Engineering inhibitors highly selective for the S1 sites of Ser190 trypsin-like serine protease drug targets.

BACKGROUND: Involved or implicated in a wide spectrum of diseases, trypsin-like serine proteases comprise well studied drug targets and anti-targets that can be subdivided into two major classes. In one class there is a serine at position 190 at the S1 site, as in urokinase type plasminogen activator (urokinase or uPA) and factor VIIa, and in the other there is an alanine at 190, as in tissue type plasminogen activator (tPA) and factor Xa. A hydrogen bond unique to Ser190 protease-arylamidine complexes between O gamma(Ser190) and the inhibitor amidine confers an intrinsic preference for such inhibitors toward Ser190 proteases over Ala190 counterparts. RESULTS: Based on the structural differences between the S1 sites of Ser190 and Ala190 protease-arylamidine complexes, we amplified the selectivity of amidine inhibitors toward uPA and against tPA, by factors as high as 220-fold, by incorporating a halo group ortho to the amidine of a lead inhibitor scaffold. Comparison of K(i) values of such halo-substituted and parent inhibitors toward a panel of Ser190 and Ala190 proteases demonstrates pronounced selectivity of the halo analogs for Ser190 proteases over Ala190 counterparts. Crystal structures of Ser190 proteases, uPA and trypsin, and of an Ala190 counterpart, thrombin, bound by a set of ortho (halo, amidino) aryl inhibitors and of non-halo parents reveal the structural basis of the exquisite selectivity and validate the design principle. CONCLUSIONS: Remarkable selectivity enhancements of exceptionally small inhibitors are achieved toward the uPA target over the highly similar tPA anti-target through a single atom substitution on an otherwise relatively non-selective scaffold. Overall selectivities for uPA over tPA as high as 980-fold at physiological pH were realized. The increase in selectivity results from the displacement of a single bound water molecule common to the S1 site of both the uPA target and the tPA anti-target because of the ensuing deficit in hydrogen bonding of the arylamidine inhibitor when bound in the Ala190 protease anti-target.

Exploiting subsite S1 of trypsin-like serine proteases for selectivity: potent and selective inhibitors of urokinase-type plasminogen activator.

A nonselective inhibitor of trypsin-like serine proteases, 2-(2-hydroxybiphenyl-3-yl)-1H-indole-5-carboxamidine (1) (Verner, E.; Katz, B. A.; Spencer, J.; Allen, D.; Hataye, J.; Hruzewicz, W.; Hui, H. C.; Kolesnikov, A.; Li, Y.; Luong, C.; Martelli, A.; Radika. K.; Rai, R.; She, M.; Shrader, W.; Sprengeler, P. A.; Trapp, S.; Wang, J.; Young, W. B.; Mackman, R. L. J. Med. Chem. 2001, 44, 2753-2771) has been optimized through minor structural changes on the S1 binding group to afford remarkably selective and potent inhibitors of urokinase-type plasminogen activator (uPA). The trypsin-like serine proteases(1) that comprise drug targets can be broadly categorized into two subfamilies, those with Ser190 and those with Ala190. A single-atom modification, for example, replacement of hydrogen for chlorine at the 6-position of the 5-amidinoindole P1 group on 1, generated up to 6700-fold selectivity toward the Ser190 enzymes and against the Ala190 enzymes. The larger chlorine atom displaces a water molecule (H(2)O1(S1)) that binds near residue 190 in all the complexes of 1, and related inhibitors, in uPA, thrombin, and trypsin. The water molecule, H(2)O1(S1), in both the Ser190 or Ala190 enzymes, hydrogen bonds with the amidine N1 nitrogen of the inhibitor. When it is displaced, a reduction in affinity toward the Ala190 enzymes is observed due to the amidine N1 nitrogen of the bound inhibitor being deprived of a key hydrogen-bonding partner. In the Ser190 enzymes the affinity is maintained since the serine hydroxyl oxygen O gamma(Ser190) compensates for the displaced water molecule. High-resolution crystallography provided evidence for the displacement of the water molecule and validated the design rationale. In summation, a novel and powerful method for engineering selectivity toward Ser190 proteases and against Ala190 proteases without substantially increasing molecular weight is described.

Development of serine protease inhibitors displaying a multicentered short (<2.3 A) hydrogen bond binding mode: inhibitors of urokinase-type plasminogen activator and factor Xa.

Novel scaffolds that bind to serine proteases through a unique network of short hydrogen bonds to the catalytic Ser195 have been developed. The resulting potent serine protease inhibitors were designed from lead molecule 2-(2-hydroxyphenyl)1H-benzoimidazole-5-carboxamidine, 6b, which is known to display several modes of binding. For instance, 6b can recruit zinc and bind in a manner similar to that reported by bis(5-amidino-2-benzimidazolyl)methane (BABIM) (Nature 1998, 391, 608-612).(1) Alternatively, 6b can bind in the absence of zinc through a multicentered network of short (<2.3 A) hydrogen bonds. The lead structure was optimized in the zinc-independent binding mode toward a panel of six human serine proteases to yield optimized inhibitors such as 2-(3-bromo-2-hydroxy-5-methylphenyl)-1H-indole-5-carboxamidine, 22a, and 2-(2-hydroxybiphenyl-3-yl)-1H-indole-5-carboxamidine, 22f. Structure-activity relationships determined that, apart from the amidine function, an indole or benzimidazole and an ortho substituted phenol group were also essential components for optimal potency. The affinities (K(i)) of 22a and 22f, for example, bearing these groups ranged from 8 to 600 nM toward a panel of six human serine proteases. High-resolution crystal structures revealed that the binding mode of these molecules in several of the enzymes was identical to that of 6b and involved short (<2.3 A) hydrogen bonds among the inhibitor hydroxyl oxygen, Ser195, and a water molecule trapped in the oxyanion hole. In summation, novel and potent trypsin-like serine protease inhibitors possessing a unique mode of binding have been discovered.

A novel serine protease inhibition motif involving a multi-centered short hydrogen bonding network at the active site.

We describe a new serine protease inhibition motif in which binding is mediated by a cluster of very short hydrogen bonds (<2.3 A) at the active site. This protease-inhibitor binding paradigm is observed at high resolution in a large set of crystal structures of trypsin, thrombin, and urokinase-type plasminogen activator (uPA) bound with a series of small molecule inhibitors (2-(2-phenol)indoles and 2-(2-phenol)benzimidazoles). In each complex there are eight enzyme-inhibitor or enzyme-water-inhibitor hydrogen bonds at the active site, three of which are very short. These short hydrogen bonds connect a triangle of oxygen atoms comprising O(gamma)(Ser195), a water molecule co-bound in the oxyanion hole (H(2)O(oxy)), and the phenolate oxygen atom of the inhibitor (O6'). Two of the other hydrogen bonds between the inhibitor and active site of the trypsin and uPA complexes become short in the thrombin counterparts, extending the three-centered short hydrogen-bonding array into a tetrahedral array of atoms (three oxygen and one nitrogen) involved in short hydrogen bonds. In the uPA complexes, the extensive hydrogen-bonding interactions at the active site prevent the inhibitor S1 amidine from forming direct hydrogen bonds with Asp189 because the S1 site is deeper in uPA than in trypsin or thrombin. Ionization equilibria at the active site associated with inhibitor binding are probed through determination and comparison of structures over a wide range of pH (3.5 to 11.4) of thrombin complexes and of trypsin complexes in three different crystal forms. The high-pH trypsin-inhibitor structures suggest that His57 is protonated at pH values as high as 9.5. The pH-dependent inhibition of trypsin, thrombin, uPA and factor Xa by 2-(2-phenol)benzimidazole analogs in which the pK(a) of the phenol group is modulated is shown to be consistent with a binding process involving ionization of both the inhibitor and the enzyme. These data further suggest that the pK(a) of His57 of each protease in the unbound state in solution is about the same, approximately 6.8. By comparing inhibition constants (K(i) values), inhibitor solubilities, inhibitor conformational energies and corresponding structures of short and normal hydrogen bond-mediated complexes, we have estimated the contribution of the short hydrogen bond networks to inhibitor affinity ( approximately 1.7 kcal/mol). The structures and K(i) values associated with the short hydrogen-bonding motif are compared with those corresponding to an alternate, Zn(2+)-mediated inhibition motif at the active site. Structural differences among apo-enzymes, enzyme-inhibitor and enzyme-inhibitor-Zn(2+) complexes are discussed in the context of affinity determinants, selectivity development, and structure-based inhibitor design.

CVT-313, a specific and potent inhibitor of CDK2 that prevents neointimal proliferation.

The activity of cyclin-dependent kinase 2 (CDK2) is essential for progression of cells from G1 to the S phase of the mammalian cell cycle. CVT-313 is a potent CDK2 inhibitor, which was identified from a purine analog library with an IC50 of 0.5 microM in vitro. Inhibition was competitive with respect to ATP (Ki = 95 nM), and selective CVT-313 had no effect on other, nonrelated ATP-dependent serine/threonine kinases. When added to CDK1 or CDK4, a 8.5- and 430-fold higher concentration of CVT-313 was required for half-maximal inhibition of the enzyme activity. In cells exposed to CVT-313, hyperphosphorylation of the retinoblastoma gene product was inhibited, and progression through the cell cycle was arrested at the G1/S boundary. The growth of mouse, rat, and human cells in culture was also inhibited by CVT-313 with the IC50 for growth arrest ranging from 1.25 to 20 microM. To evaluate the effects of CVT-313 in vivo, we tested this agent in a rat carotid artery model of restenosis. A brief intraluminal exposure of CVT-313 to a denuded rat carotid artery resulted in more than 80% inhibition of neointima formation. These observations suggest that CVT-313 is a promising candidate for evaluation in other disease models related to aberrant cell proliferation.

Active-site topology of bovine cholesterol side-chain cleavage cytochrome P450 (P450scc) and evidence for interaction of tyrosine 94 with the side chain of cholesterol.

Combining site-directed mutagenesis with analysis of the active-site topology of bovine cholesterol side-chain cleavage cytochrome P450scc (P450scc), we have investigated the roles of tyrosine residues 93 and 94 on substrate binding. Four single mutants (Y93A, Y93S, Y94A, and Y94S) and one double mutant (Y93S/Y94S) were examined. The largest increase in Ks was observed for binding of cholesterol and 25-hydroxycholesterol to the Y94S mutant (approximately 5.5-fold), with a smaller increase (< 2.5-fold) for binding of 22-hydroxycholesterol. Mutation of Y94 thus appears to influence the interaction with cholesterol, 25-hydroxycholesterol, and possibly 22-hydroxycholesterol. Y93 is not involved in binding of 22- and 25-hydroxycholesterol but may interact with cholesterol. The active-site topologies of P450scc and its mutants were probed by reaction with three arylhydrazines. The N-arylprotoporphyrin IX regioisomer patterns obtained with phenyl- and 2-naphthylhydrazine indicate that the active site is primarily open above pyrrole ring A and suggest that a region some distance above pyrrole ring D is also open. The single mutations Y93S, Y93A, Y94A, and Y94S do not detectably alter the regioisomer patterns obtained with the phenyl- and 2-naphthyl probes, but a small, reproducible change is observed with the 2-naphthyl probe for the Y93S/Y94S double mutant. The conformational alteration implied by this change could not be detected by titration with 22- and 25-hydroxycholesterol but is detectable by titration with cholesterol. The results indicate that cholesterol binds over pyrrole ring D of the heme in bovine P450scc, strongly suggest that Y94 interacts with the side chain of cholesterol, and provide evidence that the side chains of 22- and 25-hydroxycholesterol bind to a different region of the active site than the side chain of cholesterol.

Relationship of active site topology to substrate specificity for cytochrome P450terp (CYP108).

Earlier studies have shown that the reactions of cytochrome P450 with arylhydrazines yield aryl-iron complexes, and that oxidative migration of the aryl groups to the pyrrole nitrogens of the heme provides information on the active site topology. Comparison of cytochromes P450terp (CYP108), P450cam (CYP101), and P450BM-3 (CYP102) by this method suggests that the active site of P450terp is effectively more sterically restricted than those of the other two enzymes and is primarily open above pyrrole ring D of the heme group. This experimental model of the P450terp active site differs from that deduced by x-ray crystallography, which shows that pyrrole ring C is also relatively open. The results suggest that aryl shifts can be used to probe conformations of the active site other than that trapped in the crystal state. Identification of the product formed from alpha-terpineol by P450terp shows that the enzyme exclusively hydroxylates the most sterically accessible, allylically activated position. The enzyme also oxidizes substituted thioanisoles and styrenes unrelated to alpha-terpineol to the corresponding sulfoxides and epoxides. In the case of 4-methylthioanisole and 4-methylstyrene, methyl hydroxylation competes effectively with sulfoxidation and epoxidation in the reaction catalyzed by P450terp but not those catalyzed by P450BM-3 or P450cam. Comparison of the stereoselectively of thioanisole sulfoxidation and styrene epoxidation by P450terp, P450cam, and P450BM-3 shows that P450terp is the most, and P450BM-3 the least, stereospecific. The stereospecificity of thioanisole sulfoxidation by P450terp depends on the electronic nature of the para-substituent and rises from an (R):(S) ratio of 20:80 for p-MeO to a value of < 01:99 for p-CN. The (R):(S) ratio for the epoxides produced by P450terp is approximately 90:10 for the two substituents investigated. Cytochromes P450cam and P450BM-3 are much less stereoselective. A model is suggested by the stereochemical and topological data for the binding of substrates in P450terp.