Specific inhibitors of urokinase plasminogen activator for treatment of cancers; In silico approach

Invasion, metastasis and angiogenesis are fundamental processes in the development of solid cancers. All these depend on the proteolysis where plasminogen activation system (PAS) is the prominent component. Plasmin of PAS, with a broad spectrum of proteins lysed, is secreted as a pro-enzyme and then activated by urokinase (uPA) or tissue plasminogen activator (tPA). Urokinase is in control of pericellular proteolysis while tPA mediates intravascular fibrinolysis. The most effective way to reduce excessive activity of plasmin in cancers is by inactivation of its activators (PAs). Inhibition of PAs reduces tumor size of cancers in vivo. However, for successful targeted anticancer therapy it is essential to find specific uPA inhibitors and to protect normal function of tPA. Unfortunately, most known inhibitors are unspecific, acting on both PAs. PAs are highly homologous enzymes with extreme similarities in their active sites so large numbers of chemicals need to be tested to find novel specific uPA inhibitors. Methods: As the availability of 3D protein structures determined experimentally by X-ray crystallography is growing, computational methods are ever more used in targeted drug discovery. AutoDock Vina molecular docking, exploring binding of small molecules to target protein using a Monte Carlo technique and scoring function, is gaining popularity due to its excellent prediction accuracy. Results: Using AutoDock we have found many inhibitors of PA from our 3D database of 6170 compounds which bind in silico to X-ray structures of the specificity pocket (B187-B197, B212-B229) of both PAs preventing activation of plasminogen. However we were successful in identifying a few molecules which are specific for uPA. For example: both amiloride and chrysin bind to the specificity pocket of uPA but not to tPA. We have found that they bind to other parts of tPA, distant from the specificity pocket, thus preserving the tPA enzymatic activity while being effective toward uPA. Conclusion: In silico search yields specific inhibitors of uPA which, when verified for safety and efficiency by in vivo testing, could be used as novel therapeutics to limit metastasis and angiogenesis in anticancer therapy.

Invasion, metastasis and angiogenesis are fundamental processes in the development of solid cancers. All these depend on the proteolysis where plasminogen activation system (PAS) is the prominent component. Plasmin of PAS, with a broad spectrum of proteins lysed, is secreted as a pro-enzyme and then activated by urokinase (uPA) or tissue plasminogen activator (tPA). Urokinase is in control of pericellular proteolysis while tPA mediates intravascular fibrinolysis. The most effective way to reduce excessive activity of plasmin in cancers is by inactivation of its activators (PAs). Inhibition of PAs reduces tumor size of cancers in vivo. However, for successful targeted anticancer therapy it is essential to find specific uPA inhibitors and to protect normal function of tPA. Unfortunately, most known inhibitors are unspecific, acting on both PAs. PAs are highly homologous enzymes with extreme similarities in their active sites so large numbers of chemicals need to be tested to find novel specific uPA inhibitors. Methods: As the availability of 3D protein structures determined experimentally by X-ray crystallography is growing, computational methods are ever more used in targeted drug discovery. AutoDock Vina molecular docking, exploring binding of small molecules to target protein using a Monte Carlo technique and scoring function, is gaining popularity due to its excellent prediction accuracy. Results: Using AutoDock we have found many inhibitors of PA from our 3D database of 6170 compounds which bind in silico to X-ray structures of the specificity pocket (B187-B197, B212-B229) of both PAs preventing activation of plasminogen. However we were successful in identifying a few molecules which are specific for uPA. For example: both amiloride and chrysin bind to the specificity pocket of uPA but not to tPA. We have found that they bind to other parts of tPA, distant from the specificity pocket, thus preserving the tPA enzymatic activity while being effective toward uPA. Conclusion: In silico search yields specific inhibitors of uPA which, when verified for safety and efficiency by in vivo testing, could be used as novel therapeutics to limit metastasis and angiogenesis in anticancer therapy. urokinase | tissue plasminogen activator | inhibitor | molecular modeling P roteolysis is defined as the degradation of proteins by numerous proteolytic enzymes that plays a vital function in physiology and pathology of life forms. Proteolysis is closely controlled on the level of expression, activation, and inhibition in disease-free organisms. However, under-expression or over-expression of proteolytic activity is frequently observed in various diseases (1)(2)(3)(4). One of the fundamental and common properties of malignant tumors is activation of proteolysis (5). In numerous malignancies, activated proteolytic enzymes are in control of local invasion, metastasis and angiogenesis (6)(7)(8)(9). One of the members of the proteolytic family overexpressed in cancers is the plasminogen activation system (PAS), which include: (i) Plasminogen: Secreted as a pro-enzyme that is cleaved by urokinase (uPA) or tissue plasminogen activator (tPA) and converted into its active form, plasmin. Plasmin digests a range of proteins and can activate other latent proteolytic enzymes (5,6). Overactive plasmin is responsible for proteolysis in tissue remodeling, tumor invasion, development of distant metastasis, angiogenesis, and fibrinolysis (5,9).
(ii) Activators -uPA and tPA: These enzymes are weak proteases with high structural similarities. Both possess the primary function of activating plaminogen by proteolytic cleavage; uPA is in control of the activation of pericellular proteolysis while tPA mainly mediates intravascular fibrinolysis (7,10,11).
(iii) The other members of the PAS such as the binding site of urokinase, called uPA receptor (uPAR), and inhibitors of plasminogen activators including PAI-1, and PAI-2, are not relevant to this paper (8,12).
Since current therapeutic options for cancer patients are limited, novel tactics are needed (9,13,14). The aberrant PAS activity in cancers can be explored to limit progression, metastasis, and angiogenesis in the potential therapeutic approach of PAS-targeted cancer therapies. In the activation cascade of the PAS, catalytic cleavage of plasminogen to plasmin is initiated by its activators uPA and tPA (6,9,14,15). Thus, to control plasmin activity it is logical to normalize activity of plasminogen activators. In many cancers, abnormal plasmin activity is driven by elevated activity of uPA (16)(17)(18). uPA and tPA are highly homologous serine protease with extreme similarities in their active sites and with similar efficiency (8,(19)(20)(21)(22)(23).
Most known inhibitors of plasminogen activators are unspecific, acting not only on tPA and uPA but also to some extent on other serine proteases such as thrombin, trypsin and others (24-27). Thus, for successful targeted anticancer therapy, it is imperative to find specific uPA inhibitors, and at the same time protect normal function of other serine proteases of analogous structures, such as tPA. As the number of protein structures determined experimentally by X-ray crystallography and nuclear magnetic resonance spectroscopy is growing, computational methods are ever more used as a tool in targeted drug discovery (28-32). Among them, AutoDock Vina molecular docking (MD) methodology that predicts binding of small molecules to a target protein using a Monte Carlo sampling technique and scoring function, is gaining popularity due to its considerable improvement in prediction accuracy and docking time (33-38). By screening many molecules against uPA and tPA, predicted specific inhibitors of uPA should likely be identified.

Ligand source
Chemical structures of potential plasminogen activator inhibitors were obtained from our own internal databases created manually or extracted from PDB protein structures. Additionally, once the preliminary screening of our own databases was complete, the online PubChem chemistry database of the National Institutes of Health (46) was searched for compounds similar in structure to potential inhibitors identified from our databases, and the 3D structures of these similar compounds were downloaded in SDF format. Overall, 6,170 ligands were selected for screening.

Ligand preparation
When necessary, the two dimensional (2D) chemical structures of ligands were converted to three dimensional (3D) SDF structures using the "Baloon" (46) plugin for BioVia Draw (47), 3D structures of different formats (MOL and SDF), were converted to .pdbqt format using OpenBabel 2.4.1 functionality within PyRx (44,45) in preparation for docking.

Molecular docking
PyRx's AutoDoc Vina program (38,45) was utilized to dock ligands to the uPA and tPA enzymes. Whenever applicable both uncharged and partially charged ligands were screened at both the active sites and entire enzymes periphery to determine all possible locations of minimum binding energy. The entire enzyme structure of both uPA and tPA were selected to screen molecules. An "exhaustiveness" of 8 was selected within PyRx to utilize all 8 cores of the CPU (38,45). A computer running the Windows 10 Professional, (10.0.18362 Build 18362) operating system with an Intel Core i7-8550u CPU, and 16.0 GB of physical memory was utilized for all functions of this project.

Data Analysis
An SDF file containing the 3D coordinates and predicted binding affinities of the docked ligands was exported from PyRx, then converted into a XYZ Cartesian coordinate file system by Open-Babel 2.4.1 (48) then opened with Microsoft Excel for quantitative analysis. Ligands with a difference of greater than 1.5 kcal/mol predicted affinity between binding to uPA versus tPA were further visualized. All ligands that bound within 10 Angstroms of the specificity site of uPA and greater than 15 Angstroms of the tPA specificity site were also further visualized.
In order to roughly determine the 3D straight line distance of docked molecules to the enzyme active site for both uPA and tPA, the key amino acids of the catalytic triad for each enzyme were identified, and the 3D molecular coordinates for the nitrogen in the first position of the imidazole ring of His57 was determined. Each molecule that was docked to an enzyme also had their 3D molecular coordinates for every atom extracted and then averaged to give 3D point roughly in the center of the molecule. The following formula was used to calculate distance: where: X,Y,Z are coordinates of His57 atom and center of inhibitor.
This provided a rough idea of the distance from the active site a molecule was predicted by AutoVina to dock, allowing for rapid identification of molecules that were docked within 10 Angstroms to His57 on uPA, but greater than 15 Angstroms from His57 of tPA.

Visualization
PyMol (48) was utilized for 3D visualization of the docking to compare the location of the molecule bound to uPA and tPA. Lig-Plot (49) was utilized to generate 2D schematics of selected ligandprotein binding.

Results and Discussion
Amiloride, known as a specific inhibitor of uPA, binds to the urokinase specificity pocket, but does not inhibit tPA, as it binds distant to its specificity site. To validate the Vina Autodock docking protocol we used this molecule as the positive control (50,51,52).
As expected amiloride was docked in the predicted position within RMSD<2.0 Angstroms to uPA crystallographic structure (data not shown) (40) with strong calculated affinity (-8.8 kcal/mol). Calculations done for tPA and amiloride showed very weak affinity (-4.3 kcal/mol), and what is of great importance, the binding site of amiloride to tPA was distant from specificity pocket (26.9 Angstrom) further corroborating accuracy of established protocol.
It must be emphasized that in the course of the recurrent in silico simulations and generating of 3D structures, there are small variations in the affinity scores as well as 3D structure locations within protein structure. However, occasionally one or two simulations might be locked in a local minimum energy producing unusually different results from the rest of simulations. Therefore, each simulation presented in this paper was run ten times and only results having consistently small and negligible differences were accepted and presented (53)(54)(55). An ideal specific inhibitor could be one binding with high affinity to the specificity pocket of uPA, but in great distance from the specificity pocket of tPA, in a fashion that does not interfere with the binding of tPA to its substrates (plasminogen, PAI-1 or PAI-2).
Inhibiting serine proteases (including uPA and tPA) by blocking the specificity pocket is considered very effective since amino acids are major determinants the substrate specificity. Thus chemicals binding to uPA at the specificity pocket would less likely inhibit serine proteases family members such trypsin, chymotrypsin, elastase or others. However achieving inactivation of uPA but not tPA in that way remains elusive. This is due to the highly homologous nature of uPA and tPA especially in the specificity pocket where they are almost identical (40).
We decided to search the entirety of the uPA and tPA enzymes in hope that some ligands will have a higher predicted binding affinity for distant sites to the tPA specificity pocket, offering uPA inhibi-tion only. As expected, due to the homologous nature of uPA and tPA, the majority of screened compounds bound non-preferentially to both uPA and tPA within the specificity site.   However, by comparing differences in predicted binding affinity as well as the predicted binding location of the screened compounds on both uPA and tPA, we can identify potential compounds that would act as specific inhibitors of uPA, while retaining enzymatic activity of tPA.
Of the 6170 molecules screened as potential specific inhibitors, fourteen molecules (including amiloride) were identified as binding close to the center of the urokinase specificity pocket, but far from the center of the tPA specificity site (Table 1). In addition to amiloride, are least three additional compounds (CHY, NHA, BAC) display greater affinity to uPA and are bound to tPA at a distance greater that 20 Angstroms from the tPA specificity pocket. Chrysin, a flavone found in honey, propolis, and passion flowers, has been investigated for anticancer activity and many possible pathways have been proposed (56). It is interesting that only baicalein decreased the activity of the urokinase but not expression of uPA as it was reported in case of CHY and NHA (59). Specific inhibition of uPA, unknown until now can synergically increase anticancer activity of these compounds. Baicalein, a flavone, originally was isolated from the roots of Scutellaria baicalensis and Scutellaria lateriflora. This flavonoid has been shown to inhibit certain types of lipoxygenases and act on many other enzymes having curable effects on human diseases.
We were not able to find any references on inhibition of tPA in the literature for all of these three compounds promising that this approach could achieve significant selectivity for uPA inhibition.

Conclusion
Therapy of malignancies by preventing invasion, metastasis and pathological angiogenesis should include downregulation of the variety of proteolytic enzymes. Examples include inhibitors of urokinase or metalloproteinase which could provide alternatives to existing anticancer therapies. However, at the same time we must preserve activity of physiologically important enzymes such as the tissue plasminogen activator. This enzyme is critical in normal coagulation but is highly homologous to urokinase, and therefore most of known inhibitors of uPA are also inactivators of tPA. This work provides evidence that it is possible to find highly specific inhibitors of urokinase while preserving activity of tPA for novel anticancer therapy.