Contribution of Uptake and Efflux Transporters to Oral Pharmacokinetics of Furosemide.

Furosemide is a widely used diuretic for treating excessive fluid accumulation caused by disease conditions like heart failure and liver cirrhosis. Furosemide tablet formulation exhibits variable pharmacokinetics (PK) with bioavailability ranging from 10 to almost 100%. To explain the variable absorption, we integrated the physicochemical, in vitro dissolution, permeability, distribution, and the elimination parameters of furosemide in a physiologically-based pharmacokinetic (PBPK) model. Although the intravenous PBPK model reasonably described the observed in vivo PK data, the reported low passive permeability failed to capture the observed data after oral administration. To mechanistically justify this discrepancy, we hypothesized that transporter-mediated uptake contributes to the oral absorption of furosemide in conjunction with passive permeability. Our in vitro results confirmed that furosemide is a substrate of intestinal breast cancer resistance protein (BCRP), multidrug resistance-associated protein 4 (MRP4), and organic anion transporting polypeptide 2B1 (OATP2B1), but it is not a substrate of P-glycoprotein (P-gp) and MRP2. We then estimated the net transporter-mediated intestinal uptake and integrated it into the PBPK model under both fasting and fed conditions. Our in vitro data and PBPK model suggest that the absorption of furosemide is permeability-limited, and OATP2B1 and MRP4 are important for its permeability across intestinal membrane. Further, as furosemide has been proposed as a probe substrate of renal organic anion transporters (OATs) for assessing clinical drug-drug interactions (DDIs) during drug development, the confounding effects of intestinal transporters identified in this study on furosemide PK should be considered in the clinical transporter DDI studies.

Introduction

Furosemide is one of the most commonly used loop diuretics,[1] which continues to be an important first-line drug in the treatment of edema associated with kidney impairment, liver cirrhosis, hypertension, and chronic heart failure.[2,3] It acts by blocking the Na+–K+–2Cl– cotransporter in the thick ascending limb of the loop of Henle, thus causing excessive excretion of water along with sodium, chloride, and potassium.[4−6] Furosemide is generally safe and routinely prescribed to adult as well as pediatric patients[2,3,7−10] in the normal dose ranges, i.e., orally 20–80 mg/day, which may be increased but not to exceed 600 mg/day in adults and 0.5 to 2 mg/kg/dose every 6 to 24 h in pediatrics.[11−13] However, the higher doses of furosemide can lead to profound water and electrolyte depletion, and in some cases, it can lead to potential toxic effects like ototoxicity.[14−19] Therefore, the food and drug administration (FDA) recommends a careful monitoring of furosemide to adjust its dose and dosing schedule according to the need of individual patients in certain vulnerable populations such as patients receiving doses higher than 80 mg/day for prolonged periods or those using other antihypertensive drugs.[20]

The average bioavailability of furosemide is ∼60%;[21−23] however, it is highly variable and can range from 10–100% with significant inter- and intra-subject variability.[2,24] Such erratic drug absorption is generally caused by factors that influence solubility or permeability, such as differences in pH, dosage form, gastric emptying, food intake, transporter activity, etc. However, furosemide exhibits linear pharmacokinetics (PK) in the dose range (10–80 mg).[25,26] Majority of furosemide dose is eliminated unchanged renally through active secretion (fraction excreted unchanged in urine, fe ≈ 65%), which is primarily mediated by organic anion transporters (OATs), i.e., OAT3 and OAT1, in concert with multidrug resistance-associated protein 4 (MRP4), and the rest by metabolism (fraction metabolized, fm ≈ 35%).[2,27−29] Moreover, furosemide is practically insoluble in water and acidic pH, i.e., solubility < 0.1 mg/mL,[30] but its solubility is >20–100-fold greater in intestinal pH at ∼6.8, which allows complete dissolution of an 80 mg tablet in the intestinal fluid.[31] On the other hand, permeability of furosemide determined using in vitro systems is low, i.e., apparent permeability (Papp) ≤ 2.0 × 10–6 cm/s.[32,33] To better characterize the variability in the PK of furosemide tablet formulation, we first integrated its physicochemical, in vitro dissolution, permeability, distribution, and the elimination parameters of furosemide in a physiologically-based pharmacokinetic (PBPK) model and compared the simulated data with the observed PK values. However, the reported low passive permeability failed to capture the observed oral absorption and underpredicted the plasma concentration–time profiles. To mechanistically justify this discrepancy, we hypothesized that transporter-mediated uptake contributes to the oral absorption of furosemide in conjunction with passive permeability.

In general, drug permeability can be influenced by efflux or uptake transporters besides passive diffusion. High basolateral to apical permeability in comparison to apical to basolateral permeability in Caco-2 monolayers[34] and the excised rat jejunum model[35] has been reported, indicating that it is a substrate of efflux transporter(s). However, the involvement of individual intestinal transporters such as organic anion transporting polypeptides (e.g., OATP2B1), P-glycoprotein (P-gp), breast cancer resistance protein (BCRP), and multidrug resistance-associated proteins (MRPs) is understudied. Therefore, we elucidated the mechanisms of intestinal efflux and uptake transport that regulate furosemide absorption using validated in vitro models, i.e., membrane vesicles and transfected cells. We then characterized the kinetic parameters, i.e., maximum transport rate (Jmax) and substrate affinity (Km). These data were then normalized using the transporter abundance data (generated by quantitative proteomics) in vesicles and cells to in vivo conditions.

Finally, the oral model was refined by considering additional transporter-mediated uptake in conjunction with passive diffusion, defined as “net uptake” with in vitro dissolution, passive permeability, metabolism, and renal excretion to accurately describe oral furosemide PK under the fasting and fed conditions.

Results

Furosemide Intravenous PBPK Model Development and Verification

The RAF values of 65.49 and 70.3 were derived for OAT3- and MRP4-mediated disposition by parameter estimation module and the clinical data.[21,23,28,36−38] UGT1A9-mediated clearance of 42 μL/min/mg was back-calculated by subtracting renal clearance from total systemic clearance.[27−29] The model adequately simulated furosemide PK at two intravenous bolus doses, i.e., 40 and 80 mg (Figure ), where all the observed data points[21,23,28,36−38] were within the 95th and 5th percentiles of all simulations. The cumulative urinary excretion–time profiles of respective intravenous doses were in accordance with the observed data (Figure S1).[23,39] Corresponding simulated versus observed PK parameters are shown in Table , which further confirmed that furosemide disposition was well captured by the model. The final model simulated percent contributions of the metabolism and excretion, i.e., an fm of ∼35% and an fe of ∼65% were in accordance with the literature values.[27,28] The developed model successfully captured the effect of probenecid on furosemide PK (Figure ).[28] The simulated plasma concentration profiles of probenecid were also in accordance with the observed data points (Figure S2). Current PBPK simulation results were within the two-fold acceptance criterion (Table ). The evaluation for our model verification relied only on the two-fold criterion, as it was the most commonly used reference criterion for model acceptance.

Figure 1

Simulated and observed concentration–time profiles for furosemide given intravenously at doses 40 (A) and 80 mg (B) in healthy volunteers. Included in each plot is the simulated mean, 95th percentile, and 5th percentile concentration range. The open circles represent the mean observed data.[21,23,28,36−38]

Table 1

Model Verification: Comparison of Simulated and Observed PK Data of Furosemide in Adults at Different Doses and Routes of Administration

intravenous bolus for 3 min (healthy subjects)a
study ID	dose (mg)	parameter	observed mean (O)	simulated mean (S)	S/O ratio	two-fold acceptance criterion
Kelly 1974[21]	80	AUC (μg × h/mL)	9.66	8.50	0.88	pass
		Vss (L/kg)	0.07	0.11	0.64	pass
		CL (L/kg)	0.12	0.12	1.04	pass
		t1/2 (h)	0.54	0.99	1.83	pass
Verbeeck 1982[38]	80	AUC (μg × h/mL)	8.55	8.50	0.99	pass
		Vss (L/kg)	0.12	0.11	0.95	pass
		CL (L/kg)	0.13	0.12	0.94	pass
		t1/2 (h)	1.00	0.99	0.99	pass
Andreasen 1977[36]	40	AUC (μg × h/mL)	4.02	4.20	1.04	pass
		Vss (L/kg)	0.18	0.11	0.61	pass
		CL (L/kg)	0.14	0.12	0.85	pass
		t1/2 (h)	1.20	0.99	0.83	pass
Smith 1980[28]	40	AUC (μg × h/mL)	4.20	4.20	1.00	pass
		Vss (L/kg)	0.12	0.11	0.93	pass
		CL (L/kg)	0.14	0.12	0.89	pass
		t1/2 (h)	1.37	0.99	0.72	pass
Keller 1981[37]	40	AUC (μg × h/mL)	3.83	4.20	1.10	pass
		Vss (L/kg)	0.18	0.11	1.64	pass
		CL (L/kg)	0.15	0.12	0.81	pass
		t1/2 (h)	0.85	0.99	1.16	pass
Hammarlund 1984[23]	40	AUC (μg × h/mL)	4.20	4.20	1.00	pass
		Vss (L/kg)	0.13	0.11	0.85	pass
		CL (L/kg)	0.15	0.12	0.80	pass
		t1/2 (h)	1.17	0.99	0.85	pass

Parameter estimates obtained from references.[21,23,28,36−38] Estimates based on PBPK model simulations using the Simcyp simulator.

Parameter estimates obtained from references.[28]

Parameter estimates obtained from references.[23,41,42]

Figure 2

Simulated and observed concentration–time profiles of furosemide given intravenously at 40 mg dose along with probenecid (OAT1/3 inhibitor). The continuous green (solid) line is the simulated profile of furosemide given a single dose of 40 mg intravenously in the absence of the OAT1/3 inhibitor and the red (dashed) line is the simulated profile of furosemide given as a single dose of 40 mg intravenously along with the OAT1/3 inhibitor, i.e., probenecid administered (1000 mg twice orally)13 and 1 h prior to furosemide administration, respectively. The open circles represent the mean observed data reported.[28]

ATP-Dependent Vesicular Transport of Furosemide

The initial ATP-dependent vesicular transport screening results confirmed BCRP and MRP4 as the main efflux transporters of furosemide. P-gp, MRP2, and MRP3 expressing vesicles did not show any active transport of furosemide (Figure S3). The mean derived kinetic parameters, i.e., Jmax and Km from the kinetic curves (Figure ) were 11.3 nmol/min/mg and 20.9 μM (BCRP) and 0.55 nmol/min/mg and 27.96 μM (MRP4), respectively (Table ).

Figure 3

ATP-dependent transport kinetics of furosemide in BCRP (A) and MRP4 (B) vesicles. The kinetic experiments were conducted at various concentrations (i.e., 1 to 100 μM) with 25 μg vesicle proteins for 30 s. Differences between the ATP and AMP groups (net ATP-dependent transport rates) were calculated, and the Michaelis–Menten equation (J = Jmax ×[S]/([S] + Km)) was fitted to the data and represented as the mean (SD (n = 3)). The kinetic constants (Jmax and Km with the 95% CI) for the studied efflux transporters are presented in Table .

Table 2

Kinetic Parameters (Km and Jmax) for the Transport of Furosemide Derived Using BCRP and MRP4 Overexpressing Vesicles Using the Michaelis–Menten Modela

transporter	Km, μM (95% CI)	Jmax, pmol/min/mg (95% CI)	normalized average Jmax (pmol/min/pmol of transporter protein)
BCRP	20.9 (16.4 to 25.3)	11,296 (10,437 to 12,155)	3037
MRP4	27.96 (19.5 to 36.5)	554.2 (489.1 to 619.3)	9085.2

The values in parentheses represent 95% confidence intervals (CI).

Time- and Concentration-Dependent Uptake of Furosemide in hOATP2B1-Transfected Cells

The uptake of furosemide by OATP2B1-transfected cells was >2-fold higher than the uptake by mock cells at 15 min (Figure A). The OATP2B1-mediated uptake was non-saturable up to the highest concentration of furosemide, i.e., 500 μM (Figure B).

Figure 4

Time- and concentration-dependent uptake of furosemide by mock and OATP2B1-transfected MDCK-II cells. Uptake was measured at 37 °C over specified time (i.e., 2, 5, 10, and 15 min, as shown in (A) and in the concentration range (i.e., 0.5 to 500 μm for 15 min, as shown in (B)). The OATP2B1-specific rate of uptake in concentration-dependent studies was obtained by subtracting the uptake in mock cells (MOCK) from that of OATP2B1-transfected MDCK-II cells. The OATP2B1-specific rate was linear in the measured concentration range. Each data point represents the mean ± SD (n = 3).

Protein Abundance of Efflux and Uptake Transporters

The protein abundances of BCRP and MRP4 in vesicles were 11.4 and 0.3 pmol/mg protein, respectively. As P-gp, MRP2, and MRP3 did not transport furosemide in the initial screening assay, the protein abundance data for these transporters are not shown. The protein abundance of OATP2B1 in the transfected MDCK-II cells was 4.25 pmol/mg protein (Table S1). The % of inside-out vesicles was 33% (BCRP) and 21% (MRP4).[40]

Refined Furosemide PBPK Oral Model Development and Verification

The optimized net uptake was able to capture the observed oral absorption data[23,41,42] (Figure ). Corresponding simulated versus observed PK parameters are shown in Table . The consideration of delayed gastric emptying time in the fed-state model was able to predict decreased plasma exposure of furosemide[23] (Figure ). Our oral PBPK simulation results were also within the two-fold acceptance criterion (Table ).

Figure 5

Simulated and observed concentration–time profiles of furosemide given orally at a dose of 40 mg in fasting (A) and fed (B) conditions to healthy volunteers. Included in each plot is the simulated mean, 95th percentile, and 5th percentile concentration range. The open circles represent the mean observed data reported.[23,41,42]

Discussion

Oral absorption of furosemide is highly variable, which differs markedly across different dosage forms or formulation types and with or without food.[2] Furosemide is a biopharmaceutics classification system (BCS) class IV drug, and its poor solubility and permeability are considered to be the major determinants of its variable bioavailability.[43] Furosemide is a weak acid and exhibits pH-dependent solubility, and the dissolution studies suggest that furosemide solubility is ∼70-fold higher in fed-state simulated gastric fluid as compared to the fasted-state condition.[44] However, the higher solubility in fed-state dissolution media does not correlate with the decreased plasma exposure.[23,45,46] This indicates that furosemide absorption from IR tablets is not solubility-limited, a phenomenon that corroborates with the shorter mean dissolution time (MDT) in comparison to the mean absorption time (MAT) established by moment analysis of clinical data.[23] Moreover, the reported MAT of oral solution is significantly longer than the mean residence time (MRT) after intravenous bolus dose, indicating the flip-flop and absorption-rate-limited kinetics.[23] Using the moment analysis of clinical data, it is suggested[23] that furosemide absorption is limited by transport across gastro-intestinal epithelium or gastric emptying.

To mechanistically explain furosemide absorption and clinical PK, we investigated the role of intestinal transporters in furosemide absorption. We identified that BCRP and MRP4 are the major efflux transporters, and OATP2B1 is the key uptake transporter that govern furosemide absorption. Modulation of these transport mechanisms by drug interactions or genetic polymorphisms could influence furosemide absorption. Our data showed that furosemide is not a substrate of P-gp consistent with the Flanagan et al. data.[47] However, these data contradict with another observation, which showed an efficient inhibition of furosemide transport by the typical P-gp inhibitor, verapamil (200 μg/mL) in excised rat jejunum experiment.[35] The latter is likely an artifact because of the high verapamil concentration that can inhibit other efflux transporters, e.g., BCRP. Similar to our findings, Flanagan et al. confirmed that MRP1 and MRP2 are not involved in the secretion of furosemide.[47] Although Lin et al. showed decrease in the basolateral to apical permeability and the efflux ratio of furosemide in the presence of MRP2 inhibitor, MK-571,[48] this effect is likely a result of the BCRP inhibition.[49−52]

Although furosemide was previously shown to be a substrate of BCRP and MRP4 using in vitro and in vivo models,[53] we characterized, for the first time, the kinetic parameters of BCRP- and MRP4-mediated uptake in membrane vesicles, which were further extrapolated using quantitative proteomics data to predict normalized Jmax, i.e., pmol of substrate/mL/pmol of transporter protein. These data can be used in the in vitro to in vivo extrapolation of tissue-specific transporter activity using the proteomics-based approach established previously.[40] Furosemide is a substrate of OATP2B1 that works in tandem with MRP4 in the enterocytes to facilitate its rapid absorption. These results explain the moderate to high oral bioavailability of furosemide,[23,41,42] which otherwise cannot be simulated using the low passive permeability data.[32,33] Although the role of hepatic OATPs (OATP1B1 and OATP1B3) in furosemide uptake is known,[33] the contribution of intestinal OATP2B1 in furosemide absorption is a novel finding. While there are no reported PK studies where furosemide was co-administered with known OATP2B1 inhibitors, known OATP1B inhibitors such as aliskiren, lesinurad, valsartan, and sacubitril significantly reduce furosemide absorption.[54−56] These clinical findings suggest that these compounds likely inhibit intestinal OATP as well. Fruit juices such as grapefruit, apple, and orange juices can modulate OATP2B1 activity[57−59] and may explain the reduced absorption of furosemide with orange juice.[45] Nevertheless, further studies are warranted to confirm the role of OATP2B1 in the absorption of furosemide and other organic anions.

The role of MRP4 in furosemide uptake is not only limited to oral absorption but can also be applied to other organs such as the kidneys. Furosemide is a known potent substrate of OATs expressed in the basolateral membrane of the kidney proximal tubule, which work in concert with MRP4 at the apical side. Thus, MRP4 could play a crucial role in regulating intracellular concentration of furosemide in the proximal tubules, i.e., the site of its pharmacodynamic action.[60] This corroborates with the fact that the therapeutic response of furosemide is related to drug concentration in urine,[61] as opposed to that in plasma, and the diuretic activity is effectively blocked by the OAT1/3 inhibitor, probenecid. Both the urinary excretion rate and natriuretic response to furosemide was significantly reduced when it was given concomitantly with probenecid compared to furosemide administered alone in the first 2 h of dosing.[62] It is noteworthy that probenecid is also an inhibitor of some UGT enzymes; however, UGT1A9 inhibition by probenecid is unlikely, considering its reported high inhibition constant (Ki) value.[63,64] This implies that the diuretic activity of furosemide can be influenced by both OAT and MRP4 functions. MRP4 genetic polymorphism (g.33387C > A) has been shown to be associated with weight loss when furosemide is given in decompensated heart failure patients.[65] This finding together with our in vitro data indicates that the altered MRP4 function due to drug interactions, aging, and pathophysiological factors could alter the safety or efficacy of furosemide.

With the emerging regulatory significance of the renal drug transporters, furosemide has been proposed as a probe substrate of renal OATs for assessing clinical DDIs during drug development.[42] Particularly, furosemide was used in recent cocktail in vivo transport studies.[66,67] As furosemide is identified as a substrate for BCRP, MRP4, and OATP2B1 in this study, the confounding effects of these transporters on furosemide PK should be considered while interpreting clinical DDI studies that use furosemide as a OAT probe.

Methods

Materials

Membrane vesicles from HEK293 cells overexpressing the human BCRP, P-gp, MRP2, MRP3, and MRP4 as well as the stably transfected hOATP2B1-MDCK-II cells were provided by Solvo Biotechnology (Budapest, Hungary). Cell culture medium, i.e., Dulbecco’s modified Eagle’s medium (DMEM), fetal bovine serum, 1% GlutaMAX-1, 100 U/mL penicillin, and 100 μg/mL streptomycin were obtained from Sigma-Aldrich (St. Louis, MO). Furosemide, diclofenac, estrone-3-sulfate, rosuvastatin, adenosine 5′-triphosphate (ATP) disodium salt, adenosine 5′-monophosphate (AMP) monohydrate, glutathione, tris[hydroxymethyl]aminomethane (tris-base), MgCl2, NaCl, sucrose, and 3-[N-morpholino] propane sulfonic acid (MOPS) were purchased from Sigma-Aldrich (St. Louis, MO). Bovine serum albumin (BSA), membrane protein extraction kit, and trypsin digestion reagents were obtained from Thermo Fisher Scientific (Rockford, IL). Hanks’ balanced salt solution (HBSS), phosphate buffer saline (PBS), acetonitrile, and formic acid were purchased from Thermo Fisher Scientific (Rockford, IL). The synthetic unlabeled and stable labeled peptides were purchased from New England Peptides (Boston, MA) and Thermo Fisher Scientific (Rockford, IL), respectively. Multiscreen HTS Vacuum Manifold and 96-well filter plates with class B glass fiber filters were purchased from EMD Millipore (Billerica, MA).

Workflow of Furosemide PBPK Model and Model Evaluation

A systematic approach used for furosemide PBPK model development and verification is discussed in the following section. The intravenous model was first developed to describe furosemide elimination and distribution, which is followed by developing the oral absorption model including the food effect. A simplified PBPK modeling workflow is presented in Figure S4.

Development of the Intravenous PBPK Model

A mechanistic PBPK model of furosemide was developed in the virtual Caucasian population using Simcyp simulator Ver. 17 (Certara, Princeton, NJ). A total of 80 virtual individuals (10 trials × 8 individuals; age range between 25 and 50 years) were used in the simulations with an equal proportion of males and females. The virtual study designs were as close as possible to the reported furosemide clinical studies. The clinical data were collected from the peer-reviewed literature for the PBPK model development and verification, and the quantitative data were digitized using WebPlotDigitizer (https://automeris.io/WebPlotDigitizer/).

The intravenous furosemide PBPK model was developed using the reported physicochemical, biopharmaceutical, and PK parameters.[21,23,28,36−38,68,69] The volume of distribution was predicted using the method established by Rodgers and Rowland.[70] The permeability-limited liver model and the mechanistic kidney model were used for the prediction of furosemide disposition, incorporated with known mechanisms of furosemide metabolism and excretion, i.e., UGT1A9-mediated metabolism in the liver and kidney, and active secretion into urine mediated by OAT1 and OAT3 in conjunction with MRP4.

A step-by-step middle-out approach was used to address the uncertainty of the parameter estimation. In the first step, the total systemic clearance (hepatic plus renal clearance) of furosemide was estimated using the top-down approach, where the observed intravenous pharmacokinetics data were utilized. The “filtration plus secretion” clearance was considered as the total renal clearance. The renal filtration was then estimated by multiplying the fraction unbound in plasma (fu) of furosemide to the glomerular filtration rate (GFR). In the next step, the renal secretion clearance was estimated by using clinical probenecid–furosemide drug-interaction data[28] assuming that probenecid completely inhibits the OAT-mediated renal secretion (Table S2). Although both OAT3 and OAT1 mediate the renal uptake of furosemide, we assumed that the secretion is entirely OAT3-dependent due to the lack of data on the relative contribution of these transporters. OAT-mediated active secretion was considered to be the rate determining step for the renal uptake of furosemide,[28,42,62] whereas MRP4 was assumed to regulate the urine and the renal tissue concentration.[53] Further, as MRP4 is the only apical transporter responsible for furosemide elimination in the kidneys, the urinary excretion data were used to estimate the MRP4 contribution in furosemide renal clearance. Briefly, the reported in vitro kinetic parameters of OAT3-mediated furosemide transport[71] were incorporated and the relative activity factor (RAF) was optimized to match the in vivo disposition.[27] MRP4-mediated active secretion into urine was also incorporated by using the kinetic parameters generated in house using the vesicular assay. The RAF for MRP4-mediated transport at the apical side was optimized so that the simulated cumulative urinary excretion–time profiles became consistent with the observed cumulative urinary excretion–time profiles.[23,39]

The hepatic clearance was estimated as the total systemic clearance minus the total renal clearance. Both the sinusoidal net uptake and the intrinsic clearance of UGT1A9 were incorporated as the hepatic clearance component to match the in vivo fraction metabolized. In addition, the fraction metabolized (fm) by UGT1A9 was back-calculated from the observed total systemic clearance and the renal clearance values,[27,28] using the retrograde enzyme kinetics module of Simcyp. Detailed input parameters used for the model development are summarized in Table . The IV disposition model was used to predict the PK of furosemide at two different intravenous bolus doses, i.e., 40 and 80 mg.

Table 3

Input Parameters Used for Furosemide PBPK Model Development

PBPK parameter	value				method/reference(s)
1. physico-chemical and binding
molecular mass (g/mol)	330.74				PUBCHEM
log P	2.29				PUBCHEM
pKa	3.9				PUBCHEM
B/P ratio	0.6				reported[68]
fu, plasma	0.03				reported[69,76]
2. absorption
model	ADAM
Ptrans,0 (×10–6 cm/s)	215.15				estimateda
net uptake Ptrans,0 (×10–6 cm/s)	4500				optimizedb
dissolution data	% dissolved				reported[44]
	time (hr)	fasting	fed
		FaSSIF	FeSSGF	FeSSIF
	0.04	37	7.5	27
	0.09	60	17.9	46.3
	0.17	78.2	34.7	63.7
	0.25	84	46.3	79.7
	0.33	90.2	56.6	88.7
	0.5	96.5	68.3	96.6
	0.77	100	80.1	98
	1.00		86.9	100
	1.5		94.5
	2.0		97.6
3. distribution
model	full PBPK
Vss (L/kg)	0.1				method 2[70]
4. elimination
CLint, UGT1A9 (μL/min/mg protein)	42				estimatedc
permeability limited liver model
sinusoidal net uptake CLint, T (μL/min/106cells)	9				optimizedb
permeability-limited kidney model
Jmax, OAT3 (basolateral) (pmol/min/106cells)	54.17				reported[71]
Km(μM)	12.95				reported[71]
RAF	65.5				optimizedb
Jmax, MRP4 (apical) (pmol/min/106cells)	118.23				experimentald
Km (μM)	27.96				experimentald
RAF	70.3				optimizedb

Estimated using the MechPeff model: Mechanistic passive regional permeability predictor of Simcyp.

The parameters were optimized utilizing the parameter estimation function of Simcyp using the observed data.[21,23,28,36−38]

Estimated by back calculation from in vivo systemic and renal clearance,[28] using retrograde enzyme kinetics in Simcyp to match in vivo fraction metabolized (i.e., fm ≈ 35%).[27]

Experimental: in vitro experimental value generated in house using overexpressing membrane vesicles.

Development of the Oral PBPK Model

Once the renal and hepatic clearances were optimized and verified using the intravenous data (with and without probenecid), the furosemide absorption model was developed using the advanced dissolution absorption and metabolism (ADAM) model. This model treats the gastro-intestinal tract as eight compartments and integrates absorption kinetics data along with in vitro experimental data, i.e., dissolution data,[44] and the permeability. Initially, the effective permeability in humans (Peff) was predicted using the MechPeff model (mechanistic passive regional permeability predictor) of Simcyp, which yielded the value of intrinsic transcellular permeability (Ptrans,0) of ∼215.15 × 10–6 cm/s. Further, the Peff value estimated based on the experimentally determined passive permeability by the Caco-2 cell-based assay,[33] i.e., Papp ≈ 2.0 × 10–6 cm/sec, was used for simulating the oral absorption at a 40 mg dose. Both the Peff values underpredicted the plasma concentration–time profiles when compared to the observed data[23,41,42] (Figure S5). Therefore, we hypothesized that transporter-mediated uptake contributes to the oral absorption of furosemide in conjunction with the passive permeability.

Model Evaluation

To assess the robustness of the PBPK model, a visual comparison of the observed and simulated plasma concentration–time profiles was performed. The simulations were considered acceptable when the observed data points were within the simulated 5th and 95th percentiles. Further, the predictive performance was quantitatively assessed by comparing the simulated parameters with literature-based clinical data through two-fold criteria, wherein the simulated PK parameters were required to be within two-fold of the observed sclinical data.[72]

Vesicular Transport Assay

A previously validated vesicular transport assay[40,73] was used to investigate the mechanisms of furosemide efflux transport. Briefly, the furosemide transport was studied using an initial screening assay employing the vesicles overexpressing BCRP, P-gp, MRP2, MRP3, and MRP4. The assay was carried out in 96-well polystyrene plates by incubating 10 μM furosemide with 4 mM ATP or AMP at 37 °C in the following assay buffers: (i) 40 mM MOPS-Tris (pH 7.0), 70 mM KCl, and 7.5 mM MgCl2 for MRP2 or (ii) 10 mM Tris–HCl, 10 mM MgCl2, and 250 mM sucrose for 1 min. The control membrane vesicles were also used to estimate passive diffusion. The transport was quenched by the addition of 200 μL of cold wash buffer (40 mM MOPS-Tris, pH 7.0, 70 mM KCl), and the solution was transferred to a 96-well filter plate. The filter plate was washed with 5 × 200 μL of ice-cold wash buffer under vacuum filtration. The substrate contained in the vesicles was eluted with 100 μL of 1:1 acetonitrile:0.2% formic acid containing internal standard (diclofenac, 500 nM) and subjected to LC–MS/MS analysis. The transport kinetic analyses were then carried out for the transporters that exhibited activity in the initial screening assay. The kinetic parameters (Jmax and Km) were derived from assay conducted using a substrate concentration range of 1 to 100 μM at 37 °C for 30 s, and fitting into the Michaelis–Menten equation. The functional transport activity of the vesicles was confirmed by using probe substrates and is published elsewhere (Table S3).[40]

Time- and Concentration-Dependent Uptake of Furosemide in OATP2B1 Transfected Cells

The time course of furosemide uptake by OATP2B1-transfected MDCK-II cells was evaluated to determine the incubation time required for initial uptake rate estimates. The transport studies were initiated by seeding cells at a density of ∼2 × 105 cells per well in 24-well poly-d-lysine-coated plates. The cells were grown in DMEM containing 10% fetal bovine serum and 0.1 mg/mL streptomycin and incubated at 37 °C/5% CO2 for 24 h. The media was removed, and the cells were washed three times with PBS. After acclimatization in HBSS buffer for 10 min, the cells were treated with 20 μM furosemide for 2, 5, 10, and 15 min. Similarly, for the transport kinetic assay of furosemide, the uptake rate at varying concentrations (0.5–500 μM) was measured for 15 min at 37 °C in mock or OATP2B1-transfected MDCK-II cells. The cellular uptake was terminated by quickly removing drug solution and washing the cells three times with 500 μL of ice cold HBSS buffer. The cells were then lysed by 300 μL of 100% methanol containing the internal standard for LC–MS/MS analysis or 0.1% Triton X100 for analyzing protein content. The samples were centrifuged at 3000 × g for 10 min, and the cell supernatant was transferred to a 96-deep well plate for LC–MS/MS analysis. The functional transport activity of the OATP2B1 transfected cells was confirmed initially by using probe substrates for OATP2B1, i.e., estrone-3-sulfate and rosuvastatin at 2 μM for 2 min (Figure S6). The mock-transfected MDCK-II cells were used as the control.

Quantification of Efflux Transporters in Vesicles and Transfected hOATP2B1-MDCK-II Cells by Quantitative LC–MS/MS Proteomics

The total membrane proteins from OATP2B1-transfected cells were isolated using an optimized protocol.[74] The total protein content in the membrane samples was determined by using a BCA protein assay kit (Pierce Biotechnology, Rockford, IL), which was diluted to a working concentration of 2 mg/mL. A total of 25 μg total vesicular protein (diluted to 80 μL) and 80 μg membrane protein of hOATP2B1-MDCK-II cells (diluted to 56 μL) were incubated with 10 μL of dithiothreitol (250 mM), 30 μL of ammonium bicarbonate buffer (100 mM, pH 7.8), 20 μL of BSA (0.02 mg/mL), and 10 μL of human serum albumin (10 mg/mL) at 95 °C for 10 min. After cooling down to room temperature, 20 μL of iodoacetamide (500 mM) was added to the mixture and incubated at room temperature for 30 min in the dark. To concentrate the sample, ice-cold methanol (0.5 mL), chloroform (0.1 mL), and water (0.4 mL) were added to each sample and thoroughly mixed by vortex. After centrifugation at 16,000 × g for 5 min at 4 °C, the pellet was washed once with ice-cold methanol (0.5 mL) and centrifuged at 8000 × g for 5 min at 4 °C. The pellet was re-suspended with 60 μL of ammonium bicarbonate buffer (50 mM). Finally, the protein sample was digested with 20 μL of trypsin at a 1:10 trypsin:protein ratio (w/w) and incubated for 16 h at 37 °C with mixing at 300 rpm. The digestion reaction was quenched by 20 μL of chilled stable-labeled peptide internal standard (dissolved in 80% acetonitrile with 0.5% formic acid) and centrifuged at 4000 × g for 5 min at 4 °C. All samples were digested and processed in triplicates.

The surrogate peptides of BCRP, MRP4, and OATP2B1 were quantified in the digested samples using a validated LC–MS/MS method using an SCIEX Triple Quadrupole 6500 system (Framingham, WA) coupled to an ACQUITY UPLC system (Waters Technologies, Milford, MA).[74,75] A total of 5 μL of each sample was injected to the column (ACQUITY UPLC HSS T3 1.8 μm, C18 100A; 100 × 2.1 mm, Waters, Milford, MA). The chromatographic method was used with a gradient mobile phase (0.3 mL/min) consisted of 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B) (Table S4). Unlabeled peptides represent analytes and the corresponding stable-labeled peptides were used as internal standards (Table S5). The pooled total membrane sample isolated from liver tissue with known transporter abundance was used as a calibrator for estimation of abundances of individual transporters in vesicles and cells. The calibration curve range and linearity were verified by serial dilutions of the studied transporter peptide standards. The LC–MS/MS data were analyzed by Skyline software (University of Washington, Seattle, WA).

LC–MS/MS Analysis of Furosemide

The amount of furosemide retained in the vesicles and MDCK-II cells was quantified by LC–MS/MS using an SCIEX 6500 instrument (Framingham, WA) coupled to an ACQUITY UPLC system (Waters Technologies, Milford, MA). A total of 5 μL of sample was injected to the column (ACQUITY UPLC HSS T3 1.8 μm, C18 100A; 100 × 2.1 mm, Waters, Milford, MA) using a gradient mobile phase (0.3 mL/min) consisted of 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B) (Table S6). Furosemide and diclofenac (internal standard) were monitored using MRM transitions, 328.9 → 285 and 204.8, and 250.1 → 214 and 178, respectively.

Transporter Data Analysis

For vesicular efflux assays, the net vesicular transport was calculated as the difference of Transport (ATP) – Transport (AMP). The kinetic parameters of vesicular transport were obtained by fitting the Michaelis–Menten equation (J = Jmax × [S]/([S] + Km)), where J is the velocity in pmol/min/mg protein, Jmax is the maximal velocity, [S] is the substrate concentration (μM), and Km is the Michaelis–Menten constant (μM), using GraphPad Prism version 5.0 (La Jolla, CA). All the experiments were performed in triplicate, and the results are expressed as the mean and standard deviation. The transporter activity (pmol/min/pmol of transporter) was determined using previously reported data of the total protein abundance in vesicles (Evesicles,total in pmol/mg) normalized for the percent of inside-out vesicles (Table S1)[40] as shown in eq .

For the cellular uptake assays, the transport facilitated by OATP2B1 was determined using eq .where, UptakeMDCK-II, OATP2B1 and UptakeMDCK-II, MOCK are the uptake values (pmol/min/mg protein) obtained in OATP2B1-transfected MDCK-II cells and mock-transfected MDCK-II cells, respectively.

Refinement of Oral PBPK Model

Although we initially estimated the individual transporter contributions using the transport kinetics data, the same was not included in the model to limit overparameterization. Instead, we calculated the net uptake, i.e., contribution of the passive diffusion and transported mediated uptake, using the parameter estimation function of Simcyp by optimizing Ptrans,0 of the MechPeff model against the observed data.[23,41,42] Once the plasma concentration time profiles and PK parameters were captured, the model was extrapolated to estimate the effect of food on PK of furosemide by using fed-state physiology in Simcyp.[23]