NMR in fragment based drug discovery
Download as PPTX, PDF7 likes4,776 views
The document discusses fragment-based drug discovery (FBDD) as an alternative to high-throughput screening (HTS), highlighting the differences in compound size, binding efficiency, and potencies between the two methods. It outlines various techniques used in FBDD, including NMR methods for ligand detection and characterization, emphasizing the importance of structural reasoning and molecular efficiency in drug development. The document also details clinical advancements from FBDD, providing examples of successful drugs and their respective targets.
![Pellecchia, M.; et al. J. Biomol. NMR 2002, 22, 165.
Pellecchia, M.; Sem, D.S.; Wuthrich, K. Nature 2002, 1, 211.
12
Chemical-shift Mapping
Label the target with 15N and/or 13C and observe changes in the chemical environment
with the addition of a ligand or mixture of ligands
[13C, 1H]-HMQC of selectively labeled DHPR (13Cε/1H Met, 13Cδ/1H Ile, 13C/1H Thr)
DHPR alone: green
DHPR + : blue
DHPR + : blue
DHPR + : red](https://image.slidesharecdn.com/nmrinfragment-baseddrugdiscovery-150111185730-conversion-gate01/85/NMR-in-fragment-based-drug-discovery-12-320.jpg)
![21
“SAR by NMR” to Develop Inhibitor of Bcl-XL: First Fragment
Petros, A.M.; et al. J. Med. Chem. 2006, 49, 656.
R1 R2 R3 NMR Kd (μM)
11 F COOH H 300 + 30
12 H COOH H 1200 + 530
13 F OH H > 5000
14 H COOCH3 H > 5000
15 H CH2COOH H 2000 + 1600
16 H CH2CH2COOH H 1990 + 990
17 OCH3 COOH H 383 + 117
18 Cl COOH H 238 + 110
20 H H COOH > 5000
Initial fragment
scaffold that was
carried forward
1) Uniformly 15N-label Bcl-XL protein and purify the protein by affinity chromatography
2) [1H-15N]-HSQC NMR screening on 15N-labled protein (100 μM) in presence and
absence of small compounds (average MW of 210)
• 9373 compounds were added in increments of 10
• 66 mixtures caused a significant shift in HSQC
• The 660 compounds were then retested individually, yielding 49 compounds with
Kd values less than 5 mM.](https://image.slidesharecdn.com/nmrinfragment-baseddrugdiscovery-150111185730-conversion-gate01/85/NMR-in-fragment-based-drug-discovery-21-320.jpg)
![22
“SAR by NMR” to Develop Inhibitor of Bcl-XL: Second Fragment
[1H-15N]-HSQC NMR screening on 15N-labled protein (100 μM) in the presence of biaryl
derivative (11) and second set of small compounds
• 3472 compounds were added in increments of 5
• 60 mixtures caused a significant shift in HSQC
• The 300 compounds were then retested individually, yielding 24 compounds with
Kd values less than 5 mM.
R1 NMR Kd (μM)
21 4300 + 1600
22 5000 + 2000
23 2000 + 440
24 9000 + 2000
25 6000 + 2000
Petros, A.M.; et al. J. Med. Chem. 2006, 49, 656.](https://image.slidesharecdn.com/nmrinfragment-baseddrugdiscovery-150111185730-conversion-gate01/85/NMR-in-fragment-based-drug-discovery-22-320.jpg)
NMR in fragment based drug discovery
- 1. NMR Studies in Fragment-Based Drug Discovery Katie Strong March 20, 2013 Advisor: Dr. Dennis Liotta, Ph.D. 1
- 2. Bolten, B.M.; DeGregorio, T. Nat. Rev. Drug Discov. 2002, 1, 335. Keseru, G.M.; Makara, G.M. Nat. Rev. Drug Discov. 2009, 8, 203. Early Stage Research and Discovery 2 Hit generation is dominated by high- throughput screening (HTS) The overall success rate of HTS by measuring progression to lead optimization is 45-55% • Estimated size of drug-like compound library is 1060 compounds, while corporate chemical library is only 106 compounds • Increasing the size of the screening library does not proportionally yield more hits • Twice as likely to fail for newer targets
- 3. 3 Typical compound hit from HTS screen • Large molecule (MW between 250 – 600) • Broad surface contact with no high quality interactions in key pockets • May contain functional groups that contribute poorly to protein binding • Emphasis on potency (30 μM – nM hit activity) Alternative to HTS: Fragment-Based Drug Discovery (FBDD) Rees, D.C.; Congreve, M.; Murray, C.W.; Carr, R. Nature 2004, 3, 660. Scott, D.E.; Coyne, A.G.; Hudson, S.A.; Abell, C. Biochemistry 2012, 51, 4990.
- 4. Rees, D.C.; Congreve, M.; Murray, C.W.; Carr, R. Nature 2004, 3, 660. Scott, D.E.; Coyne, A.G.; Hudson, S.A.; Abell, C. Biochemistry 2012, 51, 4990. 4 Typical compound hit from HTS screen • Large molecule (MW between 250 – 600) • Broad surface contact with no high quality interactions in key pockets • May contain functional groups that contribute poorly to protein binding • Emphasis on potency (30 μM – nM hit activity) The idea that large molecules can be considered combinations of two or more individual fragments is a fundamental principle of fragment-based drug discovery Typical compound hits from FBDD • Smaller molecule (MW between 150 – 300) • High proportion of the functional groups involved in binding • Clearly interacts with pockets • Potency in the range of mM to 30 μM • Emphasis on efficiency and design Alternative to HTS: Fragment-Based Drug Discovery (FBDD)
- 5. Rees, D.C.; Congreve, M.; Murray, C.W.; Carr, R. Nature 2004, 3, 660. Erlanson, D.A. Top. Curr. Chem. 2012, 317, 1. Types of Fragment Elaboration Techniques 5 Elaboration of HTS hit Fragment growing Fragment linking Modified compound with higher potency Every atom that is removed or added in FBDD is modified based on structural reasoning
- 6. Lipinski’s Rule of Five for drug-like compounds • Molecular weight < 500 Da • ClogP < 5 • Number of hydrogen bond donors < 5 • Number of hydrogen bond acceptors < 10 • Number of rotatable bonds < 7 • Polar surface area < 140 Å2 Congreve, M.; Carr, R.; Murray, C.; et al. Drug Discov. Today 2003, 8, 876. Hopkins, A.L.; Groom, C.R.; Alex, A. Drug Discov. Today 2004, 9, 430. Scott, D.E.; Coyne, A.G.; Hudson, S.A.; Abell, C. Biochemistry 2012, 51, 4990 6 Fragment Criteria and Characteristics Ligand efficiency (LE): method to compare different sized fragments • LE = -ΔG/Heavy Atom Content ≈ -RTlnKd/HAC • Fragments are weakly binding, but very “atom efficient” binders • LE > 0.3 kcal/mol per HAC are considered good oral drug candidates Astex’s Rule of Three for fragments • Molecular weight < 300 Da • ClogP < 3 • Number of hydrogen bond donors < 3 • Number of hydrogen bond acceptors < 3 • Number of rotatable bonds < 3 • Polar surface area < 60 Å2
- 7. Scott, D.E.; Coyne, A.G.; Hudson, S.A.; Abell, C. Biochemistry 2012, 51, 4990. 7 Comparison of Fragment and HTS Hits
- 8. Pellecchia, M.; Sem, D.S.; Wuthrich, K. Nature 2002, 1, 211. Meyers, B.; Peters, B. Angew. Chem. Int. Ed. 2003, 42, 864. 8 NMR Methods for Fragment Discovery and Elaboration NMR methods for detecting ligand binding are divided into two categories 1) Monitor NMR signals from the protein in the presence of ligand • Chemical-shift mapping and “SAR by NMR” 2) Monitor the ligand bound to target relative to the free ligand • T2 and T1p relaxation • Transferred NOEs • Saturation transfer difference (STD) • Water-ligand Observed via Gradient Spectroscopy (Water-LOGSY) • Diffusion editing
- 9. 9 Use of Relaxation Times to Identify Ligands Relaxation Time • Small, rapidly tumbling molecules: high (longer) relaxation times • Macromolecules that move slowly through solution: low (shorter) relaxation times Shortening T2 relaxation time leads to peak broadening Pellecchia, M.; Sem, D.S.; Wuthrich, K. Nature 2002, 1, 211. Meyers, B.; Peters, B. Angew. Chem. Int. Ed. 2003, 42, 864. Molecular Weight 1000 Da T1 T2 Z X Y After resonance, where v1 = vo, magnetization relaxes back to equilibrium • T1 = relaxation of nuclear spin magnetic vector parallel to the magnetic field, Bo • T2 = relaxation of nuclear spin magnetic vector perpendicular to the magnetic field, Bo Bo
- 10. 10 Use of Relaxation Times to Identify Ligands Pellecchia, M.; Sem, D.S.; Wuthrich, K. Nature 2002, 1, 211. Meyers, B.; Peters, B. Angew. Chem. Int. Ed. 2003, 42, 864. Jahnke, W.; Perez, L.B.; Paris, G.C.; Strauss, A.; Fendrich, G.; Nalin, C.M. J. Am. Chem. Soc. 2000, 122, 7394. Distance from paramagnetic center 25 Å T1 and T2 Relaxation Time 1) SLAPSTIC (Spin labeled attached to protein side chains as a tool to identify interacting compounds): Covalently attach a spin label (TEMPO) on the target near the binding site and monitor the enhanced T2 relaxation once a ligand binds 2) Use a spin-labeled ligand to bind to an initial site on the target and then monitor the enhanced T2 relaxation once a ligand binds in the second binding site The gyromagnetic ratio of an unpaired electron is 658-fold larger than a proton γ = μ / P γ = gyromagnetic ratio μ = magnetic moment P = spin angular momentum
- 11. 11Jahnke, W.; Perez, L.B.; Paris, G.C.; Strauss, A.; Fendrich, G.; Nalin, C.M. J. Am. Chem. Soc. 2000, 122, 7394. Meyers, B.; Peters, B. Angew. Chem. Int. Ed. 2003, 42, 864. Researchers demonstrated the SLAPSTIC method with the FK binding protein by using a mixture of known binders (1) and (2) and 3 non-bonding aromatic ligands. • FKBP has lysine residues 15-20 Å away from the binding site, so all lysine residues were spin-labeled using N-hydroxysuccinimide ester 3 • Monitor the relaxation effect of the ligands in the presence of spin-labeled protein Use of Relaxation Times to Identify Ligands Kd = 1.1 mM Kd = 9.0 mM No FKPB FKBP Spin-labeled FKBP
- 12. Pellecchia, M.; et al. J. Biomol. NMR 2002, 22, 165. Pellecchia, M.; Sem, D.S.; Wuthrich, K. Nature 2002, 1, 211. 12 Chemical-shift Mapping Label the target with 15N and/or 13C and observe changes in the chemical environment with the addition of a ligand or mixture of ligands [13C, 1H]-HMQC of selectively labeled DHPR (13Cε/1H Met, 13Cδ/1H Ile, 13C/1H Thr) DHPR alone: green DHPR + : blue DHPR + : blue DHPR + : red
- 13. Shuker, S.B.; Hajduk, P.J.; Meadows, R.P.; Fesik, S.W. Science 1996, 274, 1531. 13 Chemical-shift Mapping and “SAR by NMR” Target based screening first reported by Abbott Laboratories in 1996 in an effort to find compounds to replace FK506, an immunosuppressant that binds to FKBP FK506
- 14. Shuker, S.B.; Hajduk, P.J.; Meadows, R.P.; Fesik, S.W. Science 1996, 274, 1531. 14 Chemical-shift Mapping and “SAR by NMR” Kd = 2 μM Kd = 0.8 mM Kd = 0.1 mM Kd = 19 nM
- 15. 15 Drugs from FBDD to Reach Clinical Trials Erlanson, D.A. Top. Curr. Chem. 2012, 317, 1. Drug Company Target Phase PLX-4032 (Vemurafenib) Plexxikon B-Raf V600E FDA Approved ABT 263 Abbott Bcl-2/Bcl-xL Phase 2 ABT869 Abbott VEGF and PDGFR Phase 2 AT9283 Astex Aurora Phase 2 AT5719 Astex CDKs 1,2,4,5 Phase 2 LY-517717 Lilly/Protherics Fxa Phase 2 Indeglitazar Plexxikon PPAR agonist Phase 2 VER-52296 Vernalis/Novartis Hsp90 Phase 2 ABT-518 Abbott MMP-2 and MMp-9 Phase 1 ABT-737 Abbott Bcl-2/Bcl-xL Phase 1 AT13387 Astex Hsp90 Phase 1 LP-261 Locus Tubulin Phase 1 PLX-5568 Plexxikon Kinase Phase 1 Using variety of techniques, a handful of drugs developed by FBDD have entered the clinic
- 16. 16 Bcl-2 (B-cell lymphoma) Family Proteins Tait, S.W.G.; Green, D.R. Nature Rev. 2010, 11, 621. Bcl-2 proteins are regulators of programmed cell death, and anti-apoptotic proteins are typically overexpressed in cancer cells
- 17. 17 Bcl-2 Family Proteins and Role in Apoptosis Youle, R.J.; Strasser, A. Nature Rev. 2008, 9, 47. • In healthy cells, Bax and Bak predominately exists in the cytosol, but when under stress will move to the mitochondria and activate apoptosis • In a mechanism that is not entirely understood, anti- apoptotic proteins can bind and retrotranslocate Bax from the mitochondria, inhibiting apoptosis
- 18. 18 Protein-Protein Interactions as a Classically “Difficult” Target Wells, J.A.; McClendon, C.L. Nature 2007, 450, 1001. Bauer, R.A.; Wurst, J.M.; Tan, D.S. Curr. Opin. Chem. Biol. 2010, 14, 308. Overington, J.P.; Al-Lazikani, B.; Hopkins, A.L. Nat. Rev. Drug Discov. 2006, 5, 993. It has been proposed that no class of interaction rivals the complexity of protein-protein interactions, and targeting these interactions has been regarded as “difficult.” • Contact surface area is typically very large at approximately 1500-3000 Å2 • Binding pockets are often flat, featureless, and lack well-defined grooves • Lack a natural small-molecule partner, so difficult to find a suitable starting lead • Often a HTS is dominated by compounds that have been used for classic drug targets, and each protein-protein interaction may require a different starting compound • Protein-protein interactions are key to intracellular signaling pathways
- 19. 19 Binding Site of Bcl-xL and BAX Sattler, M. et al. Science 1997, 275, 983. Petros, A.M.; et al. J. Med. Chem. 2006, 49, 656. PDB 1G5J Bcl-xL α1 α7 α3 α5 α6 α4 α8 α2 BH3 BH1 BH2 The Bcl-XL protein consists of 8 α helices with a deep hydrophobic pocket formed by the BH1, BH2, and BH3 domains
- 20. 20 Binding Site of Bcl-xL and BAX Sattler, M. et al. Science 1997, 275, 983. Petros, A.M.; et al. J. Med. Chem. 2006, 49, 656. PDB 1G5J Bcl-xL α1 α7 α3 α5 α6 α4 α8 16 residue portion from BH3 domain of BAX The binding site of Bcl-XL is a deep hydrophobic pocket and only approximately 500 Å2 of the protein surface is involved α2 BH3 BH2 BH1 The Bcl-XL protein consists of 8 α helices with a deep hydrophobic pocket formed by the BH1, BH2, and BH3 domains
- 21. 21 “SAR by NMR” to Develop Inhibitor of Bcl-XL: First Fragment Petros, A.M.; et al. J. Med. Chem. 2006, 49, 656. R1 R2 R3 NMR Kd (μM) 11 F COOH H 300 + 30 12 H COOH H 1200 + 530 13 F OH H > 5000 14 H COOCH3 H > 5000 15 H CH2COOH H 2000 + 1600 16 H CH2CH2COOH H 1990 + 990 17 OCH3 COOH H 383 + 117 18 Cl COOH H 238 + 110 20 H H COOH > 5000 Initial fragment scaffold that was carried forward 1) Uniformly 15N-label Bcl-XL protein and purify the protein by affinity chromatography 2) [1H-15N]-HSQC NMR screening on 15N-labled protein (100 μM) in presence and absence of small compounds (average MW of 210) • 9373 compounds were added in increments of 10 • 66 mixtures caused a significant shift in HSQC • The 660 compounds were then retested individually, yielding 49 compounds with Kd values less than 5 mM.
- 22. 22 “SAR by NMR” to Develop Inhibitor of Bcl-XL: Second Fragment [1H-15N]-HSQC NMR screening on 15N-labled protein (100 μM) in the presence of biaryl derivative (11) and second set of small compounds • 3472 compounds were added in increments of 5 • 60 mixtures caused a significant shift in HSQC • The 300 compounds were then retested individually, yielding 24 compounds with Kd values less than 5 mM. R1 NMR Kd (μM) 21 4300 + 1600 22 5000 + 2000 23 2000 + 440 24 9000 + 2000 25 6000 + 2000 Petros, A.M.; et al. J. Med. Chem. 2006, 49, 656.
- 23. 23 “SAR by NMR” to Develop Inhibitor of Bcl-XL: Chemical-Shift Mapping Black: 15N-labeled Bcl-XL Red: Bcl-XL and Green: Bcl-XL, fragment 11, and Petros, A.M.; et al. J. Med. Chem. 2006, 49, 656.
- 24. 24 “SAR by NMR” to Develop Inhibitor of Bcl-XL: Linking the Fragments The ortho position of fragment 9 was the most direct linker to the second site Petros, A.M.; et al. J. Med. Chem. 2006, 49, 656.
- 25. “SAR by NMR” to Develop Inhibitor of Bcl-XL: Linking the Fragments 2nd site ligand 1st site ligand (11) Petros, A.M.; et al. J. Med. Chem. 2006, 49, 656.
- 26. 26 “SAR by NMR” to Develop Inhibitor of Bcl-XL: Linking the Fragments FPA Ki = 1.4 μM Trans-linker interacts with phenylalanine and prevents fragment from binding deep in the pocket Petros, A.M.; et al. J. Med. Chem. 2006, 49, 656.
- 27. 27 “SAR by NMR” to Develop Inhibitor of Bcl-XL: Revisiting First Fragment The carboxylate of the first fragment was replaced with an acylsulfonamide and 120 analogs were synthesized using commercially available sulfonamides Petros, A.M.; et al. J. Med. Chem. 2006, 49, 656. NMR Kd = 300 μM NMR Kd = 320 μM
- 28. 28 “SAR by NMR” to Develop Inhibitor of Bcl-XL: Revisiting First Fragment Bcl-XL FPA Ki = 0.245 μM Petros, A.M.; et al. J. Med. Chem. 2006, 49, 656. Initial most potent sulfonamide fragment
- 29. 29 “SAR by NMR” to Develop Inhibitor of Bcl-XL: Revisiting the Second Fragment 125 additional compounds that maintained the acylsulfonamide and nitrophenyl moieties were prepared Petros, A.M.; et al. J. Med. Chem. 2006, 49, 656. Bcl-XL FPA Ki = 36 nM
- 30. 30 “SAR by NMR” to Develop Inhibitor of Bcl-XL: Optimizing the Final Compound Petros, A.M.; et al. J. Med. Chem. 2006, 49, 656. Bcl-XL FPA Ki = 0.245 μM Bcl-XL FPA Ki = 36 nM
- 31. 31 “SAR by NMR” to Develop Inhibitor of Bcl-XL: Optimizing the Final Compound Despite the potency, 44 had poor aqueous solubility and tight binding to human serum albumin (HSA) Petros, A.M.; et al. J. Med. Chem. 2006, 49, 656. Bcl-XL FPA Ki = 0.245 μM Bcl-XL FPA Ki = 36 nM
- 32. 32 “SAR by NMR” to Develop Inhibitor of Bcl-XL: Optimizing the Final Compound Bruncko, M.; et al. J. Med. Chem. 2007, 50, 641. Petros, A.M.; et al. J. Med. Chem. 2006, 49, 656. Oltersdorf, T.; et al. Nature 2005, 435, 677. Certain portions of 44 were exposed to lipophilic residues in the complex with HSA, and these were modified with polar substituents Basic 2-dimethylaminoethyl group Basic piperazine and biphenyl substituents
- 33. 33 “SAR by NMR” to Develop Inhibitor of Bcl-XL: Optimizing the Final Compound Oltersdorf, T.; et al. Nature 2005, 435, 677. Bruncko, M.; et al. J. Med. Chem. 2007, 50, 641. ABT-737 Bcl-XL FPA Ki < 0.5 nM Bcl-2FPA Ki < 1.0 nM In optimizing the final compound, the first dual inhibitor of Bcl-XL and Bcl-2 was discovered While no longer binding to HAS, ABT-737 is not orally available and low aqueous solubility makes intravenous delivery challenging
- 34. 34 “SAR by NMR” to Develop Inhibitor of Bcl-XL: Optimizing the Final Compound Park, C-M.; et al. J. Med. Chem. 2008, 51, 6902. ABT-737 Bcl-XL EC50 = 7.7 nM Bcl-2EC50 = 30 nM 10% HS HI46 EC50 = 87 nm AUC = 0.28 μM . h
- 35. 35 “SAR by NMR” to Develop Inhibitor of Bcl-XL: Optimizing the Final Compound 50 Bcl-XL EC50 = 0.60 nM Bcl-2EC50 = 0.90 nM Park, C-M.; et al. J. Med. Chem. 2008, 51, 6902. ABT-737 Bcl-XL EC50 = 7.7 nM Bcl-2EC50 = 30 nM 10% HS HI46 EC50 = 87 nm AUC = 0.28 μM . h
- 36. 36 “SAR by NMR” to Develop Inhibitor of Bcl-XL: Optimizing the Final Compound 50 Bcl-XL EC50 = 0.60 nM Bcl-2EC50 = 0.90 nM 51 10% HS HI46 EC50 = 40 nm Park, C-M.; et al. J. Med. Chem. 2008, 51, 6902. ABT-737 Bcl-XL EC50 = 7.7 nM Bcl-2EC50 = 30 nM 10% HS HI46 EC50 = 87 nm AUC = 0.28 μM . h
- 37. 37 “SAR by NMR” to Develop Inhibitor of Bcl-XL: Optimizing the Final Compound ABT-737 Bcl-XL EC50 = 7.7 nM Bcl-2EC50 = 30 nM 10% HS HI46 EC50 = 87 nm AUC = 0.28 μM . h 50 Bcl-XL EC50 = 0.60 nM Bcl-2EC50 = 0.90 nM 51 10% HS HI46 EC50 = 40 nm 52 AUC = 1.16 μM . h Park, C-M.; et al. J. Med. Chem. 2008, 51, 6902.
- 38. 38 “SAR by NMR” to Develop Inhibitor of Bcl-XL: Optimizing the Final Compound Erlanson, D.A. Top. Curr. Chem. 2012, 317, 1. ABT-737 Bcl-XL EC50 = 7.7 nM Bcl-2EC50 = 30 nM 10% HS HI46 EC50 = 87 nm AUC = 0.28 μM . h ABT-263 Bcl-XL EC50 = 5.9 nM Bcl-2EC50 = 4.2 nM 10% HS HI46 EC50 = 87 nm AUC = 6.26 μM . h Currently, ABT-263 is in Phase 2 clinical trials for lymphoid malignancies, chronic lymphocytic leukemia, and small cell lung cancer
- 39. 39 “SAR by NMR” to Develop Inhibitor of Bcl-XL: Fragments Leading to Inhibitor ABT-263 Bcl-xL Ki < 0.5 nM LE = 0.20 MW = 975 Erlanson, D.A. Top. Curr. Chem. 2012, 317, 1. Kd = 300 μM LE = 0.30 MW = 216 Kd = 6000 μM LE = 0.23 MW = 170 Ki = 1.4 μM LE = 0.27 MW = 394 Bcl-xL K i = 36 nM LE = 0.27 MW = 552 ABT-737 Bcl-xL Ki = 0.6 nM LE = 0.22 MW = 813
- 40. Protein-peptide interaction Bcl-XL and 26-residue of BAD Protein-small molecule interaction Bcl-XL and ABT-737 40 Bcl-XL Bound to Natural Peptide Partner and Inhibitor Wells, J.A.; McClendon, C.L. Nature 2007, 450, 1001. Molecular Mass (Da) Bcl-xL Ki (nM) LE (kcal/mol/HAC) BAD-derived peptide 3,110 0.6 0.16 ABT-737 813 0.6 0.23
- 41. 41 Protein-Protein Interactions as a Classically “Difficult” Target Wells, J.A.; McClendon, C.L. Nature 2007, 450, 1001. It has been proposed that no class of interaction rivals the complexity of protein-protein interactions, and targeting these interactions has been regarded as “difficult.” 1) Binding pockets are often flat, featureless, and lack well-defined grooves • Contact surfaces are adaptable • APT-737 bound to Bcl-XL causes a more puckered conformation 2) Lack a natural small-molecule partner, so difficult to find a suitable starting lead • Small molecule and natural partner possibly have comparable affinities • ABT-737 and BAD both have 0.6 nM affinity, and ABT has higher LE 3) The molecular size of many compounds that interact with protein-protein interfaces are too large • The criteria for defining drug-like characteristics is based on known drugs • HTS may not be successful for more “difficult” targets because libraries are typically composed of scaffolds for traditional targets. • ABT-263 breaks 3 rules from Lipinkski’s Rule of 5, but is still orally bioavailable.
- 42. 42 Summary • FBDD provides an alternative to HTS where small ligands are found and elaborated in a process based on design and efficiency • NMR techniques can be based on observing changes to the protein or by observing changes to the bound ligand relative to the free ligand • Chemical-shift mapping and “SAR by NMR” • T2 and T1p relaxation • The first dual inhibitor of Bcl-XL and Bcl-2, developed by “SAR by NMR” is orally available and has entered Phase 2 clinical trials for types of leukemia • FBDD design is helping to develop modulators for targets that have been classically regarded as difficult and challenging • Protein-protein interactions: Bcl-XL, heat shock protein Hsp90 • RNA polymerase: HCV NS5B RNA-dependent polymerase • DNA-binding proteins: E2 transcription factor from human papillomavirus Coyne, A.G.; Scott, D.E.; Abell, C. Curr. Opin. Chem. Biol. 2010, 14, 299.
Editor's Notes
- #3: Before I introduce fragment based methods for drug discovery I think is important to introduce how the majority of drugs that enter the clinic and the market place are found. Once a hit is found, 45-55% of those hits actually become leads that are carried forward. In a study in 2006, 29 HTS labs were surveryed. These are labs that look at 100,000 wells a week and over the course of the year they generated 120 leads all together. Last month Eli Lily reported that is costs 1.3 billion dollars to bring a new drug to market, but the average developed by a major pharmaceutical company costs at least $4 billion, and it can be as much as $11 billion. http://www.forbes.com/sites/matthewherper/2012/02/10/the-truly-staggering-cost-of-inventing-new-drugs/ The portion of the process that I want to focus in on though is the early drug discovery stage. Discovery typically begins with target identification: A target is generally a single molecule, such as a gene or protein, which is involved in a particular disease. Even at this early stage in drug discovery it is critical that researchers pick a target that is “drugable,” i.e., one that can potentially interact with and be affected by a drug molecule. Target confirmation: confirm that the target is actually involved in the disease state. Hit generation is actually finding compounds that interact with the drug. results in a set of compounds that interact with the target and then then lead generation is optimizing the hits for preclinical and safety studies. -hit discovery and hit-to-lead optimization is relatively short and inexpensive – 12 months and about 10million dollars. -nonetheless the majority of discovery projects are terminated due to lack of leads. One of the paramount hit identification methods is HTS - Its overall success rate as measured by progression to lead optimization is only 45-55%. -In the early years of HTS, it was believed that the limited accuracy of the screening and the low quality of the leads would be compensated by the large number of compounds investigated. This doesn’t work well though because chemical space is so big. It has been estimated to be 10^60 compounds, so a few million compounds really makes no difference in sampling. Since most lead compounds tend to maintain the core of the structure, this is an important part of the process that has significant ramifications later. 10^60 compounds of HTS size
- #4: In a HTS, typically over 1 million compounds can be screened in well plates and there are a variety of assays and targets that this can be done with. A Typical hit from HTS will look like this though - Without any structural information, often the chemist does not know which areas are contributing poorly and which ones to fix – often systematically fix all of the areas which is resource and time consuming
- #5: In a HTS, typically over 1 million compounds can be screened in well plates and there are a variety of assays and targets that this can be done with. A Typical hit from HTS will look like this though - Without any structural information, often the chemist does not know which areas are contributing poorly and which ones to fix – often systematically fix all of the areas which is resource and time consuming The past 15 years has seen fragment based drug discovery evolve from a niche topic to a more widely used drug discovery process. Fragments deliver a high hit rate because typically with large compounds you get a great chance of mismatch
- #6: Without any structural information, often the chemist does not know which areas are contributing poorly and which ones to fix – often systematically fix all of the areas which is resource and time consuming. Once an appropriate hit has been discovered from a HTS, the compound will be optimized and since many of the compounds already exhibit drug like properties there isn’t much room to grow and often go backwards in terms of property – make a compound that is bigger and is less liphophlic. Fragment linking is joining two fragments that are already known to bind to two sites
- #7: The logP value of a compound, which is the logarithm of its partition coefficient between n-octanol and water log(coctanol/cwater), is a well established measure of the compound's hydrophilicity. Low hydrophilicities and therefore high logP values cause poor absorption or permeation. It has been shown for compounds to have a reasonable propability of being well absorbt their logP value must not be greater than 5.0. The distribution of calculated logP values of more than 3000 drugs on the market underlines this fact (see diagram)
- #8: Typically a HTS will yield hits that are in the low microM to high nanomolar region that have higher MW. The lead optimization window is not very large, but this will usually lead to a lead compound that may or may not be below 500 Da. Fragment hits on the other hand start with much smaller MW, there is a large window for optimization and then this can lead to a drug candidate. Often in a HTS, the hits will be drug-like, but not lead-like. Often times it can be a challenge to optimize a hit that is already very drug like because they are already very large and lipophilic. Typically in the lead optmization stage, MW is increased by approximately 80 and 1 logP unit, so if a HTS hit already exhibits drug-like characteristics, you actually end up making a compound that far exceeds those common critera. Chemical space: 10^60 possible molecules, but only 10^7 small molecules composed of 11 atoms of C,N,F,O that follow Rule of Three. Must easier to tackle the available chemical possibilities.
- #9: Before I really discuss an example of how NMR is used for a particular target, I want to introduce in more detail the three techniques highlighted here. Consider the ligand is the equilibrium process: if this process is slow then you would actually see two distinct peaks, so most of these techniques rely on the exchange being fast compared to relaxation time or the chemical shift. That correlates to a Kd < 10-5 M.
- #10: If rf energy having a frequency matching the Larmor frequency is introduced at a right angle to the external field (e.g. along the x-axis), the precessing nucleus will absorb energy and the magnetic moment will flip to its I = _1/2 state. This is known as resonance, but For nmr spectroscopy to be practical, an efficient mechanism for nuclei in the higher energy _1/2 spin state to return to the lower energy +1/2 state must exist. When a nucleus with I = ½ is subjected to the magnetic field, two spin states are present: alpha and beta. Initially these are equally present, but a population difference will arise. The nucleus will start spinning and precess about Bo at the Larmor frequency. Initially this is just around the z-axis. Then an RF v1 (90o pulse) in the x-plane forces the precess in phase and so the magnetic vector has components in the z axis and the x/y axis. Larmor frequency will equal electromagnetic radiation is resonance. After resonance the spinning will relax – T1 is spin lattice relaxation about the Z axis and T2 is spin-spin relaxation about the x/y plane. T1 - transfer to nearby molecules T2 – transfer to nearby nuclei The observation of these effects would require well separated resonance lines, and so this is of limited value because if many of the lines get really broad, it would be too difficult to quanitify.
- #11: In physics, angular momentum, moment of momentum, or rotational momentum[1][2] is a vector quantity that represents the product of a body's rotational inertia and rotational velocity about a particular axis. he magnetic moment of a magnet is a quantity that determines the force that the magnet can exert on electric currents and the torque that a magnetic field will exert on it Nuclei with a spin angular momentum will have a magnetic moments that can be measured and this is directly related to the gyromagnetic ratio. Typically we think of the proton as having the highest relative magnetic moment, which is why it is easier to conduct an NMR on protons. An electron though has an even higher magnetic moment, and an unpaired electron is 658-fold larger than a proton. As a result, relaxation of nuclear spins that are close to the unpaired electron are highly affected and this causes very sensitive line broadening. The difference in relaxation times of ligands close to a paramagnetic center can be taken advantage of because the protein can be spin-labeled with a paramagnetic center – TEMPO can be attached to residues near the binding site and then any ligands that bind to this site will experience enhanced T2 relaxation.
- #12: The average spin label number per protein was 3 - 4
- #13: DHPR is a 120-kDa protein Dihydrodipicolinate reductase is an enzyme found in bacteria that is involved in the biosynthesis of lysine and bacterial cell walls On binding with a substrate analog (PDC), many chemical shifts have been altered. Ligand binding can induce multiple changes including those due to long range protein conformational changes, which can make the interpretation difficult. For this reason, differential chemical shift perturbation method is used where spectra of the target protein is compared when it is bound to two slightly different ligands. A spectrum with the analog 4-cl-PDC shows which residues are in direct contact with the inhibitor. Use mutant-containing plasmids. Second figure is a result of small chemical shift changes solely from the binding of these particular small ligands.
- #14: When complexed to FK506, the FKBP protein binds tightly to calcineurin, which plays a role in a variety of transcription factors. Significant immuno suppression is observed, so FK506 is helpful in organ transplants, but it is also very toxic. Efforts were aimed at finding a drug that maintained potency, but was less toxic.
- #15: Compound 3 was the initial hit and carried forward without further purification. Then screen library in the presence of saturated compound 3 and came up with compound 4. This was optimized to give 5. At this point a variety of linkers of different length were synthesized and the most potent compound was #6 with a Kd of 19 nm.
- #16: In 2011 the FDA approved PLX-4032 as an oral therapy for the treatment of advanced melanoma. This was discovered by screening 20,000 scaffolds against a variety of kinaseses by X-ray cryst. It is selective for the B-Raf family of kinases and particularly the oncogenic V600E mutation of B-raf. Vemurafenib interrupts the B-Raf/MEK step on the B-Raf/MEK/ERK pathway − if the B-Raf has the common V600E mutation. When this was FDA approved, a test was too that tells patients whether or not they have this particular mutation.
- #17: Proteins in the Bcl-2 family have very similar sequence and structural similarities. BH3 proteins either activate BAX, BAK or inhibit the anti-apoptosis. Homodimers that can form heterodimers with other protein family members, so can get various combinations of anti- and pro-apoptosis heterodimers
- #18: Bcl-2 and Bcl-xL inhibit apoptosis by binding to a 16 residue alpha helix portion of the BAK or a 26 residue a-helical portion of BAD.
- #19: Study done in 2007 using high-affinity inhibitors for IL-2, Bcl-XL, HDME, and HPV E2 and small molecules in the MDL Drug Data Report (MDDR) and World of Molecular Bioactivity (WOMBAT) Protein-protein interactions are key to intracellular signaling pathways and cell-surface receptor-ligand interactions.
- #20: Bcl-xL is a protein composed of 8 alpha helices
- #21: BAX and BAD bind in a hydrophobic cleft and make key interactions.
- #22: Form of Bcl-xL that lacked the putative transmembrane helix and the loop connecting helix 1 to 2. NMR-based titration yielded the dissociation constant: Chemical shift differences were calculated as Δδ = [(ΔδH)2 + (0.2*ΔδN)2]1/2, where ΔδH and ΔδN are the observed chemical shift changes for 1H and 15N, respectively. For determination of dissociation constants, Δδ was plotted as a function of the molar ratio (nucleotide:protein) and the data for multiple peaks were fitted using the maximum shift and dissociation constant as adjustable parameters. COOH is critical for binding and it must be in the para position. The fluorine is not as important, although moving forward the group chose to stick with compound 9 This is the simplist and the cheapest form of labelling. The protein is produced by expression from bacteria which are grown on minimal medium supplemented with 15NH4Cl and wild-type (wt) glucose. Grow bacteria using minimal media containing 15N-labeled ammonium chloride or 13C labeled glucose.
- #23: It was already known from NMR structural studies that a second binding site is present and a second site screen was performed. These compounds all have lower affinities that the original hit. Moving forward researchers used compound 18, but did not rule out any compounds when deciding how to link
- #24: Shift changes are consistent with binding in the hydrophobic binding site. Amides G94, G138, and G196 are shifted upon binding
- #25: The structure of Bcl-XL and compounds 9 and 18 was obtained using NMR structural studies. This is a superposition of 7 low-energy conformations for the Bcl-Xl protein NMR models based on NOEs and then modeled – docked and then the lowest energy conformation is seen. The program is CNX. Dock the ligands into the binding groove randomly and then
- #26: Based on that assumption, a variety of linkers were used to connect the ortho position to napthyl or biaryl derivatives. Most compounds bound with an inhibition constant greater than 10μM based on FPA, but one compound in particular was a 200-fold improvement over the original biaryl fragment 1
- #27: NMR models based on NOEs and then modeled – docked and then the lowest energy conformation is seen. The program is CNX. The trans linker prevents the second fragment from binding deep in the pocket, so a different ligand and linker were investigated, so for that reason the carboxylate of fragment 1 was replaced with an acylsulfonamide and it was thought that this could act as the linker. Ki – FPA using BH3 domain from a BAD protein. This has a Kd of 20 nm. When excited with a plane-polarized light, fluorescent small molecules (small peptide) in solution rotate fast and the emitted light is depolarized (low FP). When binding to bigger molecules (a protein), the movement of the complex becomes slower and the emitted light is polarized, leading to generation of FP signal The inhibitor constant, Ki, is an indication of how potent an inhibitor is; it is the concentration required to produce half maximum inhibition. Fluorescently labeled BH3 domain peptide derived from the BAD protein, which has a Kd of 20nM. The small peptide is moving fast thru solution and so it causes depolarization of light and a low FP signal. When the peptide is bound to the Bcl-XL protein it takes on those properties and it can no longer depolarize the light, so you see a high FP signal. That is the assay. Then small compounds are exposed to a mixture of the Bcl-xL and BAD protein. If the small compounds are binding instead of the fluorescently labeled peptide, then the FP signal will again decrease because it is no longer binding to the protein and can stop the depolarization of light. Ki values were determined using an in-house written program.
- #28: A common isostere for carboxylates are acylsulfonamides because their pka are in the approximate same range of 3-5, so the carboxylate could be replaced with acylsulfonamide and this could also function as the linker. A series of sulfonamides were prepared and it appeared as these compounds had basically the same affinity as the original biaryl compound (approximately 300 microM). 120 analogs were synthesized in the synthesis shown and the most potent compounds was 35 with Ki of 0.245 microM. MP-TsOH is a useful alternative to quenching reactions with aqueous or soluble organic acids.
- #29: To guide subsequent studies, NMR studies were done with compound 30. The nitrophenyl sits in between phenylalanine 97 and tyrosine 194 forming a π-stacked arrangement and since this seemed to provide an energetically favorable way of linking the two fragments, the acylsulfonamide was continued to be used. With that the acylsulfonamide and nitrophenyl linker were maintained
- #30: Since only so many commercially available acylsulfonamides are available, a convenient method for synthesizing them was worked out. 125 compounds were synthesized using sulfonamide 37, and the most potent of these was compound 39.
- #31: The S-phenyl ring actually is bent back beneath the nitrophenyl ring and this causes the F97 to rotate downward and its ring pi stacks with the S-phenyl ring, which then pi-stacks with the nitro ring which pi-stacks with the tyrosine ring. This extensive pi-stacking is most likely why this particular ligand has such a high affinity.
- #32: The S-phenyl ring actually is bent back beneath the nitrophenyl ring and this causes the F97 to rotate downward and its ring pi stacks with the S-phenyl ring, which then pi-stacks with the nitro ring which pi-stacks with the tyrosine ring. This extensive pi-stacking is most likely why this particular ligand has such a high affinity. Human serum albumin is the most abundant protein in human blood plasma, so if binding tightly to HSA then will never leave the blood to reach the tissues or tumor
- #34: And so for that reason, researchers set out to make a few more changes and increase the oral bioavaliablity
- #35: Removing the very basic dimethyamino side chain led to huge increases in bioavalibality and oral exposure, but again decreases in potency due to HAS binding Changing the dimethyamino to morpholine resulted in 4 fold enhancement in oral exposure, but had submicrmolar activity. Although this functionality could be retained if other changes were favorable. Knowing the effect of the nitro group on the pka of the acylsulfonamide and the known propensity for toxicity, the nitro group was replaced. Replacing the nitro group with CF3 led to 16 fold increase in AUC, but large decrease in potencey. Replacing the group with anything that was not electron withdrawing led to large decreases in the potency. When the group was replaced with a SO2CF3 though, there was a slight increase in potency, but there was a slight decrease in PSA, which led to a 7 fold increase in AUC/EC50 value.
- #37: Serum is the portion of the blood that is left after the blood cells (red blood cells, white blood cells and platelets) are removed. It includes the fluid, the electrolytes and the dissolved proteins like albumin.
- #38: bioavailability is a measurement of the rate and extent to which a drug reaches the systemic circulation. The absolute bioavailability (%f) is the dose-corrected area under curve (AUC) non-intravenous divided by AUC intravenous. The CellTiter-Glo Luminescent Cell Viability Assay is a homogeneous method of determining the number of viable cells in culture based on quantitation of the ATP present, an indicator of metabolically active cells.
- #42: Binding pockets are often flat, featureless, and lack well-defined grooves Contact surfaces are adaptable and well-defined pockets that were not observed in an unbound state or bound to the natural peptide partner were seen in the presence of a small molecule. As we saw, once ABT-737 was bound, the binding surface was less flat, more puckered, and more closed in compared to the binding pocket with the peptide Cavities not observed in the static apo state can be available for binding and we can’t assume the a protein modeled in the unbound or naturally bound state is the best model for finding a drug. It may be a good place to start, but FBDD allows for those conformations to be seen as the compound grows. Lack a natural small-molecule partner, so difficult to find a suitable starting lead or natural protein partner is binding too tightly. Small molecule and natural partner possibly have comparable affinities and it may not be unreasonable that in situations the small molecule could accelerate the dissociation of a natural peptide partner. The molecular size of many compounds that interact with protein-protein interfaces are too large The criteria for defining drug-like characteristics is based on known drugs: this is changing and there are already notable exceptions. ABT-737 doesn’t have drug like characteristics although as we saw, after specific changes the oral bioavalibality was significatly improved and it is now in clinical trials. As we discover more of these compounds, it could be that the critera changes. HTS is usually a mass collection of scaffolds that have classically worked for targets such as GPCRS and ion channels, but as we saw, compounds that are well suited for these protein-protein interactions are probably not the same chemotypes as classical targets require. FBDD allows the exploration of new chemical space and we’re relying so heavily on already developed scaffolds, but instead a scaffold will be unique to the target. This can also lead us to target previously believed “undruggable” targets