Ascorbic Acid Is a Potential Inhibitor of Collagenases—In Silico and In Vitro Biological Studies

Ascorbic Acid Is a Potential Inhibitor of Collagenases—In Silico and In Vitro Biological Studies

C H A P T E R 22 Ascorbic Acid Is a Potential Inhibitor of Collagenases—In Silico and In Vitro Biological Studies Vijaya Lakshmi Bodiga*, Sreedhar Bo...

3MB Sizes 0 Downloads 7 Views


22 Ascorbic Acid Is a Potential Inhibitor of Collagenases—In Silico and In Vitro Biological Studies Vijaya Lakshmi Bodiga*, Sreedhar Bodiga† *

Department of Molecular Biology, Institute of Genetics & Hospital for Genetic Diseases, Begumpet, Osmania University, Hyderabad, India †Department of Biochemistry, Kakatiya University, Warangal, India

1 INTRODUCTION 1.1 Drug Repositioning The advent of informatics has overhauled the drug discovery process and enhanced our understanding of the pathophysiology and the underlying mechanisms. This has required identification and analysis of relevant literature and associated data. This daunting job craves the adoption of bioinformatic and computational tools to retrieve relevant data. It also requires examination of enormous databases to generate meaningful information (data mining), making the best use of the knowledge (knowledge management) in a logical manner and scrutinizing the data to devise a verifiable hypothesis with clinical relevance. These approaches, along with prediction algorithms, confirm the alignment of a particular disease to developmental or marketed drugs and natural compounds with definite structural features. Thus, drug repositioning opportunities require enormous informatics effort and appraisal of the biological and commercial viability of the repurposed drug. The informatics effort involves a computerized approach, with knowledge management assisted by the subject specialists. Initially the disease(s) is assessed and ranked in order of importance. Then a study to fulfill the experimental objectives and test the hypothesis is designed and executed. Knowledge integration requires comparison and compilation of employable enzymatic targets in a specific disease and the list of probable drug candidates available to hit the target (Harland & Gaulton, 2009). This drug-centric informatics approach requires knowledge of

In Silico Drug Design.

649 # 2019 Elsevier Inc. All rights reserved.



the drug’s mechanisms of action (MOAs) that can be obtained from various databases. Competitive intelligence databases (Prous Integrity, TrialScale, and BioPharm Insight) provide the most up-to-date indications for each drug based on the mechanism of action. This informatics strategy is target-centric and can be hooked to high-throughput data mining approaches for incorporation of known disease relevant data (Loging, Harland, & Williams-Jones, 2007). Therefore, computational challenge is about identification, retrieval, and analysis of quality and reliable databases. For additional resources that can aid in providing relevant information, please refer to Loging et al. (2007) For any drug molecule to enter into the market, clinical trial data indicating the efficacy of the drug candidates against a particular target are essential. Genetic studies with nucleotide polymorphism data and their functional implications or their association with upregulation/downregulation of a particular gene of interest would add substantial weight to the identified target. Generation of specific gene knockouts and silencing of mRNA can further confirm the role of the target in a particular disease. Identification of new targets can also be hypothetically achieved by an extensive literature search and data integration using MEDLINE (Notter, 1972). The obvious mention of a disease and its target together in a title are scored higher. A novel visualization tool termed the “target opportunity universe” is capable of analyzing the best fit of a particular drug against a target in a specific disease (Campbell et al., 2010).

1.2 Molecular Modeling in Drug Discovery A therapeutic target molecule in a disease can be any biological macromolecule with known relevance to the disease. Gene coding for a defective enzyme or protein, membrane-bound receptor molecule, or intracellular enzyme are the widely targeted biomolecules to achieve the therapeutic effect. The interaction between the drug and biomolecules most often involves noncovalent interactions, but in some cases a covalent interaction cannot be excluded. Regardless of the type of interactions, the net effect of the drug on the target can classify it as an agonist or antagonist. Agonists mimic the endogenous ligands in structure and function and produce a similar effect or better at lower concentrations. Antagonists, although structurally similar to endogenous ligands, are most likely inhibitory in nature due to altered structural elements. A higher degree of complementarity in terms of shape and structure between the drug and target results in tight binding. The geometrical complementarity can be assessed by visualizing the molecular surface of the target where the drug is supposed to bind. Polar groups of the drug tend to bind to hydrophilic groups on the target, while nonpolar elements attach to the hydrophobic pockets. The binding may sometimes involve structural changes in the target and mere binding may not elicit the desired effect. The drug has to reach its target inside the cell before the therapeutic effect is seen. The drug needs to have lipophilicity to cross the biological membranes. The knowledge of half-life of the drug and its metabolic fate will determine the efficacy. New drug discovery and proving its medicinal value require lot of time and effort. The drug is expected to exhibit the therapeutic effect with no undesired properties, and should score better than the current drugs. Two terms commonly referred to during the drug discovery are identification of hit molecules (hits) and lead series (leads). Hit refers to a molecule with consistent activity in a screening assay, while lead refers to a group of structurally




similar molecules with varied activities due to minor structural variations. Fine tuning of the structure of a lead series can result in a potent drug with the desired effect and selectivity. Finding a novel lead series can be a difficult problem. High-throughput screening (HTS) enables large numbers of compounds to be screened using highly automated, robotic techniques. Although HTS desires to test every molecule in the lead series in a biological screen, this is not always practical and feasible, due to practical considerations of cost involved. Unit costs associated with testing and the sheer number of samples add to the overall expenses. Some biological screens are not amenable to a high-throughput mode and the conventional assay consumes time. For these reasons, it becomes mandatory to limit the lead series to a reasonable number based on computational approach. Some methods rely on “2D” properties, as distinct from “3D” methods, which take into account the three-dimensional structure of a molecule. A basic approach of identifying the molecules with a defined substructure would reduce the number of molecules to be tested. For example, we might wish to identify all compounds containing a sulfonamide group. More complex queries are also possible in most systems; these would, for example, permit a query atom to match groups of atoms or features such as ring bonds or to specify stereochemistry. Sophisticated algorithms help identify the molecules containing the defined substructure but are time-consuming. To eliminate the molecules not possessing a given substructure, a bitstring binary screen is commonly employed. There are two types of binary screens in common use. Presence or absence of a particular substructure is assigned bit values of 1 or 0, respectively. The structural keys with 0-bit values are eliminated, saving time. Bitstring operated structural keys are commonly used in MACCS and Isis systems from Molecular Design.

1.3 Molecular Docking Molecular docking is a structural prediction tool that comprehensively analyzes the best fit between the drug and the target molecules. Docking is an algorithm-based tool that can produce all varied possible structures for the given drug and target molecule and assign scores to each structure. It can help assess the binding modes of drugs with biomolecules (Blaney & Dixon, 1993). However, the docking is constrained by the structural complexity and degrees of freedom of rotation and conformation of molecules. Interactive computer graphics afford manual handling of the docking problem, if the expected binding mode is discerned from the binding of the related ligand. It should be cautioned that very closely related structures may adopt different binding modes. Automated docking algorithms without manual intervention are more probabilistic and less biased. Multiple algorithms that negate the number of degrees of freedom are in place to tackle the docking problem. The earliest simple algorithm, DOCK, considers the two molecules as rigid bodies without a conformational degree of freedom (Kuntz, Blaney, Oatley, Langridge, & Ferrin, 1982). DOCK screens for molecules with a high degree of shape complementarity to the target site. A negative image of the target site from the molecular surface is produced. The image is constructed with molecular spheres of different radii, touching the surface at two points. The atomic elements in the ligand are then aligned to spheres to sort matching sets (cliques) with some degree of tolerance. The ligand is then positioned by accomplishing a least-squares fit of the atoms to the spherical centers. The alignment is examined to confirm unfavorable




steric interactions. Once the confirmation of the intermolecular complex is established, then binding energies are calculated and the binding modes are assigned scores. New assortments can be produced through matching different sets of atoms and sphere centers. The conformations with high scores are chosen for further analysis. To achieve a conformationally flexible docking, the conformational degrees of freedom have to be considered. It is important to contemplate the conformational degree of freedom of both the ligand and the macromolecule in question. Commonly used docking algorithms typically incorporate conformational freedom into the algorithm. Monte Carlo simulation is used for the molecular docking interactions, with simulated annealing (Goodsell & Olson, 1990). Multiple iterations with different rotational and conformational degrees of freedom are generated in the Monte Carlo procedure. Free energy of binding either accepts or rejects the docking and complex formation. An interesting variant on the basic Monte Carlo approach is the tabu search (Baxter, Murray, Clark, Westhead, & Eldridge, 1998). This maintains a record of those regions of the search space that have already been visited, thus ensuring that the method is encouraged to explore more of the binding site. Distance geometry can also be employed for molecular docking. However, this procedure requires producing multiple conformations of the drug within the target site. This can be achieved by assigning penalty scores for drug conformations that are pinned down to the binding site. A strategy that is frequently employed utilizes cumulative structural arrangement of the ligand (Leach & Kuntz, 1992; Rarey, Kramer, Lengauer, & Klebe, 1996; Welch, Ruppert, & Jain, 1996). This algorithm initially identifies the rational “base fragments” that are often part of a rigid component of the molecule. The base fragments are logically pinned down to the binding site and may then be bundled to get rid of identical arrangements. The base fragments are chosen as the starting points for further conformational analysis of the other part of the ligand in each intermolecular complex. Although this strategy appears time consuming, it might yield information about worthy constraints that might otherwise reduce the search and accelerate the process. An optimal docking should allow both the ligand and macromolecule to be flexible. Molecular dynamics (MD) simulation of the ligand-macromolecule complex would allow the flexibility of both molecules. These calculations, being onerous, are only used for fine tuning the docking results obtained with other methods. MD simulations limit the range of binding modes, except for small ligands, and surmounting the energy barriers while transitioning from one binding mode to the other complicates the matter. These hurdles can be overcome to some extent by limiting to side chain flexibility rather than molecular flexibility (Leach, 1994). Most docking algorithms are capable of generating an overwhelming number of fixes. Many of these fixes can be denied automatically due to energy constraints. The remaining fixes can be graded based on the chosen scoring function. When we are only interested in how a single ligand binds to the protein, then the scoring function only reads the arrangement that matches with the original structure of the complex. However, when docking a database of molecules, then the scoring should determine the appropriate binding mode for a given ligand and also grade the ligands for their affinity to the binding site. A rapid scoring function such as the binding free energy for the ligand is required, considering the overwhelming number of arrangements that could be generated during the docking. GOLD uses a genetic algorithm ( Jones, Willett, Glen, Leach, & Taylor, 1997), whereas FlexX uses an incremental construction method (Kramer, Rarey, & Lengauer, 1999). 3. EXAMPLES AND CASE STUDIES



1.4 Evaluation of Docking Results Regardless of the docking tools employed, all docking data have to be evaluated for chemical complementarity between the two molecules. Evaluation includes confirming if all the purported hydrogen bonds are involving right donors and acceptors in the ligand. One also needs to confirm if the salt linkages only involve the oppositely charged residues at an appropriate distance and not the hydrophobic entities. The docking data can be further judged by the consistency and reproducibility of the binding modes and free energy calculations derived from it. The docking would be considered a success if comparison of the all-atom root ˚ between the docked position and the X-ray mean square deviation (RMSD) falls within 2 A crystallography image. Multiple runs of the docking with different search parameters are advised when employing stochastic methods for docking. Predicted binding modes can be analyzed by compiling a matrix of pair-wise RMSD values and bundling the docked ˚ . If the predicted binding modes were similar, conformations using an RMSD threshold of 2 A the dockings would converge into one family, implying the reliability of the initial conditions and docking procedure. Absence of clusters would require repetition of the procedure with multiple iterations and increased sample size or population size. Assuming a perfect scoring function, the most stable docked conformation would be the one and should match with the crystallographically observed binding mode. Sometimes, a different binding mode from the one yielding the lowest energy is observed.

1.5 Matrix Metalloproteinases Matrix metalloproteinases (MMPs) are proteases containing zinc in the catalytic site and that target the extracellular matrix and basement membrane. Various essential physiological processes, including growth, wound healing, and tissue reorganization, involve the activity of MMPs (Massova, Kotra, Fridman, & Mobashery, 1998; Matrisian, 1990; Nagase & Woessner Jr., 1999; Shapiro, 1998). MMP activity is tightly controlled due to their synthesis as precursor zymogens and transcriptional regulation. Endogenous regulation by specific molecules, called tissue inhibitors of metalloproteinases (TIMPs), also controls the MMP activity (Nagase & Woessner Jr., 1999; Sternlicht & Werb, 2001). Dysregulation of MMP expression and activity are consistently observed in various pathophysiological conditions such as arthritis, cancer, atherosclerosis, aneurysms, nephritis, tissue ulcers, and fibrosis (Woessner, 1991). The majority of the MMPs need to be proteolytically activated in the extracellular space and highly regulated tissue-specific expression of MMPs is noted. MMPs are classified into five groups based on the substrates for proteolysis (Table 1): Collagenases, gelatinases, stomelysins, membraneassociated, and unclassified. Collagenases either occur as fibroblast type (MMP-1, collagenase 1) (Goldberg et al., 1986), the neutrophil type (MMP-8, collagenase 2) (Hasty, Jeffrey, Hibbs, & Welgus, 1987), or collagenase-3 (MMP-13) (Freije et al., 1994). MMP-1, an interstitial collagenase that targets type III collagen, is found to be expressed in human fibroblasts, keratinocytes, endothelial cells, monocytes, and macrophages. Neutrophil collagenase expression is not just confined to the polymorphonuclear neutrophils, but also is found in fibroblast-like cells in the rheumatoid synovial membrane as well as in cultured rheumatoid synovial fibroblasts and human endothelial cells (Hanemaaijer et al., 1997). MMP-8 expression in arthritic lesions indicates that chondrocytes, and synovial fibroblasts are capable of expressing the neutrophils collagenase (Shlopov et al., 2005; Tetlow, Adlam, & Woolley, 2001). Native type I and II collagens 3. EXAMPLES AND CASE STUDIES



TABLE 1 Classification, Common Names, and Substrates of the Matrix Metalloproteinases (MMP) Family Enzyme

Common Name



Collagenase-1; interstitial collagenase

Collagen type 1–3, 7, 8, 10; aggrecan; gelatin;


Gelatinase A; 72 kDa gelatinase; MMP-5

Collagen type 1–5, 7, 10, 11, 14; aggrecan; elastin; fibronectin; gelatin; laminin; MMP-9, -13


Stromelysin-A; Procollagenase; transin-1

Collagen type 2–4, 9–11; aggrecan; elastin; fibronectin; gelatin; laminin; MMP-1, -7, -8, -9, and -13


Matrilysin-1; PUMP-1

Collagen type 1–3, 5, 7, 8, and 10; aggrecan; elastin; fibronectin; gelatin; laminin; MMP-1, -2, and -9


Collagenase-2; neutrophils collagenase

Collagen type 4 and 10; aggrecan; elastin; fibronectin; gelatin; laminin


Gelatinase B; 92 kDa gelatinase

Collagen type 4, 5, 7, 10, 14; aggrecan; elastin; fibronectin; gelatin


Streomelysin-2; transin-2

Collagen type 3–4; aggrecan; elastin; fibronectin; gelatin; laminin; MMP-1 and 8



Aggrecan; fibronectin; laminin; α-1 antitrypsin


Macrophage metalloelastase

Collagen type 4; elastin; fibronectin; gelatin; laminin; vitronectin



Collagen type 4; elastin; fibronectin; gelatin; laminin; vitronectin


MT-1 MMP (Membrane Type-1 MMP)

Collagen type 1–3; aggrecan; elastin; fibronectin; gelatin; laminin; MMP-2, -13



Fibronectin; gelatin; laminin; MMP-2


MT3-MMP Ovary metalloproteinase




Fibrin; fibrinogen; TNF precursor


Xenopus MMP




Collagen type 4; aggrecan; COMP; gelatin; laminin; large tenas; nidogen firbillin



Ameloginin; aggrecan; COMP


Xenopus MMP



Gallus domesticus MMP

Casein; gelatin








MT6-MMP; leukolysin



Matrilysin-2; endometase

Collagen type 4; α-PI; fibronectin; fibrinogen; gelatin; Pro-MMP-9







are more efficiently hydrolyzed by MMP-8 when compared to MMP-1. On the other hand, a more stringent expression of MMP-13 confined to the connective tissue is observed (Vincenti et al., 1998), in addition to malignant breast tumors showing upregulation of MMP-13 (Freije et al., 1994). MMPs share basic structural features despite substrate variability. They are synthesized as Pre-Pro-MMPs. They contain an N-terminal signal predomain region that directs the protein for secretion. The propeptide domain of about 80 amino acids contains a conserved cysteine in the sequence -PRCGXPD-. Coordination of the cysteinyl sulfur atom to the active site zinc(II) ion ensures that the enzyme activity is suppressed in the proforma. Cleavage of the propeptide domain by other MMPs or proteases such as plasmid by a “cysteine switch” mechanism activates the MMPs (Nagase & Woessner Jr., 1999; Shapiro, 1998; Whittaker, Floyd, Brown, & Gearing, 1999; Woessner & Nagase, 2000). The general domain structure of an MMP is shown in Fig. 1. The catalytic domain of 170 amino acids shown in Fig. 2, is organized into a five-stranded β-sheet, three α-helices, and bridging loop structures (Nagase & Woessner Jr., 1999). One of the zinc(II) ions in this domain is buried in the protein and coordinated to one aspartate and three histidine residues in a tetrahedral geometry, playing a structural role. The other zinc(II) is required for catalytic action, i.e., peptide hydrolysis, and coordinated by three histidine nitrogen atoms in a conserved sequence -VAAHEXXGHXXXGXXH- (Babine & Bender, 1997). The rest of the coordination sphere is filled with water molecules, which are essential for the

FIG. 1 Domain structure of MMPs.

FIG. 2 Secondary structure of the catalytic domain of MMPs.




catalytic activity (Bertini et al., 2003). Upon binding of acarbonyl oxygen of a peptide to the zinc(II) ion, the amide bond is attacked by a zinc-bound water molecule that is hydrogen bonded to an adjacent glutamate residue. Proton transfer from the water molecule to the glutamate residue and then to the amide nitrogen completes the cleavage of the peptide (Babine & Bender, 1997; Whittaker et al., 1999). All MMPs (excluding MMP-7 and MMP23) possess a hemopexin-like domain consisting of 210 amino acids, important for substrate binding.

1.6 Inhibitors of MMPs The inhibition of MMPs has assumed great importance, due to the variety of diseases the MMPs are involved in. An important field of chemotherapeutics directed at suppressing MMP activity has grown. Most inhibitors (MPIs) use the same basic design (Fig. 3): a peptidomimetic backbone, coupled with metal-chelating moiety (zinc-binding group, ZBG) (Skiles, Gonnella, & Jeng, 2004; Whittaker et al., 1999). Following a substrate-based approach, early MPIs were designed to mimic natural substrates of MMPs by using a short peptide derivative attached to a ZBG (Whittaker et al., 1999). Availability of NMR and X-ray crystallographic structures of the MMPs further paved the way for this structure-based MPI design (Skiles et al., 2004), which is dictated by the shape of the active site subsites. The active site subsites and a variety of residues lining the subsites determine the substrate selectivity of various MMPs (Whittaker et al., 1999). The subsites depicted in Fig. 4 located to the left of the zinc(II) ion are termed unprimed (S1, S2, S3), while the ones located to the right of the zinc(II) ion are denoted primed subsites (S10 , S20 , S30 ). The functional groups of MPIs interacting specifically with the subsites are termed P or P0 groups (Whittaker et al., 1999), i.e., a P1 group is expected to interact with the S1 subsite. Most inhibitors of MMPs were directed to the primed subsites (right-handed inhibitors), particularly the S10 pocket. The S10 subsite is a deep, hydrophobic pocket common to all MMPs, with the exception of MMP-1 and MMP-7 (Bode et al., 1999; Lovejoy et al., 1999). The depth of the pocket varies between MMPs, and the residues in this pocket lack any significant sequence homology. The differences in the S10 pocket determine the substrate specificity and therefore it is called the specificity pocket (Bode et al., 1999; Lovejoy et al., 1999) and several MPIs have been designed to exploit this feature (Hajduk et al., 1997; Olejniczak et al., 1997), as in the case of WAY-170523, a hydroxamate-based MPI (Chen et al., 2000). This inhibitor (Fig. 5), designed by using NMR structural data of the active site, contains a large P10 group that fits tightly into the S10 pocket of MMP-13, but is too bulky for accommodation by the S10 pocket of MMP-1. Consequently, WAY-170523 displays nearly 6000-fold more

FIG. 3

Generalized structure of MPIs. The ZBG binds to the catalytic zinc(II) ion. The P substituents occupy various subsite pockets in the MMP active site.



FIG. 4


Molecular surface diagram of the active site of MMP-3 (top). The catalytic zinc(II) ion is shown in black; the enzyme is in gray. Location of the subsites are labeled. Schematic figure of the MMP active site before and after inhibition (bottom).

FIG. 5 Chemical structure of a hydroxamate-based MPI, WAY-170523.




potency for MMP-13 over MMP-1. However, in some inhibitor-protein complexes with MMP1, the S10 pocket expands, allowing for an MPI with a large P10 substituent to occupy this cavity (Lovejoy et al., 1999). The S20 pocket that is exposed to the solvent is directly positioned above the S10 pocket opening. The S20 cavity is hydrophobic in nature in most of the MMPs, except MMP-1 and MMP-7 where it contains Ser and Thr residues, respectively. The S20 subsite, therefore, offers less selectivity compared to the S10 pocket. Molecules with bulky hydrophobic P20 substituents are selective for MMP-2, MMP-3, MMP-8, MMP-9, and MMP-13 over MMP-1 and MMP-7, due to differences in S20 subsites. Other MPIs have been designed with large P20 substituents to better the pharmacokinetic properties of MPIs by preventing the amide bond hydrolysis of the inhibitor (Hirayama et al., 1997; Skiles et al., 2004; Skiles, Gonnella, & Jeng, 2001; Whittaker et al., 1999). The S30 subsite lies on the outer rim of the S10 pocket entry site and therefore is exposed to the solvent. This subsite has not been seriously considered for designing MPIs. But a recent study observed that some P30 substituents can afford specificity. Substituted benzylic groups at the 30 position lead to 1000-fold less potent inhibition of MMP-2, without altering the MMP-3 activity (Fray, Burslem, & Dickinson, 2001). The unprimed subsites are relatively less characterized and not studied for their role in selective inhibition. The unprimed region is a closely arranged group of subsites that are more exposed to the solvent than the primed sites. The S2 subsite is positioned next to the catalytic center, but the other unprimed subsites are remotely away from the catalytic center. The role of unprimed subsites has been gaining importance recently. The positioning of analogous residues in the S2 subsite of MMP-2 and MMP-9 offered substrate selectivity between the gelatinases. Substrate specificity was found to be due to the ability of Glu to form a hydrogen bond with specific substrates, while Asp cannot, due to shorter side chain length (Chen, Li, Godzik, Howard, & Smith, 2003). Few studies could still demonstrate that left-handed inhibitors can offer selectivity (Finzel et al., 1998). The S1 and S3 subsite-based inhibitors were proven to be successful in selectively targeting MMPs over collagenase (MMP-1) (Finzel et al., 1998). Thiadiazole inhibitors that bind to the unprimed subsites showed selective inhibition of MMP-3 (Ki ¼ 0.018 μM) over MMP-1 (Finzel et al., 1998). MPIs utilizing both primed and unprimed subsites have also been designed. Potent and selective inhibition of MMP-13 could be achieved using this strategy. Targeting the deep S10 pocket along with the S2 subsite of MMP-13 yielded 100-fold more selective inhibitors for MMP-13 over MMP-1 (Reiter et al., 2003). Thus, both unprimed and primed subsites have offered avenues for designing selective inhibitors.

1.7 Zinc-Binding Groups of MMP Inhibitors Zinc-binding groups (ZBGs) have also played an important role in developing a successful MPI. According to the nature of the group in the inhibitor interacting with the metal ion, MPIs are broadly categorized into six groups: hydroxamates, carboxylic acids, thiols, phosphorusbased, other ligands, and natural products. The hydroxamic acid group is the most widely used and most effective ZBG in inhibitor design. Hydroxamates interact with the catalytic zinc(II) in a bidentate manner, disallowing the substrate entry into the active site and incapacitating the catalytic zinc(II) of peptide hydrolysis. The binding of hydroxamate-based 3. EXAMPLES AND CASE STUDIES



MPIs has been confirmed by X-ray crystallography, which unambiguously displays bidentate ˚. coordination with average ZndO bond lengths of 2.0 A Carboxylate ZBGs have gained momentum next to the hydroxamate-based MPIs (Whittaker et al., 1999), apparently due to their precursor character. The carboxylate ZBG has been proposed to be a bidentate ligand (Esser et al., 1997; Natchus et al., 2001); however, detailed examination suggests that the carboxylate ZBGs prefer monodentate interaction with the catalytic zinc(II) ion. Among eight X-ray structures of MMP-inhibitor complexes with carboxylate-based compounds, the average ZndO bond lengths were found to be 1.9 and ˚ for the two oxygen atoms (Becker et al., 1995; Browner, Smith, & Castelhano, 1995; Esser 2.7 A et al., 1997; Matter et al., 1999; Natchus et al., 2001; Pavlovsky et al., 1999). The sum of the ˚ , which is significantly shorter than the covalent radii for oxygen and zinc is only 2.1 A ˚ reported 2.7 A bond. In addition, a survey of the Cambridge Structural Database (http:// identified 16 small molecule structures of zinc coordinated to three ˚ are reported, nitrogen atoms and a carboxylate; no ZndO bond distances larger than 2.5 A suggesting that longer ZndO bonds are not true coordinate bonds, or at best are very weak interactions. In the structures that were described as truly five-coordinate, the carboxylate oxygen atoms are bound nearly equidistantly, contrary to the asymmetric binding found in the MPI complexes (Darensbourg, Wildeson, & Yarbrough, 2002; Kremer-Aach et al., 1997). In addition, examination of the geometry at the metal centers in the MMP carboxylate MPI structures is most appropriately described as distorted tetrahedral. This supports the contention that the coordination sphere of the zinc(II) ion is essentially unaffected by the more distant oxygen atom, as true bidentate coordination should tend to induce a more standard five-coordinate geometry, such as trigonal bipyramidal or square pyramidal, similar to that found with the hydroxamic acid MPIs discussed previously (O’Brien et al., 2000). Finally, hydrogen bonding is frequently found between the distant carbonyl oxygen atom of the carboxylate ZBG and the protein sidechains (Becker et al., 1995; Browner et al., 1995; Natchus et al., 2001; Pavlovsky et al., 1999), which further reduces the ability of this atom to be strongly coordinated to the zinc(II) ion. As monodentate ligands, carboxylates should be more weakly bound to the zinc center, due to the loss of a ZndO bond (relative to hydroxamate-based compounds) and a loss of the chelate effect (see the following). Consistent with this description of carboxylic acid ZBGs being only monodentate ligands, most carboxylate-based MPIs are inferior to the bidentate hydroxamates (O’Brien et al., 2000). MPIs with thiol-based ZBGs were found to be effective inhibitors at subnanomolar concentrations (Baxter et al., 1997; Campbell et al., 1998; Freskos et al., 1999; Hughes, Harper, Karran, Markwell, & Miles-Williams, 1995). The thiophilic nature of zinc(II) in proteins has prompted the development of thiol-based MPIs. The “cysteine switch” self-inhibitory mechanism of MMPs (see previous discussion) has sparked a greater interest in thiol-ZBGs (Morgunova et al., 1999). Thiol-containing MPIs use a sulfhydryl group as the lone donating atom or in combination with other donor atoms, as in the case of mercaptoketones, mercaptoalcohols, and mercaptoamides. Very few X-ray structures are available for MPI-MMP complexes of thiol-based inhibitors. One of the available structures reveals that the MPI uses the thiol ˚ , resulting ZBG to bind the zinc(II) ion in a monodentate fashion with a ZndS distance of 2.2 A in a tetrahedral geometry around the metal center (Grams et al., 1995). Studies of inhibition of thermolysin (zinc-dependent protease) by a phosphorus-based compound have directed the efforts towards synthesizing and testing phosphorous-based MPIs (Komiyama, Aoyagi, Takeuchi, & Umezawa, 1975). Phosphonic or phosphinic acid 3. EXAMPLES AND CASE STUDIES



groups are responsible for chelation of active site Zn (II). Phosphorous-based MPIs have not shown inhibitory activity comparable to that of hydroxamates. Examination of the X-ray crystal structure of a phosphonic acid-based MPI bound to MMP-8 reveals the binding mode of the inhibitor (Gavuzzo et al., 2000). The ZBG is described as binding in a bidentate manner through two of the three oxygen atoms (Gavuzzo et al., 2000). The ZndO distances are 1.9 and ˚ . Again, based on the long ZndO bond length, monodentate coordination appears to be 2.7 A a more apt description of the binding. Additionally, the coordination geometry of the zinc center in this structure is clearly a distorted tetrahedron and shows little influence from the more distant oxygen atom. The structure of a phosphinic acid MPI has also been obtained (Gall et al., 2001). The structure of this MPI bound to MMP-11 displays a tetrahedral geometry at the metal center. One phosphinic oxygen atom is bound to the zinc(II) ion at a distance of ˚ . The second phosphinic oxygen is located 2.9 A ˚ from the metal center and is in hydrogen 2.4 A bonding distance to Glu220. A number of other MPIs have been synthesized that employ ZBGs that do not fall into any of the previously described categories. Among the many different ZBGs, two unconventional MPIs that utilize thiadiazole and 2,4,6-pyrimidine trione (barbituric acid) ZBGs have been structurally characterized bound to their MMP targets. The thiadiazole-derived MPIs PNU-142372 and PNU-141803 bind to the active site zinc(II) ion in a monodentate fashion ˚ . The crystal through the sulfur atom of these ZBGs with a ZndS bond distance of 2.3–2.4 A structures of two barbituric acid-based MPIs bound to MMP-3 and MMP-8 reveal their unusual mode of binding where the active site zinc atom is coordinated through the N3 nitrogen ˚ (Dunten et al., 2001). Of the two reported barbiatom with a ZndN bond distance of 2.1 A turate structures, one suggests that the ZBG is bidentate (Brandstetter et al., 2001) bound through the N3 nitrogen atom and one of the oxygen atoms of the ZBG, while the other reference describes the inhibitor as bound in a monodentate fashion. Inspection of the zinc centers in each of these structures indicates that the coordination geometry is best described ˚ away as distorted tetrahedral. The oxygen atoms of the ZBG are located approximately 3.0 A from the zinc(II) ion, suggesting they do not coordinate to the metal center. Based on the long ZndO distance and the tetrahedral coordination geometry, these compounds appear to be properly described as monodentate ligands, similar to carboxylate-based ZBGs. Beyond the many types of synthetic inhibitors that have been explored, various natural products have also been shown to inhibit MMPs. Indeed, the natural product group of inhibitors represents the only clinically approved MPI. The tetracycline antibiotic Periostat (doxycycline hyclate) is clinically used as an MPI against periodontal disease. Although tetracyclines were only modestly effective against MMPs, they showed tremendous anticancer properties. The antitumor activity of tetracyclines cannot be attributed solely to the MMP inhibition property. Exact knowledge of the interactions between tetracycline and the catalytic zinc is lacking, as there are no crystal structures of a tetracycline bound to an MMP. Binding of tetracyclines to divalent metal ions indicates possibly bidentate coordination through adjacent keto/hydroxyl oxygen atoms. Other natural products found to inhibit MMPs include derivatives of the compound futoenone. Futoenone is found in the Chinese herbal plant Piper futokadsura, which is used to treat inflammatory disease. Futoenone and its derivatives have been studied as MPIs, and they demonstrate modest inhibitory activity. The mode of binding of these compounds to MMPs is unknown; however, it is proposed that the 2-methoxyphenol moiety in these molecules is responsible for zinc chelation. 3. EXAMPLES AND CASE STUDIES



Although thousands of MPIs have been synthesized, only Periostat has been approved for therapeutic use and it does not contain a hydroxamate ZBG. One main criticism of the numerous failed MPIs is a lack of specificity and potency resulting in low in vivo activity and unwanted side effects, such as musculoskeletal syndrome (Coussens, Fingleton, & Matrisian, 2002; Hutchinson, Tierney, Parsons, & Davis, 1998). It has been proposed that inhibition of collagenase (MMP-1) is partially responsible for the development of musculoskeletal pain. However, this pain is observed in patients after treatment with prinomastat, an MPI that does not inhibit MMP-1 (Coussens et al., 2002). This finding suggests that the current method of improving the selectivity of inhibitors for specific MMPs may not completely eliminate the problem of unwanted side effects. In addition, a more alarming study has discovered that the MPI batimastat promotes the metastasis of human breast carcinoma cells in nude mice (Kruger et al., 2001). It has been suggested that less specific inhibitors may be acceptable when treating cancer if they exhibit potency, but in the remedy of more benign diseases such as arthritis, specific inhibitors with minimal side effects are preferred. Ideally, the development of potent MPIs that are more selective for MMPs over other metalloproteins could reduce the required dosages and minimize the potential side effects. Attempts to improve the potency and selectivity of MPIs are evidenced by the vast diversity of inhibitor backbones. Efforts made to develop superior ZBGs are miniscule in comparison. The consensus in the field appears to be that hydroxamic acids are satisfactory metal chelators and that MPI design should focus on the development of more effective backbones. However, an examination of MPI activity with identical backbones but different ZBGs indicates that both the ZBG and the backbone are essential for obtaining an effective inhibitor, and therefore adequate efforts should be made in the development of improved ZBGs. Despite the popularity of hydroxamic acids as a ZBG, there are significant limitations associated with its use in MPIs. Hydroxamic acids are vulnerable to rapid excretion and in vivo hydrolysis; in a study of 5-lipoxygenase inhibitors, hydroxamic acid-based inhibitors were found to rapidly hydrolyze to the corresponding carboxylic acid, which is a significantly weaker ZBG (see the preceding discussion) (Singh et al., 1995; Summers et al., 1987). The metal binding selectivity that can be obtained through the ZBGs is a largely overlooked area in MPI development, which is surprising considering that hydroxamic acids have poor selectivity for zinc(II) ions over other divalent first row transition metals. Hydroxamic acids bind tightly to a variety of metal ions in several oxidation states, as evidenced by their widespread use in bacterial siderophores where they act as strong iron(III) chelators (Farkas, Enyedy, & Cso´ka, 1999). Indeed, the only medically approved chelator for iron overload is the tris(hydroxamate) siderophore desferrioxamine (DFO) (Farkas, Enyedy, Zeka´ny, & Dea´k, 2001). In terms of hard/soft acid-base chemistry, the zinc(II) ion is often classified as intermediate and is generally regarded as softer than metals such as iron(III) or manganese(II) (Bertini, Gray, Lippard, & Valentine, 1994; Sigel & McCormick, 1970). In addition to the low selectivity of hydroxamates for various metal ions, the carbon-nitrogen bond in these compounds can undergo a cis to trans conformational change, which reduces its affinity for binding to all metal ions, relative to more rigid ligands. The different forms of collagenases are upregulated and activated by cytokines and growth factors. Use of MMP inhibitors is a major strategy in pathological conditions with aberrant expression and activation of MMPs. Development of inhibitors of MMPs for the last three decades had little success. Activation of latent forms of MMPs is associated with altered coordination of the zinc with the dSH group of a Cys residue, for water. Close contact between 3. EXAMPLES AND CASE STUDIES



Cys and zinc can activate the MMPs, and molecules that can prevent this contact through chelation of the active zinc can serve as potential MMP inhibitors. Despite the massive efforts towards developing MMP inhibitors and repeated failures during clinical trials, it is widely believed that repurposed drugs and natural compounds can emerge as therapeutic clinical entities. Thus, in the current study, we have tested the ability of ascorbic acid (vitamin C), as a potential inhibitor of MMP-8. Among the 28 types of collagenenases found so far, MMP-8 belongs to the class of secreted or membrane-associated collagenases, which target type I, II, and III collagens (Freije et al., 1994; Hasty et al., 1987). MMP-8 possesses an essential catalytic zinc-binding domain, a propeptide domain hinge region, and a C-terminal hemopexin-like domain, which is conserved in other MMPs, along with substructural zinc and two to three calcium ions that are required for stability and the collagenase activity (Visse & Nagase, 2003). The catalytic domain of MMPs is well conserved, except for the specificity pocket (S10 ) region, which is lined by Tyr 219-Leu 229 in the case of MMP-8 and is rather deep. The size of the S10 pocket strongly influences the substrate specificity (Kalva, Vadivelan, Sanam, Jagarlapudi, & Saleena, 2012). Sequence and structural alignment data reveal that the S10 loop in MMP-8 differs from other MMPs with two residues Arg 222 and Tyr 227 (Aureli et al., 2008). Because of their unique localization to the S10 pocket in MMP-8, any inhibitor molecule that binds to these residues may offer selectivity. Ascorbic acid ((5R)-[(1S)-1,2-dihydroxyethyl]-3,4-dihydroxyfuran-2(5H)-one), an essential vitamin and more than just a micronutrient, is identified for its role in antiscurvy treatment. Vitamin C–dependent enzymes prolyl and lysyl hydroxylase are required for the hydroxylation of collagen. In addition to this cofactor role, ascorbate was shown to influence cancer cell growth and expression of MMPs and TGF-β (Ma et al., 2014; Philips, Dulaj, & Upadhya, 2009). These differential dose-dependent effects suggest a rather interesting dual function of ascorbic acid in ECM remodeling. We therefore hypothesize that ascorbate may partly regulate the collagen turnover by influencing the MMP activity. It is also known that ascorbic acid forms zinc chelates at alkaline pH (Kleszczewska & Misiuk, 2000). We therefore probed the interactions of ascorbic acid with MMP-8 by molecular docking, molecular dynamic simulations, and validated these results using in vitro studies.

2 EXPERIMENTAL PROCEDURES 2.1 Molecular Docking ˚ were retrieved The crystal structures of matrix metalloproteinase with a resolution of 2.0 A from protein data bank (PDB ID: 2TCL, 1ZP5, 3ZXH and 3DNG) (Campestre et al., 2006). The retrieved protein was subjected to primary preparation, using the Protein preparation wizard, predefining major properties like specified bond orders to hydrogens, zero order bonds created to metal atoms; capping the termini and desolation was done by removing crystal˚ (Sastry, Adzhigirey, Day, Annabhimoju, & Sherman, lized free water molecules beyond 5 A 2013). Following this, the hydrogen bonds in the protein were optimized and minimized in the presence of force field Optimized Potential for Liquid Simulations (OPLS) 2005 (Shivakumar et al., 2010). Ascorbic acid was docked in the active site of collagenases using 3. EXAMPLES AND CASE STUDIES



Glide extra-precision (XP), version 5.5. Ligplot was chosen for analyzing the protein-ligand complexes (Wallace, Laskowski, & Thornton, 1995). The in silico work was executed using the Schr€ odinger suite 2014-3 on the HP Z600 workstation.

2.2 Ligand Preparation Primary information regarding GM 6001(2R)-N0 -hydroxy-N-[(2S)-3-(1H-indol-3-yl)-1(methylamino)-1-oxopropan-2-yl]-2-(2-methylpropyl)butanediamide) ascorbic acid structural coordinates was acquired from Chemspider. The retrieved compounds’ geometries were optimized through OPLS 2005 force field, employing the LigPrep module.

2.3 Ligand Docking The Glide module in the Maestro suite was applied for ligand docking. The module uses a grid-based ligand docking method with energetics that reflect the favorable interactions between the molecule and a protein as output. Prior to docking, a Grid box was generated at the centroid of the ligand with the assistance of receptor grid generation protocol. Following this, the prepared molecules were docked with the extra precision (XP) docking protocol (Friesner et al., 2006).

2.4 Molecular Dynamic Simulations The Desmond module was used to perform the molecular dynamic simulations of all the interactions between ligands and protein under the force field OPLS 2005 (Guo et al., 2010). In the present study, protein ligand complexes were subjected to a TIP3P water model in an orthorhombic periodic boundary under solvated condition using the system builder. In order to neutralize the system, Na+ ions or Cl ions were added with respect to the net charge of the system and a salt concentration of 0.15 M was also included. This prepared model system using the system builder was minimized up to a maximum of 5000 iterations and the total number of atoms present in the built system were calculated using the minimization step. Further molecular dynamic simulation studies were carried out with a periodic boundary condition in an NPT ensemble, temperature at 300 K, 1 atmospheric pressure, and finally relaxed using the default relaxation protocol integrated in Desmond. The simulation job was carried out for a time period of 10 ns with 5 ps intervals and a time step of 5 ps. This procedure was carried out for all protein ligand complexes for a time period of 10 ns. RMSD), root mean square fluctuation (RMSF), H-bond and total energy of all the complexes were studied.

2.5 Rheumatoid Synovial Fibroblast Culture for MMP-8 Conditioned Medium Freshly dispersed synovium obtained from rheumatoid arthritis subjects was used for preparation of synovial fibroblasts (Unemori, Hibbs, & Amento, 1991). Local ethics committee guidelines were strictly followed. Briefly, the outer superficial layer of synovium was removed, cut into pieces, and treated with 4 mg/mL clostridial collagenase (Worthington Biochemical, St. Louis, MO) and 0.1% DNase I (Sigma Chemicals, St. Louis, MO) in DMEM at 37°C for 60–90 min. The tissue was further treated with 0.25% trypsin for the next 30 min. 3. EXAMPLES AND CASE STUDIES



Suspension of synovial cells was washed 2 in 50% PBS-50% DMEM + 15% FBS at 37°C and seeded at 106–107 cells/100 mm tissue culture plate in DMEM-FBS. After allowing the cells to adhere, the nonadherent cells were eliminated by washing. Conditioned media were collected after treatment of the cells for 24 h with phorbol 12-myristate 13-acetate (PMA, 10 nM). A 50-kDa immunoreactive band corresponding to MMP-8 was observed.

2.6 Fluorogenic MMP Activity Measurements MMP-8 activity in culture media was measured by a SensoLyte Plus 520 MMP assay kit (AnaSpec, San Jose, CA), after centrifugation of the medium at 13,000  g for 4 min; 1 mM APMA (4-aminophenylmercuric acetate) was used for activation of the Pro-MMP to MMP. Ascorbate at the indicated concentrations was added to the harvested media aliquots and incubated for 24 h, to determine their inhibitory action against MMPs.

2.7 MMP-8 Enzyme Assay The full-length MMP-8 enzyme activity was tested against type II collagen. The enzyme was activated with 2 μg trypsin and incubated with canine type II collagen (1–12 μg) at 30°C in 0.05M Tris, 0.005M CaCl2 for 30 min. Digestion was stopped by boiling the samples for 3 min in sodium dodecyl sulfate–polyacrylamide gel electrophoresis loading buffer. The samples were then subjected to electrophoresis on 7.5% polyacrylamide gels and stained with Coomassie blue. The gels were analyzed by densitometry and digestion of collagen was quantified by measuring the amount of TCA present in each sample (Mitchell et al., 1996).

2.8 Collagen Zymography Zymography analysis also used similar aliquots that were utilized for fluorogenic enzyme activity measurements. Conditioned media appropriately incubated with APMA, GM 6001, and ascorbic acid was mixed with 2  SDS sample buffer, omitting β-mercaptoethanol and used for zymography. Briefly, aliquots of the conditioned media were electrophoresed on a 10% SDS-polyacrylamide gel containing 0.5 mg/mL collagen (Collagen Type I and Collagen Type IV, Sigma-Aldrich). The zymographic activities were revealed by staining with 1% Coomassie blue and, subsequently, destaining of the gel.

2.9 Statistical Analyses Data were analyzed by one-way analysis of variance using SigmaStat 3.5. Values are expressed as mean  standard error (SE).

3 RESULTS MMPs, either secreted or membrane bound, possess a general structure consisting of a prodomain, a catalytic domain, a hinge (linker) region, and a hemopexin domain (Nagase, Visse, & Murphy, 2006). The catalytic domain exhibits similar structure in all MMPs and is 3. EXAMPLES AND CASE STUDIES



composed of two zinc(II) and two or three calcium(II) ions. One of the two zinc ions is catalytic and the other plays a structural role. The inhibitor-protein interactions in the MMP active site are determined by the type of the catalytic zinc-coordination group, the presence of hydrogen bonds, and the hydrophobic interactions between the inhibitor and the S10 pocket residues (Nagase et al., 2006).

3.1 Docking Studies Molecular docking of GM 6001 ascorbic acid ligands against the 3DNG S11 active pocket of the protein revealed that the two molecules were seated inside the active pocket and produced hydrogen bond interactions with the important amino acids of the pocket. The compound GM 6001 produced hydrogen bond interactions with Pro 211, Gly 212, Ala 213, Asn 218, Arg 222 residues. Arg 222, the key amino acid that differentiates MMP-8 from other MMPs, formed two backbone hydrogen bonds with the GM 6001 via NH groups of the res˚ (Fig. 6A). The overall proteinidue and ¼O of the ligand with a distance of 1.81 and 2.77 A ligand complex was maintained with the help of six hydrogen bonds and produced a G-score of 7.99 and 40 Kcal/mol free energy of binding. The binding mode of the ascorbic acid– MMP-8 complex was maintained by three hydrogen bonds, two with Arg 222 and one with Ser 228. Arg 222, the key amino acid residue, formed two H bonds with two OH groups in the ligand. However, the two H bonds shared by Arg 222 with ascorbic acid were reported to ˚ respectively as depicted in Fig. 6B. The overall comexhibit bond lengths of 2.31 and 2.48 A plex maintained a G-score of 4.55 and 23 Kcal/mol of free energy of binding. Free energy of binding revealed by molecular docking was found to be lower for ascorbic acid when compared to GM6001, indicating a decrease in binding affinity for ascorbic acid. Arg 222 is observed to form two important H-bonds in the binding pockets in both the interactions (Fig. 7). The active site is mainly a hydrophobic groove with a much less positively charged region. Arg 222, the positively charged residue, is found on this small region of positive charge, making it exclusive for these interactions. The variation of the fitting in the binding pockets may be attributed to the difference in the size of the molecules. Ascorbic acid, being a smaller molecule, fit comfortably into shallow portions of the pocket, whereas the relatively bigger molecule GM 6001 does not seem to fit in the shallow groove, as depicted in Fig. 7B. Comparison of the G-scores indicates that the ascorbic acid molecule exhibits a low score when compared to standard GM 6001, again owing to the size and the difference in the number of functional groups. After the docking studies, these two complexes were further subjected to dynamic studies to observe the deviations, fluctuations, and consistency in maintaining the hydrogen bond with the important amino acid, Arg 222.

3.2 Molecular Dynamic Simulations The stability of the two docked complexes, namely 3DNG with GM6001 or ascorbic acid, was further analyzed with 20 ns molecular dynamic studies. In the simulation studies, both the ligand and protein were allowed complete flexibility, unlike in the docking protocol, where the protein was rigid and ligand was maintained in the flexible mode. Such a study involving different flexibilities was employed to understand the binding modes between 3. EXAMPLES AND CASE STUDIES


FIG. 6

Binding modes of the molecules with MMP-8 protein; (A) 3d and 2d plots of molecule GM 6001; and (B) 3d and 2d plots of ascorbic acid.



FIG. 7 (A) Interactions of GM-6001 and ascorbic acid with Arg 222; (B) Occupancy of the GM 6001 and ascorbic acid molecules inside pocket (black surface indicates the binding surface of the MMP-8).

the protein-ligand complex and also their consistency in maintaining the interactions observed during the docking studies.

3.3 Simulation Analysis of 3DNG Protein With GM6001 ˚ . During the initial simThe Cα of the protein RMSD was recorded between 1.0 and 2.25 A ˚ and graphed to ulation time scale, the deviations in the protein were found starting at 1.0 A ˚ ˚. 1.75 A by the end of 1.5 ns. From there up to 5.5 ns, the deviations depreciated to around 1.5 A From there, a trend of elevations and depressions in the deviations was observed until 18 ns had elapsed. Finally, the deviations acquired stability from 18 to 20 ns, recording an RMSD of ˚ . A maximum deviation, i.e., 2.25 A ˚ , was observed around the 13th nanosecond of around 1.6 A the trajectory run. The ligand fit over the protein RMSD displayed initial deviations from ˚ by the end of 3 ns. Until 12 ns the deviations were around 3.2 A ˚ and then spiked 1.6 to 3.2 A ˚ ˚ by the drastically to 5.4 A. As the simulation continued, the deviations were decreased to 3.2 A ˚ end of 15.5 ns. Between 16 and 17 ns, a sudden rise to 6.4 A was observed in the deviations, ˚ by the end of the simulation time. which settled to 4.0 A The ligand fit the exhibited deviations in the initial time period up to 5 ns, and then ˚ till the end of simulations. The deviations in the proteinexhibited deviations around 1.8 A 3. EXAMPLES AND CASE STUDIES



ligand RMSD is explained by the presence of more loop regions and the active pocket being a part of the loop region in the protein. The fluctuations in the protein were mainly reported in a ˚ , meaning that all the residues were within the appropriate range. Constable range, below 3 A tacts between amino acids and the ligand were analyzed broadly to check the H-bond formation with the important amino acid Arg 222. During the simulations run, the ligand was observed to form two hydrogen bonds with an interaction percentage of 34 and 72, respectively, unlike docking studies where it was a single hydrogen bond. The number of hydrogen bonds formed through the 20-ns trajectory run ranged between 6600 and 7000 widely held. The energy required by the protein-ligand complex was observed between 230 and 260 Kcal/mol. A stable energy consumption was observed at 230–240 Kcal/mol after 15 ns simulation, till which time a fluctuating trend of energy was observed. The RMSD, RMSF, H-bonds, energy, and percentage of interaction are illustrated in Fig. 8.

3.4 Simulation Analysis of 3DNG Protein With Vitamin C Extra docking exhibited two H-bond interactions by ascorbic acid with the protein active site residue Arg 222. Molecular dynamic simulations also turned in two H-bond interactions with Arg 222, strengthening the importance of Arg 222 in the interaction validation. The protein RMSD in this complex reported continuous deviations throughout the trajectory between ˚. 0.9 and 1.8 A The ligand fit RMSD also showed deviations up to 10 ns and from where persistent devi˚ . The ligand fit protein deviations were very high ations were observed between 1 and 2 A compared to the protein and ligand RMSD. Initially the deviations were reported around ˚ towards 10 ns; after that the RMSD was on continuous deviations progressing to 7.5 A ˚ 4A ˚ , was noticed at one instance, i.e., by the end of simulations. Maximum RMSD, above 8 A around 13 ns. This is due mainly to the size of the molecule and pocket in the protein. The protein pocket was large but shallow, whereas the molecule, ascorbic acid, was very small, due to which the deviations were continuous, in order to stay intact as a complex. The RMSF ˚ , maintaining an acceptable of the protein showed fluctuations in all the residues below 2.6 A range. The hydrogen bonds were observed to form throughout the simulations. The RMSD, RMSF, H-bonds, energy, and percentage of interaction are illustrated in Fig. 9.

3.5 Inhibition of MMP-8 Activity Against Collagen by Ascorbic Acid The activity of MMP-8 was tested against type II collagen. The full-length MMP-8 cleaved this substrate, yielding the expected TCA and TCB fragments (Fig. 10A). When the enzyme was tested in the presence of 5 and 10 μM ascorbic acid, the activity of MMP-8 was inhibited by 64% and 87%, respectively. We used a conditioned medium of synovial fibroblasts from rheumatoid arthroid patients treated with 10 nM PMA as a source of MMP-8. Rheumatoid fibroblast cells in culture showed a very faint band corresponding to a 50-kDa pro-MMP-8 on the zymogram. Treatment of the cells with 10 nM PMA for 24 h significantly increased the pro-MMP-8 form released into the culture medium, as shown earlier (Hanemaaijer et al., 1997). A very faint band corresponding to a 40-kDa protein also was observed in the conditioned medium concentrate, indicating it is an active form of MMP-8. To increase the level of active MMP-8, we incubated the conditioned 3. EXAMPLES AND CASE STUDIES



FIG. 8 Molecular dynamics profiles of MMP-8-GM6001 complex. (A) Protein-ligand-RMSD with respect to time in; (B) Protein RMSF plot with respect to residues; (C) H-bond plot along the trajectory; (D) Total energy variations in Kcal/mole along the trajectory; and (E) Lig-Plot of MMP-8 and GM6001 showing percentage of interaction with key residues.


FIG. 9 Molecular dynamics profiles of MMP-8-ascorbic acid complex; (A) Protein-ligand-RMSD with respect to time in; (B) Protein RMSF plot with respect to residues; (C) H-bond plot along the trajectory; (D) Total energy variations in Kcal/mole along the trajectory; and (E) Lig-Plot of MMP-8 and ascorbic acid showing percentage of interaction with key residues Arg 222 and Ser 228.



FIG. 10

(A) Collagen degradation by MMP-8 and inhibition by ascorbic acid. MMP-8 was activated with trypsin and incubated at room temperature overnight with type II collagen in the presence or absence of ascorbic acid. Each assay mixture contained 6 μg canine type II collagen in 50 μL assay buffer. The assay mixtures contained 2 pmoles of human recombinant MMP-8. In each pair of lanes, the sample on the left contained no ascorbic acid while that on the right contained 5 and 10 μM ascorbic acid. The MMP-8 cleaved the collagen to yield TCA and TCB fragments, which were quantified by densitometry. (B) Collagen zymography of MMP-8 conditioned medium from human rheumatoid synovial fibroblasts, with unstimulated conditioned medium (Lane 1), PMA-treated cells without APMA (Lane 2), and with 1 mM APMA (Lane 3); conditioned medium as in Lane 3, in the presence of 1 mM GM 6001 (Lane 4, ascorbic acid) (1 μM, Lane 5), ascorbic acid (5 μM, Lane 6) and ascorbic acid (10 μM, Lane 7). Markings depict bands that were intensified by APMA and are consistent with Pro-MMP-8 (upper arrow) and active MMP-8 (lower arrow) activity.

medium with 1 mM APMA for 24 h, which increased the active/inactive MMP-8 ratio. In parallel aliquots, 1 mM GM 6001 significantly attenuated APMA-induced formation of active MMP-8. Co-incubation of conditioned medium with APMA and increasing concentrations of ascorbic acid showed a clear reduction in the amount of the active form of MMP-8, as shown in the zymogram (Fig. 10B).

4 DISCUSSION Various physiological processes including embryogenesis, tissue repair and remodeling, and organ morphogenesis involve collagenolysis (Sternlicht & Werb, 2001). Zn2+ ion in the catalytic domain of MMPs is coordinated to a tris (histidine) motif and is required for both substrate binding and cleavage. MMP expression and activity are deregulated in multiple pathological conditions. Either attenuating the expression of endogenous TIMPs or design and development of molecules with potential for specific and selective inhibition of the MMPs is warranted. MMP inhibitor design (MMPi) requires two parts: a ZBG and a peptidomimetic backbone. The catalytic Zn2+ in the active site is surrounded by subsite pockets designated as S1, S2, S3, S10 , S20 , and S30 (Cuniasse et al., 2005). Of the different subsite pockets, targeting of the S10 pocket has provided the basis of selectivity for many MMPi (Rao, 2005).




Ascorbate (vitamin C) has the potential to influence the extracellular matrix and tumor state. At low concentrations, ascorbate stimulates MMP-1, MMP-2, and TGF-β, with the end result being the elimination of cancer cells with damage to the ECM. At high concentrations, ascorbate inhibits MMP and TGF-β expression, implicating growth and ECM advantage (Philips, Keller, & Holmes, 2007). Ascorbate inhibited digestion of type II collagen by MMP-8, but the concentration necessary for 50% inhibition was lower than the concentration of ascorbate likely to be present in human plasma (22–85 μmol/L  0.4–1.7 mg/dL) (Halliwell, Wasil, & Grootveld, 1987). Based on the simulation results it can be inferred that both molecules fit into the inhibitory pocket of the MMP-8, producing antagonistic interactions. Interaction with Arg 222, with two H-bonds, in both the molecules is a striking similarity in binding mode. However, the percentage of interaction with that amino acid varied in both complexes. The percentage interactions with H-bonds between ascorbic acid and Arg 222 are 100% and 30%, respectively, and higher than that of standard GM6001. From these results, it is evident that, even though smaller in size, the interaction of ascorbic acid with MMP-8 was strong enough and showed interactions similar to those exhibited by GM6001. A comparison between the uninhibited MMP-8 and the ascorbic acid–bound complexes reported here reveals important structural differences regarding essentially the S10 specificity loop. The first part of the loop, containing the third coordinated H207 and two consecutive β-turns (210–216) with the strictly invariant M215, remains practically unaltered. Large conformational changes are induced by both inhibitors on the segment 219–229, separating the tube-like crevice from the bulk of water. The largest Cα displacements induced by both inhibitors occur for the sequence R222-N226. As a consequence, the Y227 side-chain of both complexes has been pushed from the position occupied in the uninhibited enzyme. MMP-8 is routinely considered as a target for cancer, but could also be a hit for treatment of acute liver failure, as MMP-8 deficient mice were found to be resistant to hepatitis induced by TNF-α (Van Lint et al., 2005). MMP-8 inhibitors could also work in alleviating inflammation and tumor progression (Van Lint & Libert, 2006). Therefore, the current study paves the way for testing of vitamin C as a potential inhibitor for specific inhibition of MMP-8, which could be useful in various pathophysiological conditions. This work provides compelling evidence for the concept of repurposing currently available drugs/natural compounds in therapeutic interventions to treat aberrant MMP expression. Considering the repeated clinical trial failures for various MMP inhibitors in the past, the repurposing may open new vistas in drug research and combinatorial chemistry can push this further to avoid nonspecific actions, while providing more potency. Novelty in drug discovery does not have to depend on new chemical structures but on finding new targets for a known chemical with biological potency. This approach is likely to provide new avenues to fight a disease with the small effort of in silico simulations, rather than wasting time and effort on finding new chemicals.

Acknowledgments The authors thank Dr. Talluri Venkateswara Rao for providing access to the Schrodinger suite at KL University. This work was supported by a UGC-Startup Grant for faculty selected under the Faculty Recharge Programme (F.4-5(23-FRP)/2013(BSR)) sanctioned to SB.




References Aureli, L., Gioia, M., Cerbara, I., Monaco, S., Fasciglione, G. F., Marini, S., … Coletta, M. (2008). Structural bases for substrate and inhibitor recognition by matrix metalloproteinases. Current Medicinal Chemistry, 15(22), 2192–2222. Babine, R. E., & Bender, S. L. (1997). Molecular recognition of proteinminus signligand complexes: applications to drug design. Chemical Reviews, 97(5), 1359–1472. Baxter, A. D., Bhogal, R., Bird, J. B., Buckley, G. M., Gregory, D. S., Hedger, P. C., … Watson, R. J. (1997). Mercaptoacyl matrix metalloproteinase inhibitors: the effect of substitution at the mercaptoacyl moiety. Bioorganic & Medicinal Chemistry Letters, 7(21), 2765–2770. Baxter, C. A., Murray, C. W., Clark, D. E., Westhead, D. R., & Eldridge, M. D. (1998). Flexible docking using Tabu search and an empirical estimate of binding affinity. Proteins, 33(3), 367–382. Becker, J. W., Marcy, A. I., Rokosz, L. L., Axel, M. G., Burbaum, J. J., Fitzgerald, P. M., et al. (1995). Stromelysin-1: threedimensional structure of the inhibited catalytic domain and of the C-truncated proenzyme. Protein Science, 4(10), 1966–1976. Bertini, I., Calderone, V., Fragai, M., Luchinat, C., Mangani, S., & Terni, B. (2003). X-ray structures of binary and ternary enzyme-product-inhibitor complexes of matrix metalloproteinases. Angewandte Chemie (International Ed. in English), 42(23), 2673–2676. Bertini, I., Gray, H. B., Lippard, S. J., & Valentine, J. S. (1994). Bioinorganic chemistry. Mill Valley, CA: University Science Books. Blaney, J. M., & Dixon, J. S. (1993). A good ligand is hard to find: automated docking methods. Perspectives in Drug Discovery and Design, 1(2), 301–319. Bode, W., Fernandez-Catalan, C., Tschesche, H., Grams, F., Nagase, H., & Maskos, K. (1999). Structural properties of matrix metalloproteinases. Cellular and Molecular Life Sciences: CMLS, 55(4), 639–652. s000180050320. Brandstetter, H., Grams, F., Glitz, D., Lang, A., Huber, R., Bode, W., … Engh, R. A. (2001). The 1.8-A crystal structure of a matrix metalloproteinase 8-barbiturate inhibitor complex reveals a previously unobserved mechanism for collagenase substrate recognition. The Journal of Biological Chemistry, 276(20), 17405–17412. 10.1074/jbc.M007475200. Browner, M. F., Smith, W. W., & Castelhano, A. L. (1995). Matrilysin-inhibitor complexes: common themes among metalloproteases. Biochemistry, 34(20), 6602–6610. Campbell, D. A., Xiao, X. -Y., Harris, D., Ida, S., Mortezaei, R., Ngu, K., … Navre, M. (1998). Malonyl α-mercaptoketones and α-mercaptoalcohols, a new class of matrix metalloproteinase inhibitors. Bioorganic & Medicinal Chemistry Letters, 8(10), 1157–1162. Campbell, S. J., Gaulton, A., Marshall, J., Bichko, D., Martin, S., Brouwer, C., & Harland, L. (2010). Visualizing the drug target landscape. Drug Discovery Today, 15(1-2), 3–15. Campestre, C., Agamennone, M., Tortorella, P., Preziuso, S., Biasone, A., Gavuzzo, E., … Gallina, C. (2006). N-Hydroxyurea as zinc binding group in matrix metalloproteinase inhibition: mode of binding in a complex with MMP-8. Bioorganic & Medicinal Chemistry Letters, 16(1), 20–24. Chen, E. I., Li, W., Godzik, A., Howard, E. W., & Smith, J. W. (2003). A residue in the S2 subsite controls substrate selectivity of matrix metalloproteinase-2 and matrix metalloproteinase-9. The Journal of Biological Chemistry, 278 (19), 17158–17163. Chen, J. M., Nelson, F. C., Levin, J. I., Mobilio, D., Moy, F. J., Nilakantan, R., … Powers, R. (2000). Structure-based design of a novel, potent, and selective inhibitor for MMP-13 utilizing NMR spectroscopy and computer-aided molecular design. Journal of the American Chemical Society, 122(40), 9648–9654. Coussens, L. M., Fingleton, B., & Matrisian, L. M. (2002). Matrix metalloproteinase inhibitors and cancer: trials and tribulations. Science, 295(5564), 2387–2392. Cuniasse, P., Devel, L., Makaritis, A., Beau, F., Georgiadis, D., Matziari, M., … Dive, V. (2005). Future challenges facing the development of specific active-site-directed synthetic inhibitors of MMPs. Biochimie, 87(3), 393–402. Darensbourg, D. J., Wildeson, J. R., & Yarbrough, J. C. (2002). Solid-state structures of zinc(II) benzoate complexes. Catalyst precursors for the coupling of carbon dioxide and epoxides. Inorganic Chemistry, 41(4), 973–980. Dunten, P., Kammlott, U., Crowther, R., Levin, W., Foley, L. H., Wang, P., & Palermo, R. (2001). X-ray structure of a novel matrix metalloproteinase inhibitor complexed to stromelysin. Protein Science, 10(5), 923–926. https://dx.doi. org/10.1110/ps.48401.




Esser, C. K., Bugianesi, R. L., Caldwell, C. G., Chapman, K. T., Durette, P. L., Girotra, N. N., et al. (1997). Inhibition of stromelysin-1 (MMP-3) by P1’-biphenylylethyl carboxyalkyl dipeptides. Journal of Medicinal Chemistry, 40(6), 1026–1040.  A., & Cso´ka, H. (1999). A comparison between the chelating properties of some dihydroxamic Farkas, E., Enyedy, E. acids, desferrioxamine B and acetohydroxamic acid. Polyhedron, 18(18), 2391–2398.  A., Zeka´ny, L., & Dea´k, G. (2001). Interaction between iron (II) and hydroxamic acids: oxidation Farkas, E., Enyedy, E. of iron (II) to iron (III) by desferrioxamine B under anaerobic conditions. Journal of Inorganic Biochemistry, 83(2-3), 107–114. Finzel, B. C., Baldwin, E. T., Bryant, G. L., Jr., Hess, G. F., Wilks, J. W., Trepod, C. M., … Mitchell, M. A. (1998). Structural characterizations of nonpeptidic thiadiazole inhibitors of matrix metalloproteinases reveal the basis for stromelysin selectivity. Protein Science, 7(10), 2118–2126. Fray, M. J., Burslem, M. F., & Dickinson, R. P. (2001). Selectivity of inhibition of matrix metalloproteases MMP-3 and MMP-2 by succinyl hydroxamates and their carboxylic acid analogues is dependent on P3’ group chirality. Bioorganic & Medicinal Chemistry Letters, 11(4), 567–570. Freije, J. M., Diez-Itza, I., Balbin, M., Sanchez, L. M., Blasco, R., Tolivia, J., & Lopez-Otin, C. (1994). Molecular cloning and expression of collagenase-3, a novel human matrix metalloproteinase produced by breast carcinomas. The Journal of Biological Chemistry, 269(24), 16766–16773. Freskos, J. N., McDonald, J. J., Mischke, B. V., Mullins, P. B., Shieh, H. -S., Stegeman, R. A., & Stevens, A. M. (1999). Synthesis and identification of conformationally constrained selective MMP inhibitors. Bioorganic & Medicinal Chemistry Letters, 9(13), 1757–1760. Friesner, R. A., Murphy, R. B., Repasky, M. P., Frye, L. L., Greenwood, J. R., Halgren, T. A., … Mainz, D. T. (2006). Extra precision glide: docking and scoring incorporating a model of hydrophobic enclosure for protein-ligand complexes. Journal of Medicinal Chemistry, 49(21), 6177–6196. Gall, A. L., Ruff, M., Kannan, R., Cuniasse, P., Yiotakis, A., Dive, V., … Moras, D. (2001). Crystal structure of the stromelysin-3 (MMP-11) catalytic domain complexed with a phosphinic inhibitor mimicking the transition-state. Journal of Molecular Biology, 307(2), 577–586. Gavuzzo, E., Pochetti, G., Mazza, F., Gallina, C., Gorini, B., D’Alessio, S., … Tucker, P. A. (2000). Two crystal structures of human neutrophil collagenase, one complexed with a primed- and the other with an unprimed-side inhibitor: implications for drug design. Journal of Medicinal Chemistry, 43(18), 3377–3385. Goldberg, G. I., Wilhelm, S. M., Kronberger, A., Bauer, E. A., Grant, G. A., & Eisen, A. Z. (1986). Human fibroblast collagenase. Complete primary structure and homology to an oncogene transformation-induced rat protein. Journal of Biological Chemistry, 261(14), 6600–6605. Goodsell, D. S., & Olson, A. J. (1990). Automated docking of substrates to proteins by simulated annealing. Proteins, 8 (3), 195–202. Grams, F., Reinemer, P., Powers, J. C., Kleine, T., Pieper, M., Tschesche, H., … Bode, W. (1995). X-ray structures of human neutrophil collagenase complexed with peptide hydroxamate and peptide thiol inhibitors. Implications for substrate binding and rational drug design. European Journal of Biochemistry, 228(3), 830–841. Guo, Z., Mohanty, U., Noehre, J., Sawyer, T. K., Sherman, W., & Krilov, G. (2010). Probing the alpha-helical structural stability of stapled p53 peptides: molecular dynamics simulations and analysis. Chemical Biology & Drug Design, 75 (4), 348–359. Hajduk, P. J., Sheppard, G., Nettesheim, D. G., Olejniczak, E. T., Shuker, S. B., Meadows, R. P., … Fesik, S. W. (1997). Discovery of potent nonpeptide inhibitors of stromelysin using SAR by NMR. Journal of the American Chemical Society, 119(25), 5818–5827. Halliwell, B., Wasil, M., & Grootveld, M. (1987). Biologically significant scavenging of the myeloperoxidase-derived oxidant hypochlorous acid by ascorbic acid. Implications for antioxidant protection in the inflamed rheumatoid joint. FEBS Letters, 213(1), 15–17. Hanemaaijer, R., Sorsa, T., Konttinen, Y. T., Ding, Y., Sutinen, M., Visser, H., … Salo, T. (1997). Matrix metalloproteinase-8 is expressed in rheumatoid synovial fibroblasts and endothelial cells. Regulation by tumor necrosis factor-alpha and doxycycline. The Journal of Biological Chemistry, 272(50), 31504–31509. Harland, L., & Gaulton, A. (2009). Drug target central. Expert Opinion on Drug Discovery, 4(8), 857–872. https://dx.doi. org/10.1517/17460440903049290. Hasty, K., Jeffrey, J., Hibbs, M., & Welgus, H. (1987). The collagen substrate specificity of human neutrophil collagenase. Journal of Biological Chemistry, 262(21), 10048–10052.




Hirayama, R., Yamamoto, M., Tsukida, T., Matsuo, K., Obata, Y., Sakamoto, F., & Ikeda, S. (1997). Synthesis and biological evaluation of orally active matrix metalloproteinase inhibitors. Bioorganic & Medicinal Chemistry, 5(4), 765–778. Hughes, I., Harper, G. P., Karran, E. H., Markwell, R. E., & Miles-Williams, A. J. (1995). Synthesis of thiophenol derivatives as inhibitors of human collagenase. Bioorganic & Medicinal Chemistry Letters, 5(24), 3039–3042. Hutchinson, J. W., Tierney, G. M., Parsons, S. L., & Davis, T. R. (1998). Dupuytren’s disease and frozen shoulder induced by treatment with a matrix metalloproteinase inhibitor. Journal of Bone and Joint Surgery. British Volume (London), 80(5), 907–908. Jones, G., Willett, P., Glen, R. C., Leach, A. R., & Taylor, R. (1997). Development and validation of a genetic algorithm for flexible docking. Journal of Molecular Biology, 267(3), 727–748. Kalva, S., Vadivelan, S., Sanam, R., Jagarlapudi, S. A., & Saleena, L. M. (2012). Lead identification and optimization of novel collagenase inhibitors; pharmacophore and structure based studies. Bioinformation, 8(7), 301–308. https:// Kleszczewska, E., & Misiuk, W. (2000). Determination of chelate complexes by spectrophotometry; L-ascorbic acid with cadmium (II) and zinc (II) in alkaline solution. Acta Poloniae Pharmaceutica, 57(5), 327–330. Komiyama, T., Aoyagi, T., Takeuchi, T., & Umezawa, H. (1975). Inhibitory effects of phosphoramidon on neutral metalloendopeptidases and its application on affinity chromatography. Biochemical and Biophysical Research Communications, 65(1), 352–357. Kramer, B., Rarey, M., & Lengauer, T. (1999). Evaluation of the FLEXX incremental construction algorithm for proteinligand docking. Proteins, 37(2), 228–241. Kremer-Aach, A., Klaui, W., Bell, R., Strerath, A., Wunderlich, H., & Mootz, D. (1997). Cobalt as a probe for zinc in metalloenzyme model compounds? A comparison of spectroscopic features and coordination geometry of fourand five-coordinate complexes. Crystal and molecular structures of [Co(eta(3)-Tp(Ph))(eta(2)-Tp(Ph))], [(eta(3)-Tp (Ph))Zn(anthranilate)], and [(eta(3)-Tp(Ph))M(eta(2)-acac)] (Tp(Ph) ¼ Hydrotris(3-phenylpyrazol-1-yl)borate, acac ¼ Pentane-2,4-dionate, and M ¼ Zn, Co). Inorganic Chemistry, 36(8), 1552–1563. Kruger, A., Soeltl, R., Sopov, I., Kopitz, C., Arlt, M., Magdolen, V., … Schmitt, M. (2001). Hydroxamate-type matrix metalloproteinase inhibitor batimastat promotes liver metastasis. Cancer Research, 61(4), 1272–1275. Kuntz, I. D., Blaney, J. M., Oatley, S. J., Langridge, R., & Ferrin, T. E. (1982). A geometric approach to macromoleculeligand interactions. Journal of Molecular Biology, 161(2), 269–288. Leach, A. R. (1994). Ligand docking to proteins with discrete side-chain flexibility. Journal of Molecular Biology, 235(1), 345–356. Leach, A. R., & Kuntz, I. D. (1992). Conformational analysis of flexible ligands in macromolecular receptor sites. Journal of Computational Chemistry, 13(6), 730–748. Loging, W., Harland, L., & Williams-Jones, B. (2007). High-throughput electronic biology: mining information for drug discovery. Nature Reviews. Drug Discovery, 6(3), 220–230. Lovejoy, B., Welch, A. R., Carr, S., Luong, C., Broka, C., Hendricks, R. T., … Browner, M. F. (1999). Crystal structures of MMP-1 and -13 reveal the structural basis for selectivity of collagenase inhibitors. Nature Structural Biology, 6(3), 217–221. Ma, Y., Chapman, J., Levine, M., Polireddy, K., Drisko, J., & Chen, Q. (2014). High-dose parenteral ascorbate enhanced chemosensitivity of ovarian cancer and reduced toxicity of chemotherapy. Science Translational Medicine, 6(222), 222ra218. Massova, I., Kotra, L. P., Fridman, R., & Mobashery, S. (1998). Matrix metalloproteinases: structures, evolution, and diversification. The FASEB Journal, 12(12), 1075–1095. Matrisian, L. M. (1990). Metalloproteinases and their inhibitors in matrix remodeling. Trends in Genetics, 6(4), 121–125. Matter, H., Schwab, W., Barbier, D., Billen, G., Haase, B., Neises, B., … Weithmann, K. U. (1999). Quantitative structure-activity relationship of human neutrophil collagenase (MMP-8) inhibitors using comparative molecular field analysis and X-ray structure analysis. Journal of Medicinal Chemistry, 42(11), 1908–1920. 10.1021/jm980631s. Mitchell, P. G., Magna, H. A., Reeves, L. M., Lopresti-Morrow, L. L., Yocum, S. A., Rosner, P. J., … Hambor, J. E. (1996). Cloning, expression, and type II collagenolytic activity of matrix metalloproteinase-13 from human osteoarthritic cartilage. The Journal of Clinical Investigation, 97(3), 761–768. Morgunova, E., Tuuttila, A., Bergmann, U., Isupov, M., Lindqvist, Y., Schneider, G., & Tryggvason, K. (1999). Structure of human pro-matrix metalloproteinase-2: activation mechanism revealed. Science, 284(5420), 1667–1670.




Nagase, H., Visse, R., & Murphy, G. (2006). Structure and function of matrix metalloproteinases and TIMPs. Cardiovascular Research, 69(3), 562–573. Nagase, H., & Woessner, J. F., Jr. (1999). Matrix metalloproteinases. The Journal of Biological Chemistry, 274(31), 21491–21494. Natchus, M. G., Bookland, R. G., Laufersweiler, M. J., Pikul, S., Almstead, N. G., De, B., … Mieling, G. E. (2001). Development of new carboxylic acid-based MMP inhibitors derived from functionalized propargylglycines. Journal of Medicinal Chemistry, 44(7), 1060–1071. Notter, L. E. (1972). MEDLINE—newest service in the medical information network. Nursing Research, 21(2), 101. O’Brien, P. M., Ortwine, D. F., Pavlovsky, A. G., Picard, J. A., Sliskovic, D. R., Roth, B. D., … Hallak, H. (2000). Structure-activity relationships and pharmacokinetic analysis for a series of potent, systemically available biphenylsulfonamide matrix metalloproteinase inhibitors. Journal of Medicinal Chemistry, 43(2), 156–166. Olejniczak, E. T., Hajduk, P. J., Marcotte, P. A., Nettesheim, D. G., Meadows, R. P., Edalji, R., … Fesik, S. W. (1997). Stromelysin inhibitors designed from weakly bound fragments: effects of linking and cooperativity. Journal of the American Chemical Society, 119(25), 5828–5832. Pavlovsky, A. G., Williams, M. G., Ye, Q. Z., Ortwine, D. F., Purchase, C. F., 2nd, White, A. D., … Blundell, T. L. (1999). X-ray structure of human stromelysin catalytic domain complexed with nonpeptide inhibitors: implications for inhibitor selectivity. Protein Science, 8(7), 1455–1462. Philips, N., Dulaj, L., & Upadhya, T. (2009). Cancer cell growth and extracellular matrix remodeling mechanism of ascorbate; beneficial modulation by P. leucotomos. Anticancer Research, 29(8), 3233–3238. Philips, N., Keller, T., & Holmes, C. (2007). Reciprocal effects of ascorbate on cancer cell growth and the expression of matrix metalloproteinases and transforming growth factor-beta. Cancer Letters, 256(1), 49–55. 10.1016/j.canlet.2007.05.009. Rao, B. G. (2005). Recent developments in the design of specific matrix metalloproteinase inhibitors aided by structural and computational studies. Current Pharmaceutical Design, 11(3), 295–322. Rarey, M., Kramer, B., Lengauer, T., & Klebe, G. (1996). A fast flexible docking method using an incremental construction algorithm. Journal of Molecular Biology, 261(3), 470–489. Reiter, L. A., Mitchell, P. G., Martinelli, G. J., Lopresti-Morrow, L. L., Yocum, S. A., & Eskra, J. D. (2003). Phosphinic acid-based MMP-13 inhibitors that spare MMP-1 and MMP-3. Bioorganic & Medicinal Chemistry Letters, 13(14), 2331–2336. Sastry, G. M., Adzhigirey, M., Day, T., Annabhimoju, R., & Sherman, W. (2013). Protein and ligand preparation: parameters, protocols, and influence on virtual screening enrichments. Journal of Computer-Aided Molecular Design, 27 (3), 221–234. Shapiro, S. D. (1998). Matrix metalloproteinase degradation of extracellular matrix: biological consequences. Current Opinion in Cell Biology, 10(5), 602–608. Shivakumar, D., Williams, J., Wu, Y., Damm, W., Shelley, J., & Sherman, W. (2010). Prediction of absolute solvation free energies using molecular dynamics free energy perturbation and the OPLS force field. Journal of Chemical Theory and Computation, 6(5), 1509–1519. Shlopov, B. V., Lie, W. R., Mainardi, C. L., Cole, A. A., Chubinskaya, S., & Hasty, K. A. (2005). Osteoarthritic lesions. Involvement of three different collagenases. Arthritis and Rheumatism, 40(11), 2065–2074. Sigel, H., & McCormick, D. B. (1970). Discriminating behavior of metal ions and ligands with regard to their biological significance. Accounts of Chemical Research, 3(6), 201–208. Singh, J., Conzentino, P., Cundy, K., Gainor, J. A., Gilliam, C. L., Gordon, T. D., … Wahl, R. C. (1995). Relationship between structure and bioavailability in a series of hydroxamate based metalloprotease inhibitors. Bioorganic & Medicinal Chemistry Letters, 5(4), 337–342. Skiles, J. W., Gonnella, N. C., & Jeng, A. Y. (2001). The design, structure, and therapeutic application of matrix metalloproteinase inhibitors. Current Medicinal Chemistry, 8(4), 425–474. Skiles, J. W., Gonnella, N. C., & Jeng, A. Y. (2004). The design, structure, and clinical update of small molecular weight matrix metalloproteinase inhibitors. Current Medicinal Chemistry, 11(22), 2911–2977. Sternlicht, M. D., & Werb, Z. (2001). How matrix metalloproteinases regulate cell behavior. Annual Review of Cell and Developmental Biology, 17, 463–516. Summers, J. B., Gunn, B. P., Mazdiyasni, H., Goetze, A. M., Young, P. R., Bouska, J. B., … Carter, G. W. (1987). In vivo characterization of hydroxamic acid inhibitors of 5-lipoxygenase. Journal of Medicinal Chemistry, 30(11), 2121–2126.




Tetlow, L. C., Adlam, D. J., & Woolley, D. E. (2001). Matrix metalloproteinase and proinflammatory cytokine production by chondrocytes of human osteoarthritic cartilage: associations with degenerative changes. Arthritis and Rheumatism, 44(3), 585–594. Unemori, E., Hibbs, M., & Amento, E. (1991). Constitutive expression of a 92-kD gelatinase (type V collagenase) by rheumatoid synovial fibroblasts and its induction in normal human fibroblasts by inflammatory cytokines. Journal of Clinical Investigation, 88(5), 1656–1662. Van Lint, P., & Libert, C. (2006). Matrix metalloproteinase-8: cleavage can be decisive. Cytokine & Growth Factor Reviews, 17(4), 217–223. Van Lint, P., Wielockx, B., Puimege, L., Noel, A., Lopez-Otin, C., & Libert, C. (2005). Resistance of collagenase-2 (matrix metalloproteinase-8)-deficient mice to TNF-induced lethal hepatitis. Journal of Immunology, 175(11), 7642–7649. Vincenti, M. P., Coon, C. I., Mengshol, J. A., Yocum, S., Mitchell, P., & Brinckerhoff, C. E. (1998). Cloning of the gene for interstitial collagenase-3 (matrix metalloproteinase-13) from rabbit synovial fibroblasts: differential expression with collagenase-1 (matrix metalloproteinase-1). Biochemical Journal, 331(Pt 1), 341. Visse, R., & Nagase, H. (2003). Matrix metalloproteinases and tissue inhibitors of metalloproteinases: structure, function, and biochemistry. Circulation Research, 92(8), 827–839. 80711.3D. Wallace, A. C., Laskowski, R. A., & Thornton, J. M. (1995). LIGPLOT: a program to generate schematic diagrams of protein-ligand interactions. Protein Engineering, 8(2), 127–134. Welch, W., Ruppert, J., & Jain, A. N. (1996). Hammerhead: fast, fully automated docking of flexible ligands to protein binding sites. Chemistry & Biology, 3(6), 449–462. Whittaker, M., Floyd, C. D., Brown, P., & Gearing, A. J. (1999). Design and therapeutic application of matrix metalloproteinase inhibitors. Chemical Reviews, 99(9), 2735–2776. Woessner, J. F. (1991). Matrix metalloproteinases and their inhibitors in connective tissue remodeling. The FASEB Journal, 5(8), 2145–2154. Woessner, J. F., & Nagase, H. (2000). Matrix metalloproteinases and TIMPs. New York: Oxford University Press.