Modelling the interaction between Irinotecan and efflux transporters inhibitors: A KPD tumour growth inhibition model including interaction components.
Alexandre Sostelly (1), Léa Payen (2), Benjamin Ribba (3), Attilio Di Pietro (4), Pierre Falson (4), Pascal Girard (1,5), Michel Tod (1,6)
(1) EA3738 Ciblage Thérapeutique en Oncologie, Faculté de Médecine Lyon-Sud, Oullins France; (2) Institut des Sciences Pharmaceutiques et Biologiques, Lyon France; (3) INRIA Rhône Alpes, Project Team NUMED, Ecole Normale Supérieure de Lyon, Lyon France ; ( 4) Institut de Biologie et Chimie des Protéines, Lyon France ; (5) INSERM, Lyon France ; (6) Hospices Civils de Lyon, Lyon France
ATP Binding Cassette (ABC) transporters are known to play an important role in drug absorption and distribution, normal tissue protection and anticancer drug resistance.
Although efforts to reverse drug resistance using P-glycoprotein inhibitors failed in the past, the use of other ABC transporter inhibitors in clinic has never been tested.
Breast Cancer Resistance Protein (BCRP) is an ABC transporter involved in the efflux of a wide range of substrates, such as, SN38, the active metabolite of irinotecan (CPT-11). BCRP inhibitors have been newly designed and optimized based on acridone derivatives. For instance, MBLI87 has shown high activity against BCRP efflux in in vitro studies with the advantage of not inhibiting P-glycoprotein .
A proof of concept study has been carried out in xenografted mice and has demonstrated the efficiency of this new drug against CPT-11 BCRP mediated resistance .
In order to optimize the therapeutic regimen and effects of this new drug combination, a model is necessary to predict the effect on tumour growth given both drug exposures.
To model the interaction between MBLI87 and irinotecan (CPT-11) in Severe Combined ImmunoDefiency (SCID) mice with either CPT-11 resistant or non CPT-11 resistant xenografts and to compare the effect of MBLI87 with the reference BCRP inhibitor, gefitinib. Our model includes a KPD component which accounts for drug kinetics when pharmacokinetics information is not available and the interaction of both BCPR inhibitors on CPT-11 cytotoxic effect.
Animals: 60 SCID mice were inoculated with either CPT-11 resistant tumour cells or non CPT-11 resistant tumour cells. Each mouse was implanted on the left and right flanks with the same cells. Mice were spread into 10 treatment arms: Water, CPT-11, Gefitinib, Ethanol (Gefitinib vehicle), MBLI87, Nano-Particles (MBLI87 vehicle), CPT-11+Gefitinib and CPT-11+MBLI87.
CPT-11 was administered by intra-peritoneal route at 30 mg.kg-1 3 days a week during 2 consecutive weeks followed by a 15 days rest period. Gefitinib was administered at 15 mg.mL-1 in water by gavage for a total of 75 mg.kg-1 following the same schedule as CPT-11. MBLI87 was administered by intra-peritoneal route as a 2.4 mg.kg-1 dose for 5 days a week during 2 consecutive weeks followed by a 15 days rest period.
The 2-week period of drug administration plus its 15 days rest period were considered as one therapy cycle. Mice thus received 2 cycles over 8 weeks. Tumour-bearing mice were randomized before receiving drugs, the day after cells implantation. Tumour measurements (length and width) were assessed every two days after the first drug administration. If the volume of one tumour exceeded 1800 mm3, the entire group was euthanized at the same time, for ethical reasons.
Data: At each day of measurements, the geometric mean of the four measures (length and width of tumours at right and left flanks) of each mouse was calculated to summarize information.
When CPT-11 was administered alone or in combination to mice with non CPT-11 resistant tumours, the tumour was not measurable. These data were not analysed.
There was no significant difference (based on analysis of variance for repeated measurements) between water group, i.e. the control group, and other vehicles treaments groups. Consequently these 3 groups (water, ethanol and nano-particles) were lumped in a single control treatment group.
Thus 45 mice were spread into 6 cohorts: control (N=18), CPT-11 (N=9), Gefitinib (N=6), MBLI87 (N=6), CPT-11+Gefitinib (N=3) and CPT-11+MBLI87 (N=3).
Different kinds of models were tested to model longitudinal tumour growth measurements. Since two drugs were administered, we first used the surface response model proposed by Minto and colleagues . Then we used a sigmoid effect model considering an additive effect to describe the interaction. We also modelled directly tumour growth profiles using 3 tumour growth models: exponential tumour growth model, tumour growth inhibition model proposed by Claret and colleagues  and the one proposed by Simeoni and colleagues called the modified-Gompertz model  with some modifications.
These modifications concerned the use of a KPD model  for describing drug kinetics, since no pharmacokinetic information was available in those animals. Consequently drug effects were dependant on the amount in this compartment. An interaction parameter was introduced to quantify the action of BCRP inhibitors on CPT-11 cytotoxic effect.
Models parameters were estimated by the FOCE method implemented in NONMEM VI software.
For fitting growth curves, a sequential approach was chosen. First, the control and monotherapy data have been modelled separately. Final estimates from this analysis were used as initial estimates, in a second step, where all the cohorts were modelled simultaneously.
At each step of the model building process, goodness-of-fit plots, individual plots based on posthoc distribution, parameters confidence intervals and OFV value guided the choice of the best model. Once the final model had been established, simulation based diagnostics were performed.
In our study, only one dose level was administered therefore, we were not able to describe the surface response and Minto's model has been rejected. Sigmoid effect model has been also rejected because it was not able to describe our tumour growth profiles. Moreover, these two models appeared to be overparametrised. Among the tumour growth models, the modified-Gompertz model from Simeoni and colleagues has been preferred to the Claret and exponential ones according to OFV and Bayesian Information Criteria values.
Model Specification: The final model was a tumour growth inhibition model with KPD and interaction components, the CPT-11 effect was related to the amount of the drug in the kinetic compartment. In case of joint administration, the CPT-11 effect was related to both the amount of CPT-11 present and of BCRP inhibitors. The final model is written as follows:
dACPT-11/dt = -Ke CPT-11*ACPT-11
dAgefitinib/dt = -Ke gefitinib*Agefitinib
dAMBLI87/dt = -Ke MBLI87*AMBLI87
dΦtumour/dt = [L0*Φtumour/(1+(L0*Φtumour/L1)Ψ)1/Ψ] - [K2 X*DRX*Φtumour]With:
DRX = Ke X*AX
X = CPT-11, Gefitinib, MBLI87
In case of joint administration, K2 parameter accounted for the effect of MBLI87 and gefitinib on CPT-11 cytotoxic effect:
K2 CPT-11= K'+K"*DRinhibitors
inhibitors = Gefitinib, MBLI87
- A is the amount of drug with a constant elimination rate Ke
- L0, L1 are the modified Gompertz model parameters which describe unperturbed tumour growth
- Ψ is the parameter allowing exponential to linear growth phase switch
- K2 is the drug potency
- K" is the parameter describing the interaction
Only tumour growth parameters, L0 and L1 were described to the individual level, all the others parameters were fixed effects in the population model.
Discussion: Exponential growth parameter was estimated at 0.06 day-1 and linear growth parameter at 0.2 mg.day-1. These values were in accordance with values reported by Simeoni and colleagues .
Potency of BCRP inhibitors were estimated at 10-2 mg-1. These two molecules didn't have any effect on cell growth. Their effect only consisted in reversing CPT-11 BCRP mediated resistance. CPT-11 is still active on CPT-11 resistant xenografts; its potency has been estimated at 0.3 mg-1.
A significant synergistic effect was found between MBLI87 and CPT-11 (K"=5.3) whereas none was found between gefitinib and CPT-11 (K"=10-2). Although there was no difference in tumour size kinetics between both cohorts, the interaction model confirmed that interaction was the strongest with MBLI87.
A modified tumour growth model including interaction and KPD components was built on pre-clinical data for a new BCRP inhibitor, MBLI87, and irinotecan. Our results showed that MBLI87 was able to revert CPT-11 resistance at a 20-fold lower dose compared to gefitinib. The model was accepted thanks to standard procedures. Future use of the model will be firstly optimizing a dose finding study in mice and secondly simulating effects of this new drugs combination in humans in order to prepare the design a phase 1 study.
 Boumendjel A, Macalou S, Ahmed-Belkacel A, et al. Acridone derivatives: Design, synthesis, and inhibition of breast cancer resistance protein ABCG2. BioOrg Med Chem 2007(15):2892-97.
 Arnaud O, Boumendjel A, Gèze A, et al. The acridone MBLI87 reverses in vivo breast cancer resistance protein-mediated resistance to CPT-11. J Cell Mol Med. Submitted
 Minto CF, Schnider TW, Short TG, et al. Response surface model for anesthetic drug interactions. Anesthesiology. 2000 Jun;92(6):1603-16.
 Claret L, Girard P, Hoff PM, et al. Model-based prediction of phase III overall survival in colorectal cancer on the basis of phase II tumor dynamics. J Clin Oncol. 2009 Sep 1;27(25):4103-8.
 Simeoni M, Magni P, Cammia C, et al. Predictive pharmacokinetic-pharmacodynamic modeling of tumor growth kinetics in xenograft models after administration of anticancer agents. Cancer Res. 2004 Feb 1;64(3):1094-101.
 Jacqmin P, Snoeck E, van Schaick EA, et al. Modelling response time profiles in the absence of drug concentrations: definition and performance evaluation of the K-PD model. J Pharmacokinet Pharmacodyn. 2007 Feb;34(1):57-85.