Physiologically Based Pharmacokinetic Model For The Prediction Of Olaquindox Residues In Porcine Edible Tissues

Olaquindox (OLA), as a feed additive with antibacterial and growth-promoting effects, is widely used in pig industry in China. The existence of OLA and its metabolites in animal derived foods can rise to a series of food safety issues and pose an adverse effect on pig industry, because of their phototoxicity, nephrotoxicity, genotoxicity, potential mutagenicity and carcinogenicity. To reduce the incidence of violative residue, OLA has been only allowed to be used in pigs weighing less than 35 kg at the level of 50-100 mg/kg in China. Methyl-3-quinoxaline-2-carboxylic acid (MQCA), one of its metabolits, is designated as the marker residue. The maximum residue limit (MRL) is fixed at 50μg/kg and 4μg/kg for the liver and muscle, respectively. A 35-day withdrawal time (WDT) has been recommended for OLA premix in pigs. However, the established MRL and WDT can still not effectively prevent the violative residues of OLA in animal derived foods because of its extra-label uses in practice. In addition, the traditional residue monitoring methods may unnecessarily waste a large number of resources. It has been proved that physiologically based pharmacokinetic (PBPK) model had the ability of in vivo residue prediction. Its use in veterinary drugs residue monitoring can meet the shortfall of these traditional methods. However, most of PBPK models can neither forecast the residue depletion of metabolites in food animals nor take into account the variability in predicted results, and therefore can not meet the requirements of residue monitoring of OLA. To improve the predictive ability of the current PBPK model and ensure the safety of animal derived foods, a predictive PBPK model for OLA in pigs was developed and extrapolated to quinocetone (QCT) and cyadox (CYA).1. Development of the quantitative analysis methodSeveral high-performance liquid chromatography (HPLC) methods were developed for MQCA in porcine edible tissues and body fluids. Muscle, liver and kidney samples were handled using a modified procedure by Wu. Adipose sample was extracted with dipotassium phosphate buffer and cleaning up using Waters Oasis MAX cartridge. Urine sample was extracted with acid ethyl acetate then cleaned up using thin layer chromatography. Acid plasma sample was extracted with ethyl acetate and detected directly. The limit of quantification of these HPLC methods were 4μg/kg for plasma, muscle and fat samples,10μg/kg for liver samples,20μg/kg for kidney samples,100μg/kg for urine samples, respectively. The recovery for MQCA in all samples ranged from 70% to 120% with relative standard deviation less than 15%. Calibration curves for edible tissues and body fluids were also established. Good linear correlation (r>0.92) was achieved within the concentration range of 4μg/kg-640μg/kg. These sensitivity and precise assay methods were able to meet the requirements of the development of predictive PBPK model for OLA in pig.2. Development of the PBPK model for the prediction of OLA in pigThe plasma protein binding of MQCA was determined using an equilibrium dialysis method. Blank pig plasma 2 mL was placed in dialysis bag (molecular-weight cutoff less than 3000) and incubated with MQCA in phosphate buffer (the concentration ranging from 50 to 800μg/kg) at 4℃for 12,24,48,96 h, respectively. At each time point, phosphate buffer 150μL was analyzed directly by HPLC. Another equilibrium dialysis system without plasma was used as control to assess the nonspecific binding of MQCA to the dialysis bag. The value of plasma protein binding was equal to the amount of bound MQCA divided by the total amount added. The result showed that the equilibrium of the MQCA level inside and outside the dialysis bag was achieved after 96 h, and the measured value (the average of the three concentrations, n=9) was 27.99%.Four pigs (40±5.82 kg) were used to determine the renal clearance of MQCA. They were put into four clean metabolic cages with water and blank feed provided during the whole experiment. All pigs received a bolus injection of MQCA solution (4 mg/mL, in a total of 5 mL) via the ear vein. Blood sample (5 mL) was collected via precava before and five hours after administration. Urine samples were collected during the whole experiment (a period of 10 hours after dosing) with their volumes recorded accurately. All samples were analyzed using the developed HPLC methods. The value of renal clearance was equal to the rate of excretion of MQCA in urine divided by the plasma MQCA concentration at the midpoint time of urine collection. The result showed that the measured renal clearance ranged from 0.051 L/(h*kg) to 0.111 L/(h*kg), and the mean value is 0.094 L/(h*kg). Four pigs (20±1.67kg) were fasted using soft cotton rope. MQCA solution (0.5 mg/mL) was infused via the ear vein at a rate of 2 mL/min, until the steady-state distribution of MQCA in the body was attained. The time required to achieve steady-state was determined by monitoring the plasma MQCA concentration. Subsequently, all pigs were sacrificed. Blood, liver, kidney, fat, muscle samples were collected and analyzed using developed HPLC methods. The result showed that the equilibrium of the distribution of MQCA in the body achieved at 60 min, and the mean values of four pigs were 1.02 (for the liver),4.02 (for the kidney),0.39 (for the adipose) and 0.23 (for the muscle).Thirty-seven pigs were randomly divided into test (n=32) and control (n=5) group. The later group was subdivided into the group for model simulation (n=20) and the group for model validation (n=12). After the experiment began, the test group was treated with 100 mg/kg OLA in feed for 30 days; the control group was given feed without quinoxaline compounds at the same time. Four treated (from the group for model simulation) and one control animals were slaughtered at 0.5,3,10,17,28 days postmedication. Four treated pigs from the group for model validation were slaughtered at 7,21,23 days postmedication. Muscle, liver, kidney, fat and plasma samples were collected and analyzed using a validated HPLC method. Results showed the level of MQCA in tissues was liver> kidney> fat> muscle at almost all time points. At the last time point (28 days posttreatment), the MQCA concentrations in all liver and fat samples were above the LOQ of the method, while three out of four kidney samples, two out of four plasma samples and three out of four muscle samples had not enough MQCA to be quantified using the HPLC method.Based on these results mentioned above, a flow-limited PBPK mode was developed to predict the residue depletion of MQCA in pig. In these models, gastric emptying rate and absorption rate constant were used to describe the oral absorption of OLA from gastrointestinal; an assumption of "one-step metabolism" and a first-order equation were used to simplify the metabolism from OLA to MQCA; liver, gastrointestinal and kidney were assumed to be the majority metabolic site of OLA in pig; physiological and other compound-specific parameters were obtained from literatures directly, or by use of UN-SCAN-IT (version 5.1.6) and WinNonlin (version 5.1.2) software; all equations were solved by ACSL xtreme (version 1.4, Aegis Technologies Group Inc, Huntsville, Ala) software. In addition, a targeted adjustment to model parameters and a "step by step" strategy for parameter estimation were performed during the course of model simulation. Judged by the measured data and existing studies, the liver model was selected as the most suitable one to describe the disposition of OLA in pig. The validated liver model underestimated the plasma OLA concentration slightly, also underestimated level of MQCA in all compartments at 12 h postdosing, but well predicted the residue depletion of the marker residue in edible tissues at 3-36 d postdosing.Sensitivity coefficients of all model parameters were calculated using central difference formula. The contribution of each parameter to model was evaluated by normalized sensitivity coefficient. Results showed that compound-specific parameter like tissue/plasma partition coefficients, clearance and metabolic rate constant were very sensitive to the predicted results, and large difference in sensitivity of parameters existed among different model structures.Uncertainty analysis was also performed use Monte-Carlo algorithm. Cardiac output and tissue/plasma partition coefficients were selected as the target parameters for Monte-Carlo algorithm, and their variability were described by use of the lognormal distribution. These two parameters were assigned a random value in a specified range with 99% confidence interval, and run the model for 1000 times. Results showed that all the observed data were within the predicted range of the liver model and gastrointestinal model, while greater deviations between the observed and predicted value were found in the kidney model.3. The extrapolation of PBPK model from OLA to QCT and CYAThe developed gastrointestinal model for OLA was selected to extrapolate tissue residue data of the marker residues of QCT and CYA. Pharmacokinetic parameters of the parent drug (such as bioavailability and absorption rate constant) were from literature, while the apparent metabolic rate constant of the marker residue was simulated based on the tissue residue data of MQCA and urinary excretion data of QCA. Compared with experimental data, the extrapolated model for QCT overestimated the concentration of MQCA in kidney and adipose at early time, but achieve good accuracy at later time points. With respect to the liver and muscle, the model gave an accurate prediction of the residue depletion of MQCA during the period of 0 d-15 d postdosing. The extrapolated model for CYA significantly underestimated the level of QCA in edible tissues at early time postdosing, while good fitness between the predicted and observed values achieved at 7 d-21 d postdosing. In the present study, a predictive PBPK model for OLA in pig was developed and extrapolated to QCT and CYA for the first time. These models enable us predict the residue depletion of these three food additives in porcine edible tissues accurately, which would help ensure the safety of animal derived foods. Meanwhile, the current studies also basically solved the problems that the present PBPK model could not predict the residue depletion of metabolites in food animals, and improved the predictive ability of PBPK models for veterinary drug residues in group of animals.