Transcript presentation

Advances in Computational Modeling of Enzyme
Inhibition: Predicting Potency and Drug Resistance
Targeting HIV-1 Reverse Transcriptase
QM/MM Simulations on Macrophomate Synthase
Diels-Alder vs Michael-aldol Reaction Mechanism
Marina Udier
William L. Jorgensen
Yale University Chemistry Department
Workshop on Continuing Challenges in Free Energy Calculations
Cecam, Lyon May 12-14 2004
Computer-Aided Drug Design
Molecular
biology
gene cloning, protein expression
Computer-Aided
Drug Design
Structural
biology
docking candidates from large databases
>25,000 entries in PDB
information on shape and
size of binding pockets
improving potency, selectivity
binding
de novo ligand design
Lead
Discovery
Lead
Optimization
Clinical
Trials
iterative process between rational ligand
design, synthesis, biological evaluation,
and structure determination
Enzyme Catalysis in a Nutshell
•TS stabilization (electrostatic), Warshel et al.
•Near-attack conformers concept (NAC, preorganization), Bruice et al.
DGnon‡
DGcat‡
DGrxn0
Adults and Children Estimated
to be Living with HIV/AIDS as of end 2003
Western Europe
Eastern Europe &
Central Asia
520 000 – 680 000 1.2 – 1.8 million
790 000 – 1.2 million
East Asia & Pacific
North Africa
700 000 - 1.3 million
& Middle East
Caribbean
South
470 000 – 730 000 & South-East Asia
350 000 - 590 000
4.6 – 8.2 million
Latin America
sub-Saharan
1.3 - 1.9million
Australia
Africa
North America
25.0-28.2 million
& New Zealand
12 000-18 000
Global Total: 34 - 46 million
4.2 – 5.8 million new HIV infections in 2003
2.5 - 3.5 million deaths due to HIV/AIDS in 2003
Cumulative number of deaths due to HIV/AIDS: 30 million
Molecular Structure of HIV
surface glycoprotein gp120
transmembrane glycoprotein gp41
viral envelope
membrane associated
matrix protein p17
capsid protein p24
viral core
RNA
reverse transcriptase
HIV Life Cycle
attachment and fusion
release of genetic information
uncoating
reverse transcription
transcription
integration
translation
envelope processing
extrusion
viral maturation
assembly
© 1999, Johns Hopkins University Division of Infectious Diseases and AIDS Service.
Two Classes of HIV-1 RT Inhibitors
Nucleoside
binding site (NRTI)
Non-nucleoside
binding site (NNRTI)
chemically diverse, hydrophobic, noncytotoxic
fingers
palm
thumb
RNase H
Estimation of Binding Free Energies
BOMB
BOMB score
thousands of cmpds
seconds/cmpd
single conformation
rigid receptor
no solvation or implicit
initial screening
simulation setup
absolute DGbind
ELR
ELR score (RT, kinase)
hundreds of cmpds
day/cmpd
structural diversity
statistical sampling (MC)
explicit solvent
absolute DGbind
structural rationalization
of simulation results
FEP
 statistical perturbation theory
 few cmpds
 days-week/cmpd
 explicit solvent
 relative DDGbind
 detailed interpretation of
computed quantities obtained
by sampling low energy
comfigurations
Metropolis Monte Carlo Method
 is a random
number between (0,1)
1
exp(–DE/RT)
2
Reject move if
2 > exp(–DEi/RT)
1
Accept move if
1 < exp(–DEi/RT)
Reject
Always accept
move if
DE < 0
Accept
0
DE
Figure adapted from Allen, M. P.; Tildesley, D. J. Computer Simulations of Liquids;
Clarendon Press: Oxford, UK, 1987.
DEi
Potential Energy Function (OPLS-AA)
E bond   K r r - ro 
2
Eangle   Kθ θ - θ0 
2
E torsion  
Vn
1  cosnφ - γ 
2
12
6

 σij 
 σij 
qiq je



 
 4ε
-  


r  
rij
 rij 
 ij  

2
E non -bond
E
E
r, 
bond, angle
E

torsion
rij
non-bond
Single Topology Framework FEPs
MD
H
DM
DM
H
H
C H
H
H
H
H
H
H
H
H
H
DG(A→B) = -kBT ln exp[-(EB-EA)/kBT]A
H
Zwanzig
Conversion through many intermediate unphysical states along imaginary reaction
coordinate l, bond-stretching, angle-bending, torsional and non-bonded parameters of
the initial functional group () smoothly mapped into params of the second group as a
function of l so that at any intermediates state between two endpoints l=0 and l=1
they are described by l=l1+(1-l)0.
If change in functionality involves change in the number of atoms, real atoms perturb
to/from dummy (DM) atoms (no non-bonded params).
Stretching and bending params for DM atoms set equivalent to their “real” counterparts.
Good convergence (depends on similarity of initial and final states).
MD
H
DM
DM
H
H
C H
H
H
H
H
H
H
H
H
H
H
“Dual Topology” Framework
O H R
N
H
O
O
N
H
H
N
H
OH
Both topologies present in initial and final states (not interacting with each other).
Stretching and bending parameters for dummy (DM) atoms in both topologies
present throughout the course of the simulation. Non-bonded params (charges
and Lennard-Jones) are gradually disappeared for the first topology and
appeared for the second. Potential energy function same as in single topology.
Allows transformation between any two endpoints (structurally different
functionalities, closed to open ring).
“Dual Topology” Framework Advantage
Bonds to DM Atoms Reduced (0.2-0.4Å)
More gradual introduction of DM atoms to a system
Look into the Future, Automation of in silico Mutagenesis Studies
FoLd resistance Automated ScHeme (FLASH)
Udier-Blagovic, M. and Blagovic, D., FLASH version 1.0
Principal Point Mutations That Confer Resistance to
Non-nucleoside HIV-1 RT Inhibitors
efavirenz in green
L100I
K103N
V106A/I/L
Y181C
Y188C/H/L
Mutations related to drug resistance (even in the absence of selection by a drug) are a consequence of
statistical distribution of mutations along the HIV genome.
Diminished anti-HIV activity against the K103N
HIV-1 RT is attributed to reduced favorable
interactions in the binding pocketb
Tyr181A
Trp229A
Trp229A
Tyr181A
Glu138B
Glu138B
Lys101A
Lys101A
Lys103A (wt)
b Ren
et al., Structure, 2000, 8, 1089-94.
Asn103A (mutant)
“The tale of two crystals”
Coordinates of Efavirenz-K103N RT complexes
pdb entries 1fkob and 1ikvc
Trp229A
Tyr181A
Tyr188A
Glu138B
Lys101A
Asn103A
b Ren
et al., Structure, 2000, 8, 1089-94.
c Lindberg et al., Eur. J. Biochem., 2002, 269, 1670-66.
Objectives
Which structure leads to better agreement with the biological activity
data?
What are the structural origins of resistance of efavirenz and its analogs
caused by K103N mutation?
What are the structural reasons for improved resistance profile for
quinazolinone analogs of efavirenz against K103N mutant?
Cl
F3C
Cl
O
N
H
O
F3C
NH
N
H
O
Thermodynamic Cycle Used to Compute Relative
Fold Resistance Values
NH2
NH2
DGWT
DGA
NH2
O
Inhibitor A
DGB
DGMUT
NH2
O
Inhibitor B
DDGFR(AB) = DGB-DGA=DGMUT-DGWT
Computed Relative Fold Resistance Energies (DDG)
in kcal/mol for K103N HIVRT
DDGFR  RT (ln IC BN103 / IC BK103 - ln IC AN103 / IC AK103 )
Cl
efavirenz
F3C
DPC 083
NH
0.00
NH
N
H
-0.66
0.29, -1.60)e
2.49±0.2f
-2.20±0.3g
-1.2 ± 0.4h
-1.25
0.35, -2.15)e
0.52±0.2f
-2.10±0.3g
-3.5 ± 0.4h
O
F3C
e
0.00
O
F3C
N
H
Cl
calc
O
N
H
Cl
exp activity
O
DPC 961
Experimental relative fold resistance energies (DGFR) estimated from fold resistance (FR) values
normalized to efavirenz.
f, g Computed relative fold resistance energies (DDG) from inhibitor mutations for K103N HIVRT
using crystal structures 1fko and 1ikv.
h
Computed relative fold resistance energies from mutation of a side chain.
Computed Structures from the MC Simulations for
DPC 083 Bound to K103N RT (1ikv, left and 1fko, right)
Tyr181A
Glu138B
Trp229A
Trp229A
Glu138B Tyr181A
2.77Å
1.85Å
1.79Å 2.03Å
2.00Å
Tyr188A
2.00Å
2.27Å
1.91Å
1.66Å
Lys101A
Tyr188A
3.17Å
Lys101A
2.17Å
Asn103A
Asn103A
DPC 083
Indicated Distances Averaged Over MC Simulations.
Computed Structures for DPC961 and Efavirenz Bound to K103N
RT. Water bridge not Observed in WT Efavirenz-RT Complex
Trp229A
Trp229A
Tyr181A
Tyr181A
Glu138B
2.23Å
Tyr188A
1.89Å
Glu138B
Tyr188A
2.16Å
1.72Å
2.46Å
2.47Å
1.67Å
1.66Å
Lys101A
2.66Å
1.71Å
2.40Å
2.18Å
Asn103A
Asn103A
Lys101A
DPC 961
i
efavirenz
M. Udier-Blagović, J. Tirado-Rives, and W. L. Jorgensen J. Med. Chem., 2004, 47, 2389-2392.
“The tale of no Crystal”
TiboTec NNRTI (Etravirine)
TMC125
•
•
•
•
EC50 = 1.4 nM
EC90 = 2.9 nM
CC50 = >100 µM
SI = >71,429
•
•
•
•
EC50(100I) = 3.3 nM
EC50(103N) = 1.2 nM
EC50(181C) = 7 nM
EC50(188L) = 4.6 nM
Rapid decay in viral load observed after only 1 week of therapy in patients with high level of NNRTI
resistance, Gazzard et al., AIDS, 2003, 17, 49-54.
Objectives
Compute a structure for the TMC125/HIVRT complex (BOMB).
Validate the structure through computation of the effects of key
mutations on the binding of TMC125 vs. nevirapine and efavirenz
(MC/FEP).
Elucidate the origins of improved resistance profile for TMC125.
S-1153 (capravirine) in NNRTI binding pocket
Coordinates from 1ep4j
Trp229A
Leu100A
Tyr188A
Lys101A
Tyr181A
Lys103A
S-1153
j Stammers
et al., J. Biol. Chem., 2000, 275, 14316-14320.
TMC125
Proposed Binding Mode for the
TMC125/HIVRT Complex
Trp229A
Tyr318A
Tyr188A
1.86 Å
Lys103A
Tyr181A
3.02 Å
2.21 Å
Glu138B
k
Lys101A
M. Udier-Blagović, J. Tirado-Rives, and W. L. Jorgensen J. Am. Chem. Soc., 2003, 125, 6016-17.
Surface Representation of TMC125/RT Complex
Bound conformation is identical to the global energy minimum from gas
phase conformational search
Computed Relative Fold Resistance Energies (DDG)
in kcal/mol for HIVRT Mutations
H
N
N
DDG (L100I)
O
N
exp actl
N
nevirapine
TMC125
0.00
-0.88
DDG (Y181C)
exp actl
calc
0.00
-2.4±0.4
0.00
-2.55
DDG (K103N)
Cl
F3C
exp actl
O
N
H
0.00
-1.4±0.3
DDG (Y188L)
exp actl
calc
calc
O
efavirenz
TMC125
0.00
-2.28
0.00
-2.0±0.4
0.00
-2.39
- FR
= mutant /WT activities
- DGFR = RT ln FR in kcal/mol
DDGFR  RT (ln IC
l
calc
N 103
B
/ IC
K 103
B
0.00
-2.1±0.3
CH3
Br
O
CN
- ln IC
N 103
A
/ IC
K 103
A
)
CH3
NH2
N
N
NH
CN
Ludovici, D. W.; De Corte, B. L.; Kukla, M. J.; Ye, H.; Ho, C. Y.; Lichtenstein, M. A.; Kavash, R. W.; Andries, K.;
de Bethune, M. P.; Azijn, H.; Pauwels, R.; Lewi, P. J.; Heeres, J.; Koymans, L. M.; de Jonge, M. R.; Van Aken,
K. J.; Daeyaert, F. F.; Das K.; Arnold, E.; Janssen, P. A. Bioorg. Med. Chem. Lett. 2001, 11, 2235-2239.
Convergence in Tyr  Leu FEP for TMC125
Phe  Leu
“dual topology”
DG (kcal/mol)
Tyr  Phe
single topology
l
- Running average after 2M 4M 6M 8M 10M
for Tyr  Phe
- Running average after 5M 10M 15M 20M 30M 40M for Phe  Leu
Structural Explanations of the Improved Resistance
Profile for TMC125
 L100I: hydrogen bonds to K101A and E138B retained.
flexible “U-shaped” conformation allows for accommodation of
Ile g-methyl group.
 Y181C: less specific interactions with Y181A (analogous to efavirenz).
hydrophobic contacts with W229A, Y188A, Y318A.
hydrogen bond to K101A, E138B.
K103N: diminished potency of efavirenz might originate in an unfavorable
electrostatic interactions between Hd2 of N103A and
benzoxazin-2-one ring hydrogen of efavirenz.
average distance between ring hydrogen and backbone carbonyl of
Lys101A is 1.76Å for TMC125 and 2.13Å for efavirenz.
 Y188L: reduces favorable aryl-p interactions.
not drastic size reduction as Y181C, flexibility allows for adjustment.
TMC125 gains more from interactions with L188 than efavirenz.
Balanced flexibility and preorganization for binding
Advances in Computational Modeling of Enzyme
Inhibition: Predicting Potency and Drug Resistance
Targeting HIV-1 Reverse Transcriptase
QM/MM Simulations on Macrophomate Synthase
Diels-Alder vs Michael-aldol Reaction Mechanism
O-
2+
 Macrophomate synthase (MPS) of the
O
Mg
CO2-
-O
Mg2+
STEP 1
-
OMe
O
O
-CO2
O
oxalacetate
Me
O
O
 Examination of the kinetic parameters indicates
the third step of the overall transformation as the
rate-determining step.
Me
42
STEP 2
 MPS is a Mg2+-dependent enzyme with 339
residues for each of the six monomers belonging
to a hexameric structure.
O
OMe
O
Me
HO2C
Me
macrophomate
STEP 3
-CO2
-H2O
O
MeO
OH
Me
phytopathogenic fungus Macrophoma
commelinae catalyzes the transformation of
2-pyrone derivatives into the corresponding
benzoate analogues, like macrophomate.
 The second chemical step occurs via an inverseelectron demand Diels-Alder reaction, making
MPS an example of natural Diels-Alderase.
CO2Me
O
H
Sakurai, I.; Miyajima, H.; Akiyama, K.; Shimizu, S.; Yamamoto, Y. Chem. Pharm. Bull. 1988, 36, 1988, 2003-2011
Watanabe, K.; Mie, T.; Ichihara, A.; Oikawa, H.; Honma, M. J. Biol. Chem. 2000, 49, 38393-38401
Ose, T.; Watanabe, K.; Mie, T.; Honma, M.; Watanabe, H.; Yao, M.; Oikawa, H.; Tanaka, I. Nature 2003, 422, 185-189
Diels-Alder vs Michael-aldol Reaction
Mechanism for Step 2
O-
2+
O
-
Mg
-
2+
Mg
OMe
O
O
Me
Diels-Alder
route
-
O
O
O
CH3
O
O
O
Me
O
OCH3
Michael-aldol route
O-
Mg2+
O
OMe
O
O
Me
-
O
O
Me
O
Quantum Mechanics/Molecular Mechanics (QM/MM)
 electronic structures of atoms involved in the bond-making/breaking
process treated with QM, the rest of the system represented with
non-polarizable force fields.
 advantage: computational efficiency (MM, FE simulations of proteins).
computational accuracy (QM - semiempirical, ab initio, DFT).
 reactions in condensed phase and in enzymes. Pmf calculations follow rxn
coordinate from a reference geometry, yielding free energy barriers for rxns.
 critical issues: treatment of boundary.
 Requirements:
- reproduce correct geometry of the QM part.
- electronic structure (charge distribution and electrostatic potential)
retained in QM fragment.
- torsional PES about the bond connecting QM and MM consistent
with results from pure QM and MM.
Modified QM/MM Interface
H
N
O
H
C
H
R''
O
R
H
N
N
H
H
O
H
C
H
R'
O
N
H
EQM / MM 
  q q
onQM onMM
i
i
j
/ rij  4 ( ij12 / rij12 -  ij6 / rij6 )

j
 The link atoms are hydrogens but their bonds to the QM region are elongated by ~0.5 Å so that the
hydrogens and the C atoms overlap.
 The link atom and C form a single atom, collectively interact with the system.
 The interactions with the QM region are done through the link atoms, included in the SCF calculation.
 The non-bonded interactions with the MM region (treated classically) are also done through the link
atoms, which have quantum mechanical charges obtained from the SCF calculation and LJ parameters of a
carbon atom. C is chargeless and does not have LJ parameters.
 The bonded interactions (bond stretching, angle bending and torsions) with the MM region are defined
through C.
 To avoid duplicating interactions of the QM with the MM region, computed both by bonded and nonbonded force field terms, all 1-2, 1-3, and 1-4 non-bonded interactions between these regions are
excluded.
The Near Attack Conformation (NAC) Concept
Free Energy/kcalmol-1
TS
DGTS
‡
ES
NACs
DG ‡
DG ‡  DGN  DGTS
0
DG N
0
DG N  - RT ln
0
ES
non-NACs
products
Reaction Coordinate
Bruice et al., Acc. Chem. Res., 1999, 32, 127-136.
‡
PNAC
Pnon- NAC
The Near Attack Conformation Concept
‡
DGTS  DGNAC  NAC (min)  DGNAC (min) TS
DG NAC  NAC (min)  - RT ln PNAC (min)
DGallsubset  - RT ln Psubset
Probability of sampling the subset of configurations displaying the free energy
minimum geometry for reaction coordinate R is computed from radial
distribution function g(R) and volume element of the configuration space
corresponding to coordinate R.
Psub(R)=g(R)*4pR2
Bruice et al., Acc. Chem. Res., 1999, 32, 127-136. Kollman et al., Acc. Chem. Res. 2001, 34, 72-79.
Validation of PM3 for QM Region, 28 Atom Model
Bicyclic Product
1.87
1.89
1.91
1.87
2.10
2.09
2.16
2.13
1.91
1.86
PM3
2.21
2.01
B3LYP/3-21+G
Michael Intermediate
PM3
1.86
2.08
2.19
1.86
1.87
1.91
B3LYP/3-21+G
2.11
2.01
2.07
2.2
1.91
1.90
1.59
1.56
Diels-Alder “TS”
Mg Coordination
from crystal structure
Glu185
Asp211
Glu185
Asp211
1.86 1.88
1.87
1.90
1.90
2.08
1.89
2.12
2.17
2.19
2.38
2.27
2.02
2.06
Michael TS
PM3
B3LYP/3-21+G
1.86 1.88
1.91
2.06
2.09
1.87
2.09
1.91
2.22
1.88
2.40
Michael TS
Diels-Alder “TS”
Michael Intermediate
Byciclic Product
PM3
0.0
11.43
The most stable species –32.45
–10.05
2.15
2.08
2.50
B3LYP/6-31+G(d)//3-21+G
0.0
–
–26.59
–17.35
QM Models (77 Atoms)
MPS Binding Site Model Surrounded
by a 22 Å Water Cap (coordinates from 1izc)
2085 atoms
145 residues
794 water molecules
● ●
●
●
●
●
PMF Calculations Inside the MPS
Searching for DA and Michael-aldol TS
Rcc2
2.38
2.02
DA TS
Rcc1
Aldol TS
Michael TS
Bicyclic Product
2.40
Rcc2
Michael
Intermediate
Rcc1
PMF Curve for 1st Step in Michael-aldol Reaction Path
DG/P(Rcc1)*10
2.08
1.60
DG(TS1,2.08;int,1.60)=11.46 kcal/mol
Rcc1/Å
P(1.60, int) = 23.6% ; DGall→1.60= -RT lnP = 0.85 kcal/mol
DG‡total = 11.46 kcal/mol + 0.85 kcal/mol = 12.31 kcal/mol
PMF Curve for 2nd Step in Michael-aldol Reaction
Path
DG/P(Rcc2)*10
2.20
3.48
Rcc2/Å
P(3.48, int) = 10.2% ; DGall→3.48= -RT lnP = 1.35 kcal/mol
DG‡total = 16.57 kcal/mol + 1.35 kcal/mol = 17.92 kcal/mol
-
2+
Mg
-
Free Energy Surface
O
O
O
OCH3
O
O
O
90M SCF calculations
CH3
CH3
-
Mg2+
O
O
OCH3
-d
-
Mg2+
O
O
O
-d
O
O
CH3
-d
O
O
OCH3
O
CH3
O
O
CH3
CH3
-
Mg2+
O
O
Mg2+
-
O
OCH3
O
CH3
Mg2+
O
OCH3
-d
-d
O
-
-
O
O
-
O
O
CH3
O
O
OCH3
O
CH3
-d
O
O
CH3
O
CH3
O
O
CH3
Energetics for Diels-Alder vs Michael-aldol Reaction Path
DATS
12.07
MTS2
5.61
MTS1
12.31
reactants
products
M. intermediate
 Relative fold resistance energies reproduce experimental trends if coordinates from 1ikv
complex are used. Improved resistance profile for quinazolinones DPC961 and DPC083
against K103N HIVRT might originate in a more favorable water bridge between their
additional amido hydrogen and the side chain of E138.
 Binding of TMC125 to HIV RT involves hydrogen bonding to K101 and E138 and extensive
hydrophobic contacts with surrounding residues. Potentially damaging mutations at positions
138 and 318 were not tested and might cause diminished potency of this inhibitor.
Global energy minimum for TMC125 from gas phase MC conformational search is identical
to the bound conformation, indicating negligible reorganization penalty for binding.
 Superior resistance profile of TMC125 against known RT mutants is attributed to
balanced drug flexibility; accommodate mutation while retaining favorable interactions.
Mixed quantum and molecular mechanics (QM/MM) calculations combined with Monte
Carlo simulations revealed that the mechanism for the second step of transformation
catalyzed by MPS might occur via stepwise Michael-aldol rather than concerted Diels-alder
mechanism. The enzyme is not necessarily a Diels-Alderase.
Prof. William L. Jorgensen
Prof. Andrew D. Hamilton
Prof. Martin Saunders
Prof. Karen S. Anderson
Dr. Julian Tirado-Rives
Dr. Jayaraman Chandrasekhar
Dr. Daniel Severance
Dr. Cristiano R. W. Guimarães
Dr. Robert C. Rizzo
Dr. Yukio Tominaga
Dr. Matt Repasky
Ivan Tubert-Brohman
Steve, Dennis, Melissa, Dan, Deping, Al
Patrick, Jakob, Orlando, Juliana, Theresa
Paty Morales
Jorgensen lab
Anderson lab
$$$
National Institute of Allergy and Infectious Diseases