Rational Drug Design

Bernd Wendt
EMBL, Heidelberg
Large and small molecules
• Large molecule
• X-ray structure
• Protein
• Target
• “Lock”
• COX-1

• Small molecule
• 3D model
• Ligand
This is not • “Key”
the keyhole! • Aspirin
• surface representation
Large and small molecules
• Look inside the protein
• Display of the protein fold
• Active site in green dots
• Aspirin
• Heme-group

This is the
keyhole!
• ribbon representation
From serendipity to rational drug design: Aspirin

• 400 BC: Hippocrates: extracts from willow tree (containing salicylic acid) for pain
relieve
• 1890’s: synthesis of acetylsalicylic acid (Aspirin)– improvement on side effects
(irritation of stomach)
• 1970’s: Aspirin target identified as cyclo-oxygenase (COX)
• 1980’s: Two forms of COX identified:
• COX-1, constitutive
• COX-2, induced at sites of inflammation
• Mid 1990’s: X-ray structures of COX-1 and COX-2 available
• Rational design of selective COX-2 inhibitors
• 1998/99: Celebrex and Vioxx approved by FDA for osteoarthritis and rheumatoid
arthritis
• Mid 2000’s: Randomized placebo-controlled trial showed an increased risk of heart
attack and stroke
• Vioxx withdrawn from market
• Celebrex with warning label
A detailed look at Aspirin and its target

COX-1 with natural substrate (arachidonic acid) COX-1 inhibited by Aspirin

COX-1 with non-selective inhibitor (fluorbiprofen) COX-2 with selective inhibitor

Drug Discovery Process

Target Target Lead Lead Clinical Market

Identification Validation Identification Optimization Phases I-III Launch

Develop-
Biology Chemistry
ment

• Duration: 12-15 years

• Costs: 500-800 Million $
Rational Approaches: Biological Universe

• Biological Space (large molecules)

• Number of predicted genes:
• ~30.000
• Number of proteins in proteome: 5%2%2%
other
7% Proteases
• 21,688 Protein kinase
10% 42%
GPCR
• Number of estimated druggable Ion channels

targets: 16%
Transporters
Cytochrom P450
16%
• 3,051 NHR

• Number of targets of current

drugs: The druggable genome
• 400-500
Rational Approaches: Chemical Universe
• Chemical Space (small molecules)
• Estimated number of possible drug-like molecules:
• 1060
• Number of synthesized molecules:
• 107
• Typical size of corporate compound collection:
• 105 to 107
• Parallel/Combinatorial Chemistry
• Linking of chemical components A*B*C
• Library sizes of 102 to 104
• Virtual Chemistry
• Size of AllChem database: 1020
Rational approaches: Filtering on properties
• Prescreening of chemicals for biological relevance and
‘druglikeness’
• Use computer-calculated properties for in silico screening
• Lipinski’s rule of ‘5’:
• Poor absorption or permeation more likely when:
• There are more than 5 H-bond donors
• The MWT is over 500
• MLOGP is over 4.15 ( or CLOGP is over 5)
• The sum of N‘s and O‘s is over 10
• Further properties to consider
• Not more than 10 rotatable bonds
• Limited polar surface area
Rational Approaches: Docking

• Use binding pocket from X-

ray structure
hydrophilic
• Define required interactions:
• H-bonds
• Hydrophobic interactions
• Metal- and water atoms
• Find complementary poses of
test compounds
• Score those poses
• Rank all test compounds
hydrophobic
according to score
X-ray structure of human carboanhydrase
Rational Approaches: Docking
• Carboanhydrase: X-ray-
~180.000 compounds structure as template
Filtering • Commercially available
compound database as pool
• 13 compounds selected for
purchase and screening
• 3 subnanomolar, 1 nanomolar
Shape-
and 7 micromolar inhibitors
Matching
found

Proof-of-
Docking
Concept:
X-ray structure
of bound ligand
13 compounds
Rational Approaches: Fragment-based discovery

• Chemical Microarrays
(Graffinity, Heidelberg)
• ~100,000 fragments linked onto
chip
• parallel detection of weak,
specific interactions
• Confirmation of hits by functional
testing and X-ray
• Thrombin-example
• Novel fragment identified that
binds to S1-pocket (usually
occupied by benzamidine-group)
S1
• Confirmed by X-ray
Rational Approaches: Comparative Molecular Field
Analysis (CoMFA)

Contour
Maps

Bio PLS
QSAR
equation
QSAR Table = SYBYL MSS
Rational Approaches: Comparative Molecular Field
Analysis (CoMFA)
• Example Steroids:
• Dataset of 31 ligands
• Measured activity data
(CBG-Affinity)
• CoMFA-steps
• Build molecules in 3D
• Align all molecules on top of
each other
• Calculate fields
• Run statistical analysis
• Display results in from of
contour maps
• Use regression equation for
prediction of untested
compounds
Rational Drug Design at EMBL: Aurora A

• Cell-cycle and cancer:

• Family of 3 related kinases with central role in cell cycle
• Aurora A (localized and activated by TPX2)
• Aurora B
• Aurora C

• Aurora A has elevated expression in

> 50% colorectal , ovarian, gastric cancers
> 95% invasive adenocarcinomas (breast cancer)

Amplification of locus correlates strongly with poor clinical prognosis

 Brief History of Aurora

• 1997: Aurora A gene overexpressed in breast tumors

• 1998: Aurora A established as an oncogene
• 2002: first X-ray structure of inactive Aurora A resolved
• 2003: X-ray structure of activated Aurora A resolved by
Elena Conti et al. (EMBL)
• 2004: Aurora-kinase inhibitor suppresses tumor growth in
vivo
• To-date: most major pharmaceutical companies with
discovery programs on Aurora-kinases targeting ATP site
• Novel approach targeting the allosteric site
 Aurora A: Binding Sites

Aurora-A overview of binding sites Aurora-A: TPX2-site

Aurora-A: ATP binding site with bound ATP Aurora-A: ATP-binding site blocked by inhibitor
 Aims for Aurora A Project

• Identify compounds active against Aurora using couple

ATP consumption assay

• Use leads to dissect the function of Aurora in the cell

cycle

• Optimize for efficacy and potency

• Proof-of concept in in-vivo tumor model

18
 Aurora A: Characterisation of Compounds

• 91 (from ~53000) in total had an IC50 < 50 µM

• 89 compounds fall predominantly into the ATP NON-

Competitive class
• Structurally diverse
• IC50s from 200nM to high micromolar
• Potential allosteric inhibitors

• Two novel ATP competitive inhibitors found as unique

compounds

19
 Aurora A:
Comparison of ATP versus allosteric inhibitor
ATP competitive non-competitive
123671 103444
Inhibition (%) Inhibition (%)
110
80
80
60
40 50

20 20
0
-10
1 10 100 1 10 100
Compound Concentration [µM] Compound Concentration [µM]

ATP [µM] IC50 [µM] r² Hill-Slope ATP [µM] IC50 [µM] r² Hill-Slope
20 6.68 0.997 1.163 20 1.7 0.996 1.537
200 49.8 0.997 1.06 200 1.9 0.993 1.759
1000 >100 0.975 0.42 1000 0.84 0.978 2.165

The majority of compounds tested fall into non-ATP competitive category

20
 Allosteric inhibitors of Aurora A:
TPX2 competition
IC50 Aurora A
103421:02
IC50 = 4.8 µM; Slope = 1.33; r2 = 0.995 Of ~100 compounds tested
90
Inhibition (%)

60

30 61 showed a strong shift to less potent

0 IC50s- potentially allosteric inhibitors
1 10 100
Concentration (µM) at the TPX2 site
w/o TPX2

IC50 Aurora A + TPX2

103421:02 4 showed no significant shift of IC50
110
IC50 = 78.8 µM; Slope = 5.9; r2 = 0.988 i.e. not competed by TPX2
Inhibition (%)

80

50

20

-10
1 10 100
Concentration (µM)
with TPX2

21
Aurora A Project: Summary
• Rational design of allosteric
inhibitors
• Examine H-bond interactions
• Hydrophobic interactions
• Water molecules
• Fit active compounds into
binding pockets
• Improve hypothesis
• Docking and scoring of
untested compounds
• Help guiding synthesis and
medicinal chemistry efforts One of the TPX2-binding sites
Rational Drug Design:
Unsolved problems and limitations
• Experimental data in early stages is inaccurate
• X-ray structures give static picture
• Conformational changes upon binding
• Role of water molecules and protonation states of functional
groups in binding pocket often unknown
• Limited experimental data in later stages of development
• Complexity of biological systems
• Too many targets (whole proteome)
• Proteome variability
• Short-term versus long-term effects

• Computers need data to build models

Summary

• Rational drug design

• Replaced traditional approaches
• Numerous successful examples
• (still) holds great promises (rightly so!)
• Limitations
• challenging
Acknowledgements

• Elena Conti
• Friedrich Reinhard (Chemical Biology Core Facility)
• Joe Lewis (Chemical Biology Core Facility)