Mechanistic Modeling of in vitro Assays to Improve in vitro/in vivo Extrapolation

Transcription

1 Mechanistic Modeling of in vitro Assays to Improve in vitro/in vivo Extrapolation Grace Fraczkiewicz Viera Lukacova Jim Mullin 1

2 OVERVIEW MembranePlus a software platform for simulation of in vitro transport assays: Mechanisms Inputs Models Case Studies In vitro model validation In vitro/in vivo extrapolation 2

3 MECHANISMS: TRANSWELL 3

4 MECHANISMS: SANDWI C H HEPATOCYTES Collagen is assumed to not affect transport processes Model is also applicable for plated hepatocytes when bile volume is not considered 4

5 MECHANIMS: SUSPENDED HEPATOCYTES 5

6 I N P UTS: COMPOUND PROPERT I ES Physicochemical properties (in vitro or in silico) Enzyme and transporter settings applicable only to cell based assays 6

7 I N P UTS: EXPERIMENTA L S E T T I NGS Experimental Setup specific for selected Membrane type (cell monolayers, PAMPA, sandwich hepatocytes, suspended hepatocytes) Additional compound/assay specific settings 7

8 MODELS: MEMBRANE ENTRY/ E X I T R AT E STRUCTURE BASED MODEL Observed vs. Predicted on 44 training datasets V Intercept C1 logp C2 M _ RNG C3 HBDH HBD C4 HBAo log i All experimental time points to 2 hr Membrane entry and exit rates for anions and cations are determined based on logd vs. ph profile 8

9 MODELS: LYSOSOMAL TRAPPING Lipophilic Amines LogP > 1 pka > 6.5 Lysosome V i and V o correspond to the membrane/water entry and exit rate ph = 4-5 Lysosome membrane Kazmi F., Drug Metab. Disp. 41(3):897 (2013) Cell ph = 7.4 9

10 MODELS: ENZYMES AND TRANSPORT E RS Kinetics of carrier-mediated transport and metabolism is calculated using Michaelis-Menten kinetics: u metab = å i æ ç è i V max i K m intracell c u ö intracell + c u ø efflux i i V K max i m c c intracell u intracell u influx i V max units: mm/s (mmol/l/s) K m units: mm (mmol/l) General units converter allows converting these into different types of units Transporter types: Influx and Efflux Transporter locations: Apical and Basolateral (where applicable) i V max i Km c c buffer u buffer u 10

11 Case Studies: in vitro Model Validation 11

12 Case Study 1: Sodium Taurocholate Uptake into Bile 12

13 S ODIUM TAUROCHOLAT E Guo, ISSX 2014 OST alpha/beta 13

14 MEMBRANEPLUS MODEL ASSUMPTIONS Assumptions No protein binding for Na Taurocholate No stirring Complete monolayer of cells (100% viability) ADMET Predictor values for properties and transport model parameters calculated using calibration Literature values for K m were used as a starting point for building the model. K m values are similar across species Swift-Mol-Pharm (2) Transporter K m (mm) Cells Literature Source Cell Assay Inputs Feed Solution Conc. 1,2.5 mm BSA 4 % Well size 24 well Volume 0.3 ml Cell Vol pl Cell Layer Thickness 18.6 micron Cell Den 0.4 Mcell/well OST alpha/beta 25.8 Human Swift-Mol-Pharm (2) Overall Uptake 19 Rat Schwarz-Eur-J-Biochem NTCP 6 Human J-Exp-Biol BSEP 5 Rat J-Exp-Biol

15 S I MULAT I ON RES ULT Experimental K m values utilized from literature and V max values fit to data (all remaining properties were predicted by ADMET Predictor). Transporter Type K m (Exp) (mm) V max (mmol/s/l) BSEP Efflux E-03 OST alpha/beta Efflux E-02 NTCP Influx E-02 Efflux Phase 15

16 Case Study 2: Quantification of Influx Transport vs. Metabolism Statin Compounds Suspended Hepatocytes 16

17 Q UA NTIFY THE RELAT I VE IMPORTANCE OF INFLUX T R ANSPORT (OAT P 1 B1) VS. METABOLISM Paine, DMD (2008) 36: Media and whole cell concentration data for atorvastatin, cerivastatin, and indomethacin. Used a simple compartmental model to extract clearance values 17

18 MEMBRANEPLUS MODEL RES ULTS I Atorvastatin Cells Media Media Cells Cerivastatin Cells For this compound, lipid bilayer permeability model was fitted due to poor prediction from structure Cells Media Media 18

19 Metabolism Rate (ul/min/10 6 cells) MEMBRANEPLUS MODEL RES ULTS II Influx Rate (ul/min/10 6 cells) Metabolism Rate (ul/min/10 6 cells) Influx Rate (ul/min/10 6 cells) Paine, DMD (2008) 36: Influx Influx Metabolism Atorvastatin Results Metabolism Cerivastatin Results V max 3A4 (umol/s/l) V max 3A4 (umol/s/l) 3.74E-02 V max OATP1B1(umol/s/L) 31.9 V max OATP1B1(umol/s/L) Cl met (ul/min/mcell) 63 (68 Lit.) Cl met (ul/min/mcell) 20 (17 Lit.) Cl upt (ul/min/mcell) 471 (375 Lit.) Cl upt (ul/min/mcell) 704 (413 Lit.) 19

20 MODEL COMPARISON The mechanistic model in MembranePlus achieved similar result as the simpler compartmental model with fewer fitted parameters Compartmenal model* Membrane Plus Intracellular volume Fitted System parameter Cell membrane volume Fitted System parameter Membrane/water partitioning (kmem) Fitted Predicted from compound properties Active uptake Fitted Fitted Passive diffusion Fitted Predicted from compound properties (atorvastatin only) Metabolism Fitted Fitted *Paine, DMD (2008) 36:

21 Case Studies: in vitro to in vivo Extrapolation 21

22 Case Study 3: Metabolic IVIVE 22

23 P ROPAFENONE HUMAN HEPATOCYTE DATA Reports indicate it is a CYP2D6 substrate and has saturable dose dependent kinetics. In vitro data from literature was used to fit intracellular unbound Km and Vmax Komura, Drug metabolism and disposition 33.6 (2005):

24 F I T Km AND Vmax TO i n v i t r o DATA K m = mm V max = 9.27E-02 mmol/s/l cytosol (Converts to: 2.17E-02 nmol/min/10^6 cells) K m fitted in mechanistic model was lower than the one reported in paper which is indicative of K m based on unbound intracellular concentration 24

25 P R E DICT in vivo PK PBPK model: Kps predicted using default (Lukacova) method in GastroPlus CYP2D6 clearance extrapolated from fitted in vitro values (shown on previous slide) 70 mg IV bolus dose Data: Hollmann, Cardiac Arrhythmias, Springer

26 Case Study 4: Transporter IVIVE 26

27 D I GOX I N: FIT INTRACELLULAR Km i n v i t r o Published data on nonlinear Papp vs. donor concentration for Digoxin were used to fit Pgp Km (intracellular unbound) and Vmax Data from: Troutman and Thakker, Pharm. Res., Vol. 20, No. 8:

28 Papp [10^6 x cm/s] D I GOX I N: MODEL VERIFICAT I ON Km obtained from fitting to a published dataset was used to predict concentration-time profiles from another dataset Vmax was adjusted to account for different expression levels of Pgp in different systems Donor (Apical) Concentration [um] 28

29 D I GOX I N: PREDICT i n v i v o A B S ORPTION A B C A: Observed (symbols) vs. predicted plasma conc. (blue) and urinary excretion (red) of digoxin (Ochs, 1978). B: Observed (symbols) vs. predicted plasma conc. (blue) of digoxin for a PO formulation with 6.5 mm radius particle size (Jounela, 1975). C: Observed (symbols) vs. predicted plasma conc. (blue) of digoxin for a PO formulation with 51 mm radius particle size (Jounela, 1975). All simulations are using the fitted intracellular unbound P-gp Km value of 95.3 mm 29

30 Case Studies 5: Lysosomal Trapping and Absorption 30

31 LYSOSOMAL TRAPPING OF LIPOPHILIC CAT I ONS Drug Log P Basic pka T max (h) Lipophilic Amines LogP > 1 pka > Kazmi F., Drug Metab. Disp. 41(3):897 (2013) 31

32 LYSOSOMAL TRAPPING OF LIPOPHILIC CAT I ONS Hepatic perfusion experiment with and without H+ ionophore Monensin Empty symbols represent controls and solid symbols represent monensin treatment. Dashed and solid lines stand for fitted data in control and treatments, respectively. Drug Antipyrine Atenolol Propranolol Monensin Effect No Effect Minor Effect Strong Effect Siebert GA, JPET 308(1):228 (2004)) 32

33 LYSOSOMAL TRAPPING OF LIPOPHILIC CAT I ONS Caco-2 Apical-to-Basolateral permeability experiment Propranolol Propranolol + Bafilomycin Ibuprofen Testosterone Solid squares Apical compartment Solid triangles Basolateral compartment Empty squares Cell monolayer Lines represent fitted model results Heikkinen AT, JPET 328: 882 (2009) 33

34 CONSEQUENCES i n v i v o : D ESIPRAMI NE Desipramine physicochemical properties Samant T et al. CPT: PSP, 6(5): , High permeability and intestinal solubility Late Tmax Kurtz D.L. et al. CPT 1997, 62:

35 50 mg dose CONSEQUENCES i n v i v o : D ESIPRAMI NE Fu ent = 0.55% Fu ent = 100% 100 mg dose Samant T et al. CPT: PSP, 6(5): ,

36 50 mg dose CONSEQUENCES i n v i v o : D ESIPRAMI NE Fu ent = 0.55% Fu ent = 100% 100 mg dose Samant T et al. CPT: PSP, 6(5): ,

37 METHYLPHENIDAT E MEMBRANEPLUS SIMULAT I ON Lysosome S+LogP = 2.02 (AP 7.2) S+pKa = 8.56 (Base) (AP 7.2) Cytosol Cell Membrane Total cell Lysosome Membrane Apical Compartment Basolateral Compartment 37

38 METHYLPHENIDAT E MEMBRANEPLUS SIMULAT I ON S+LogP = 2.02 (AP 7.2) S+pKa = 8.56 (Base) (AP 7.2) Cytosol Fu = 4.37% Total cell 38

39 METHYLPHENIDAT E GASTROPLUS SIMULAT I ON Fu,ent = 100% Fu,ent = 4.37% 39

40 S UMMARY Mechanistic cellular simulations: Help to separate drug and system parameter inputs (similar to in vivo PBPK models) for easier translation between systems Are important to assess critical mechanisms affecting drug absorption, distribution and elimination pathways: Contributions of passive and active drug transport Interplay between drug metabolism and cellular uptake Disposition in bile Distribution within cellular structures (lipid bilayers, lysosomes) Facilitate extraction of variety of relevant parameters for more accurate IVIVE 40

41

42 Additional Slides 42

43 COMPOUND X - C L ASS II/IV Lipophilic (log P > 4) Moderate base (pka 3.2 and 6.2) Low (0.001 mg/ml), ph dependent solubility Moderate gut permeability (1.48 x 10-4 cm/s) Estimated bioavailability ~30% Are the different (fitted) precipitation and gastric emptying times under fasted & fed conditions masking something else in the model? Zhang et al. AAPS PharmSciTech 2014 January 17 43

44 COMPOUND X FA STED STAT E MODEL DEVELOPMENT Measured & ACAT Default Model Parameters Default precipitation Optimized precipitation Dissolved Absorbed Enterocyte Portal Vein Systemic Circulation Absorption through enterocyte to portal vein is too fast predicting Precipitation low can t Tmax account even for with delayed onset alone precipitation. Hence, need to optimize gastric emptying time 44

45 COMPOUND X : GASTROPLUS SIMULAT I ONS WITH MEMBRANEPLUS fu, E NTEROCYTE = 3.47% Optimized precipitation & simulated Fu, Enterocyte Dissolved Absorbed Enterocyte Portal Vein Systemic Circulation The lag between absorption into enterocyte and basolateral clearance into portal vein captures the extended Tmax No changes to default GI physiology required 45

46 COMPOUND X FOOD EFFEC T PREDICTIONS ACROSS D OSES simulated MembranePlus Fu, Enterocyte input Optimized precipitation from low dose/fasted state PK data >> predicted remaining 3 study arms Default ACAT fasted/fed physiology parameters 46

47 MEMBRANEPLUS Data Analysis Instant permeability output from molecular structure or experimental data. Unbound Intracellular concentrations (Membrane, cytosol, lysosome, etc.) In vitro K m and V max for enzymes and transporters Parameter sensitivity analysis (on logp, shaking rate, ph etc.) Transwell, sandwich hepatocyte, and suspended hepatocyte models Assay Prediction Permeability Concentrations in cellular structures (Membrane, cytosol, lysosome) 47