2013_Sept_29_Hsu_Renal_TransporterPBPK_Supplemental

Supplementary Materials for

Towards Quantitative Modeling of the Effect of Renal Impairment and Probenecid Inhibition

on Kidney Uptake and Efflux Transporters using PBPK

Vicky Hsu, Manuela de L T Vieira, Ping Zhao, Lei Zhang, Jenny Huimin Zheng, Anna Nordmark,

Eva Gil Berglund, Kathleen M. Giacomini, and Shiew-Mei Huang

1. PBPK models for substrates

1.1 Oseltamivir Carboxylate

Drug-dependent parameters for oseltamivir carboxylate (OC, active metabolite of oseltamivir phosphate), including logP, compound type, acid/base pKas, blood-to-plasma partition ratio, and fraction unbound in plasma, were obtained from Parrott et al [1]. Distribution parameters Vss and tissue-to-plasma partition coefficients were predicted using methods published by Rodgers et al based on drug physicochemical and blood binding properties [2,3]. OC has an absolute bioavailability of ~80% after oral administration of oseltamivir phosphate, and it is exclusively renally cleared, with significant contribution by active secretion [4]. To independently model OC PK after oral administration of oseltamivir phosphate, conversion to OC was assumed with the appearance of plasma OC approximated by fraction absorbed (fa, used to estimate the extent of conversion) and by a first order absorption rate constant (ka, used to estimate rate of conversion). Fa was therefore derived from OC’s bioavailability, while ka was optimized based on a single oral dose study [4].

Due to simultaneous availability of both plasma and urine over time profiles, the healthy, control arm (n=5) of an oseltamivir renal impairment multiple dose study [5] was used as a reference for obtaining transporter-mediated intrinsic clearances (CLint,T) for the net basolateral uptake transporter “Tup,b” (based on plasma data) and the net apical efflux transporter “Teff,a” (based on urine data). The CLint,T by “Tup,b” and oral absorption lag time were derived simultaneously via parameter estimation based on plasma profile observed from a multiple dose renal impairment study design of oseltamivir phosphate administered as 100 mg single dose (day 1), 100 mg twice daily (days 2 to 5), and 100 single dose (day 6) (Figure S1a). Since basolateral and apical passive diffusion clearances were assumed to be negligible, a sensitivity analysis on a range of CLint,T by “Teff,a” was conducted and compared to the observed urine profile. The results showed a “Teff,a” CLint,T base value of 0.001 µL/min/106 cells may sufficiently recover the observed cumulative fraction of drug excreted unchanged over time, with a higher-fold value used in the model to assure complete efflux of drug into urine (Figure S4a). This final model was then evaluated against separate observed clinical studies not used during mechanistic kidney model development for performance verification (Figure S1b).

Figure S1a Simulated OC Cp-t profile following 100 mg oseltamivir phosphate PO multiple dose (single dose on day 1, twice daily on days 2 to 5, single dose on day 6), with observed data points from reference RI normal study [5].

Figure S1b Performance of simulated OC Cp-t profile following 150 mg oseltamivir phosphate PO single dose (5 virtual trials, n=18 each) against observed data from separate studies [4,6,7,8]

1.2 Cidofovir

Drug dependent parameters are summarized in Table 1 in the main text. The cidofovir model was developed using the population representative of the healthy volunteer virtual population, assuming drug PK is similar between healthy volunteers and HIV patients (see below).

Human pharmacokinetic data are available in HIV patients receiving intravenous infusion of cidofovir over 1 hour [9]. Mean values of systemic clearance (CL), renal clearance (CLr), urinary excretion over time profile, and volume of distribution at steady state (Vss) from this PK study, along with the mean unbound fraction in plasma (fu,p) form the basis of developing cidofovir PBPK model. Based on physicochemical parameters provided to the PBPK software (Table 1), Vss was predicted, along with a modification of an empirical scaling factor (Kp Scalar, Table 1) to be 0.49 L/kg using the algorithm by Rodgers et al [2,3]. Optimization of Kp scalar was performed for all tissues equally. The predicted Vss value matches the Vss observed in HIV patients receiving 1, 3, and 10 mg/kg intravenous infusion [9].

In order to define clearance pathways for cidofovir PBPK model, several steps were taken sequentially.

(1). A mean CL of 12.8 L/h, a mean CLr of 11.4 L/h were calculated from the above mentioned cidofovir study at three infusion doses (Cundy, 1995). The mean CL and CLr values of each dose were 130 and 129, 152 and 129, and 148 and 124 mL/h/kg, respectively, and the mean values of fraction excreted in the urine at 24 hours were 98.5, 83.9, and 94.7%, respectively, at 1, 3, and 10 mg/kg (mean body weight of these doses were 89.2, 76.5, 67.3 kg, respectively).

(2). The residual, non renal clearance of cidofovir was calculated using the retrograde function of the PBPK software. Mean CL and CLr of 12.8 L/h and 11.4 L/h were used, and liver was assumed responsible for the non renal clearance. This retrospectively derived hepatic intrinsic clearance is based on the activity in human liver microsomes [10]. The intrinsic clearance value for microsomes was further translated to the value that is based on human S9 fraction in the PBPK software using default protein concentration in mg per gram liver for both subcellular fractions (39.8 and 120.8 mg/g for microsomes and S9, respectively). The residual hepatic CL was assigned to S9 because metabolic enzyme-specific and pathway-specific information was not available for cidofovir. In addition, microsomal protein concentration (39.9 mg/g) is fixed in the software, whereas the value can be altered for S9, providing flexibility of further defining S9 protein concentration in specific populations.The resulting intrinsic clearance will be responsible for extrapolating in vivo, non renal clearance of cidofovir when the model is used for simulations.

(3). The apparent CLr was further defined in the mechanistic kidney module of the PBPK software (Figure 1 of the main text). Assuming passive diffusion is negligible (major assumptions of the main text), the non filtrational clearance of cidofovir was defined using a net basolateral uptake (Tup,b) and a net apical efflux (Teff,a) clearance terms. Parameter estimation was conducted on the intrinsic clearance of uptake transporter at the tubular cell level (CLint,T in µL/min/106 cells) using plasma data at all three doses (1, 3, and 10 mg/kg intravenous infusion). Briefly, the PBPK model of cidofovir was used as structural model. In the absence of detailed patient information, mean demographics such as mean body weight (89, 77, and 67 kg), mean age of 39 years and mean serum creatinine level of 70 mol/L for all dose groups, and the gender of male were defined for each dose level. The mean plasma concentration time data set of each dose was simultaneously used to estimate CLint,T by the net uptake transporter (Tup,b). A maximum a posteriori estimation method was used along with both proportional and additive errors. The initial estimate of uptake CLint,T was 3.0 L/min/106 cells. A uniform distribution was assumed. A total of 10 iterations were used. The individual and population predicted (IPRED and PPRED) concentration time profiles were plotted against the observed data (Figure S2a, upper panel). The observed versus model predicted concentrations (IPRED and PPRED) are shown in


Figure S2a Upper panel: Individual and population predicted (IPRED and PPRED) cidofovir concentration-time profiles using PBPK model with CLint,T of renal uptake transporter estimated from the mean plasma PK data (observed) of three doses (Cundy, 1995) [9]. Panels from left to right are for 1, 3, and 10 mg/kg doses, respectively; Lower panel: Observed versus model predicted cidofovir concentrations (IPRED and PPRED) using PBPK model.

(4). The PBPK model of cidofovir in healthy volunteers was further developed to include CLint,T of net apical efflux transport (Teff,a) in the kidney. A sensitivity analysis was conducted using urine accumulation data from the study by Cundy et al [9] The values of 20, 0.2, 0.02 and 0.002 L/min/106 cells for the efflux CLint,T were tested manually in the PBPK model to predict urinary collection of cidofovir over time. The simulated urinary drug accumulation-time profiles under different efflux CLint,T after 3 mg/kg intravenous infusion of cidofovir were compared to the mean profiles of each dose (Figure S2b). Although the observed mean drug accumulation profile of 3 mg/kg dose group appeared to have a lower total accumulation at 24 hours and a slightly slower appearance than those of 1 and 10 mg/kg groups, a CLint,T of >0.2 L/min/106 cells can provide reasonable description of the urinary collection of cidofovir. As the model can not further confirm a value of >0.2 L/min/106 cells, we chose 20 L/min/106 cells in the final model to ensure effective accumulation of the drug in the urine compartment.


Figure S2b Simulated (at 3 mg/kg intravenous infusion of cidofovir over 1 hour) and observed cidofovir accumulation in the urine. Values are efflux intrinsic clearance (in L/min/106 cells) used for PBPK model of cidofovir. Observed data are the mean urinary drug accumulation for each intravenous dose [9].

1.3 Cefuroxime

Cefuroxime physicochemical parameters were obtained from Chemspider ( Royal Society of Chemistry, Cambridge, United Kingdom), the protein binding affinity from ex vivo data [11] and the blood/plasma ratio was in silico predicted based on the drug physicochemical properties using ADMET PredictorTM (Simulation Plus®, Lancaster, CA, USA). The Vss and organ partition coefficients were predicted by the method of Rodgers et al [2,3] based on the drug physicochemical properties and blood binding affinity. Sensitivity analysis was performed on the value of Kp scalar to adequately adjust the predicted Vss, to the average value of 0.19 L/kg observed in healthy subjects after IV dosing [11]. Optimization of Kp scalar was performed for all tissues equally. Cefuroxime is exclusively renally eliminated: 99% of dose administered is recovered unchanged in 24h urine. The contribution of net secretion of the total clearance is around 45% in a healthy adult [11]. The role of basolateral uptake transporter in the tubular secretion of cefuroxime is demonstrated by the in vivo observed 40% reduction on cefuroxime’s clearance by competitive inhibition with probenecid [12] and the increased in cefuroxime AUC by 27% by co-administration of the OAT substrate NX-059 [13]. The intrinsic clearance of a renal basolateral uptake transporter, “Tup,b”, was estimated (software PE function) using intravenous plasma data of three ascending doses (0.25, 0.5 and 1g IV bolus) [11]. The basal and apical passive diffusion clearance was assumed to be negligible as a function of the drug hydrophilicity and observed net secretion. Sensitivity analysis of the range of values of the intrinsic clearance of a renal apical efflux transporter, “Teff,a”, was conducted based on urine data over time from three ascending doses (0.25, 0.5 and 1g IV bolus) [11]. A value of Teff,a Clint,T (µL/min/106cells) higher than 0.1 was able to recover the cumulative fraction of drug excreted unchanged in urine over 12 hours (Figure S4c), with a higher-fold value of 10 being used in the model to assure complete efflux of drug into urine.

PK data and parameters from three ascending bolus doses in healthy male subjects [11] were used for model building (Figure S3a and S3b). Cefuroxime PBPK model was verify by two independent data sets: Plasma concentration–time profile of cefuroxime 1.5 g infused over 30 min [13] (Figure S3c) to a population of healthy young subjects, and 0.75 and 1.5 g infused over 20 min [14] (Figure S3d).

Figure S3 Performance of cefuroxime PBPK model by comparing the simulated Cp-t profile with observed data across different studies and dosing scenarios: a) 1.0 g cefuroxime IV bolus over 3 min [11] b) 0.25, 0.5 and 1.0g cefuroxime IV bolus over 3 min [11] c) 1.5g cefuroxime infusion over 30 min [13] and d) 0.75 and 1.5 g cefuroxime over 20 min infusion [14].

Figure S4 Simulated cumulative fraction of drug excreted unchanged for a) oseltamivir carboxylate, b) cidofovir, and c) cefuroxime.

2. PBPK model for probenecid

Drug-dependent parameters for probenecid are summarized in Table S1. In silico predictions included PubChem ( National Institutes of Health, Bethesda, MD, USA) and ADMET PredictorTM (Simulation Plus®, Lancaster, CA, USA) for logP and blood-to-plasma partition coefficient, respectively. The remaining physicochemical and blood binding properties were based on literature information [15,16,17]. In vivo clearance was calculated based on an observed clinical study (n=6) in which a single low intravenous dose of 0.5 g was administered [18]. In terms of elimination, general renal clearance (in contrast to mechanistic kidney model used in the model substrate drugs) was applied, such that approximately 10% of dose is excreted renally unchanged [19]. Using the retrograde model to derive intrinsic clearances, 70% of the total clearance was assigned to a hepatic CYP, with the remaining 20% assigned to a UGT enzyme [19]. Distribution parameters Vss and tissue-to-plasma coefficients were predicted using methods published by Rodgers et al (Rodgers), and optimized using parameter estimation based on observed IV profile [18]. Oral absorption parameters fa was estimated based on observed near complete oral absorption [15], and the ka was optimized to fit oral profiles observed from two separate studies [18,20]. Due to known saturation at higher oral doses, Km and Vmax values were assigned to the hepatic CYP enzyme and optimized according to the CYP CLint derived from the IV data of probenecid. Intravenous and oral probenecid model results are shown in Figure S5. The main source of probenecid pharmacokinetic nonlinearity is due to saturable metabolism process. Therefore, despite slight observed changes in fu,p, a constant fu,p value was assumed for probenecid (less than 10% change in fu,p in the probenecid concentration range simulated if compared to observed fu,p changes) [18].

Table S1 Drug-dependent parameter summary table for probenecid

Parameters / Probenecid
Molecular weight (g/mol) / 285a
LogP / 3.2b
Compound Type / Monoprotic Acidc
Acid pKa / 3.4c
Base pKa / N/A
B/P / 0.60d
Fu,p / 0.10e
Vss (L/kg) / 0.10 (predicted using method 2)f
Kp scalar / 0.30 (optimized)g
CLiv (L/h) / 1.03h
CLr (L/h) / 0.10i
CLint (µL/min/106 cells) by CYP / 0.036 (IV, retrograde analysis)i
CYP Vmax (pmol/min/pmol of isoform) / 0.816 (PO, optimized based on CYP CLint)g
CYP Km (µM) / 22.8 (PO, optimized based on CYP CLint)g
CLint (µL/min/106 cells) by UGT / 0.53 (retrograde analysis)i
fa / 0.95 (PO)e
ka (1/h) / 0.45 (PO, optimized)g

LogP, partition coefficient; pKa, dissociation constant; B/P, blood-to-plasma partition ratio; Fu,p, fraction unbound in plasma; Vss, volume of distribution at steady state; Kp, tissue-to-plasma partition coefficient; CLiv, in vivo clearance; CLr, renal clearance; CLint, in vitro intrinsic clearance; CYP, cytochrome p450 enzyme; Vmax, maximum rate of metabolite formation; Km, Michaelis-Menten constant; fa, fraction available from dosage form; ka, first-order absorption rate constant.

aChemSpider ( Royal Society of Chemistry, Cambridge, United Kingdom)

bPubChem ( National Institutes of Health, Bethesda, MD, USA)

cShore et al [17]

dADMET PredictorTM (Simulation Plus®, Lancaster, CA, USA)

eDayton et al [15]

fRodgers et al [2,3]

gOptimization involves manual or automated sensitivity analysis, or parameter estimation techniques

hEmanuelsson et al [18]

iCunningham et al [19]

Figure S5 Simulated probenecid Cp-t profile following a) 0.5 g probenecid IV infusion over 15 m, with observed data [18], b) 1.0 g probenecid PO single dose, with observed data points [18,20], and c) 2.0 g probenecid PO single dose, with observed data points [18,20].

3. Probenecid dosing and predicted concentrations in drug-drug interaction (DDI) studies

3.1. Oseltamivir

Oseltamivir-probenecid DDI study [6]: “single 150-mg dose of oseltamivir, administered during treatment with probenecid, in which 500 mg of probenecid was given every 6 h for 16 doses beginning 23 h before the dose of oseltamivir.” Predicted probenecid concentration at time of substrate dosing is provided below in Figure S6.

Figure S6 Predicted probenecid Cp-t profile in OC-probenecid DDI study [6].

3.2. Cidofovir

Cidofovir-probenecid DDI study [9]: Cidofovir dosed “with concomitant oral probenecid given as a 4 g course (comprising 2 g at 2 h before a dose of cidofovir and 1 g at each of 2 and 8 h after the dose).”Predicted probenecid concentration at time of substrate dosing is provided below in Figure S7.

Figure S7 Predicted probenecid Cp-t profile in cidofovir-probenecid DDI study [9].

3.3. Cefuroxime

Cefuroxime-probenecid DDI study [14]: “750 mg cefuroxime with 1 g of probenecid given orally 3 h before the cefuroxime infusion.”Predicted probenecid concentration at time of substrate dosing is provided below in Figure S8.


Figure S8 Predicted probenecid Cp-t profile in cefuroxime-probenecid DDI study [14].

4. PBPK model files

It is important to note that whenever possible, we always strive to use reliably measured parameters as a primary means of building PBPK models. Only in situations in which parameters are not readily available do we rely on in silico prediction or parameter estimation to derive a parameter value. All PBPK model files using Simcyp Simulator® (e.g.,”.cmp”, “.lbr”, and “.wks”) are available upon request to the authors.

5. Simulated cellular cidofovir drug amount-time profile over 24 hours after intravenous infusion of drug


Figure S9 Amount of cidofovir in kidney cells vs. time profile following 3.0 mg/kg cidofovir IV infusion over 1 h with and without probenecid hypothetical inhibition(s).

References

1. Parrott N, Davies B, Hoffman G, Koerner A, Lave T, Prinssen E, Theogaraj E, Singer T. Development of a physiologically based model for oseltamivir and simulation of pharmacokinetics in neonates and infants. Clin Pharmacokinet. 2011;50:613-23.

2. Rodgers T, Leahy D, Rowland M. Physiologically based pharmacokinetic modeling 1: predicting the tissue distribution of moderate-to-strong bases. J Pharm Sci. 2005;94:1259-76.

3. Rodgers T, Rowland M. Physiologically based pharmacokinetic modelling 2: predicting the tissue distribution of acids, very weak bases, neutrals and zwitterions. J Pharm Sci. 2006;95:1238-57.

4. Study of the pharmacokinetics and absolute bioavailability of the neuraminidase inhibitor Ro 64-0796/Ro 64-0802 (Protocol NP15719). Clinical Pharmacology and Biopharmaceutics Review from Drugs@FDA(1999).

5. Multiple ascending oral dose study of the pharmacokinetics, tolerability, and safety of the neuraminidase inhibitor Ro 64-0796 in subjects with renal impairment. Clinical Pharmacology and Biopharmaceutics Review from Drugs@FDA(1999).

6. Hill G, Cihlar T, Oo C, Ho ES, Prior K, Wiltshire H, Barrett J, Liu B, Ward P. The anti-influenza drug oseltamivir exhibits low potential to induce pharmacokinetic drug interactions via renal secretion-correlation of in vivo and in vitro studies. Drug Metab Dispos. 2002;30:13-9.

7. Jhee SS, Yen M, Ereshefsky L, Leibowitz M, Schulte M, Kaeser B, Boak L, Patel A, Hoffman G, Prinssen EP, Rayner CR. Low penetration of oseltamivir and its carboxylate into cerebrospinal fluid in healthy Japanese and Caucasian volunteers. Antimicrob Agents Chemother. 2008;52:3687-93.

8. Snell P, Oo C, Dorr A, Barrett J. Lack of pharmacokinetic interaction between the oral anti-influenza neuraminidase inhibitor prodrug oseltamivir and antacids. Br J Clin Pharmacol. 2002;54:372-7.