Productive and promiscuous hits Biophysics/label free assays in hit discovery and verification Johannes Ottl Novartis Institutes for BioMedical Research Basel Center for Proteomic Chemistry Label Free Technologies SBS Dresden June 2008 1
Agenda Why biophysics in drug discovery? Typical problems of a lead discovery project to be addressed Capabilities of our main biophysics technologies Biophysics in the lead discovery flow chart Biophysics technologies in hit evaluation Remarks / Outlook 2
Drug Discovery A continuous fight against bottlenecks and attrition rate Discovery Preclinical Development Clinical Development Drug Registration Drug Marketing Clinical Application IND Investigational New Drug Application NDA New Drug Application Appr. FDA Approval Launch Product Launch Drug Research and Discovery Process: Target Identification Assay/ Tool Development Hit Finding Hit Verification Hit Optimization Lead Optimization Focus area over the last decade optimized, automated, established Screening of >1.000.000 compounds yields 100s to 1000s of potential drug candidates Evolving new focus area high attrition rate, limited throughput, lack of streamlined processes How to identify the few good drug candidates?!?!? 3
Why biophysics? Increase of hit quality in drug discovery With biophysics assays we want to understand the mechanisms of binding and activity for our hit lists and drug candidates Goal: Advance hits to generate drug candidates more efficient Minimize wasted resources Identify & discard bad compounds early Confirm and characterize weaker hits and singletons Enable proper follow up for all hits Solid pre-filter for structural efforts Bona fide hits for chemistry Better selection for late stage biology, structural efforts, CADD, and chemistry Most biophysics methods are difficult, resource intensive, and yet quite low throughput! Biophysics are COMPLEMENTING and NOT REPLACING other existing and established assays and methods (e.g. orthogonal assays)! 4
Biophysics/label free toolbox Currently used tools at Novartis technology type of information Dynamic Light Scattering (DLS) Mass Spectrometry (MS) HT-SPR-like, NMR Protein folding assays (e.g. DSC) Surface Plasmon Resonance (SPR) Isothermal Titration Calorimetry (ITC) X-ray, NMR compound aggregation / solubility Yes/No binding; covalent?; activity assay Yes/No binding, K d Protein stabilization (ΔT structure ) stoichiometry, promiscuous binding?, K d,k on,k off stoichiometry, ΔH, ΔS 3D structure and dynamics, mechanisms different information content different resource/protein needs DSC: Differential Scanning Calorimetry DLS: Dynamic Light Scattering ITC: Isothermal Titration Calorimetry MS: Mass Spectrometry NMR: Nuclear Magnetic Resonance SPR-like : Surface Plasmon Resonance; Resonant waveguide grating; Biolayer Interferometry 5
Biophysics technologies in flowchart Examples for approaches applied Compound properties (DLS) Yes/No binding (HT- SPR-like, MS, NMR) Protein stabilization Binding affinities, K d, K i (HT- SPR-like, MS) Protein stabilization [Binding thermodynamics (ITC)] Binding kinetics: k on, k off (SPR) Protein Structure (NMR, X-ray) Binding thermodynamics (ITC) Hits from HTS, etc Including orthogonal assay and cpd LC-MS Validated hit (binding + activity) Hit expansion, optimization Validated hit series and analogues Drug candidates non-target binder promiscuous compounds (e.g. aggregating or irreversible) DSC: Differential Scanning Calorimetry DLS: Dynamic Light Scattering ITC: Isothermal Titration Calorimetry MS: Mass Spectrometry NMR: Nuclear Magnetic Resonance SPR-like : Surface Plasmon Resonance; Resonant waveguide grating; Biolayer Interferometry 6
Compound aggregation artifacts Dynamic Light Scattering (DLS) Compound aggregation is published as serious screening problem causing false positive hits Compound aggregation is per se not yet a problem, but has to be put in perspective with the compound activity! Aggregation/solubility of compounds is strongly assay and buffer dependant: e.g. aggregation of Tetraiodophenolphthalein starts at 0.5 or at 50 μm - depending on your buffer conditions! aggregation dynamic light scattering autocorrelation function 100 50 25 12.5 6.25 Frequent hitter and aggregator aggregates > 12.5μM inactive in NR assay N + N + 100 50 25 12.5 6.25 100 50 25 12.5 6.25 hit compound 1 aggregates >25 μm NR EC 50 = 0.7μM hit compound 2 NR EC 50 = 12μM aggregates > 12.5 μm 7
Compound aggregation and frequent hitter: Frequent hitter versus normal screening library Promiscuous compounds frequently active in HTS assays are often linked with the compound aggregation phenomena Normal screening library: 83% of compounds clean 7% aggregating 10% strongly aggregating / insoluble Light scattering 100<x=1000; 5% 20<x=100; 5% 10<x=20; 2% 3<x=10; 5% Sg/Bg x<3; 83% Frequent hitter library: 43% of compounds clean 20% aggregating 37% strongly aggregating / insoluble 100 < x = 1000; 22% 20 < x = 100; 14% 1000 < x; 2% Sg/Bg x = 3; 43% 10 < x = 20; 7% 3 < x = 10; 13% 8
Aggregator or insoluble? Aggregation is simply a transition from soluble to precipitating compound Particle size distribution 1.65μM 3.125μM 12.5μM 50μM Particle size distribution 1 10 10 2 10 3 10 4 10 5 R (nm) I O O I I I O 1 10 10 2 10 3 10 4 10 5 R (nm) 1 10 10 2 10 3 10 4 10 5 R (nm) 1 10 10 2 10 3 10 4 10 5 R (nm) 1 10 10 2 10 3 10 4 10 5 R (nm) O N HN N NH O Some compounds are large particles (precipitates), others are distributed over size ranges from aggregates to precipitates Compound aggregation/insolubility is an important, but of course not the only compound artifact! HO HO HO 0.1 O 0.1 H HO O H H 1 10 10 2 10 3 10 4 10 5 R (nm) O H O Chiral 1 10 10 2 10 3 10 4 10 5 R (nm) 9
Direct binding of compounds on proteins High throughput still challenging! HTS-compatible techniques are evolving (e.g. SPR-like : surface plasmon resonance; resonant waveguide grating; biolayer interferometry) SPR-like technologies measure binding events on immobilized target molecules (e.g. via changes in refractive index) Typical problems/ challenges however: Sensitivity: binding of compounds [e.g. MW 200] on protein [MW 40000] only 0.5% mass increase Immobilization: good immobilization strategy and high quality protein are pre-requisites for success Robustness: slight changes of components (e.g. Δ0.1% DMSO) and aggregating/insoluble or unspecific compounds falsify true binding signals binding 100 80 60 40 20 What is the correct K d fit range??? How to reliably resolve biphasic binding data? 0 DRC full range 1x10-6 1x10-5 0.0001 cpd conc (M)? binding 35 30 25 20 15 10 5 0-5 10 DRC zoom 1x10-6 1x10-5 cpd conc (M)
Direct binding of compounds on proteins SPR-like technologies we have already looked at Biorad Proteon XPR36 surface plasmon resonance, flow cell injection from 96w compound plate 6 compounds on 6 protein channels in parallel Biacore T100 (A100) surface plasmon resonance flow cell injection from 96w/384w compound plate 1 (4) compound on 4 (5) protein channels in parallel Fortebio Octet Red biolayer interferometry sensor tip with protein dips in 96w compound plate 8 compounds on 1 protein in parallel Corning EPIC resonant waveguide grating sensor with protein on bottom of 384w plate 384 compounds on 1 protein and 1 reference zone RU Response 40 35 30 25 20 15 10 5 0-5 pm shift 38 32 26 20 14 8 2-4 -100 0 100 200 300 400 500 600 700 800 Time 1x10-8 1x10-7 1x10-6 1x10-5 cpd conc (M) s 11
Feasibility study for HTS hit list follow up 2 Examples for SPR-like technologies HTS hit list (690 compounds) for direct protein binding true active binder false positives known artifact compounds from HTS ( challenge compounds ) non-binder From the biophysics data we want to judge if the hits are target binders non-binders promiscuous compounds 5% 3% Both results look very promising, but data need close inspection for judgments! correct result 8% wrong result EPIC (Corning) 91% 10% non conclusive result 12 A100 (Biacore( Biacore) High throughput Medium throughput full dose response single data point only Lower sensitivity High sensitivity Sg/Bg can be borderline weaker compounds found Plate-based sensor Flow-cell sensor compounds can t harm compounds harm each other higher protein low protein consumption consumption 81%
Affinity Mass Spectrometry Screening Qualitative and quantitative affinity screening No protein immobilization and protein size limitations as with SPR-like binding assay FluphenazineLuna_A01_080228085145 #247 RT : 1.26 AV: 1 NL: 1.67E7 T: IT MS + c ESI sid=10.00 Full ms [200.00-1200.00] 438.18 100 90 incubation Relative Abundance + 80 rapidsec 70 60 50 40 30 20 10 460.22 217.06 241.09 279.36 324.26 384.25 406.35 0 200 250 300 350 400 Compounds Target protein Protein-binder complex 450 500.21 538.97 586.35 614.8 500 550 600 Ion trap MS (ESI mode) Qualitative: Identify target Yes/No binder hits from mixtures of test compounds Compound 10 0-10 incubation rapid SEC Reporter1 Reporter2 -(% Marker) -20-30 -40-50 -60-70 -80-90 -100 Target protein Protein-binder complex -7 1x10-6 1x10 cpd conc (M) -5 1x10 0.0001 Ion trap MS (ESI mode) Ki = 0.487 Quantative: Quantify reporter compounds and derive from them mechanistics and affinity for the test compound 13
Example for biophysics data lead compound characterization Example for a fully characterized active agonist compound: 5.5 shift in the protein stabilization assay in presence of compound Saturating binding signals in the Proteon XPR36 SPR and in the EPIC system, K d, on- and off-rate determined 1:1 stoichiometry confirmed in Isothermal Titration Calorimetry (ITC) Proteon XPR36 Proteon XPR36 EPIC ITC k d (um) g pm shift 38 32 26 20 14 8 2-4 EPIC 1x10-8 1x10-7 1x10-6 1x10-5 cpd conc (M) k on (1/Ms) k off (1/s) 0.851 8296 0.00706 0.638 / / 0.810 / / ITC 1000 500 0-500 -1000-1500 -2000-2500 -3000-3500 -4000 30 35 40 45 50 55 60 65 70 75 80 protein stabilization assay 14
Remark/Outlook We want to apply biophysics technologies systematically to annotate large numbers of hit compounds with biophysics information BUT: Moving away from dealing with few dozens of well characterized, freshly dissolved, and soluble compounds towards hundreds or thousands of uncharacterized archive solutions of potentially insoluble compounds is still a challenge! However the technologies, methods, and processes are developing We are in a learning curve Probably there will be a mix of technologies to be applied depending on the compound numbers and information needed 15
Acknowledgement Label Free Technologies Group Christian Bergsdorf Danielle Barlier Oliver Esser Pascal Bernet Vincent Acker Many others at Novartis: Andreas Boettcher Lukas Leder Marco Meyerhoefer Anke Blechschmidt Lorenz Mayr Peter Fuerst Sylvain Cottens Hans Widmer Ulrich Hommel George Addona Eric Martin Travis Stams Peter Fekkes Corning Volker Eckelt Jack Fang Tony Frutos Gordon Shedd Ute Verspermann Biorad Tsafrir Bravman Tamar Telem-Shafir Thomas Krueger 16