Quantification of free ligand conformational preferences by NMR and their relationship to the bioactive conformation Charles Blundell charles.blundell@c4xdiscovery.com www.c4xdiscovery.com
Rigid: single conformation Flexibility requires ensembles All data satisfied by a single conformation Flexible: two+ rapidly interchanging conformations Fitting data to single conformation produces unrealistic virtual conformations Replace MEAN averaging MODAL averaging Data needs to be fit to an ensemble of conformations as a better model of reality
mean angle Ensemble generation
Ensemble generation mean angle libration 0º 0º 270º 90º 270º 90º 180º large libration 180º small libration
Ensemble generation mean angle libration 3 rd minor mode 0º 0º 270º 90º 270º 90º 2 nd mode with its own mean & libration 180º large libration 180º small libration Single conformation model requires 1 variable per torsion C4XD dynamic model requires up to 8 variables per torsion (3 means, 3 librations, 2 population partitions) Avoids the problems of ensemble generation via comp chem Therefore ~10x more data is needed to solve the structure
Data and fitting 1) Combine multiple kinds of data into one coherent solution: NOESY/ROESY! Scalar couplings, Residual Dipolar Couplings Chemical shifts, H-bonds etc. etc. 2) Make the most of what you can measure:!! Explicitly calculate exact value of all observables! NOE x-peaks heights including nonoes (height = zero) Avoid approximations and assumptions inherent in e.g. converting NOES into distances, 3 JHH into torsional angles 3) Fitting function similar to that used in X-ray crystallography:!2 least-squares measure!! goodness of fit between measured data and theoretical prediction allowing for both measurement and prediction errors iterative rounds varying dynamic model until best fit found for fewest number of variables Quantification of free ligand conformational preferences by NMR and their relationship to the bioactive conformation Blundell CD, Packer MJ and Almond A, Bioorganic & Medicinal Chemistry, 21 (17), 4987 4976..
Software implementation NMR data Proprietary software suite 4D-structure 2-3 weeks
Applicable to all ligands Lisinopril Carazolol DRUGS Hyaluronan CARBOHYDRATES Losartan Ivermectin Streptomycin PEPTIDES Methotrexate AngiotensinII TRH COFACTORS Acetyl-CoA
Streptomycin solution structure Data type No. of restraints NOESY NOEs 176 no-noes 93 Residual dipolar couplings 4.5-6% gel 35 Scalar couplings 5 -------------------------------------------- TOTAL 309 31 restraints per rotatable bond Composition of NOE restraints Classification NOEs nonoes All Intraresidue S1 9 5 14 R2 19 0 19 G3 32 1 33 ----------------------------------------------------------------- TOTAL 60 6 66 Data recorded in water at physiological temperature and ph Interresidue S1-R2 39 19 58 R2-G3 56 17 73 S1-G3 21 51 72 ------------------------------------------------------------------ TOTAL 116 87 203
4D-structure Mean positions for each conformational mode Full ensemble showing range of libration adopted in solution
2 conformational families Glycosidic linkage between ribose and glucosamine adopts two distinct conformations * * MAJOR 62 ± 16 % MINOR 38 ± 16 %
Linkage geometry distribution "#$%!"&$%!'$% $% '$% "&$% "#$ "#$%!"&$%!'$% $% '$% "&$% "#$%
Crystal structures Ribosome (natural target) 10 structures Transcriptional regulator TcaR 2 structures RNA aptamers 2 structures
Free vs primary bioactive "#$%!"&$%!'$% $% '$% "&$% "#$% "#$%!"&$%!'$% $% '$% "&$% "#$ Family 1 "#$%!"&$%!'$% $% '$% "&$% "#$%
Free vs other bioactives RNA aptamers 2 structures family 1 & family 2 Transcriptional regulator TcaR 2 structures family 1 & family 2 "#$%!"&$%!'$% $% '$% "&$% "#$ Family 2 Family 2 "#$%!"&$%!'$% $% '$% "&$% "#$%
Free vs other bioactives RNA aptamers 2 structures family 1 & family 2 Transcriptional regulator TcaR 2 structures family 1 & family 2 Di!erences between free and bioactive conformations impact observed binding a"nity Free solution structure predicts the diverse range of bioactive forms "#$%!"&$%!'$% $% '$% "&$% "#$ Family 2 Family 2 "#$%!"&$%!'$% $% '$% "&$% "#$%
Modes vs Bioactive Conformation(s) BrefeldinA Carazolol Streptomycin Ivermectin Lisinopril Imatinib Hyaluronan
In-house drug discovery Currently running several programs HI, H2L and LO phases Type A & B GPCRs, ion channels, PPI With and without co-crystal support Free Ligand Conformational Populations in Solution A Powerful Drug Discovery Tool!"#$%&'()*+,-.$/+0$1&2+/'&)/$&/$1,34$1&'()*+,Jonathan Byrne*, Thorsten Nowak, Barrie Martin & Charles Blundell C4X Discovery Ltd., Unit 310 Ducie House, 37 Ducie Street, Manchester, M1 2JW, UK; Contact email: Jonathan.Byrne@C4XDiscovery.com The dynamic interchange of accessible low energy conformations for ligand molecules has traditionally been predicted computationally or extrapolated from small molecule crystal structure data. Here we present a new quantitative and precise NMR methodology that provides experimentally-derived detailed conformational information for free ligands in aqueous solution. These 4D-structures give unprecedented insight into how to exploit and control ligand conformational preferences in drug design. The detailed conformational information on ligand molecules contained in 4D-structures can be applied to all stages and settings of the drug discovery process. In Hit Identification, accurate 3D-pharmcophore information derived from reference compounds permits the rapid identification of chemical equity independently of other screening approaches (e.g., VS, HTS). In Lead-Generation and Lead-Optimisation, 4D-structures can significantly impact compound design separately or in conjunction with traditional structure-based drug design approaches. Specific examples for a range of targets are presented, demonstrating how 4D-structures can be used to great effect in drug discovery. Case study Orexin-1 selective compounds 9 Hit-to-Lead *+,-+./+0!'(1 2+3+./+0!'(1 4,05%!'() #$%&!'() 7 6 Target Area 5 " 6 67+!8-0$3-$0+8!9:!-758!,;<+%!5%%$8-0;-+!-7+!39<3+,-!;<=!=9!<9-!0+,0+8+<-!-7+!37+.58-0&!=+>+%9,+=!;-!?@A!25839>+0& Consensus Ph4 Shape-o-phore + Project Start, 6 4D structures -> VS 38 compounds tested, 5 < 1µM & de novo design! Lead Generation Problem Solving 7 8 pic50 Orexin 1 9 Potent and Selective Compounds Design exploiting Ph4 & Shape-o-phore 86 compounds, 1 Synth. Chem., 1 Med. Chem., 9 months Where in drug design do ligand 4D-structures add value? 15% 85% Libration Major Modes 4K5Y1 Consensus Ph4 Elements Conformational Dynamics Complementing SBDD Ph4 Positive impact on: Project initiation, SBDD, Comp. Chem., potency optimisation; Sel. Improvements, Med. Chem. problem solving Shape-o-phore defines the steric and electronic factors which determine the conformational dynamics & selection in solution which related to affinity Comparing and contrasting bound v free solution structure impacts positively on SBDD How do we measure ligand 4D-structures? bimodal population 50:50 unimodal angle: -77±8 libration: 25±7 trimodal 56:38:6 NOE, ROE RDC, 3J,!" Small Molecule NMR coupled with proprietary software generates 243 modes Occupancy Developed 3 novel series of <10 nm potent and 1000-fold orexin-1 selective antagonists 86 compounds 1 Synth. Chem. 1 Med. Chem. 9 months pic50 Orexin 2 8 85% in 1 of 27 modes 45% in 1 of 9 modes Conformation index an experimentally determined dynamic solution structure which is predictive of the bioactive conformation NMR-based 4D solution structures map the complete conformational space a small molecule naturally oscillates through in solution Dynamic solution structures are resolved into prioritised!idealised" conformers (shapes) 56 7)88+/'9+&/:$;<:$;+=/:$><:$?),9)8=9):$.<:$!@+/4:$A<$;<$B<:$%),C:$.<$D<:$>=E=-+,&:$.<:$F=,'@=88:$G<$7<$HIJ5K6:$!"#$%&:$'(()HL"MN6:$"KOP"K< I6?83/1+88:$!<$%<:$Q=(R+,:$F<$><:$S$.82)/1:$.<$HIJ5K6:$*+,,%-".+/)0)1&2+/+."3)/4&1+5#%6!"!78)H5L6:$"NOLP"NLT<$ C4X Discovery information 2013. All rights reserved. Worked examples see Poster 18
Using 4D structures in design In the absence of co-crystal data 4D-structures provide an estimate of the bioactive conformation Comparison with di"erent sca"olds reduces uncertainties Potency optimisation by tracking changes vs activity until Ph4 is known Synergistic with co-crystal data Di"erences between free and bioactive conformations directly address potency 0º 0º 270º 90º 270º 90º 180º adjust populations 180º adjust means Database of solution conformational behaviour ~1000 bonds, ~100 structures de novo conformational design and isostere replacement
Summary Accurate quantification of free ligand conformational preferences by NMR New ensemble description minimising number of new variables Increased amounts of experimental data Software supported : 2-3 weeks per structure Applicable to all classes of ligands Their relationship to the bioactive conformation 4D-structures frequently predict the bioactive conformation Can therefore be used to drive HI, H2L and LO processes Can be used independently of or synergistically with co-crystal data charles.blundell@c4xdiscovery.com www.c4xdiscovery.com