 nM antibody for 30 min at room temperature in HBS-P. At various times after the addition of 5 nM plasmin, samples were removed and analysed by reducing SDS/PAGE and immunoblotting with rabbit polyclonal anti-uPA antibody F1609.
Complex formation between pro-uPA and PAI-1 followed by SDS/PAGE
The effect of antibodies on the complex formation between single-chain uPA and PAI-1 was analysed by pre-incubating pro-uPA with or without a 2-fold molar excess of mAb-PUK, mAb-112 or mAb-101 for 30 min at room temperature in HBS-P with 1 μg/ml aprotinin added. The reaction was started by addition of PAI-1 in a 1.5-fold molar excess compared with pro-uPA. After incubation for various time points, the reaction was stopped by addition of PMSF to a final concentration of 1 mM and boiling for 10 min. The reaction products were analysed by SDS/PAGE under non-reducing conditions. The resulting protein bands were visualized with Coomassie Brilliant Blue staining. Relative amounts of pro-uPA alone and in complex were quantified by densitometric scanning using Quantity One software (Bio-Rad Laboratories). Reaction rates were evaluated as kobs, the pseudo-first-order rate constant describing the monoexponential approach to complete inhibition, by fitting the data points to eqn (1), where [AB]t is the amount of complex formed between pro-uPA and PAI-1 at time t and [AB]max is the maximal amount of complex that can be formed: (1)
Cell-surface-associated plasminogen activation
uPA-catalysed plasminogen activation was assayed with pro-uPA or active uPA bound to uPAR at the surface of U937 cells. U937 cells were cultured as described previously , washed, and resuspended for 3 min in 50 mM glycine/HCl, pH 3, containing 0.1 M NaCl to remove endogenous pro-uPA and uPA from the cell surface. After washing, the cells were resuspended in HBS-B and distributed into the wells of non-transparent 96-well microtitre plates (Nunc) yielding a final cell density of 5×106 cells/ml. Cells were preincubated for 15 min at 37 °C with 200 nM plasminogen and 400 nM α2-antiplasmin, to restrict plasmin activity to the cell surface, prior to initiating the reactions by adding 1 nM pro-uPA or uPA pre-incubated alone or with mAb-101, mAb-112, mAb-PUK or mAb-12E6B10 at various concentrations and the fluorogenic plasmin substrate VLK-AMC (H-D-Val-Leu-Lys-7-amido-4-methylcoumarine; 200 μM). The increase in fluorescence was monitored in a Spectromax Gemini fluorescence plate reader (Molecular Devices) using a λex of 390 nm and a λem of 480 nm. The approximate IC50 values of the inhibition of plasminogen activation were calculated by plotting the relative fluorescence development after incubation for 60 min against the antibody concentration and fitting the curves to a hyperbolic decay equation.
In vitro invasion assays
For the cancer cell invasion assays, PC-hi/diss cells were used . In Matrigel invasion assays, the upper side of a 6.5 mm insert with 8 μm pore Transwell membranes (Fisher Scientific) was pre-coated with 2 μg of Matrigel (BD Biosciences). As an attractant for cell migration and invasion, CM-CEF (conditioned medium from chicken embryonic fibroblasts) was used. CM-CEF was prepared by growing chicken embryonic fibroblasts in serum-containing medium to confluence with approximately 5×104 cells/cm2. The cells were washed and overlaid with 6 ml of serum-free medium for up to 48 h before CM-CEF was harvested. CM-CEF was diluted 1:1 with SF-DMEM (serum-free Dulbecco's modified Eagle's medium) and placed in the lower chamber. PC-hi/diss cells (1×105) were plated in 100 μl of SF-DMEM on top of the Matrigel in the upper chamber and invasion was allowed for 48 h at 37 °C. After incubation, adherent cells from the underside of the membrane were detached with trypsin/EDTA, combined with non-adherent cells from the lower chamber and counted. Monoclonal antibodies were added to both the upper and lower chambers at 333 nM final concentrations. Aprotinin was used as a positive control for the inhibition of invasion at a concentration of 0.1 TIU (trypsin inhibitory units)/ml.
For the three-dimensional tumour escape assay, a drop containing 2.1 mg/ml collagen (type I from rat tail; BD Biosciences) and 25×103 PC-hi-diss cells in SF-DMEM was prepared. After gelation, this drop was then immersed into a mixture containing 3 mg/ml fibrin (fibrinogen from bovine plasma; Sigma–Aldrich; activated by thrombin), 1.5 mg/ml collagen and 20 ng/ml human EGF. Once the droplet-containing matrix had polymerized, AIM-V medium containing 0.01% chicken serum and monoclonal antibodies (final concentrations of 167 nM) or aprotinin (final concentration of 0.1 TIU/ml) was overlaid. Cells were allowed to escape into the surrounding fibrin/collagen matrix for 10 days at 37 °C. The AIM-V medium containing inhibitors was exchanged on days 3 and 7. Images were captured on day 10 with an Olympus microscope equipped with an infinity1 at 10× original magnification. The number of escaped cells and distance invaded from the original spheroid were quantified by ImageJ (http://rsbweb.nih.gov/ij/).
Chicken embryo intravasation assay
The effect of anti-uPA monoclonal antibodies on the intravasation of human cancer cells in the chicken embryo was analysed as described previously . Briefly, SPAFAS White Leghorn embryos (Charles River) were allowed to develop for 10 days, after which the CAM (chorioallantoic membrane) was dropped and 2.5×106 PC-hi/diss cells in 25 μl were applied to the CAM. On days 2 and 4 after tumour cell grafting, 25 μg of monoclonal antibodies in 100 μl PBS, pH 7.4 (Dulbecco's PBS; Lonza), with 5% DMSO were added topically to the primary tumours formed on the CAM. On day 7, primary tumours were excised and weighed and the lower CAM harvested for analysis of human cancer cell intravasation quantified by qPCR (quantitative PCR) on primate-specific Alu-sequences, as described previously .
Immunocytochemical detection of uPA or pro-uPA on the surface of PC-hi/diss cells
To discriminate between human pro-uPA and active uPA on the surface of PC-hi/diss cells, the conformation-sensitive antibodies mAb-112 and mAb-12E6B10 were employed in immunofluorescence microscopy. To this end, 3×105 PC-hi/diss cells were seeded on to coverslips pre-coated with collagen. The cells were allowed to grow in AIM-V medium alone or in the presence of 0.1% chicken serum with or without 0.1 unit/ml aprotinin. After 2 days of growth, the cells were washed in PBS, fixed with 2% paraformaldehyde and incubated with PBS containing 2% BSA and 2% goat serum to block any non-specific binding. A total of 4 μg/ml of mAb-112, mAb-12E6B10, mouse IgG or anti-CD44 mAb 29-7 (conditioned hybridoma medium from clone 29-7 diluted 1:1) was added to the cells and allowed to bind overnight at 4 °C. After washing with PBS, FITC-conjugated goat anti-mouse antibody was added and allowed to bind for 1 h at room temperature. The cell nucleus was stained with 1 μg/ml DAPI (4′,6-diamidino-2-phenylindole). Pictures were acquired with an immunofluorescence microscope using the same exposure time for all samples. In parallel, samples of conditioned medium from the cells grown under the various conditions were taken and analysed for the relative amounts of pro-uPA and uPA by reducing SDS/PAGE followed by Western blotting.
Choice of conformation-specific monoclonal antibodies
Three monoclonal antibodies, showing preferential binding affinity towards either pro-uPA or active uPA, were selected for functional characterization. Two of the antibodies have previously been described: mAb-112 was isolated after immunization of mice with human pro-uPA , and mAb-12E6B10 was isolated after immunization of mice with human active uPA . A commercially available monoclonal antibody, mAb-PUK, was also characterized. A fourth antibody, mAb-101, binding to the catalytic domain but without preference for either pro-uPA or active uPA, was selected to serve as a control. The binding kinetics of the selected antibodies towards pro-uPA and active uPA was determined by SPR analysis. As reported previously , mAb-112 has a clear preference towards pro-uPA with a Kd that is over 300-fold lower than that for active uPA (Table 1). mAb-PUK demonstrated highly selective binding towards pro-uPA. In agreement with a previous report , mAb-12E6B10 showed selective binding towards two-chain active uPA. The control antibody mAb-101 bound equally well to both forms of uPA (Table 1). mAb-112 competed with mAb-PUK for binding to pro-uPA and with mAb-12E6B10 for binding to active uPA, in the SPR analysis, but none of these three antibodies competed with mAb-101 (results not shown).
Effects of the monoclonal antibodies on plasminogen activation
The functional effects of the antibodies on uPA were investigated in a coupled plasminogen activation assay, in which plasminogen is incubated with pro-uPA or active uPA and plasmin generation scored currently by a chromogenic plasmin substrate. We chose this assay, although it does not allow a strict quantitative analysis, due to the complicated nature of the events taking place. When using pro-uPA, the reaction is initiated by trace amounts of plasmin activating pro-uPA and/or trace amounts of active uPA activating plasminogen. The plasmin formed will activate more pro-uPA and the active uPA formed will generate more plasmin. Moreover, plasmin may convert Glu-plasminogen to Lys-plasminogen, which is a better substrate for active uPA than Glu-plasminogen [29–31]. In spite of the qualitative nature of the assay as employed in the present study, it did allow us to distinguish between antibodies inhibiting pro-uPA activation and antibodies inhibiting active two-chain uPA directly. The latter type of antibodies will inhibit the assay initiated with pro-uPA as well as with active uPA, whereas antibodies binding only to pro-uPA will not inhibit the assay initiated by active uPA.
Pre-incubation of pro-uPA with mAb-112, mAb-12E6B10 or mAb-PUK inhibited the generation of plasmin (Figure 1). Plasminogen activation catalysed directly by active two-chain uPA was also reduced when uPA was pre-incubated with mAb-112 or mAb-12E6B10, but not with mAb-PUK. However, whereas mAb-12E6B10 was equally effective in inhibiting plasmin generation in assays initiated by pro-uPA or active uPA, much higher concentrations of mAb-112 were needed to inhibit active uPA (Figures 1A, 1B, 1E and 1F). At high concentrations, mAb-101 showed a small stimulating effect (Figures 1G and 1H). Stimulation of plasminogen activation by anti-uPA antibodies has previously been reported and presumed to be due to formation of a ternary complex between antibody, uPA and plasminogen, antibody–plasminogen binding mediated by binding of the C-terminal lysine residues of the antibody to the kringles of plasminogen . In contrast, no inhibition of plasminogen activation was observed using a control antibody, anti-PAI-1 mAb-2, regardless of whether the reaction was initiated by pro-uPA or uPA (Figures 1I and 1J).
Effects of the monoclonal antibodies on cell-surface-associated plasminogen activation
Cell-surface-associated plasminogen activation catalysed by uPAR-bound uPA was assessed using U937 cells, which express high levels of uPAR. By including α2-antiplasmin in the assay, the plasmin activity is restricted to the cell surface where plasmin is protected from inhibition by this serpin. The same inhibitory functionalities that were found in the coupled assay without cells were recapitulated in the cell-dependent assay (Figure 2). Thus the antibodies were also able to distinguish between pro-uPA and active uPA in a cell-based setting. When pro-uPA was incubated with mAb-PUK or uPA was incubated with mAb-12E6B10 or mAb-112, IC50 values of ~10 nM were found. When pro-uPA was incubated with mAb-112, an IC50 value of ~1 nM was found. At high concentrations, mAb-101 displayed an inhibitory effect on plasminogen activation at the cell surface. Given the fact that mAb-101 has a Kd for binding to pro-uPA and uPA, similar to those for mAb-PUK and mAb-12E6B10, this effect is likely to be non-specific.
The effects of the monoclonal antibodies on in vitro and in vivo cell invasion
To address the potential of the antibodies for interfering with uPA-dependent cell invasion, two different invasion assays were employed. The first was the Matrigel invasion assay, where cells must proteolytically degrade basement membrane proteins in order to cross an extracellular matrix barrier. The second in vitro assay was a three-dimensional invasion model, where cancer cells are incorporated into a collagen droplet and allowed to escape into a surrounding matrix composed of collagen and fibrin. In these invasion assays, we used a highly disseminating cell variant of the PC-3 prostate carcinoma, i.e. PC-hi/diss, dissemination of which in the chicken embryo was shown to be dependent on the uPA/plasmin system . In the present study we found that mAb-PUK, mAb-12E6B10 and mAb-112 were all able to significantly inhibit cell invasion in the Matrigel assay by 35–40% (Figure 3A). Also in the cell escape assay, all three monoclonal antibodies inhibited tumour cell invasion, however, mAb-112 was a more efficient inhibitor than the two other antibodies (Figure 3B).
We then analysed the capacity of the monoclonal antibodies to inhibit dissemination of PC-hi/diss cells in vivo using the CAM model system. Whereas none of the antibodies affected in vivo tumour growth (Figure 3C), mAb-112 significantly reduced the level of tumour cell dissemination to 40% of the control, whereas the other antibodies had no effect (Figure 3D).
Detection of uPA and pro-uPA on the surface of cultured cells using the monoclonal antibodies
The ability of the antibodies to bind pro-uPA or uPA prebound to uPAR on the surface of a CM5 sensor chip was tested using SPR analysis. mAb-101, mAb-112 and mAb-12E6B10 showed unaltered binding to pro-uPA and uPA when these were prebound to uPAR. Surprisingly, mAb-PUK showed >10-fold reduced binding to uPAR-bound pro-uPA compared with the binding to free pro-uPA (results not shown).
Subsequently, the capability of the conformation-specific antibodies to bind to cell-surface-associated pro-uPA or active uPA was tested by immunofluorescence analysis of non-permeabilized PC-hi/diss cells. When the cells were grown in serum-free medium, uPA was primarily present in the pro-form, as indicated by Western blot analysis of the conditioned medium (Figure 4A). Correspondingly, under serum-free conditions, PC-hi/diss cells could be stained with mAb-112 (58.9±41.7% positively stained cells), which preferentially binds to the pro-uPA, whereas no fluorescence was detected with mAb-12E6B10, which exclusively recognizes active uPA (Figure 4B). In contrast, when the cells were grown in the presence of 0.1% chicken serum, which allows for the plasmin-mediated generation of two-chain active uPA (Figure 4A), the surface of PC-hi/diss cells could be stained with mAb-12E6B10 (24.1±8.3% positively stained cells), whereas very little staining was observed with mAb-112 (4.7±4.3% positively stained cells) (Figure 4B). Both the processing of pro-uPA to two-chain uPA and the positive cell surface staining with mAb-12E6B10 were abrogated if PC-hi/diss cells were grown in 0.1% chicken serum, but in the presence of aprotinin, a potent inhibitor of plasmin (2.1±3.6% positively stained cells with mAb-12E6B10), whereas the cells under these conditions could be stained with mAb-112 (76.2±21.2% positively stained cells) (Figure 4). mAb-PUK was also tested in the staining analyses, but its use did not result in positive staining, which is in agreement with the much reduced binding of mAb-PUK to uPAR-bound pro-uPA (results not shown). In these binding assays, mAb 29-7 binding to CD44 was used as a positive control and mouse IgG antibody as a negative control.
Mechanisms of inhibition of pro-uPA activation
In view of the fact that mAb-PUK was only effective when pro-uPA was the initiator in the coupled plasminogen activation assay (Figure 1), we evaluated the effect of this antibody on the activation of pro-uPA using a more direct approach. The amount of uPA generated over time by plasmin or matriptase cleaving pro-uPA was measured in the presence or absence of mAb-PUK. mAb-PUK caused an almost complete inhibition of pro-uPA activation by both plasmin and matriptase (Figure 5A). We previously reported the same findings in the case of mAb-112 . To directly evaluate whether mAb-PUK influenced the rate of cleavage of single-chain pro-uPA into two-chain active uPA, pro-uPA was incubated with plasmin in the presence or absence of the antibody for different periods of time before being subjected to SDS/PAGE and immunoblotting analysis. In the presence of excess mAb-PUK, the plasmin-catalysed proteolytic cleavage of pro-uPA was significantly delayed, in agreement with the above described inhibition of plasminogen activation (Figure 5B). We previously reported the same findings in the case of mAb-112 . Furthermore, neither mAb-PUK nor mAb-12E6B10 showed any direct inhibitory activity towards S-2444 cleavage by either pro-uPA or active uPA (results not shown). In contrast, we previously reported that mAb-112 is a non-competitive inhibitor of S-2444 cleavage by uPA with a Ki similar to the Kd for mAb-112 binding to active two-chain uPA .
Effect of the antibodies on the interaction between PAI-1 and pro-uPA or uPA
The very slow reaction between pro-uPA and PAI-1 can be followed by the formation of a covalent complex of Mr ~100000 in SDS/PAGE. In agreement with previously published data, mAb-112 completely blocked the complex formation . mAb-PUK caused an approximately 3.5-fold reduction of the pseudo-first-order rate constant (kobs) (Figure 6). mAb-12E6B10 was previously shown to strongly delay the reaction between active uPA and PAI-1 . Conversely, mAb-101 did not inhibit the reaction of PAI-1 with either pro-uPA or uPA (Figure 6 and results not shown).
We next analysed whether the antibodies bind to preformed covalent uPA–PAI-1 or pro-uPA–PAI-1 complexes. Such complexes were therefore captured on the surface of a Biacore sensor chip by the anti-PAI-1 antibody mAb-7, which binds to the α-helix D-β-strand 2A loop of PAI-1, distant from the expected position of uPA or pro-uPA in the complexes . It was found that only mAb-101 was able to bind measurably to the covalent complexes between uPA and PAI-1 or pro-uPA and PAI-1, with Kd-values equal to that for the free proteins, whereas the other antibodies showed no measurable binding to the complexes (results not shown).
The epitopes of the conformation-specific antibodies are all localized to the activation and/or autolysis loop
We mapped the epitopes for mAb-101, mAb-12E6B10 and mAb-PUK by using alanine-scanning mutagenesis with uPA variants expressed in HEK-293T cells on pro-form when used to test for binding to mAb-PUK or on activated form when used to test for binding to mAb-12E6B10 (see the Materials and methods section). Since mAb-112 was found to be able to compete with mAb-PUK for binding to pro-uPA and with mAb-12E6B10 for binding to active uPA, residues in the vicinity of the epitope for mAb-112  were chosen for the analysis of the epitopes for mAb-12E6B10 and mAb-PUK. Binding of both antibodies was found to be dependent on specific residues within the activation and/or autolysis loop in the activation domain (Table 2 and Figure 7). Importantly, we found that the binding of mAb-PUK and mAb-12E6B10 (as well as mAb-112) depended on the same three residues within the autolysis loop, namely Glu144, Tyr149 and Tyr151. Furthermore, mAb-PUK and mAb-12E6B10 also showed reduced binding when specific residues in the activation loop were mutated. For mAb-PUK, this involved the residues Lys15, Ile16 and Ile17, and consequently, mAb-PUK proved to have an epitope spanning the peptide bond that is cleaved upon activation of pro-uPA. Mutagenesis of the buried residue Asp194 decreased the binding to mAb-12E6B10 and mAb-112, but not to mAb-PUK. In general, this residue changes conformation during the activation of serine protease zymogens as it makes a salt bridge to the new N-terminus in the two-chain form . Effects of mutating Asp194 are therefore expected to be caused by changes in the conformation of the activation domain.
mAb-101 was found to bind away from the activation domain, involving residues in the 110-loop, explaining the inability of this antibody to interfere with events involving the activation domain of uPA (Table 2 and Figure 7).
Previously, a number of monoclonal anti-uPA antibodies inhibiting the enzyme activity have been described in the literature. The first monoclonal anti-uPA antibody, selected by its ability to inhibit uPA-catalysed plasminogen activation, was reported in 1982 , but the epitope or mechanism of action of this antibody was never reported. Later, a series of inhibitory antibodies binding pro-uPA and active uPA equally well were found to have epitopes in the 37- and/or 60-loop of uPA and to inhibit uPA-catalysed plasminogen activation by steric hindrance . Since then, an antibody inhibiting mouse uPA-catalysed plasminogen activation was reported to be able to impair uPA-mediated hepatic fibrinolysis and delay skin wound healing in tPA (tissue plasminogen activator)-deficient mice. However, the epitope and detailed mode of inhibition has not been reported [35,36].
To increase our understanding of structure–function relationships for regulatory mechanisms of serine proteases, we report in the present study further data on the inhibitory mechanisms of three previously developed anti-human uPA antibodies, mAb-112, mAb-PUK and mAb-12E6B10, each having differential binding affinity to single-chain pro-uPA and two-chain active uPA. Furthermore, a fourth antibody, mAb-101, also binding the catalytic domain of uPA but distant from the active site, is reported. This antibody binds pro-uPA, active uPA, and the uPA–PAI-1 complex equally well, yet it does not inhibit pro-uPA activation or enzyme activity of uPA and does not perturb formation of the uPA–PAI-1 complex reaction.
In the present study, we placed particular emphasis on characterizing the epitopes and inhibitory mechanisms for two conformation-specific antibodies, mAb-PUK and mAb-12E6B10, in comparison with a previously partially characterized pro-uPA-specific antibody, mAb-112 , and a schematic representation of the point of action displayed by the antibodies is shown in Figure 8. The first, mAb-PUK, is a commercial antibody known to have preference for pro-uPA over two-chain uPA. Previously reported pro-uPA-specific antibodies, generated by immunizing mice with full-length pro-uPA, lack a detailed functional characterization and epitope mapping . We demonstrate in the present study that mAb-PUK is able to inhibit plasmin- and matriptase-catalysed cleavage of single-chain pro-uPA into two-chain active uPA (Figure 8). Also, we show that it has an epitope spanning the cleavage site at Lys15–Ile16 as well as additional contacts within the autolysis loop. Therefore, it apparently inhibits pro-uPA cleavage by blocking the access of activating proteases to this peptide bond. The ability to inhibit pro-uPA cleavage is common to mAb-PUK and mAb-112, in good agreement with their overlapping epitopes. Nevertheless, although the epitopes for the two antibodies are overlapping, they are clearly different. Thus, Lys15, Ile16 and Ile17 are not part of the epitope for mAb-112 . This difference in the epitopes is in good agreement with the fact that mAb-112 inhibits the conformational change following cleavage , whereas mAb-PUK can no longer bind after cleavage has occured.
mAb-12E6B10, previously shown to be selective for two-chain uPA over pro-uPA , was found to bind an epitope comprising the autolysis loop and activation loop. Thus, mAb-12E6B10 has an epitope different from previously characterized antibodies inhibiting the activity of two-chain uPA. In spite of the overlap of the epitopes for mAb-12E6B10 and mAb-112, they do not inhibit two-chain uPA by the same mechanism. In contrast with mAb-112, mAb-12E6B10 only inhibits the plasminogen activation activity of uPA, and not the hydrolysis of the small chromogenic substrate S-2444, showing that mAb-12E6B10 does not induce a conformational change in the active site, but rather inhibits plasminogen activation by sterically interfering with access of plasminogen to uPA (Figure 8). Interestingly, it has previously been reported that the autolysis loop of tPA forms an exosite for plasminogen . The localization of the epitope for mAb-12E6B10 and the ability to inhibit plasminogen activation by a steric mechanism is therefore in agreement with the hypothesis that the autolysis loop of uPA can also make an exosite interaction with plasminogen. Thus targeting the autolysis loop represents a new way of inhibiting plasminogen activation by uPA. The number of residues implicated in the epitope of mAb-12E6B10 is much larger than seen for mAb-112 and mAb-PUK. This finding agrees well with the fact that the binding of mAb-12E6B10 is extremely sensitive to conformational changes, rather than the result of all residues being directly implicated in binding. Asp194 is buried in the activation domain and directly associated with the catalytic site. Mutating this residue caused a significantly decreased binding of mAb-112 and mAb-12E6B10, again demonstrating the sensitivity of these antibodies to conformational changes. mAb-PUK, on the other hand, was unaffected by this mutation and the action of this antibody therefore seems to be more dependent on the preservation of the Lys15–Ile16 bond rather than on the conformation of the autolysis loop.
The three-dimensional structure of pro-uPA remains to be reported, but from the three-dimensional structures of trypsinogen and trypsin, the autolysis loop is part of the activation domain which changes conformation after cleavage of trypsinogen . The localization of the epitopes in the autolysis loop is therefore in good agreement with the preference of these antibodies for either pro-uPA or active uPA. Even some specific amino acid residues, i.e. Glu144, Tyr149 and Tyr151, are part of the epitopes for all three antibodies. Given the differential reactivities of the antibodies towards the different forms of u-PA, this observation indicates that these residues are differently oriented in pro-uPA and u-PA, thereby providing evidence for structural changes within this loop upon zymogen activation. In previous studies, mAb-112 was shown to be a non-competitive inhibitor of active uPA cleaving a small substrate  and able to hinder the effect of dipeptides stimulating the activity of pro-uPA . This indicates that the autolysis loop bound by mAb-112 is restrained in a conformation that does not allow the active site to mature. Thus, mAb-112 obviously shifts the equilibrium between the pro and the active conformation in favour of the pro-conformation. In contrast, mAb-12E6B10 and mAb-PUK are totally selective for active uPA and pro-uPA respectively and do not seem to change the equilibrium between active and inactive conformations, on the basis of the fact that neither of them induced or inhibited the ability of pro-uPA or uPA to catalyse hydrolysis of S-2444.
The ability of mAb-PUK, mAb-12E6B10 and mAb-112 to inhibit formation of the pro-uPA–PAI-1 complex or active uPA–PAI-1 complex and inability of the antibodies to bind to the preformed complexes is likely to be caused by competition with PAI-1 for the same binding site on pro-uPA and uPA. The recently determined three-dimensional structure of a Michaelis complex between uPA and PAI-1, representing the initial encounter between inhibitor and enzyme, shows that the autolysis loop is engaged in hydrogen-bonding interactions with PAI-1 . Occupation of the autolysis loop by an antibody would therefore be incompatible with formation of this complex (Figure 9A). The stable covalent complex between uPA and PAI-1 has not yet been reported. However, on the basis of structures of homologous protease–serpin complexes, a model of the complex can be built (Figure 9B). The model shows that the autolysis loop is close to the body of the serpin and unlikely to be accessible to an antibody. Importantly, the two tyrosine residues implicated in all three epitopes are within a distance of a few Å (1 Å=0.1 nm) from PAI-1.
The conformational antibodies characterized in the present study were evaluated for their ability to inhibit cell invasion in two different invasion assays. Although they are simplified systems, these assays partially recapitulate the escape of cancer cells from a primary tumour and their invasion into the surrounding stroma. All three antibodies proved equally able to inhibit cancer cell invasion in the Matrigel assay. However, in a more complex in vitro setting, namely tumour cell escape into fibrin-enriched three-dimensional collagen, mAb-112 was the most efficient in the inhibition of cancer cell invasion. Furthermore, in the in vivo setting in the chick embryo CAM model, only mAb-112 proved to be an effective inhibitor of cancer cell dissemination. The affinity of mAb-112 for pro-uPA was approximately 10-fold higher compared with the affinity of mAb-PUK and mAb-12E6B10 to pro-uPA and uPA respectively. The same difference was observed for the inhibition of plasminogen activation on the cell surface of U937 cells, where a 10-fold lower IC50 value for mAb-112 was observed when pro-uPA initiated the reaction. Therefore, the reason for the higher potency of mAb-112 in the in vivo inhibition of cancer cell dissemination is probably the result of a higher affinity and/or the bi-functional inhibitory mechanism displayed by mAb-112 (Figure 8).
The relative amounts of active two-chain uPA and inactive pro-uPA in the tumour microenvironment could have a great impact on tumour growth and metastasis, but, to the best of our knowledge, differential detection of each of these two forms directly on the cell surface has not been carried out previously. Therefore, the selective binding ability of the antibodies was exploited to detect cell-surface-bound pro-uPA or active uPA. In agreement with the presence of either pro-uPA or uPA in the culture, the PC-hi/diss cells were positively stained with either mAb-112 or mAb-12E6B10 respectively. Therefore, this pair of monoclonal antibodies might represent a new tool for evaluating the activation status of uPA in the tumour microenvironment and to study whether pro-uPA or active uPA or both are relevant targets.
In conclusion, a functional characterization of conformation-specific monoclonal antibodies demonstrates that the autolysis loop of uPA is highly implicated in the conformational changes that form a mature active site when pro-uPA is converted to active uPA. The efficacy of targeting plasminogen activation by interfering with regions on uPA outside the highly conserved catalytic site has been shown in the present study to be a feasible way of hindering cancer cell invasion and dissemination in vivo. Finally, the conformation-specific antibodies proved to be able to discriminate between pro-uPA and uPA on the surface of live cells.
Kenneth Botkjaer was involved in all the experiments carried out, as well as in the preparation of the manuscript. Sarah Fogh performed epitope mapping for mAb-12E6B10 and mAb-101, the cell-based enzyme assay and prepared alanine mutants of uPA. Erin Bekes was involved in the Matrigel assay, three-dimensional invasion assay and CAM assay. Zhuo Chen was involved in the cell-based enzyme assay. Grant Blouse was involved in the preparation of alanine mutants of uPA and theoretical design of biochemical assays. Janni Jensen generated the monoclonal antibodies mAb-101 and mAb-112. Kim Mortensen generated the monoclonal antibodies mAb-101 and mAb-112. Mingdong Huang provided structural information on the location of the autolysis loop in the Michaelis complex between uPA and PAI-1. Elena Deryugina was involved in the Matrigel assay, 3D invasion assay, CAM assay and immunocytochemical analysis, and the theoretical design of these. James Quigley was involved in the theoretical design of the Matrigel assay, 3D invasion assay, CAM assay and immunocytochemical analysis. Paul Declerck generated the monoclonal antibody mAb-12E6B10. Peter Andreasen was the co-ordinator of the work performed and manuscript preparation, as well as theoretical design of biochemical assays.
This work was supported by grants from the Danish Cancer Society [grant numbers DP 07043 and DP 08001]; the Danish National Research Foundation [grant number 26-331-6]; the National Natural Science Foundation of China [grant numbers 30811130467, 30973567 and 30770429]; the Danish Research Agency [grant number 272-06-0518]; and the Novo Nordisk Foundation [grant number R114-A11382].
Abbreviations: CAM, chorioallantoic membrane; CM-CEF, conditioned medium from chicken embryonic fibroblasts; DAPI, 4′,6-diamidino-2-phenylindole; EGF, epidermal growth factor; HEK-293T, human embryonic kidney-293 cells expressing the large T-antigen of simian virus 40; PAI-1, plasminogen activator inhibitor-1; qPCR, quantitative PCR; RCL, reactive centre loop; RU, response units; S-2251, H-D-Val-Leu-Lys-p-nitroanilide; S-2444, H-D-Glu-Gly-Arg-p-nitroanilide; SF-DMEM, serum-free Dulbecco's modified Eagle's medium; SPR, surface plasmon resonance; TIU, trypsin inhibitory units; tPA, tissue plasminogen activator; uPA, urokinase-type plasminogen activator; uPAR, uPA receptor; wt, wild-type
- © The Authors Journal compilation © 2011 Biochemical Society