Maurizio

Targets looking for drugs: a versatile method for the development of structure-based pharmacophoric models Prof. Maurizio Botta, Ph.D. University of Siena

Virtual Screening Application of computational filters to screen a large library of compounds in order to obtain quickly and rationally a restricted number of compounds to be submitted to biological evaluation Virtual Library Design

VL

M

e h C

ay br

k n Ba DB m

id ge

Synthesis

Virtual Screening (pharmacophore, docking, QSAR etc.) Selection of few compounds

Ex n I s A

Available compounds

Purchase

Biological evaluation

Hit identification

What Is A Pharmacophore? “A pharmacophore is the ensemble of steric and electronic features that is necessary to ensure the optimal supramolecular interactions with a specific biological target and to trigger (or block) its biological response”

C.-G. Wermuth et al., Pure Appl. Chem. 1998, 70:1129-1143

How To Build Pharmacophore Models? There are two basic approaches for pharmacophoric model generation, which depend on the available information about the 3D structure of the target under study. unknown known

3D structure of the target

Structure-based pharmacophores (LigandScout, MOE)

Ligand-based Pharmacophores (DISCO, Catalyst, GASP)

Receptor-Based Pharmacophores In view of the rapidly growing number of protein structures that are now available, receptor-based pharmacophore generation methods are becoming more and more used. Almost all of these procedures require the knowledge of the 3D structure of the ligandprotein complex and they convert interaction pattern into pharmacophore point location. (G. Wolber et al. J. Chem. Inf. Model. 2005, 45, 160-169.)

Receptor-Based Pharmacophores These methods have some limitations: • they can not be applied when no compounds targeting the binding site of interest is known • one pharmacophore model accounts for one binding mode.

Receptor-Based Pharmacophores The generation of receptor-based pharmacophores starting from the knowledge of the sole target has received little attention and only few applications have been reported so far. In this context, we have applied a versatile computational procedure able to generate receptorbased pharmacophoric models starting from the analysis of the sole protein target.

Receptor-Based Pharmacophores The Approach Described Here: • Does not require the knowledge of the 3D structure of the ligand-protein complex. • Allows the generation of structure-based pharmacophoric models also for receptor sites which have never been targeted before and for which no ligands are known. • Could be applied together with other techniques such as docking studies, molecular dynamic simulations and conformational analysis. • Could be integrated in a virtual screening procedure to improve the drug discovery process.

Method I step: Calculation of the MIFs (Molecular Interaction Fields) for the binding site (GRID program). The software initially builds a grid over the region of interest, then computes at each grid point the interaction energies between the protein and several probes. At least three different probes should be used: a hydrophobic probe (DRY, hydrophobic probe; C3, methyl group; C1=, aromatic carbon); a hydrogen bond acceptor (HBA) probe (O, that is a sp2 carbonyl oxygen); a hydrogen bond donor (HBD) probe (N1, a neutral flat NH group).

Method II step: The points of minimum of MIFs were calculated, thus identifying the regions of most favorable interaction between each probe and the protein. The files corresponding to each MIF were processed by means of the Minim and Filmap programs ( GRID package), collecting all points within a certain energy value. A threshold of 0 Kcal/mol has been used to exclude points with a positive energy of interaction.

Method III step. Number of points of minimum is generally too high to allow the construction of useful pharmacophoric models, a selection of the most suitable points is necessary. The criteria for the selection are: (1) The interaction energy: for each probe, the points at lower interaction energy should be preferred. (2) The position within the binding site: points located within cavities are more suited than points exposed on the surface.

Method (3) The distance from other minima: - Too close, could not be simultaneously mapped by the ligand moieties (minimum distance between points 1 Å) - Too distant, could force the selection of only large ligands (maximum distance 15 Å).

(4) The presence (close to the point of minimum of N1 and O probes) of a protein functional group able to act as HBA or HBD, respectively.

Method IV step. The coordinates of the minima were converted on pharmacophoric features: - O probe => HBA features - N1 probe => HBD features - hydrophobic probes => hydrophobic features HBD and HBA features were directed toward the atoms of the protein which act as HB acceptors or donors, respectively (Catalyst 4.10.)

Method V step. Excluded volume spheres are added to the pharmacophore to mimic the boundary of the active site. They represent regions that can not be occupied by any ligand portion.

The use of excluded volumes allows to codify information about the shape and the steric hindrance of the binding site itself into the pharmacophore

Outline • Structure-based virtual screening for the identification of potential inhibitors of Mycobacterium tuberculosis thioredoxin reductase. • Molecular modeling, synthesis and biological studies of novel HIV-1 integrase inhibitors. • A target for new antiviral drugs: Inhibition of the Reverse Transcriptase Dimerization.

Outline • Structure-based virtual screening for the identification of potential inhibitors of Mycobacterium tuberculosis thioredoxin reductase. • Molecular modeling, synthesis and biological studies of novel HIV-1 integrase inhibitors. • A target for new antiviral drugs: Inhibition of the Reverse Transcriptase Dimerization.

The Target: Thioredoxin Reductase • Thioredoxin reductase (TrxR) is part of the thioredoxin system, which is responsible for maintaining reducing conditions inside the cell. • TrxR catalyzes the transfer of electrons between NADPH and thioredoxin, promoting catalysis via FAD and a redox-active disulfide. NADPH

NADP

+

FADox

FADred

SH SH

S

S

SH SH

S

Thioredoxin Reductase

S

Reductive cellular processes

Thioredoxin

• In Mycobacterium tuberculosis the glutathione system is absent, thus the thioredoxin system plays a crucial role for the pathogen’s resistence to the oxidative stress exerted by the innate immune response. • No inhibitors of M.tuberculosis TrxR have been reported so far.

The “Ball-And-Socket” Motion

The proper functioning of TrxR involves the switching between two different conformations. The twisting of the domains relative to one another resembles the twisting of the ball within the socket.

NADPH domain

Cys138

Trx Cys138

FAD

FAD

NADPH NADP+

FO conformation

FAD domain

FR conformation

• FO conformation: the redox-active disulfide (Cys135-Cys138) is buried. • FR conformation: the active site disulfide is exposed and can interact with thioredoxin (Trx).

Available Crystallographic Data

• FO conformation of M. tuberculosis TrxR (PDB entry: 2a87) • FO and FR conformation of E. coli TrxR (PDB entry: 1tde and 1f6m) • Sequence identity M. tuberculosis / E. coli TrxR = 45% • Despite the moderate sequence identity, “the overall structure of M. tuberculosis TrxR is quite similar to that of E. coli TrxR” (M. Akif et al. Acta Crystallogr. D Biol. Crystallogr. 2005, 61, 1603-11).

Two Different Strategies Aim of our work Elaborating a structure-based virtual screening protocol to identify compounds able to inhibit the TrxR activity by: 1) Targeting the TrxR/Trx interaction 2) Targeting the NADPH binding site NADPH domain

Cys138

Trx Cys138 FAD FAD NADPH NADP+

FO conformation

FAD domain

FR conformation

1st Approach: Targeting The TrxR/Trx Interaction Aim of our work Elaborating a structure-based virtual screening protocol to identify compounds targeting the exposed active site of TrxR and able to inhibit its activity by affecting the TrxR/Trx interaction, possibly alkylating the active site Cys138.

NADPH domain

Cys138

Trx Cys138 FAD FAD NADPH NADP+

FO conformation

FAD domain

FR conformation

To reach our goal, it was necessary to model the FR conformation of M. tuberculosis TrxR, in order to bring the active site disulfide to the surface of the enzyme.

Computational protocol • Homology modeling of the FR conformation of M. tuberculosis TrxR using the FR conformation of E. coli TrxR as template. • Creation of the Michael acceptor (MA) pharmacophoric feature. • Binding site analysis by means of GRID software and identification of points of minimum in GRID molecular interaction fields for probes DRY, C3, N1 and O. • Design of receptor-based pharmacophoric models based on GRID points of minimum and also containing the MA feature. • Elaboration of a virtual screening procedure based on pharmacophoric models and docking studies.

Definition of the Michael acceptor feature O

O

• Compounds reported to undergo Michael addition by cysteine residues were collected from literature. • Fragments extracted from such compounds were used to create a new pharmacophoric feature, called MA (Michael Acceptor).

N NO2

O N

MA

NO2

O

S

O [C,O,N]

CN

O N N

Fragments (together with corresponding centroids) used for the assembly of the new pharmacophoric feature.

Receptor-based pharmacophoric model Pharmacophoric GRID probes: features:

● Hydrophobic (DRY/Methyl) DRY ● H-bond Methyl acceptor (Carbonyl oxygen) ● H-bond (Flat NH) Carbonyldonor oxygen ● Michael Flat NH acceptor ● Excluded volume

Virtual screening procedure Asinex Database ____ • Parmacophoric search ___________________________ ____ • Rotatable bonds ≤ 10 ____________________________ ____ • Doking energy __________________________________ ____ • Analysis of complexes ____________________________

10 compounds selected

Ezymatic assays

% inhibition 50 45 40 35 30

50 µM

25

100 µM

20 15 10 5

C

TR 9 TR 10

TR 8

TR 7

TR 6

TR 5

TR 4

TR 3

TR 2

on tr ol TR 1

0

2nd Approach: Targeting The NADPH Binding Site NADPH domain

Aim of our work Elaborating a structure-based virtual screening protocol to identify compounds able to inhibit the activity of TrxR by competing with NADPH.

Cys138

Trx Cys138 FAD FAD NADPH NADP+

FO conformation

FAD domain

FR conformation

For this kind of approach, it was possible to use the available crystallographic structure of the FO conformation of M. tuberculosis TrxR, since in this case the binding site of interest was exposed.

Computational Protocol

• Analysis of NADPH binding site by means of GRID software and identification of points of minimum in GRID molecular interaction fields for probes DRY, C3, N1 and O. • Design of receptor-based pharmacophoric models based on GRID points of minimum. • Elaboration of a virtual screening procedure based on pharmacophoric models and docking studies.

Receptor-Based Pharmacophoric Model GRID probes: features: Pharmacophoric

● Hydrophobic DRY Methyl (Methyl) ● H-bond Methyl acceptor Carbonyl oxygen (Carbonyl oxygen) ● H-bond Carbonyl Flat NH donor oxygen (Flat NH) ● Excluded Flat NH volume

Virtual Screening Procedure Asinex Database • Parmacophoric search ___________________________ _____ • Rotatable bonds ≤ 10 ____________________________ _____ • Docking energy _______________________________ _____ • Analysis of complexes ___________________________ _____

9 compounds selected

Virtual Screening Procedure

The structure-based pharmacophoric models were applied as first filter to screen the Asinex Database and led to the selection of 22290 compounds. Next, the number of rotatable bonds and the Best Fit value were used to remove chemicals either too flexible or poorly fitting the pharmacophore models, respectively.

Virtual Screening Procedure

The resulting 9156 compounds were subjected to docking (software GOLD) in two consecutive runs to optimize the balance between the quality of docking and the calculations time: 1st. Search efficiency parameter set to 50% to speed up the process. 2st. Search efficiency was set to 100% on the top 1000 scored hits.

Virtual Screening Procedure

47 compounds for which the binding mode is in agreement with the pharmacophoric features, were chosen. In fact, both pharmacophore and docking are aimed at hypothesizing the bioactive conformation of ligands, so compounds for which a consensus between the two methods occurs should be preferred to those for which do not occur.

Virtual Screening Procedure

9 molecules was finally selected considering: • predicted binding energy • different structural classes

Enzymatic Assays % inhibition 100 90 80 70

TR12, IC50 = 16 µM

60

50 µM

50

100 µM

40 30 20 10

C

on tr ol TR 11 TR 12 TR 13 TR 14 TR 15 TR 16 TR 17 TR 18 TR 19

0

Assay conditions: Compound: 50/100 µM NADPH: 225 µM

Active Compound O O

HN O

N H

O

O

O O

a) Alignment of the active compound (orange) on pharmacophore. b) Binding mode of the active compound within the NADPH binding site of MTB TrxR. There is a good agreement between pharmacophore mapping and docking results, the main difference being the location of the methoxy substituent on the phenyl ring which is not predicted to be involved in any interaction with the binding pocket.

Conclusions

• Structure-based virtual screening procedures for the search of inhibitors of M. tuberculosis TrxR resulted in the identification of TR1 and TR12 as hit compounds. • The candidate compounds inhibited the TrxR activity through two distinct putative mechanisms of action. • The biological activities of TR1 and TR12 make them worth of further development (already in progress).

Outline • Structure-based virtual screening for the identification of potential inhibitors of Mycobacterium tuberculosis thioredoxin reductase. • Molecular modeling, synthesis and biological studies of novel HIV-1 integrase inhibitors. • A target for new antiviral drugs: Inhibition of the Reverse Transcriptase Dimerization.

Anti-HIV-1 Therapy AIDS (Acquired Immune Deficiency Syndrome) is an infection disease caused by a virus known as the Human Immunodeficiency Virus (HIV). Highly active antiretroviral therapy • Nucleoside analog (RT inhibitors) • Non-nucleoside analog (RT inhibitors) • Protease inhibitor drugs • Entry inhibitors

Role Of Integrase The Integrase catalyses two temporally and spatially separated reactions known as 3’-processing and strand transfer. The 3’-processing occurs in the cytoplasm where integrase cleaves a dinucleotide from the 3’-ends of the double-stranded viral DNA. The protein-DNA complex is then transported into the nucleus where the strand transfer reaction takes place and the 3’-ends of the viral DNA are covalently linked to the 5’-ends of the host DNA. 3’ 5’ 5’ 3’

3’ 5’

5’ 3’

3’-processing

3’

5’

Strand transfer

5’

Integrase Inhibitors Strand-transfer inhibitors

3’-processing inhibitors

IC50 3P= 20.5 µM ST= 0.059 µM HIV-1 (IIIB) = 5.5 µM

O

S

O O

O

IC50

OH

β- Diketo acids (L-708, 906)

V-104 (R=Cl) 3P= 0.9±0.5 ST= 201.3±119.7

O

HN

OO

R

NH N H

O

N H

V-165 (R=NO2) 3P= 0.9±0.4 ST= 16.1±0.7

S

Pyrano-dipyrimidines

DNA

HOOC

N OH

Mg2+

OH

IC50

OH

3P= 0.4±0.5

OH

Styrylquinolines Mg2+

Pyrano-dipyrimidines and Styrylquinolines inhibit 3’-processing and display moderate antiviral activity

Aim Of The Work

Lys 156 Lys 159

Loop 140-149

Asn 155 Hys 67

The residues involved in the interaction with the DNA are red. The loop near the active site is represented in green.

Protein engineering and X-ray crystallography have shown that the mobility of the HIV-1 active site loop is correlated with DNA recognition. On this basis, following our intent to discover integrase inhibitors affecting the binding of integrase with viral DNA, we have looked for molecules able to block the loop in an open conformation.

Conformational Analysis Of The Loop Conformational analysis of the loop close to the active site of the enzyme was performed using the software Macromodel. 50000 conformations were generated. Among them, only the conformations with an energy that fell within a 10 Kcal/mol range above the global minimum were kept. Such conformations were submitted to cluster analysis using an rmsd of 0.8 Å (referred to backbone atoms) and the selected ones were used in all the following calculations.

Pharmacophores Generation Protocol • Analysis of the binding site by means of GRID software, taking into account the conformations selected as previously described. • Identification of points of minimum in GRID molecular interaction fields for probes C1=, C3, N1 and O. • Design of receptor-based pharmacophoric models based on GRID points of minimum. • Elaboration of a virtual screening procedure based on pharmacophoric models and docking studies.

Receptor-Based Pharmacophores

Virtual Screening Workflow 1. Receptor based pharmacophoric models 2. Rotatable bonds number < 10

Asinex Gold Collection Database – 200000 - ca. 500 - 10 -1-

5. Biological assay

3.Evaluation of the energy of interaction 4. Docking studies on selected compounds

The pharmacophoric models were in turn used to screen the Asinex Gold Collection Database. Since Hypo1 and Hypo3 failed to retrieve hits, several simplified 5-feature pharmacophores were generated from the original models. About 20 000 compounds were selected. Next, only molecules with less than 10 rotatable bonds were kept while the molecules that poorly mapped the pharmacophoric models (Best Fit value < 1) were removed.

Virtual Screening Workflow 1. Receptor based pharmacophoric models 2. Rotatable bonds number < 10

Asinex Gold Collection Database – 200000 - ca. 500 - 10 -1-

5. Biological assay

3.Evaluation of the energy of interaction 4. Docking studies on selected compounds

Next, the binding energy of the selected compounds was evaluated. Each ligand identified by a particular pharmacophore was superimposed to it and the integraseligand complexes were minimized. Then, the interaction energy between the ligands and the target was calculated through four different scoring functions: the scoring function of Autodock 3.0 and those implemented in the XSCORE package.

Virtual Screening Workflow 1. Receptor based pharmacophoric models 2. Rotatable bonds number < 10

Asinex Gold Collection Database – 200000 - ca. 500 - 10 -1-

5. Biological assay

3.Evaluation of the energy of interaction 4. Docking studies on selected compounds

Finally, docking studies were performed on the retrieved compounds (GOLD software). Ten compounds were chosen for biological investigations on the basis of: • the matching between the compound interactions and the pharmacophoric features; • the consensus score between pharmacophoric matching and docking; • the scaffolds diversity.

Hit compound As a results of the biological investigations, one compound (MAS 10018) was identified showing significant inhibitory potency against IN in overall integration assay. Such compound was also able to inhibit HIV-1 replication in cells. Further biological studies on compound MAS 10018 have shown that it is active both on 3’-processing and strand transfer reactions. MAS 10018 IC50 (overall integration assay) = 9 µM 3P= 25 µM ST= 3 µM EC50 (IIIB)= 30 µM EC50 (NL43wt)= 8 µM

Binding Mode Of The Hit Compound Loop 140-149 Green: HB acceptor Cyan: Hydrophobic Magenta: HB donor Gray: Exclusion volumes

Binding Mode Of The Hit Compound Loop 140-149

Lys 159 Val 150 Glu 152 Ile 151

Pro 142 Hys 67

Ile 141

Conclusions A structure-based virtual screening protocol that taken into account the flexibility of the loop close to the active site was applied to discover new IN inhibitors •

• As a results of the biological investigations, one compound (MAS 10018) was identified showing significant inhibitory potency against IN in overall integration assay. •Such compound was also able to inhibit HIV-1 replication in cells. •Further biological studies on compound MAS 10018 have shown that it is active both on 3’-processing and strand transfer reactions. • Synthetic strategies are in progress to optimize the antiviral activity of the hit compound.

Outline • Structure-based virtual screening for the identification of potential inhibitors of Mycobacterium tuberculosis thioredoxin reductase. • Molecular modeling, synthesis and biological studies of novel HIV-1 integrase inhibitors. • A target for new antiviral drugs: Inhibition of the Reverse Transcriptase Dimerization.

Role Of Reverse Transcriptase Reverse Transcriptase Central role in the viral replication cycle by generating a double-stranded DNA copy of the single-stranded RNA genome. Then, DNA is integrated into the host cell genome to enable expression of viral proteins for the production of new virions.

Reverse Transcriptase Inhibitors Nucleoside inhibitors (NRTIs) Non-nucleoside inhibitors (NNRTIs)

Are Are there there new new possible possible RT RT binding binding sites sites as as target target for for antiviral antiviral therapy? therapy?

Reverse Transcriptase Structure RT is a multifunctional heterodimer consisting of two subunits of identical sequence named p66 p51.

p66 subunit adopts an “open” catalytically competent conformation to accomodate a nucleic acid template.

p66

Thumb

RNase H

Thumb

Fingers Palm

Fingers

p51 subunit is closed and is assumed to play a largely structural role in the heterodimer strucuture.

p51

Reverse Transcriptase Structure p66 subunit contains the catalytic site and both polymerase and RNase H activities. The shorter p51 subunit lacks these functions but it is essential for loading the p66 subunit on the template primer (Harris et al., 1998).

Three major contact areas in the dimer interface: Fingers subdomain of p51

Palm subdomain of p66

Thumb subdomain of p51

RNase H domain of p66

Connection subdomain of p51

Connection subdomain of p66

RT Dimerization Process The p66 and p51 subunits of the mature RT heterodimer are initially expressed as part of larger polyprotein precursor (Pr160Gag-Pol). Two models have been proposed for the formation of RT heterodimer (Sluis-Cremer et al., 2004).

1. Concerted model In the concerted model, the two subunits are cleaved from two separate precursors, followed by their association and assembly into the p66/p51 heterodimer.

Gag-Pol

p66

p51

+

p66/p51

RT Dimerization Process 2. Sequential model Gag-Pol

p66

p66

+

In the sequential model, the p66 subunit is cleaved from separate Gag-Pol molecules, which then forms a p66/p66 homodimer intermediate, prior to formation of the p66/p51 heterodimer. Following the formation of a p66/p66 homodimer intermediate: p66 +

p66/p66

p66/p51

p66

p66/p66 (inactive)

p66/p66 (active)

p66/p51

one of the subunits in the homodimer is then further cleaved to form p66/p51 heterodimer RT by an HIV-1 protease. This protease can cleave one of the two p66 subunits in the homodimer which shows a structural asymmetry with respect to RNase H domain to render this region accessible to the protease in only one of the subunits.

Structural Structural analysis analysis favours favours the the sequential sequential model. model.

RT Connection Subdomain Connection

p66

The first interaction between the two subunits, when the RT is forming, occurs in a TRP-rich hydrophobic cluster located in the connection subdomain of the two subunits.

Connection subdomain: residues 312-425

Trp - - Trp Trp - - - Trp - - - Trp - - - Trp p51

398

401 402

406

410

414

Trp402

- Mutagenesis of Trp401 and Trp414 in p66 impairs dimerization with p51 by altering the proper positioning of structural elements (for example W402 and W410) in between these residues that make contacts with p51. - Mutagenesis of Lys331 in p51 results in impaired RT dimerization.

Trp410 Asn363 Asp364 Lys331 Tyr405

Trp401

Aim Of The Work The dimerization is essential for a fully functional RT and this process offers an excellent opportunity to inhibit the enzyme by a different mechanism by disrupting a key protein-protein interaction.

The aim of this work was to find compounds binding to the connection subdomain of p66 subunit and preventing the coupling between the two subunits.

We proposed to develop structure-based pharmacophoric models for the connection subdomain of p66 to screen a commercial available database and applying docking studies to selected compounds.

p66 Pharmacophoric Models Three different three-dimensional pharmacophoric models of the p66 connection subdomain were generated taking into account the receptor flexibility. 1. p66 subunit was first submitted to Molecular Dynamics simulation.

p66 subunit (taken from the crystallographic structure of the HIV-1 RT, entry 1RTH of the PDB) was first submitted to MD calculations by means of the software package NAMD (version 2.5) with CHARMM27 force field. Hydrogen atoms were added by means of the psfgen package.

The p66 subunit was embedded in a sphere of water molecules (60 Å radius) applying spherical boundary conditions. The starting structure was optimized with 1000 steps of conjugate gradient energy minimization to remove unfavorable contacts.

p66 Pharmacophoric Models MD simulation was carried out at 310 K for 1 ns, collecting snapshot structures every 1 ps. The Langevin Dynamics procedure, with a dumping factor of 5 ps-1, was used to control the temperature. 3,5 3

RMSD

2,5 2 1,5 1 0,5 0 0

100

200

300

400

500

600

Time (ps)

700

800

900 1000

From MD trajectory, several snapshots were chosen on the basis of different conformations of relevant residues of the connection subdomain (in particular, W402 and W410).

K395

W410 W402

RMS fluctuation 0,14 0,12

nm

W410

W402

0,1 0,08 0,06 0,04 0,02 0 380

385

390

395

400

405

Residue number

Green denotes more mobiles residues and red denotes residues which moved less during simulation.

410

415

420

p66 Pharmacophoric Models 2.

For each selected snapshot, molecular interaction fields (MIFs) were computed to describe hydrophobic interactions and hydrogen bond contacts, using three different probes (hydrophobic, DRY; hydrogen bond donor, N1; hydrogen bond acceptor, O). The best interaction points between the selected probes and the residues of the connection subdomain were computed. 3.

The best interaction points identified by computing MIFs with DRY, N1 and O probes (left) were converted, respectively, into hydrophobic, hydrogen bond donor and acceptor features of pharmacophoric models. Excluded-volumes corresponding to residues W398, W402, W410 and W414 were also added (right).

p66 Pharmacophoric Models 4. The different structure-based pharmacophoric models, generated for the selected MD snapshots, were merged into three pharmacophoric hypotheses on the basis of the distance tolerance value. These hypotheses were derived to use as a search query during the next virtual screening procedure.

Frame n. 680

Frame n. 730

Frame n. 780

Frame n. 810

Frame n. 870

Frame n. 760

Database Virtual Screening The ASINEX Gold Collection was screened and compounds satisfying at least one of the pharmacophoric hypotheses and fulfilling Lipinski’s Rule of Five were retrived and docked into the p66 connection subdomain. This approach identified ten compounds that were tested in biological assays.

ca. 200000 cpd Structure-based Pharmacophoric search

ca 51000 Lipinski’s Rule of Five filter Fit value filter

ca 40000 110

10 Biological assays

Docking studies

In Vitro Subunit Association Assay The reassociation of RT subunits was initiated in the absence or presence of MAS compounds by a 15-fold dilution into acetonitrile Free standard buffer with a final enzyme concentration of 1.3 µM and followed by HPLC.

Reassociation-Assay RT Dimer

Dissociation 12 % ACN

RT Subunits

Reassociation

RT Dimer

+ +M

AS

+ RT Subunits

In Vitro Subunit Association Assay Heterodimeric form of HIV-1 RT was reversibly dissociated by the addition of acetonitrile. Upon reduction of the acetonitrile concentration to 0.8 % by a simple dilution of the samples, reassociation of RT was initiated and followed by size-exclusion chromatography or increase of the enzymatic activities. O H 3C O

N

OH

N N N CH3

MAS0

N

Et

N

Et N

S O

OH

MAS1

NH2

Applying this procedure, MAS0 and MAS1 showed a dose-dependent inhibition of the dimerization process. MAS0 proved to be about 5 times more active than MAS1 in this assay.

b) Dose-dependent inhibition of RT reassociation by MAS compounds. Data were fitted to a hyperbolic equation. 50% inhibition was observed at ca. 150 µM for MAS0 (○) and ca. 750 µM for MAS1 ( ). Prior to HPLC size-exclusion chromatography samples were incubated for 16 h at 20ºC.

a) Reassociation reaction of acetonitrile treated HIV-1 RT heterodimers in the absence (□) or presence of MAS0 (○) and MAS1 ( ) compounds (1 mM each).

Inhibition of RT Activities by MAS0 A) Dose-dependent, simultaneous inhibition of both the polymerase (○) and the RNase H activity ( ) by MAS0, yielding IC50 values of 155 and 111 µM, respectively.

B) Effect of RT/MAS0-preincubation time on polymerase (○) and the RNase H activity ( ). Inhibition of RT could only be observed following a preincubation of enzyme and MAS0 with t1/2 of about 2 h. During preincubation no significant reduction of enzymatic activities was observed in the absence of the compound (data not shown).

Specificity of inhibition by MAS0 To analyze the specificity of MAS for HIV-1 enzyme, polymerase assays were performed with HIV-2 RT, Avian Myeloblastosis Virus (AMV) RT, and E. coli DNA Pol I (Klenow fragment, KF).

MAS0 decreased the polymerase activity of HIV-2 by about 60%, whereas the activity of the AMV RT and KF remained unaffected.

The different levels of inhibition could be explained by some minor sequence variation in this region. The likewise heterodimeric AMV RT, on the other hand, is not affected by MAS0. Even though all three RTs (HIV-1, HIV-2 and AMV) do presumably share a similar overall folding (the X-ray structure of AMV RT is not known), only enzymes containing a tryptophan cluster (AMV RT has no such motif) are affected.

Conclusion Taking into account biological assays results, MAS0 is supposed to interfere with the dimer equilibrium by shifting the equilibrium from an active to an inactive dimer by trapping the enzyme in the inactive conformation. As the structures of the intermediate inactive dimer and the monomeric species are not known, the structures shown here for illustration purposes are all derived from the structure of an active heterodimer.

This study represents the first successful rational screen for a small molecule HIV RT dimerization inhibitor, which may serve as an attractive hit compound for the development of novel therapeutic agents.

Acknowledgments • University of Leuven, Belgium: Myriam Witvrouw Zeger Debyser • Universität Lubeck, Germany : Tobias Restle Dina Grohmann • MOLISA GmbH, Magdeburg: Leopold Flohé Timo Jaeger We thank Gabriele Cruciani for the access to the GRID code

!! ! n o i t n e t t a d n i k r u o y r o Th a n k s f

From Eastern Roman Empire…

…to the Western one

Arrivederci in Siena:

VII EWDD Seventh European Workshop in Drug Design May 24 - 30, 2009 – University of Siena (Italy) www.unisi.it/EWDD