pharmacophore and counter pharmacophore ...

6 downloads 0 Views 775KB Size Report
imported into CATALYST version 4.10 (Accelrys: Bur- lington; MA) and conformational models of all molecules for TACE and MMP-1 were generated using ''best ...
Strategy for generation of new TACE inhibitors: pharmacophore and counter pharmacophore modeling to remove nonselective hits Malkeet Singh Bahia & Om Silakari

Medicinal Chemistry Research ISSN 1054-2523 Volume 20 Number 6 Med Chem Res (2011) 20:760-768 DOI 10.1007/ s00044-010-9385-3

1 23

Your article is protected by copyright and all rights are held exclusively by Springer Science+Business Media, LLC. This e-offprint is for personal use only and shall not be selfarchived in electronic repositories. If you wish to self-archive your work, please use the accepted author’s version for posting to your own website or your institution’s repository. You may further deposit the accepted author’s version on a funder’s repository at a funder’s request, provided it is not made publicly available until 12 months after publication.

1 23

Author's personal copy MEDICINAL CHEMISTRY RESEARCH

Med Chem Res (2011) 20:760–768 DOI 10.1007/s00044-010-9385-3

ORIGINAL RESEARCH

Strategy for generation of new TACE inhibitors: pharmacophore and counter pharmacophore modeling to remove non-selective hits Malkeet Singh Bahia • Om Silakari

Received: 21 October 2009 / Accepted: 29 June 2010 / Published online: 9 July 2010 Ó Springer Science+Business Media, LLC 2010

Abstract Present study describes the ligand-based molecular modeling along with virtual screening (VS) approach for generation of new tumor necrosis factor-a converting enzyme (TACE) inhibitors. In ligand-based molecular modeling, two statistically reliable HipHop pharmacophore models Hip1 and counter pharmacophore (CP1) were generated using training set of 3 and 2 molecules, respectively. CP1 was generated using inhibitors of MMP-1 (matrix metalloproteinase-1), an important enzyme involved in musculoskeletal degradation. VS was performed with model Hip1 in in-house database of 1.2 million molecules. In addition, the retrieved molecules were screened with CP1. The combination of both models helped for generating new improved TACE inhibitor molecules. Keywords TACE inhibitors  Rheumatoid arthritis  Pharmacophore modeling  Counter pharmacophore

Introduction TNF-a converting enzyme (TACE, ADAM 17, or CD156q) (Black, 2002) is a first ADAM (A disintegrin and metalloprotease containing enzyme) protease to process known physiological substrate and inflammatory cytokine pro TNF-a (tumor necrosis factor alpha) to its mature soluble form sTNF-a. Thus, TACE inhibition receives an attractive approach for affecting circulating sTNF-a level to treat inflammatory diseases. Since last some decades TACE M. S. Bahia  O. Silakari (&) Department of Pharmaceutical Science and Drug Research, Punjabi University, Patiala, Punjab 147002, India e-mail: [email protected]

123

inhibitors have shown good efficacy in treatment of inflammatory bowel disease (IBD) (van Deventer, 2002), RA (rheumatoid arthritis), acute pancreatitis, osteoarthritis, stroke, Crohn’s disease, and reduce brain ischemic injury, etc. (Amin, 1999; Kim and Remick, 2007; Lucas et al., 2006; Maeda et al., 2003). Therefore, currently in pharmaceutical industry inhibition of TACE to block overproduction of TNF-a is an important therapeutic target for inflammatory disorders (Levin, 2004; Nelson and Zask, 1999). TNF-a (Aggarwal et al., 1985; Bemelhans et al., 1996) is a multifunctional pro-inflammatory cytokine involved in inflammation, apoptosis, cell survival, and immunity acting via two receptors (Bazzoni and Beutler, 1996; Marcia et al., 2001) TNF-R55 and TNF-R75 but cause severe damage when produced in excess amount. It is also a strong inducer of other pro-inflammatory cytokines (Brennan et al., 1992; Maini et al., 1993; Rink and Kirchner, 1996; Sekut and Connolly, 1996) such as IL-1b, IL-6, and IL-8. It is produced by activated monocytes/macrophages and other type of cells. Pro-inflammatory cytokines are necessary for survival of the human species but as they may cause significant inflammatory damage so their production and secretion needed to be tightly regulated. Hence, targeting TNF-a release serves as potential approach for various disease conditions. At cellular levels TNF-a is synthesized as membrane anchored precursor but not active as such. Two-third of this precursor is released as active soluble form comprising C-terminal into extra cellular space by limited proteolysis. Proteolysis at Ala76–Val77 bond of membrane bound pro form (pro TNF-a) to convert it into active soluble form is supervised by a protease TACE (Black et al., 1997; Letavic et al., 2002; Moss et al., 1997). Therefore, of the many possible strategies to regulate TNF-a production, TACE inhibition has long been

Author's personal copy Med Chem Res (2011) 20:760–768

viewed as one of the most promising molecular targets (Newton et al., 2001). TACE shares many active site similarities with many other matrix metalloproteinases (MMPs) and ADAMs (Lukacova et al., 2005; Maskos et al., 1998). As a result, many of the early TACE inhibitors suffered from a lack of selectivity (Renkiewicz et al., 2003). This broad spectrum MMP inhibition has been the suspected cause of musculoskeletal side effects observed with developed inhibitor molecules. In present study, ligand-based molecular modeling approach was applied to develop new selective TACE inhibitors. The ligand-based HipHop pharmacophore modeling approach (Ananthula et al., 2008) is able to reveal the common chemical structural feature requirement of multiple ligands and is not restricted to the bound conformation of the ligand in the crystalline complex. Further, highly reliable pharmacophore model was employed for virtual screening (VS) (Greene et al., 1994; Opera, 2002) in in-house database to find new molecules having improved TACE inhibitory activity. Finally, during VS an additional validated screening filter, i.e., counter pharmacophore (Ananthula et al., 2008) model for MMP-1 inhibitors was used to achieve improved inhibitors for TACE.

Materials and methods Data set For modeling study, a data set of molecules having inhibitory activity against TACE and MMP-1 was selected from the literature (Levin et al., 2001a, b, 2002, 2003, 2005, 2006; Levin, 2004; Nelson et al., 2002; Yamamoto et al., 1998; Zask et al., 2003; Zask et al., 2005). For the generation of pharmacophore models for TACE and MMP1, training set of 3 and 2 molecules was selected on the basis of structural diversity (Table 1). Molecular modeling The geometry of each compound was built using builder module of Cerius2 version 4.10 (Accelrys: Burlington; MA) on a silicon graphics Octane2 workstation. All the structures were minimized using smart minimizer after applying CFF95 force field. All the molecules were imported into CATALYST version 4.10 (Accelrys: Burlington; MA) and conformational models of all molecules for TACE and MMP-1 were generated using ‘‘best quality’’ conformational search option within the Catalyst ConFirm module using the ‘‘Poling algorithm’’ (Smellie et al., 1995). A maximum of 250 conformations were generated for each compound to ensure maximum coverage in the conformational space within an energy threshold of

761

20.0 kcal/mol above the global energy minimum. For the generation of Catalyst pharmacophore not only a single lowest energy conformation was used, but also all conformational models for each molecule in training set was used. Hence, this search will probably identify the best 3D arrangement of chemical functionalities explaining the activity variations among the training set. Pharmacophore modeling CATALYST includes two separate pharmacophore modeling modules, namely: HypoGen and HipHop. HypoGen enables automatic pharmacophore construction by using a collection of at least 16 molecules with bioactivities spanning over four orders of magnitude. On the other hand, HipHop generates common feature pharmacophores regardless to the activities of the training set compounds. Still, both modules generate 3D pharmacophres that can be used as search queries to virtually screen 3D structures. However, in present study, TACE inhibitors were having activity in range of two orders. Hence, we avoided using HypoGen for generation of pharmacophore, which requires the bioactivity spread of inhibitors to be in four logarithmic units. HipHop identifies 3D spatial arrangements of chemical features that are common to active molecules in training set. These configurations are identified by a pruned exhaustive search, starting from small sets of features and extending them until no longer common configuration is found. Active training set members are evaluated on the basis of the types of chemical features they contain, along with the ability to adopt a conformation that allows those features to be superimposed on a particular configuration. The user defines how many molecules must map completely or partially to the hypothesis via the ‘‘Principle’’ and ‘‘MaxOmitFeat’’ parameters. These options allow broader and more diverse hypotheses to be generated. HipHop generates 10 hypotheses in a single run and hypotheses are ranked as they are built. The ranking is measure of how well the active training molecules map onto the proposed pharmacophores, as well as the rarity of the pharmacophore model. If a particular pharmacophore is rare then it will less likely map to inactive molecules and therefore will be given higher rank. Pharmacophore validation Ligand-based pharmacophore modeling is a theoretical technique so generated model must be statistically validated. In this study, molecules biological activity range is of two log orders so HipHop pharmacophore modeling approach was used. For the model generation training set comprises of three molecules along with their conformational space was submitted to HipHop module of Catalyst.

123

Author's personal copy 762

Med Chem Res (2011) 20:760–768

Table 1 Training set molecules for pharmacophore models

Model Hip1 generated for TACE inhibitors Sr. No.

Structures of molecules

IC 50 (nM) Principle MaxOmitFeat

HO NH

1

O O N S O

S

3

2

0

4

1

1

4

1

1

O

O N

2

H HO N N O

HO NH

3

O

O S O

O

O HN S O

O

Model CP1 generated for MMP-1 inhibitors (counter pharmacophore)

O

H N

HOHN

4

O

O

5

O

H N

HOHN O

S

N H

0.3

2

0

N H

0.4

1

1

O

S

Ten chemical featured pharmacophoric spaces were resulted in a single run and best model (Hip1) was selected on the basis of their assigned ranks. Selected best model (qualitative) has to be used for virtually screen the database to generate new improved TACE inhibitors so it needed to be validated for its ability to successfully pick the active TACE inhibitor molecules. For validation of best model to identify active and inactive molecules, an in-house data set of 1,000 molecules was prepared which was spiked with known TACE inhibitor molecules. Then model Hip1 was challenged to

123

identify molecules from the total data set. Percentage retrieval of active molecules and enrichment factor was calculated. Enrichment factor was calculated using Eq. 1. E ¼ Ha=Ht  A=D;

ð1Þ

where Ht is the number of retrieved hits, Ha is the number of actives in the hits retrieved, A is the number of active molecules present in the database, and D is the total number of molecules in the database. The ability of model to well retrieve active molecules from spiked database confirms high reliability of model to

Author's personal copy Med Chem Res (2011) 20:760–768

be used for VS. Hence, it is supposed that validated model will successfully pick active molecules when used as 3D query for VS the database. Counter pharmacophore Broad spectrum MMP inhibition of early TACE inhibitors has been reported as suspected cause of musculoskeletal side effects observed with developed inhibitor molecules. Among MMPs, MMP-1 inhibition is reported as main cause of this side effect in literature. Therefore, it was aimed in study to remove molecules capable of MMP-1 inhibition from molecule pool retrieved VS the main database with Hip1, to generate improved molecules. Hence, a counter pharmacophore model (CP1) was generated for MMP-1 inhibitor molecules. Qualitative model CP1 was generated with training set of two highly active MMP-1 inhibitor molecules (4 and 5) using ‘‘HipHop’’ module of CATALYST software. However, in our study, MMP-1 inhibitors were having activity in range of two orders. Hence, we avoided using HypoGen for generation of pharmacophore, which requires the bioactivity spread of inhibitors to be in four logarithmic units. HipHop identifies 3D spatial arrangements of chemical features that are common to active molecules in training set. Training set molecules along with their conformations were used to generate hypotheses. Ten hypotheses were generated in a single run. Among 10 hypotheses, best hypothesis was selected on the basis of pharmacophore rank. Further, model was validated for its ability to pick highly active and leave least active MMP-1 inhibitory molecules from known database. Virtual screening An in-house database of about 1.2 million compounds was used in the VS process. The validated pharmacophore model Hip1 was used as 3D query for performing VS in database. The ‘‘best flexible search’’ method in CATALYST is used for the searching of database. The hits obtained were further filtered using Lipinski’s rule of five and with fit score of zero. Finally, hits were subjected to counter pharmacophore model for screening to get selective compounds.

Results and discussion Pharmacophore study A total data set of 198 molecules having inhibitory activity spanning over two log orders (3–1,000 nM) against TACE was collected from literature (Levin et al., 2001a, b, 2002,

763

2003, 2004, 2005, 2006; Nelson et al., 2002; Zask et al., 2003, 2005). Biological activity data for all molecules was calculated using same assay method (Jin et al., 2002). The three dimensional structures were constructed and energy minimized using cerius2 programming package version 4.10. The structures were imported into Catalyst software. To generate model conformational space of each inhibitor molecule was extensively sampled using poling algorithm employed within ConFirm module of Catalyst. A set of structurally different molecules (three molecules 1–3) was selected as training set for HipHop pharmacophore modeling. Training set molecules associated with their conformations was submitted to HipHop module of Catalyst and it was instructed to explore up to four featured pharmacophoric space for the following possible features: hydrogen bond acceptor (HBA), hydrogen bond donor (HBD), hydrophobic (HY), and ring aromatic (RA). Furthermore, the number of features of any particular type was allowed to vary from 0 to 4. In case of highly active molecule, 2 and 0 were assigned as Principle and MaxOmitFeat values, respectively, and molecule 1 was considered as highly active whereas molecules 2 and 3 were considered as moderately active and value 1 was assigned as Principle and MaxOmitFeat values. Ten hypotheses were generated as a result of single HipHop run. Among 10 models Hip1 was selected as best hypothesis on the basis of highest rank of 36.30 and its ability to pick maximum number of known active hits spiked in database. Model Hip1 consisted of four features, i.e., two HBA, one HBD, and one HY features. Stereoview of Hip1 along with its intra feature distances is shown in Fig. 1. Mapping of highest active molecule 1 of training set over Hip1 is shown in Fig. 2. HBD group of model is mapped to OH of hydroxamic acid group of compound, two HBA groups are mapped to sulphur of thiomorpholine ring and ether ‘‘O’’ of molecule, and HY group is mapped to terminus of but-2-ynyl chain. Ability (validation) of Hip1 to identify active TACE inhibitory molecules was checked using a data set of 1,000 (D) molecules which was spiked with 195 (A) known TACE inhibitor molecules. In 195 known TACE inhibitors 96 are highly active against TACE (3–50 nM) and 99 are moderate to inactive range (51–1,000 nM). Hip1 was challenged to identify this in-house database of 1,000 molecules, as result 213 molecules were retrieved as hits (Ht). Among these hits 179 were known actives (96 highly actives and 83 moderate to inactive). Thus, an enrichment factor (as per Eq. 1) was found to be 4.31, indicating that it is 4.31 times more probable to pick an active compound from the database than an inactive one. Hip1 retrieved 91.79% of actives from the database indicating that it is able to identify the active TACE inhibitor molecules. The drawback of this model is: it also retrieved 34 false positives.

123

Author's personal copy 764

Fig. 1 Top scoring HipHop pharmacophore Hip1: (a) stereoview of ˚ ). The chemical structure features and (b) intra feature distances (A hypothesis features are: hydrogen bond acceptor (HBA); hydrophobic (HY); hydrogen bond donor (HBD)

Med Chem Res (2011) 20:760–768

HipHop module of CATALYST and it was instructed to explore up to five featured pharmacophoric space for the following possible features: HBA, HBD, HY, and RA. Furthermore, the number of features of any particular type was allowed to vary from 0 to 5. In case of highly active molecule, 2 and 0 were assigned as Principle and MaxOmitFeat values, respectively, and molecule 4 was considered as highly active whereas molecule 5 was considered as moderately active and value 1 was assigned as Principle and MaxOmitFeat values. Ten hypotheses were generated as a result of single HipHop run. Best hypothesis (CP1) was selected on the basis of highest rank of 26.03 among generated hypotheses. Hypothesis CP1 consists of five features, i.e., two HBA and three HY features. Stereoview of CP1 along with its intra feature distances is shown in Fig. 3. For validation of CP1 to identify active and inactive molecules, a data set of 100 molecules of known MMP-1 inhibitory activity was prepared. This data consisted of 25 highly active (1–100 nM), 56 moderate active (100–10,000 nM) and 19 least active ([10,000 nM) molecules. When CP1 was challenged to identify molecules from the data set it picks 80 molecules as active hits. Among picked 80 molecules, 22 were from highly active range (88% retrieval), 39 were from moderately activity range (69.6% retrieval) and there was no molecule retrieved from least active range. These data shows that there is high reliability on CP1 to pick active

Fig. 2 Pharmacophore mapping of the most active training set compound 1 onto Hip1

These data shows that there is high reliability on Hip1 to pick active molecules from molecule database. Hence, it is supposed that it will successfully pick active TACE inhibitor molecules when be used as 3D query searching in-house molecule database. Validation of counter pharmacophore model For the generation of counter pharmacophore two highly active molecules 4 and 5 against MMP-1 were selected from literature (Yamamoto et al., 1998). Both molecules associated with their conformations were submitted to

123

Fig. 3 Top scoring HipHop counter pharmacophore CP1: (a) stereoview of chemical structure features and (b) intra feature ˚ ). The hypothesis features are: hydrogen bond acceptor distances (A (HBA); hydrophobic (HY)

Author's personal copy Med Chem Res (2011) 20:760–768

765

(highly and moderate active) molecules and refuse least active molecules from database. Hence, it is supposed that it will successfully pick active molecules when used as counter pharmacophore model onto VS retrieved data and improved molecules against TACE can be obtained as final hits. Virtual screening For the VS pharmacophore model Hip1 was employed as 3D search query to screen the in-house database (1.2 million compounds). Initially, Hip1 was used for VS in database, and then molecules retrieved by this model were further screened with counter pharmacophore (CP1) to avoid the interactions of retrieved molecules with MMP-1, which leads to side effects. Total 689 molecules were retrieved as hits from VS employing pharmacophore Hip1. Hits are defined as those compounds that possess chemical groups that spatially overlap (map) with corresponding features within the pharmacophore model. The hits were subsequently fitted against Hip1 and those having fit values of zero were excluded from subsequent processing. A low fit value indicates that although the chemical features of particular hit overlap with the corresponding pharmacophoric features, the centers of the functional groups (of the hit) are displaced from the centers of the corresponding pharmacophoric features. Furthermore, 689 molecules retrieved from Hip1 were screened with CP1. Finally, 176 molecules were generated as selective inhibitors for TACE querying the database.

interactions with protein TACE were docked into the active site of protein MMP-1 (PDB ID 996C) to remove the molecules showing interactions with protein MMP-1. This combined docking filter remove all the false positives and finally provided the molecules having interactions with active site amino acids of protein TACE but not showing interactions with active site amino acids of protein MMP-1. Thus, the finally retrieved 57 molecules will be active and selective for TACE over MMP-1. Structure of best and highly selective molecule Hip1-1 (mandatory code) is shown in Table 2. Pharmacophore mapping of molecule Hip1-1 with Hip1 and CP1 is shown in Fig. 4 and the docking of same molecule with active site

Validation of virtually screened molecules Those 176 molecules which are retrieved by VS may have false positives. Therefore, these molecules were further subjected to validation using docking method (Software used for docking is Discovery Studio 2.1). The retrieved 176 molecules were docked into the active pocket of protein TACE (PDB ID 3edz) to remove the molecules showing false TACE activity. Then, the selected molecules showing

Table 2 Best hit obtained from virtual screening employing pharmacophore model Hip1

Fig. 4 Overlapping of virtual screening hit molecule Hip1-1. (a) Onto Hip1 and (b) onto CP1

Code of molecules

Structures of retrieved hits

Hip1-1

H2N

O

Estimated fit value with Hip1

O

3.417

H N O

CH3

123

Author's personal copy 766

Med Chem Res (2011) 20:760–768

Fig. 5 Docking of virtual screening hit molecule Hip1-1. (a) With protein TACE (PDB ID 3edz) and (b) with protein MMP-1 (PDB ID 966C)

amino acids of protein TACE and MMP-1 are shown in Fig. 5. The whole VS scheme is mentioned in Fig. 6.

Concluding remarks A statistically consistent and highly reliable HipHop pharmacophore model Hip1 is generated consisted of four structural features, i.e., two hydrogen bond acceptor, one HBD and HY each. Hip1 is well validated for its ability to identify active TACE inhibitor molecules. Furthermore, a counter pharmacophore model CP1 is generated for MMP-1

123

inhibitors to remove the molecules with MMP-1 inhibitory activity from the VS retrieved molecular pool employing Hip1. The molecular pool retrieved by employing both pharmacophores was further validated by docking method to remove the false positives. This combined study of two pharmacophore models and docking with TACE and MMP1 explores important information on requirements for activity and improvement/selectivity of TACE inhibitors. Therefore, this knowledge can successfully be used in two ways to generate new molecules with improved TACE inhibitory activity: (1) for designing of novel compounds and (2) as 3D query in molecule database for VS.

Author's personal copy Med Chem Res (2011) 20:760–768

Fig. 6 Step by step virtual screening scheme to retrieve active and selective inhibitor molecules for TACE over MMP-1 Acknowledgments The authors thank Dr. A. K. Tiwari, Head, Department of Pharmaceutical Science and Drug Research, Punjabi University, Patiala, for his steady advice and Dr. J. A. R. P. Sarma, Director, Bioinformatics Division, GVK Bioscience Pvt. Ltd., for providing software facilities and giving a great chance to work there.

References Aggarwal BB, Kohr WJ, Hass PE, Moffat B, Spencer SA, Henzel WJ, Bringman TS, Nedwin GE, Goeddel DV, Harkins RN (1985) Human tumor necrosis factor production, purification and characterization. J Biol Chem 260:2345–2354 Amin AR (1999) Regulation of tumor necrosis factor-a and tumor necrosis factor-a converting enzyme. Osteoarthr Cartil 7:392– 394 Ananthula RS, Mutteneni RK, Pramod AB, Madala KK, Mahmood SK (2008) Strategies for generating less toxic P-selectin inhibitors: pharmcophore modeling, virtual screening and counter pharmacophore screening to remove toxic hits. J Mol Graph Model 27:546–557 Bazzoni F, Beutler B (1996) The tumor necrosis factor ligand and receptor families. N Engl J Med 334:1717–1725 Bemelhans MHA, Tits LJV, Burrman WA (1996) Tumor necrosis factor: function, release and clearance. Crit Rev Immunol 16: 1–11 Black RA (2002) Tumor necrosis factor alpha converting enzyme. Int J Biochem Cell Biol 34:1–5 Black RA, Rauch CT, Kozlosky CJ, Peschon JJ, Slack JL, Wolfson MF, Castner BJ, Stocking KL, Reddy P, Srinivasan S, Nelson N, Boiani N, Schooley KA, Gerhart M, Davis R, Fitzner JN, Johnson RS, Paxton RJ, March CJ, Cerretti DP (1997) A metalloprotease disintegrin that releases tumor necrosis factor alpha from cells. Nature 385:729–733 Brennan FM, Maini RN, Walport M (1992) TNF-a: a pivotal role in rheumatoid arthritis? Br J Rheumatol 31:293–298 CATALYST software, version 4.10; Accelrys, Burlington, MA CDOCKER docking programme (CHARMm-based docker) of Discovery Studio 1.7, Accelrys, Burlington, MA Cerius2 software, version 4.10; Accelrys, Burlington, MA

767 Greene J, Kahn S, Savoj H, Spargue P, Teig S (1994) Chemical function queries for 3D database search. J Chem Inf Comput Sci 34:1297–1308 Jin G, Huang X, Black R, Wolfson M, Raunch C, McGregor H, Ellestad GA, Cowling R (2002) A continuous fluorimetric assay for tumor necrosis factor alpha converting enzyme. Anal Biochem 302:269 Kim J, Remick DG (2007) Tumor necrosis factor inhibitors for treatment of asthma. Curr Allergy Asthma Rep 7:151–156 Letavic MA, Axt MZ, Barberia JT, Carty TJ, Danley DE, Geoghegan KF, Halim NS, Hoth LR, Kamath AV, Laird ER, LoprestiMorrow LL, McClure KF, Mitchell PG, Natarajan V, Noe MC, Pandit J, Reeves L, Schulte GK, Snow SL, Sweeney FJ, Tan DH, Yu CH (2002) Synthesis and biological activity of selective pipecolic acid based TNF-a converting enzyme (TACE) inhibitors. Bioorg Med Chem Lett 12:1387–1390 Levin JI (2004) The design and synthesis of aryl hydroxamic acid inhibitors of MMPs and TACE. Curr Top Med Chem 4: 1289–1310 Levin JI, Chen J, Du M, Hogan M, Kincaid S, Nelson FC, Venkatesan AM, Wehr T, Zask A, DiJoseph J, Killar LM, Skala S, March CJ, Skotnicki JS (2001a) The discovery of anthranilic acid based MMP inhibitors. Part 2: SAR of the 5-position and P1’ groups. Bioorg Med Chem Lett 11:2189–2192 Levin JI, Chen JM, Du MT, Nelson FC, Wehr T, DiJoseph JF, Killar LM, Skala S, Sung A, Sharr MA, Roth CE, Jin G, Cowling R, Di L, Sherman M, Xu ZB, March CJ, Mohler KM, Black RA, Skotcicki JS (2001b) The discovery of anthranillic acid based MMP inhibitors. Part 3: incorporation of basic amines. Bioorg Med Chem Lett 11:2975–2978 Levin JI, Chen JM, Du MT, Nelson FC, Killer LM, Skala S, Sung A, Jin G, Cowling R, Barone D, March CJ, Mohler KM, Black RA, Skotnicki JS (2002) Anthranilate sulfonamide hydroxamate TACE inhibitors. Part 2: SAR of the acetylenic P1’ group. Bioorg Med Chem Lett 12:1199–1202 Levin JI, Chen JM, Cheung K, Cole D, Crago C, Santos ED, Du X, Khafizova G, MacEwan G, Niu C, Salaski EJ, Zask A, Cummons T, Sung A, Xu J, Zhang Y, Xu W, Ayral-Kaloustain S, Jin G, Cowling R, Barone D, Mohler KM, Black RA, Skotnicki JS (2003) Acetylenic TACE inhibitors. Part 1. SAR of the acyclic sulfonamide hydroxamates. Bioorg Med Chem Lett 13:2799–2803 Levin JI, Nelson FC, Santos ED, Du MT, MacEwan G, Chen JM, Ayral-Kaliustian S, Xu J, Jin G, Cummons T, Barone D (2004) Benzodiazepine inhibitors of the MMPs and TACE. Part 2. Bioorg Med Chem Lett 14:4147–4151 Levin JI, Chen JM, Laakso LM, Du M, Du X, Venkatesan AM, Sandanayaka V, Zask A, Xu J, Xu W, Zhang Y, Skotnicki JS (2005) Acetylenic TACE inhibitors. Part 2: SAR of sixmembered cyclic sulfonamide hydroxamates. Bioorg Med Chem Lett 15:4345–4349 Levin JI, Chen JM, Laakso LM, Du M, Schmid J, Xu W, Commons T, Xu J, Jin G, Barone D, Skotnicki JS (2006) Acetylenic TACE inhibitors. Part 3: thiomorpholine sulfonamide hydroxamates. Bioorg Med Chem Lett 16:1605–1609 Lucas ML, Rothwell NJ, Gibson RM (2006) The role of inflammation in CNS injury and disease. Br J Pharmacol 147:S232–S240 Lukacova V, Zhang Y, Kroll DM, Raha S, Comez D, Balaz S (2005) A comparison of the binding sites of matrix metalloproteinases and tumor necrosis factor-a converting enzyme: implications for selectivity. J Med Chem 48:2361–2370 Maeda K, Hitora M, Kimura Y, Inoue K, Kuwata K, Ohmuraya M, Gawa M (2003) The effect of TNF-a converting enzyme inhibitors on cytokine response in acute pancreatitis. Int Congr Ser 1255:173–175 Maini RN, Brennan FM, Williams R, Chu CQ, Cope AP, Gibbons D, Elliot M, Feldmann M (1993) TNF-a in rheumatoid arthritis and

123

Author's personal copy 768 prospects of anti-TNF therapy. Clin Exp Rheumatol 11: S173–S175 Marcia LM, Judith MW, Millard HL, Robert CA (2001) TACE and other ADAM proteases as targets for drug discovery. Drug Discov Today 6:417–426 Maskos K, Fernandez-Catalan C, Huber R, Bourenkov G, Bartunik H, Ellestad G, Reddy P, Wolfson M, Rauch C, Castner B, Davis R, Clark H, Peterson M, Fitzner J, Cerratti D, March C, Paxton R, Black R, Bode W (1998) Crystal structure of the catalytic domain of human tumor necrosis factor-a converting enzyme. Proc Natl Acad Sci USA 95:3408–3412 Moss ML, Jin SLC, Milla ME, Burkhart W, Carter HL, Chen WJ, Clay WC, Didsbury JR, Hassler D, Hoffman CR, Kost TA (1997) Clonning of a disintegrin metalloprotease that posses precursor tumor necrosis factor alpha. Nature 385:733–736 Nelson FC, Zask A (1999) The therapeutic potential of small molecule TACE inhibitors. Expert Opin Invest Drugs 8:383–392 Nelson FC, Santos ED, Levin JI, Chen JM, Skotnicki JS, DiJoseph JF, Sharr MA, Sung A, Killar LM, Cowling R, Jin G, Roth CE, Albright JD (2002) Benzodiazepine inhibitors of MMPs and TACE. Bioorg Med Chem Lett 12:2867–2870 Newton RC, Solomon KA, Covington MB, Deccico CP, Haley PJ, Friedman SM, Vaddi K (2001) Biology of TACE inhibition. Ann Rheum Dis 60:417–426 Opera TI (2002) Virtual screening in lead discovery: a viewpoint. Molecules 7:51–62

123

Med Chem Res (2011) 20:760–768 Renkiewicz R, Qiu L, Lesch C, Un X, Devalaraja R, Cody T, Kaldjian E, Welgus H, Baragi V (2003) Broad spectrum matrix metalloproteinase inhibitor marimastat induced musculoskeletal side effects in rats. Arthr Rheum 48:1742–1749 Rink L, Kirchner H (1996) Recent progress in tumor necrosis factor-a field. Int Arch Allergy Immunol 111:199–209 Sekut L, Connolly KM (1996) Pathophysiology and regulation of TNF-a in inflammation. Drug News Perspect 9:261–269 Smellie A, Teig SL, Towbin P (1995) Poling: promoting conformational coverage. J Comput Chem 16:171–187 van Deventer SJH (2002) A place of TACE. GUT 51:5–6 Yamamoto M, Tsuijishita H, Hori N, Ohishi Y, Inoue S, Ikeda S, Okada Y (1998) Inhibition of membrane type 1 matrix metalloproteinase by hydroxamate inhibitors: an examination of the subsite pocket. J Med Chem 41:1209–1217 Zask A, Gu Y, Albright JD, Du X, Hogan M, Levin JI, Chen JM, Killar LM, Sung A, DiJoseph JF, Sharr MA, Roth CE, Skala S, Jin G, Cowling R, Mohler KM, Barone D, Black R, March C, Skotnicki JS (2003) Synthesis and SAR of bicyclic heteroaryl hydroxamic acid MMP and TACE inhibitors. Bioorg Med Chem Lett 13:1487–1490 Zask A, Kaplan J, Du X, MacEwan G, Sandanayaka V, Eudy N, Levin J, Jin G, Xu J, Cummons T, Barone D, Aayral-Kaloustian S, Skotnicki J (2005) Synthesis and SAR of diazepine and thiazepine TACE and MMP inhibitors. Bioorg Med Chem Lett 15:1641–1645