Combination therapy of P. aeruginosa with special reference to modeling of polymyxins in vitro and to preliminary animal models

Transcription

1 Combination therapy of P. aeruginosa with special reference to modeling of polymyxins in vitro and to preliminary animal models April 20 th, 2010 Jürgen B. Bulitta, PhD Authors Copyright All rights reserved.

2 Target Goals Anti-infective therapy Maximize Killing Genotypic resistance Pre- vent Link Adaptive resistance Phenotypic resistance (incl. persisters) Pre-existing mutants De novo mutation(s) during therapy Authors Copyright All rights reserved. 2

3 Sometimes a Single Drug (Man) just cannot Achieve the Target Goals Most problematic infections: 1. Pre-existing resistant bacteria present in a high initial inoculum. 2. De novo formation of resistant mutants during long therapy or due to error prone replication. 3. Phenotypic tolerance of bacteria at the infection site (CSF, CF / mucus). 4. Sequestered infection sites. 5. Immuno-compromised patients. Sisyphos by Franz von Stuck, 1920 Authors Copyright All rights reserved. 3

4 Best PK/PD index: T>MIC, AUC/MIC, C max /MIC Beta-lactams Glycopeptides Inhibition Cell wall Quinolones Nitroimidazoles DNA RNA cell membrane Tetrahydrofolic acid synthesis misreading Proteins Polymyxins Stimulation of loss Replication of cell Sulfonamides Trimethoprim Rifampicin PK/PD indices for cell kill and for prevention of resistance differ within the same drug! Authors Copyright All rights reserved. Aminoglycosides Tetracyclines, Chloramphenicol, Macrolides, Linezolid, Clindamycin Ambrose PG et al. CID 2007, 44: Gumbo T et al. AAC 2007, 51: Louie A et al. AAC 2008, 52:

5 Rapid killing and inoculum effect of colistin in vitro 5

6 Log10 (CFU/mL) Inoculum effect of colistin vs. P. aeruginosa PAO1 Inoculum: 10 6 CFU/mL 10 8 CFU/mL 10 9 CFU/mL Individual curve fits Initial inoculum: 10 4 CFU/mL 10 6 CFU/mL 10 8 CFU/mL 10 9 CFU/mL Time (h) Plates with no colonies plotted as zero Population predictions Colistin sulfate (x MIC) Leave-one-inoculum-out internal cross-validation No colony on plate plotted as 0. Bulitta JB et al. Antimicrob Agents Chemother, 2010 March. 6 Funding : R01AI079330, NIAID.

7 Structural model for colistin vs. P. aeruginosa Growth half-life: t 1/2,S t 1/2,I t 1/2,R Susceptible population Intermediate population Resistant population 1 st -order natural death k d k 2S k 2I k 2R Inhibition Imax Kill, IC 50 k d Synthesis Signal molecules Synthesis k d 2 nd -order killing by colistin 1 st or 2 nd -order process Synthesis of signal molecules Inhibitory effect Illustration from: Storm DR et al. Ann Rev Biochem k deg Ca 2+ Mg 2+ Target site model Competitive inhibition Kd Cations Kd Colistin Membrane binding sites occupied by Ca 2+ or Mg 2+ EC 50, Hill Colistin in Colistin (C Colistin,eff ), fast broth (C Colistin ) active at target site Bulitta JB et al. Antimicrob Agents Chemother, 2010 March. Funding : R01AI079330, NIAID.

8 Adaptive resistance to colistin and inter-conversion of sub-populations 8

9 Translation to 1-compartment infection model: Colistin vs. P. aeruginosa ATCC Colistin half-life set to 4h. Data: Bergen PJ et al., ICAAC 2008, A Modeling: Bulitta JB et al., American Conf. on Pharmacometrics, Funding : R01AI079330, NIAID.

10 Translation to 1-compartment infection model: Colistin vs. P. aeruginosa ATCC Therapia magna (sterilisans) by colistin Paul Ehrlich ( ) Colistin half-life set to 4h. Therapia magna sterilisans: Eradication therapy with ONE large dose. Therapia fractionata sterilisans: Eradication therapy with fractionated doses. Data: Bergen PJ et al., ICAAC 2008, A Modeling: Bulitta JB et al., American Conf. on Pharmacometrics, Funding : R01AI079330, NIAID.

11 Mechanism-based model for colistin vs. P. aeruginosa ATCC Sub-population dynamics model with four sub-populations; formation of one intermediate sub-population is induced by colistin Target site model Mg 2+ Ca 2+ Competitive inhibition Kd Cations Kd Colistin Colistin conc. in broth (C Col ) Membrane binding sites occupied by Ca 2+ or Mg 2+ EC 50, γ Induction: Smax SI, Susceptible population (S) pre-existing Less susceptible population (L) pre-existing fast Active Colistin (C Col,eff ) at target site Colistin Loss SC 50,SI Fr IS Fr RL Fr SL Fr LS Inducible intermediate pop. (I) not pre-existing Fr LR Resistant population (R) not pre-existing Bulitta JB et al., American Conf. on Pharmacometrics, Bergen PJ et al., ICAAC 2008, A st-order, 2nd-order, or saturable process Inhibitory effect Competitive inhibition Funding : R01AI079330, NIAID.

12 Mathematical modeling methods Nonlinear mixed-effects modeling using the state-of-the-art Monte Carlo Parametric Expectation Maximization (MC-PEM) algorithm in S-ADAPT (version 1.56) parallelized on a computer cluster or pooled analysis in NONMEM VI. LSODA differential equation solver that can handle both stiff and non-stiff systems. Life-cycle model [1] to describe bacterial replication. All viable counts (including plates with no colonies) for each antibiotic alone and for the combination fitted simultaneously. Additive error on log-scale for CFU counts 100 CFU/mL. Low CFU counts were fit on linear scale as number of colonies per plate. Poisson error was included for these low colony counts. 1: Bulitta et al. Antimicrob Agents Chemother 2009, 53: : Bulitta & Yang et al. Antimicrob Agents Chemother 2010 Mar 8. 12

13 Parameter estimates from nonlinear mixed-effects modeling (S-ADAPT) and a pooled fit (NONMEM) Parameter Symbol Unit Estimate (%SE) 5-95% percentile from leave 20% out NONMEM S-ADAPT cross-validation Log 10 (Initial inoculum) Log 10 CFUo 6.14 (3.9%) 6.16 (2.8%) 6.14 [ ] Half-life of growth lag-time Ln(2) / k lag min 31.5 (60%) 26.8 (13%) 31.7 [ ] Mean generation time at low signal molecule conc. MTT 12 = k 12-1 Bulitta JB et al., American Conf. on Pharmacometrics, min 20.5 (12%) 23.5 (22%) 20.5 [ ] Doubling rate constant k 21 h (fixed) 50 (fixed) 50 (fixed) Maximum population size CFU max CFU/mL 7.93 (0.9%) 7.99 (0.8%) 7.94 [ ] Ratio of transfer rate constant (k 12 ) from state 1 to state 2 relative to the susceptible pop. for less susceptible population frc 12,L (13%) (32%) [ ] for resistant population frc 12,R 1 (fixed) 1 (fixed) 1 (fixed) for inducible intermediate pop. frc 12,I 1 (fixed) 1 (fixed) 1 (fixed) Second order killing rate constants relating colistin (base) concentrations at the target site to the rate of killing for susceptible population k 2S L/(mg h) 30.1 (12%) 27.8 (34%) 29.3 [ ] for less susceptible population k 2L L/(mg h) (16%) (49%) [ ] for resistant population k 2R L/(mg h) 0 (fixed) 0 (fixed) 0 (fixed) for inducible intermediate pop. k 2I L/(mg h) 1.03 (16%) (63%) 1.04 [ ] Log 10 fraction of cells converting from one population to another during one growth cycle from population L to S Log 10 Fr LS (26%) (29%) [-8.73 to -0.46] from population R to L Log 10 Fr RL (10%) (13%) [-0.88 to -0.47] Log 10 (Fr SL / Fr LS ) (2.6%) (9.0%) [-7.27 to -6.06] Log 10 (Fr LR / Fr RL ) (23%) (7.9%) [-11.9 to -4.28] from population I to S Log 10 Fr IS (5.1%) (26%) [-0.57 to -0.44] Maximum fraction of cells converting from pop. S to I Colistin (base) conc. causing with 50% of Smax SI Log 10 Smax SI (66%) (63%) [ to ] SC 50,SI mg/l 50 (fixed) 50 (fixed) 50 (fixed) Both estimation methods (programs) yielded consistent results. Bergen PJ et al., ICAAC 2008, A-1671.

14 Sometimes, single agent therapy just can t get the job done WHAT ABOUT COMBINATION THERAPY AND PREVENTION OF RESISTANCE? 14

15 T>MIC, AUC/MIC, C max /MIC How can these indices be applied to optimize drug combinations? Case I: Drug A: AUC/MIC Drug B: AUC/MIC Combination: Sum of AUC/MIC? Quo vadis? Case II: Drug A: T>MIC Drug B: AUC/MIC Combination:??? Applying PK/PD indices to combination therapy is difficult. Many antibiotics bind to more than one receptor. Mechanistic knowledge about the relationship between receptor occupancy and bacterial responses (incl. resistance) is critical. Authors Copyright All rights reserved. 15

16 Unique Receptor Occupancy Patterns can be used to Rationally Optimize Combination Chemotherapy DRUG A DRUG B Receptor 1 Receptor 2 Killing via pathway A Killing via pathway B Bacterial killing Enhances killing Receptor 3 Inhibit (upregulation) Phenotypic resistance mechanism(s) Eradicate Genotypically susceptible bacteria Genotypically intermediate bacteria Mutation Spontaneous or error-prone mutation Genotypically resistant cells Pre-existing vs. de novo formation Phenotypically resistant nonreplicating persisters Mechanism-based modeling integrates time course & probabilities Authors Copyright All rights reserved. 16

17 Sub-population synergy Drug A kills the resistant sub-population of drug B & vice versa. Susceptible to A Susceptible to B Resistant to B Killed by Drug A >99% of cells susceptible to drug A and B Drug A Drug B Resistant to A Drug B Example of sub-population synergy: Imipenem & colistin vs. P. aeruginosa Bergen PJ et al., ICAAC 2009, poster A

18 Colistin and imipenem alone & in combination against Pseudomonas aeruginosa at two initial inocula A: Colistin alone B: Imipenem alone C: Combination Legend for Panel C: Colistin / Imipenem (xmic) Log 10 (CFU/mL) Legend for Panels A & B / / / / 16 4 / / 4 4 / 16 8 / / 4 8 / / / 4 16 / 16 Time (h) Bergen PJ et al., ICAAC 2009, poster A Funding : R01AI079330, NIAID. 18

19 Mechanism-based Synergy for Antibacterial Combinations Susceptible sub-population for colistin Growth Killing by colistin Resistant sub-population for colistin Killing by rifampicin alone not shown in the diagram. Bulitta JB et al., ICAAC 2009, poster A Enhance EC 50,Rif = 0.01 mg/l (fixed) Emax = 244 Rifampicin Funding : R01AI079330, NIAID. Li J et al., ICAAC 2009, poster A

20 Log (Colony Forming Units /ml) Rifampicin Enhances Rate of Killing by Colistin time-kill studies Acinetobacter baumannii A: Colistin alone B: Rifampicin alone C: Combination Legend for Panels A, B, & D (mg/l) Pseudomonas aeruginosa D: Colistin E: Combination Time (h) 32 mg/l 32 mg/l Rifampicin (16 mg/l) showed <0.5 log net killing in monotherapy. Col / Rif (mg/l) 0 / / / / 16 4 / / 2 4 / 16 8 / / 2 8 / 16 Funding : R01AI079330, NIAID. Time (h) Colistin / Rif. (mg/l) 0 / / / / / / 16 2 / / 1 2 / 2 2 / 8 2 / / / 1 20

21 Mechanistic synergy: Colistin increases the effective intracellular concentration of ciproflox. potentially via interference with efflux transporters Susceptible subpopulation for CIP Killing by colistin Growth Killing by ciprofloxacin Resistant subpopulation for CIP Killing by colistin Colistin Colistin reduces the EC50 of ciprofloxacin Bulitta JB et al., ECCMID Funding : R01AI079330, NIAID. 21

22 Curve Fits: Colistin + ciprofloxacin Colistin susceptible strain vs. P. aeruginosa CIP (mg/l) CIP (mg/l) COL (mg/l) COL (mg/l) COL / CIP 0 (mg/l) / / / / / 8 2 / 0 2 / / 1 2 / 8 16 / 0 16 / / 1 16 / 8 COL / CIP 0 / / / / / 4 4 / 0 4 / / / 4 8 / 0 8 / / / 4 Bulitta JB et al., ECCMID Funding : R01AI079330, NIAID. 22

23 Transition to man 23

24 In vivo protein binding a truly exciting story for colistin Dudhani RV. et al. Antimicrob Agents Chemother 2010; 54:

25 PK of colistin (base) in mice Dudhani RV. et al. Antimicrob Agents Chemother 2010; 54:

26 PK/PD indices in neutropenic animals Dudhani RV. et al. Antimicrob Agents Chemother 2010; 54:

27 PK/PD parameter estimates in mice Dudhani RV. et al. Antimicrob Agents Chemother 2010; 54:

28 PK/PD index value for certain killing endpoints Dudhani RV. et al. Antimicrob Agents Chemother 2010; 54:

29 Population PK of colistin in CF-patients and in Cystic Fibrosis and Critically ILL patients CMS CMS Colistin Colistin Li J et al. JAC 2003; 52: Plachouras D. et al. AAC 2009; 53:

30 Modeling the impact of the immune system 10 4 to 10 5 CFU/g Log 10 CFU/g or Conc. (mg/l) Hope WW, Drusano GL, et al. AAC 2007, 51: Time (h) 30

31 Conclusions 1. Colistin is a very promising component of our armamentarium against MDR gram-negatives. 2. The rapid killing and rapid emergence of resistance to colistin in vitro suggests administering a large initial dose of colistin and a short duration of therapy. 3. PK in special patient groups needs to be considered. 4. Synergy in cell kill and prevention of resistance of colistin with a variety of compounds in vitro warrants studies in vivo and in the hollow fiber system. 5. Rational development of combination regimens with colistin supported by mathematical modeling holds great promise. 31

32 A Global Team Approach Team of Alan Forrest, Brian T. Tsuji (Buffalo, NY, USA) and Jurgen Bulitta (Albany, NY) Roger Nation s and Jian Li s Team in Melbourne, Australia And a series of other collaborators, including our colleagues (David Z D Argenio & Robert J Bauer, et al.) writing the mathematical software tools. Funding: Colistin work supported by R01AI from NIAID, NIH.