Physiologically based pharmacokinetic prediction …dmd.aspetjournals.org/.../2014/08/04/dmd.114.058461.full.pdfDMD #58461 1 Physiologically based pharmacokinetic prediction of telmisartan

DMD #58461

1

Physiologically based pharmacokinetic prediction of telmisartan in human

Rui Li, Avijit Ghosh, Tristan S. Maurer, Emi Kimoto, Hugh A. Barton

Department of Pharmacokinetics, Dynamics, and Metabolism, Worldwide Research and Development, Pfizer Inc.,

Cambridge, MA (R.L., A.G., and T.S.M.) and Groton, CT (E.K. and H.A.B.).

This article has not been copyedited and formatted. The final version may differ from this version.DMD Fast Forward. Published on August 4, 2014 as DOI: 10.1124/dmd.114.058461

at ASPE

T Journals on June 10, 2018

dmd.aspetjournals.org

Dow

nloaded from

http://dmd.aspetjournals.org/

DMD #58461

2

Running title: Physiologically based PK prediction of telmisartan in human

Corresponding author: Hugh A. Barton

Email address: [email protected]

Telephone number: +1 860 715 1020

Fax number: +1 860 686 5665

Number of text pages: 20

Number of tables: 3

Number of figures: 7

Number of references: 41

Number of words in the Abstract: 187

Number of words in the Introduction: 531

Number of words in the Discussion: 1500

List of nonstandard abbreviations

IVIVE, in vitro to in vivo extrapolation; Kp, tissue to plasma partition coefficient; HBSS, Hanks’ balanced salt

solution; HLM, human liver microsome; LC-MS/MS, liquid chromatography-tandem mass spectrometry; OATP,

organic anion-transporting polypeptide; PBPK, physiologically based pharmacokinetic; PET, positron emission

tomography; PK, pharmacokinetic(s); SCHH, sandwich culture human hepatocytes; SF, scaling factors; UGTs,

UDP-glucuronosyltransferase.


at ASPE



Dow

nloaded from


DMD #58461

3

Abstract

A previously developed physiologically based pharmacokinetic (PBPK) model for hepatic transporter

substrates was extended to an organic anion transporting polypeptide (OATP) substrate, telmisartan. Predictions

utilized in vitro data from sandwich culture human hepatocyte (SCHH) and human liver microsome (HLM) assays.

We have developed a novel method to calibrate partition coefficients (Kps) between non-liver tissues and plasma

based on published human positron emission tomography (PET) data in order to decrease the uncertainty in tissue

distribution introduced by in silico predicted Kps. With in vitro data-predicted hepatic clearances, published

empirical scaling factors (SFs), and PET-calibrated Kps, the model could accurately recapitulate telmisartan PK

behavior before 2.5 hours. Reasonable predictions also depend on having a model structure that can adequately

describe the drug disposition pathways. We showed that the elimination phase (2.5 to 12 hours) of telmisartan PK

could be more accurately recapitulated when enterohepatic recirculation of parent compound derived from intestinal

deconjugation of glucuronide metabolite was incorporated into the model. This study demonstrated the usefulness of

the previously proposed PBPK modeling approach for purely predictive IV PK simulation and identified additional

biological processes that can be important in prediction.


at ASPE



Dow

nloaded from


DMD #58461

4

Introduction

Predicting plasma and tissue pharmacokinetic profiles in human is a critical step during drug discovery.

Several attempts have been made to predict drug disposition through physiologically based pharmacokinetic (PBPK)

modeling incorporating in vitro data. PBPK models offer the opportunity to characterize the underlying biological

processes, and hence, have the potential to improve prediction accuracy. Typically, generic PBPK models assume

flow-limited distribution and well-stirred tissue kinetics. Successful predictions using these PBPK models have been

made for lipophilic compounds whose absorption and distribution are governed by passive processes (Jones et al.,

2006; Rostami-Hodjegan and Tucker, 2007; Jones et al., 2011). However, these models would not be expected to

provide reasonable predictions for compounds that are actively transported. To improve predictions for transporter

substrates, in vitro assays assessing active processes have been developed, and the active uptake rates estimated

from these assays have been integrated into PBPK models (Poirier et al., 2009; Watanabe et al., 2009; Jamei et al.,

2014). In these models, reasonable predictions for substrates of liver transporters are achieved only when empirical

in vitro to in vivo extrapolation (IVIVE) scaling factors (SFs) are applied. Previously, our group has developed a

liver transporter PBPK model and a unified set of SFs for hepatic uptake, passive diffusion, biliary and metabolic

clearances, which adequately describe the human PK of seven structurally diverse OATP substrates, based only on

in vitro sandwich cultured human hepatocytes (SCHH) and human liver microsome (HLM) data (Jones et al., 2012;

Li et al., 2014).

The aim of this study is to predict the human PK of an OATP substrate not previously evaluated, by

deriving the relevant hepatic clearance parameters from in vitro data and using the previously estimated empirical

SFs for those clearances within the PBPK framework. Telmisartan, a potent, long-acting, small molecule antagonist

of the angiotensin II type-1 receptor (Wienen et al., 2000), was chosen as the compound in this study because it is an

OATP substrate with clinical intravenous infusion data not included in our previous work (Jones et al., 2012; Li et

al., 2014). It has been reported to be predominantly transported into liver by OATP1B3 (encoded by the SLCO1B3

gene) (Ishiguro et al., 2006) and metabolized by UDP-glucuronosyltransferase (UGTs) 1A3 (Ieiri et al., 2011), with

negligible biliary and renal clearance of parent (Stangier et al., 2000a). Multiple clinical data (Stangier et al., 2000a;

Stangier et al., 2000b) are available to evaluate predictions. To help with the analysis, we split the time course into

the infusion and distribution phase (from 0 to 2.5 hour) and the elimination phase (after 2.5 hour).


at ASPE



Dow

nloaded from


DMD #58461

5

During the infusion and distribution phase, Kp values, governing tissue distribution, had a major impact on

plasma PK prediction. To estimate Kp, we translated the total energy absorbed by human tissues as reported in the

PET study (Shimizu et al., 2012) into tissue concentrations, as the total energy emitted from radiolabeled compound

is a function of the amount in the tissues. Human plasma PK predictions with in silico Kp values were compared

with the prediction using the PET Kp values. Deconjugation of glucuronide metabolite and enterohepatic

recirculation of parent were added to the PBPK model to account for the long half-life observed in telmisartan

clinical data.


at ASPE



Dow

nloaded from


DMD #58461

6

Materials and Methods

Modeling and simulations of plasma PK during the infusion and distribution phase

In vivo Model.

A previously published intravenous PBPK model (Li et al., 2014) was used to model the in vivo data. All

compartments are connected by the circulating blood system. Telmisartan distribution in the tissues except liver, and

the two tissues with the largest volume (i.e., adipose and muscle), is assumed to be perfusion limited. Telmisartan

distribution in the liver, adipose, and muscle is assumed to be permeability limited. The liver is modeled as a tissue

with 5 pairs of sub-compartments for liver blood (extracellular) and tissue (intracellular), whereas the adipose and

muscle are modeled with one pair of extracellular and intracellular sub-compartments. Renal clearance was set to

zero based on a previous report (Stangier et al., 2000a). In vitro unbound active uptake clearance (CLint, u, act),

unbound passive distributional clearance (CLint, u, pass), and unbound biliary clearance (CLint, u, bile), for telmisartan, as

well as unbound biliary clearance for telmisartan glucuronide (CLint, u, bile, glu) were estimated from SCHH using a

mechanistic model described in the following section. In vitro unbound metabolic clearance (CLint, u, met) was

estimated using HLM data. These in vitro clearances were physiologically scaled as described previously (Houston,

1994). Empirical scaling factors (SFs) for OATP substrates were those estimated in the previous study, in which the

active (SFact), passive (SFpass), metabolic (SFmet), and biliary (SFbile) scaling factor are 55, 0.092, 0.11, and 0.019,

respectively (Li et al., 2014). The product of physiologically scaled clearance and the previously derived empirical

SF is the in vivo clearance applied for in vivo simulation.

The following equations were added to the above PBPK model to predict the concentration of glucuronide

in liver tissue (CIC, glu) and residual liver bile (Cbile, glu), and the concentration of telmisartan in the residual liver bile

(Cbile, tel):

,, , , , , ,

, ,, , , , , , , , , , , ,

bile telbile bile u bile IC tel u liver tel liver bile bile tel

IC i glu gluIC met u met IC i tel u liver tel bile glu u bile glu IC i glu u liver glu

tel

bbile

dCV SF SCL C f Q C

dtdC MW

V SF SCL C f SF SCL C fdt MW

dCV

⋅ = ⋅ ⋅ ⋅ − ⋅

⋅ = ⋅ ⋅ ⋅ ⋅ − ⋅ ⋅ ⋅

⋅ ,, , , , , , , ,

ile glubile glu u bile glu IC glu u liver glu liver bile bile gluSF SCL C f Q C

dt= ⋅ ⋅ ⋅ − ⋅

(1)


at ASPE



Dow

nloaded from


DMD #58461

7

where fu, liver, tel and fu, liver, glu are the unbound fraction of telmisartan and glucuronide in the liver tissue. SCLu,bile,

SCLu,met, and SCLu, bile, glu are the physiologically scaled CLint, u, bile, CLint, u, met of unbound telmisartan, CLint, u, bile, glu of

telmisartan glucuronide. SFbile, glu is the biliary scaling factor for the glucuronide. Its value is either assumed to be the

same as SFbile, or was calibrated by fitting the total amount of telmisartan and telmisartan glucuronide in the liver

observed in the PET study (Shimizu et al., 2012). Vbile and Qliver, bile are the volume of bile ducts in the liver and the

bile flow rate. Vbile and Qliver, bile were set to previously reported values, i.e., 0.318% of liver volume (Casali et al.,

1994) and 350 mL/day (Davies and Morris, 1993), respectively. MWglu and MWtel are the molecular weight of

telmisartan glucuronide (690.76 g/mol) and telmisartan (514.62 g/mol).

Kp Calculation

Kp is defined as

( )

( )tissue

tissueplasma

C tKp

C t≈ (2)

Due to the rapid equilibrium in the post-distributional phase, the ratio between tissue concentration (Ctissue (t)) and

plasma concentration (Cplasma (t)) at any time t is assumed to be Kptissue.

We calculated tissueKp from the previously published absorbed dose in each tissue (Atissue) in the PET study

(Shimizu et al., 2012) and the plasma concentration time course. The time dependent molar concentration of drug in

plasma, ( )plasmaC t , is estimated using a previously published three-compartment population pharmacokinetic model

(Wallenstein et al., 1999).

The mean administered dose in the PET study is 1.05 µg (2.043×10-9 mole). If one assumes that every telmisartan

compound had been labeled with one 11C, the number of 11C disintegrations per unit time or total radioactivity ( Ω )

of that dose would be 6.972×105 MBq:

dose AN N λ=Ω ⋅ ⋅ (3)


at ASPE



Dow

nloaded from


DMD #58461

8

where doseN is molar amount of the dose, AN is Avogadro’s number, and λ is the decay constant of 11C. As the

observed mean injected activity was 106.2 MBq, this indicates that most of the dose is in fact unlabeled. The ratio of

labeled telmisartan at t = 0 to total dosed telmisartan is:

4

5

(0)(0)

106.21.52 10

6. 10972

labeled

dose

NR

N

−

=

= = ××

(4)

where R is the ratio between Ndose and amount of radiolabeled molecules (Nlabeled). This value is similar to a reported

typical value of only 1 in 1000 molecules labeled with 11C for PET studies (Schlyer, 2004). The amount of

radiolabeled compounds (Nlabeled) (and therefore the ratio ( )R t ) decreases due to radioactive decay according to the

standard 1st order radioactive decay process:

11 (0)

( )

(0)

t

dose

C

t

N eR t

N

R e λ

λ−

−

⋅=

= ⋅

(5)

By accounting for the volume of the tissue and rearranging equation (2), one can derive an expression for the total

amount of drug in the tissue at some time t :

( ) ( )

( ) ( )tissue tissue plasma

tissue tissue tissue plasma

C t Kp C t

N t Kp V C t

⋅

⋅

=

⋅= (6)

Since the decay rate is independent of tissue, this can be converted to the amount of 11C labeled drug in the tissue of

interest using eqn. (5):

11,( ) ( ) (0) t

tissue tissue plasmatissue CN t Kp V t R eC λ−= ⋅⋅ ⋅ ⋅ (7)

The total number of 11C disintegrations, D, in the tissue over the time course T of the experiment is expected to be:

11

0

,0

(

( )

( )0)

T

A tissue C

tA tissue tissue plasma

T

D N N t

N Kp V t e dt

dt

R C λ

λ

λ −

= ⋅ ⋅

= ⋅ ⋅ ⋅ ⋅ ⋅

∫

∫ (8)


at ASPE



Dow

nloaded from


DMD #58461

9

For each 11C disintegration, two positrons are generated and immediately annihilated releasing κ joules of energy,

where for dual positron annihilation, 131.637 10 Jκ −= × . This implies that the total energy per kg tissue (the

absorbed dose tissueA ) during the time course T of the experiment must be:

0

(0( )

) tA tissue tissuetissue plasma

tis

T

sue

N Kp Vt e d

RtA C

Wλλ κ −⋅ ⋅ ⋅ ⋅ ⋅

= ⋅∫ (9)

Rearranging, one can derive an expression for Kp in terms of the absorbed dose and the plasma PK:

0

1

(0) ( )

tissue tissuetissue

tA tissue plasm

T

a

W AKp

RN V t eC dtλκ λ −

⋅=

⋅ ⋅ ⋅ ⋅ ⋅∫ (10)

If one makes the assumption that tissue density is approximately 1 g / cm3, eqn. (10) can be simplified yielding:

0

1

(0) ( )

tissuetissue

tA plasm

T

aN t e dt

AKp

R C λκ λ −=

⋅ ⋅ ⋅ ⋅∫ (11)

The PET derived tissueKp values are corrected for residual blood in the tissue:

( ) ( )

,,

, /

, ,

/

tissue uncorrectedtissue uncorrected

plasma

tissue tissue blood blood tissue blood

plasma

tissue bloodtissue corrected B P

tissue blood tissue blood

tissue corrected tissue uncorr

CKp

C

C V C V V V

C

V VKp R

V V V V

Kp Kp

=

⋅ + ⋅ +=

= ⋅ + ⋅+ +

= /blood tissue blood

ected B Ptissue blood tissue

V V VR

V V V

⎛ ⎞ +− ⋅ ⋅⎜ ⎟+⎝ ⎠

(12)

where volumes of tissues (Vtissue) and residual blood (Vblood) are obtained from (Shah and Betts, 2012) and RB/P, tel is

the blood to plasma ratio of telmisartan (Table 1).

To test the models against literature PK data, the volume of distribution at steady state (Vdss) can be

estimated from the derived Kp values as follows:

/SS P E E P tissueVd V V R V Kp= + ⋅ + ⋅∑ (13)


at ASPE



Dow

nloaded from


DMD #58461

10

where VP and VE are volumes of plasma and erythrocytes (Shah and Betts, 2012); RE/P is the erythrocyte/plasma

concentration ratio (Table 1). For comparison purpose, Kp values were also predicted using a standard in silico

method (Rodgers and Rowland, 2006).

In vitro clearance estimation using sandwich-cultured human hepatocytes (SCHH) assay

Materials. In VitroGro-HT (thawing), In VitroGro-CP (plating), and In VitroGro-HI (incubation)

hepatocyte media were purchased from Celsis In Vitro Technologies Inc. (IVT) (Baltimore, MD). Hanks’ balanced

salt solution (HBSS) was purchased from Invitrogen (Carlsbad, CA). BioCoat 24-well plates and MatrigelTM were

purchased from BD Biosciences (Bedford, MA). The BCA Protein Assay Kit was purchased from Pierce

Biotechnology (Rockford, IL). HPLC grade acetonitrile (ACN) and water were obtained from Burdick & Jackson

(Muskegon, MI) and Mallinckrodt Baker (Phillipsburg, NJ), respectively. Telmisartan and telmisartan acyl-β-D-

glucuronide were purchased from Sequoia Research Products (Pangbourne, UK) and Toronto Research Chemicals

Inc. (Ontario, Canada), respectively. All other chemicals were purchased from Sigma (St. Louis, MO).

Cryopreserved human hepatocyte lot 310 was purchased from BD Biosciences (Woburn, MA), and HU4241 was

purchased from CellzDirect (Pittsboro, NC).

Experimental procedure. Plateable cryopreserved hepatocytes (310 and HU4241) were thawed and plated

as described previously (Bi et al., 2006). Briefly, hepatocytes were thawed in a water bath at 37 ºC and placed on ice.

The cells were then poured into In VitroGro-HT medium at 37 ºC at a ratio of one vial/50 mL in a conical tube. The

cells were centrifuged at 50 × g for 3 minutes and resuspended at 0.75 × 106 cells/mL in In VitroGro-CP medium.

Cell viability was determined by trypan blue exclusion. On day 1, hepatocyte suspensions were plated in collagen-

coated 24-well plates at a density of 0.375 × 106 cells/well in a volume of 0.5 mL/well. After 18 to 24 hours of

incubation at 37ºC, cells were overlaid with ice-cold 0.25 mg/mL MatrigelTM in In VitroGro-HI medium at 0.5

mL/well. Cultures were maintained in In VitroGro-HI medium that was refreshed every 24 hours. On day 5 of

SCHH, the hepatocytes were first rinsed twice with Ca2+/Mg2+ containing or free (containing 1 mM EGTA for free)

HBSS buffer, and then pre-incubated for 10 minutes with Ca2+/Mg2+ containing HBSS, Ca2+/Mg2+ free HBSS

containing 1 mM EGTA, or Ca2+/Mg2+ containing HBSS buffer in the presence of 100 µM rifamycin SV. After

aspirating the pre-incubation buffer, 0.5 mL of incubation buffer containing substrate was added in the absence or


at ASPE



Dow

nloaded from


DMD #58461

11

presence of rifamycin SV. The uptake was terminated at a designated time (0.5, 1, 2, 5, 10, 15 and 20 minutes) by

adding 0.5 mL of ice cold Ca2+/Mg2+ containing HBSS buffer after removal of the incubation buffer. Cells were then

washed three times with 0.5 mL of ice cold Ca2+/Mg2+ containing HBSS buffer. The hepatocytes were lysed with

methanol containing the internal standard for liquid chromatography – tandem mass spectrometry (LC-MS/MS)

quantification.

LC-MS/MS analysis of substrates. LC-MS/MS analysis of telmisartan and telmisartan acyl-β-D-

glucuronide was analyzed on an API-4000 triple quadrupole mass spectrometer (Applied Biosystem, Foster City,

CA) with an atmospheric pressure electrospray ionization source (MDS SCIEX, Concord, Ontario, Canada) that is

connected to a SLC-20A LC system (Shimadzu, Kyoto, Japan) and HTC PAL autosampler (LEAP Technologies,

Carrboro, NC). Samples (10 µL) were injected onto a Kinetex C18 column (2.6 µm, 100 Å, 30×2.1 mm,

Phenomenex, Torrance, CA) and eluted by a mobile phase with initial conditions of 10% solvent B for 0.2 minutes,

followed by a linear gradient of 10% solvent B to 90% solvent B over 1 minute (solvent A: 100% water with 0.1%

formic acid; solvent B: 100% ACN with 0.1% formic acid) at a flow rate of 0.5 mL/min.

SCHH data analysis. The modeling approach used to analyze SCHH data was developed based on a

previously described method (Jones et al., 2012). Briefly, as illustrated in Figure 1, the model includes

compartments representing the media, cells, and bile in the experiment with compound distributing among them

through CLint, u, act, CLint, u, pass, CLint, u, bile, CLint, u, met, and CLint, u, bile, glu. By varying the assay conditions (+/−

rifamycin SV to inhibit active uptake, +/− Ca2+ to inhibit biliary efflux back into the cell media), the individual

active clearance rates can be derived. To simplify the analysis, the amount of glucuronidated telmisartan that is

either actively transported or passively permeated back into the media from cells is assumed to be negligible. This is

consistent with previous findings that only a very limited amount of glucuronide is detected in the blood (Stangier et

al., 2000a). Binding constants (KB) are used to model nonspecific binding to cells at time zero (Jones et al., 2012). In

the presence of Ca2+/Mg2+ (Condition A in Figure 1), the data were modeled as


at ASPE



Dow

nloaded from


DMD #58461

12

int, , int, , int, ,,, , , , , ,

int, , int, , int, , int,,, , ,

u pass u act u passtel mediatel media u tel media tel cell u tel cell

media cell

u pass u act u pass utel celltel media u tel media

media

CL CL CLdXX f X f

dt V V

CL CL CL CLdXX f

dt V

+= − ⋅ ⋅ + ⋅ ⋅

+ += ⋅ ⋅ − , int, ,

, , ,

, int, ,, , ,

, int, , ,int, ,, , , , , ,

,

bile u mettel cell u tel cell

cell

tel bile u biletel cell u tel cell

cell

glu cell u bile gluu mettel cell u tel cell glu cell u glu cell

cell cell

glu bil

CLX f

V

dX CLX f

dt V

dX CLCLX f X f

dt V V

dX

+⋅ ⋅

= ⋅ ⋅

= ⋅ ⋅ − ⋅ ⋅

int, , ,, , ,

e u bile gluglu cell u tel cell

cell

CLX f

dt V= ⋅ ⋅

(14)

where Xtel, medial, Xtel, cell, and Xtel, bile are the amounts of telmisartan in media, cells, and bile, in pmoles/mg protein.

The observed telmisartan data with calcium, with or without rifamycin SV, are the sum of Xtel, cell and Xtel, bile, while

without calcium the data are just the amount in the cells, Xtel, cell. Xglu, cell and Xglu, bile are the amounts of telmisartan

glucuronide in cells and bile, in pmoles/mg protein. The observed glucuronide data with calcium, with or without

rifamycin SV, are the sum of Xglu, cell and Xglu, bile, while without calcium the data are just the amount in the cells, Xglu,

cell. fu, tel, cell, fu, tel, media, and fu, glu, cell are the unbound fraction of telmisartan in cells, telmisartan in media, and

glucuronide in cells, respectively, where fu, tel, media was assumed to be 1 due to the absence of protein in the media.

Vmedia and Vcell are the volume of media and cells. When rifamycin SV was present to inhibit active uptake clearance

(Condition B in Figure 1), an inhibited metabolic clearance (Finhibition × CLint, u, met) was used to represent potential

inhibition of glucuronidation by rifamycin SV observed in the current study and a previous study (Gan et al., 2010):

int, , int, ,,, , , , , ,

int, , int, , int, , int,,, , ,

u pass u passtel mediatel media u tel media tel cell u tel cell

media cell

u pass u pass u bile inhibitiontel celltel media u tel media

media

CL CLdXX f X f

dt V V

CL CL CL F CLdXX f

dt V

= − ⋅ ⋅ + ⋅ ⋅

+ + ⋅= ⋅ ⋅ − ,

, , ,

, int, ,, , ,

, int, , ,int, ,, , , , , ,

, int, ,

u mettel cell u tel cell

cell

tel bile u biletel cell u tel cell

cell


cell cell

glu bile u

X fV

dX CLX f

dt V

dX CLCLX f X f

dt V V

dX CL

dt

⋅ ⋅

= ⋅ ⋅

= ⋅ ⋅ − ⋅ ⋅

= ,, , ,

bile gluglu cell u tel cell

cell

X fV

⋅ ⋅

(15)

With Ca2+/Mg2+ free media (Condition C in Figure 1), the data were modeled as


at ASPE



Dow

nloaded from


DMD #58461

13

int, , int, , int, , int, ,,, , , , , ,

int, , int, , int,,, , ,

u pass u act u pass u biletel mediatel media u tel media tel cell u tel cell

media cell

u pass u act utel celltel media u tel media

media

CL CL CL CLdXX f X f

dt V V

CL CL CLdXX f

dt V

+ += − ⋅ ⋅ + ⋅ ⋅

+= ⋅ ⋅ − , int, , int, ,

, , ,

, int, , ,int, ,, , , , , ,

pass u bile u mettel cell u tel cell

cell


cell cell

CL CLX f

V

dX CLCLX f X f

dt V V

+ +⋅ ⋅

= ⋅ ⋅ − ⋅ ⋅

(16)

The model was implemented in MATLAB (version 7.11, 2010b, Mathworks, Natick, MA). With the model and

observed amounts in SCHH, CLs were estimated by simultaneously fitting telmisartan and glucuronide data under

different conditions using pattern search, a global optimization method (Global Optimization Toolbox TM, version

3.1, Mathworks, Natick, MA). Mean values of the parameters are estimated by simultaneously fitting all data. The

coefficients of variation (CV) are calculated using values estimated in individual fitting of each replicate.

Modeling and simulations of plasma PK during elimination phases

In vivo Model

The previously published PBPK model can adequately describe the human plasma elimination phase

profiles for four compounds with metabolic clearance (i.e., bosentan, fluvastatin, cerivastatin, and repaglinide) (Li et

al., 2014). The HLM CLint, u, met values for these compounds are 22, 29, 76, and 128 μL/min/mg, respectively (Jones

et al., 2012), which are smaller than the value for telmisartan (395 µL/min/mg microsomal protein with BSA (Gill et

al., 2012), 1212 µL/min/mg microsomal protein with BSA (this study)). However, their mean half-lives (3.4, 1.78, 1,

and 1.2 hours, respectively) are much shorter than that of telmisartan (20 hours) (Lindahl et al., 1996; Weber et al.,

1996; Muck et al., 1997; Stangier et al., 2000b; Niemi et al., 2005). To address the inconsistency between in vitro

and in vivo observations, we hypothesize that telmisartan glucuronide is converted by intestinal microflora into

telmisartan, and telmisartan experiences enterohepatic recirculation. Additional terms for the deconjugation of the

telmisartan glucuronide in the GI tract and enterohepatic recirculation of telmisartan were incorporated into the

previously published PBPK model (Figure 2), accounting for the slow elimination phase of telmisartan (time course

after 2.5 hours):


at ASPE



Dow

nloaded from


DMD #58461

14

( ),

, , , ,( )lumen tel tel

liver bile bile tel bile glu lumen tel

glu

a e

dX MWQ C C X

dt MWk k= ⋅ + ⋅ − + ⋅ (17)

where Xlumen, tel is the amount of telmisartan in GI lumen compartment. ka is the absorption rate, and ke is the fecal

excretion rate. The sum of ka and ke is calculated as ka divided by Fa, where Fa is the fractional absorption of active

drug. In this model, we assume that there is no glucuronidation in the GI tract, so the fraction that escapes from

metabolism in the GI tract is assumed to be 1. We assumed conversion from telmisartan glucuronide to telmisartan

in this compartment is effectively instantaneous, considering almost no glucuronide was observed in feces (Stangier

et al., 2000a) and the very fast clearance telmisartan glucuronide in a rat study (180 mL/min/kg, compared with 15.6

mL/min/kg for the telmisartan) (Ebner et al., 1999). Gallbladder emptying after meals was not included in the

modeling scheme because the feeding schedule was not available. The equation for the gut compartment in the

generic PBPK model was modified for recirculation:

, / ,, , ,

,

( )gut tel B P telgut gut a tel gut tel a lumen tel

gut tel

dC RV Q C C k X

dt Kp= − + (18)

where Vgut is the volume of gut tissue, Qgut is the blood flow, Ca, tel is the telmisartan concentration in arterial blood,

Cgut, tel is the telmisartan concentration in gut compartment, and Kpgut, tel is the tissue to plasma partition coefficient of

telmisartan.

Local sensitivity analysis

Local sensitivity analyses were conducted as described before using a numerical approach to approximate

the partial derivatives with respect to each parameter (Jones et al., 2012). Briefly, each drug specific parameter (i.e.

Kp, unbound fraction in plasma (fu, p), blood/plasma ratio (RB/P), fu, liver, fu, liver, glu, CLact, CLpass, CLmet, CLbile, CLbile, glu,

passive clearance in adipose and muscle, ka and Fa) was raised by 1% with respect to its value in the above

simulation and the values of the plasma concentration, AUCplasma, and the amount of telmisartan and glucuronide in

the liver were obtained throughout the time course. Sensitivity coefficients were normalized to both the parameter

value and the model output value, so when the output changes by 1% for a 1% change in the input parameters, the


at ASPE



Dow

nloaded from


DMD #58461

15

sensitivity coefficient is 1 or −1. Only parameters with normalized sensitivity coefficients greater than 0.3 or less

than −0.3 are reported.


at ASPE



Dow

nloaded from


DMD #58461

16

Results

Predicted and experimentally determined Kps.

The Kp values calculated from human PET data (Table 2) can be compared with those predicted from a

standard in silico method. Since the absorbed dose in adipose was not available from the published PET study, we

used the Rodger and Rowland adipose Kp for all the simulations. The PET bone Kp was calculated from the tissue

weight-weighted average of the red marrow and osteogenic cells PET data. The PET gut Kp was calculated from

the tissue weight-weighted average of the large and small intestine wall PET data. Lastly a “rest of body” Kp was

calculated from the tissue weight-weighted average of PET data in pancreas, thymus, and other tissues.

Estimated Vdss using Kp and observed plasma concentrations are given in Table 2. In comparison to the in

silico approach, using the PET Kp values leads to a Vdss that is closer to clinical values, though both sets of Kp

values result in Vdss less than the reported clinical value. This may be due to enterohepatic recirculation substantially

influencing the plasma concentrations post 2.5 hours as analyzed below (Miura et al., 2009; Ieiri et al., 2011).

In vitro SCHH Assay.

The in vitro SCHH data and the model fit are shown in Figure 3. This assay was performed with two

donors (HU4241 and BD310), but telmisartan glucuronide was not monitored in donor BD310. For this reason

CLint,u,met (3.73 μL/min/106 cells with CV of 24%) was fixed at the value estimated from donor HU4241. This value

is close to a previously published value (4.3 μL/min/106 cells) estimated from cryopreserved human hepatocytes

(donor HU4122) (Menochet et al., 2012). The estimates of the in vitro SCHH CLint, u, act, CLint, u, pass, and CLint, u, bile

are shown in Table 1 as the average value from the two lots.


at ASPE



Dow

nloaded from


DMD #58461

17

In vivo PK prediction during infusion and distribution phases

Prediction using in silico and PET Kp

With the in silico Kp values, the model overpredicts the plasma concentration before 1 hour and

underpredicts it after 1 hour for both dosing regimens (Figure 4). Mean squared error (MSE) between data and

predictions with Rodger and Rowland Kp is 0.199. The predictions for both dosing regimens were greatly improved

by using the PET Kp values (MSE = 0.048), although the model still overpredicts the plasma concentration reported

for the 40 mg 30 min infusion dosing. The performance of PET and in silico Kp values were not compared by

simulating plasma concentration after 2.5 hours, because the plasma pharmacokinetics is no longer sensitive to Kp

values as shown below.

Use of the liver PET data for validation of the PBPK model

After a micro dose, the combined total amount of telmisartan and telmisartan glucuronide in the liver

observed in the PET study (Shimizu et al., 2012) and predicted by the PBPK model are shown in Figure 5. The total

amount of telmisartan and telmisartan glucuronide in the liver was overpredicted. The prediction also reached the

maximum amount (tmax = 1.9 hour) later than the data (tmax = 0.8 hour). Given that the published SFbile was estimated

using compounds with very different structures from telmisartan glucuronide, we fixed SFact, SFpass, SFmet, SFbile for

telmisartan at previously published values (e.g. 55, 0.092, 0.11, and 0.019), and re-estimated the SFbile (0.254) for

telmisartan glucuronide by fitting the model to the data (Figure 5A). The result indicates that CLbile may be

underpredicted when using the previously published SFbile. Changing the value of SFbile for telmisartan glucuronide

did not affect the predicted time-concentration profiles of telmisartan in plasma. Using the fitted SFbile, the predicted

total liver concentration is largely glucuronide except at the earliest times (Figure 5B).

In vivo PK prediction during elimination phase

Prediction with enterohepatic recirculation


at ASPE



Dow

nloaded from


DMD #58461

18

The telmisartan PBPK model with the global optimal SFs substantially underpredicts plasma concentration

after 2.5 hours, which suggested that a recirculation model may be needed to predict the slow elimination of

telmisartan. Although including enterohepatic recirculation did not the change model prediction during the infusion

and distribution phases after IV infusion dosing or during the adsorption phase after oral dosing, the model

predictions were greatly improved during the elimination phase (i.e. after 2.5 hours) for both dosing regimens

(Figure 6). Adding enterohepatic recirculation has an impact on the amount of telmisartan and telmisartan

glucuronide in the liver, hence the value of SFbile, glu (0.094) was recalibrated by fitting the total amount of

telmisartan and telmisartan glucuronide in the liver observed in the PET study (Shimizu et al., 2012). Values of ka

(0.68 h−1) and Fa (0.72) were calibrated together with SFbile, glu by simultaneously fitting human plasma

concentration time courses and observed PET liver data, however they may also be estimated from published rat

data (ka = 0.775 h−1 and Fa = 0.61) (Wang et al., 2009; Furukawa et al., 2012). Plasma data after both IV and oral

dosing were digitized from the previous publication (Stangier et al., 2000a) up to 12 hours; large errors may be

introduced by digitizing small values from plots in normal space potentially impacting values after 2.5 hours.

Local sensitivity analysis

The sensitivities of plasma concentration profiles, AUCplasma, and the amount of parent and glucuronide in

the liver were evaluated for all compound specific parameters. A few of the parameters had normalized sensitivity

coefficients greater than 0.3 or less than −0.3 (Figure 7). The plasma concentration is sensitive to Kpmuscle, fu, p, RB/P,

fu, liver, glu, CLact, CLbile, glu, and Fa. AUCplasma is sensitive to fu, p, RB/P, CLact, and Fa. The total amount of telmisartan

and telmisartan glucuronide in the liver is sensitive to Kpgut, RB/P, fu, liver, glu, CLbile, glu, and Fa.


at ASPE



Dow

nloaded from


DMD #58461

19

Discussion

While predicting PK of transporter substrates remains a challenge, several groups have proposed PBPK

models to improve prediction by incorporating active uptake and/or biliary clearance estimated from in vitro data

(Poirier et al., 2009; Watanabe et al., 2009; Jones et al., 2012; Jamei et al., 2014). With empirical SFs, these models

provide reasonable simulations of plasma concentrations for liver OATP substrates. Here we applied our published

modeling strategy for predicating the human PK of OATP substrates to a new compound.

One purpose of this study is to identify potential weak spots in the modeling approach, such as whether our

ability to predict PK is confounded by inaccurate estimates of Kps and limitations of the model structure. Assessing

Kp values is challenging because of the lack of methods to acquire data in humans. Kp values in PBPK models

available in preclinical development, are usually estimated using in silico methods based on tissue composition

(Poirier et al., 2009; Watanabe et al., 2009; Jones et al., 2012). Early in silico methods focused on water-lipid

partitioning in tissues (Poulin and Theil, 2000). Later methods incorporate drug interactions with extracellular

proteins (Rodgers and Rowland, 2006), leading to the improved prediction accuracy for rat Kp and volume of

distribution (Graham et al., 2012). For acids, these methods may still mispredict Kp because they, ignore other

possible mechanisms including specific or nonspecific intracellular binding to protein or compound exclusion due to

membrane potential. Kp values have also been experimentally determined using rodents or predicted using an

empirical regression model based on volume of distribution and/or other physiological properties of the compound

observed in rodents (Yun and Edginton, 2013). However, these methods rely on the rodent data and may result in

misprediction in human due to potential species differences.

In this study, a PET-based Kp calculation method is proposed to test the PBPK model performance when it

is not confounded by misestimation of non-liver tissue distribution, and to validate or calibrate in silico Kp

prediction methods. The comparisons between in silico Kps and PET Kps were done by predicting human IV data

without fitting, using clearance estimated from in vitro data and SFs estimated from other OATP substrates (Li et al.,

2014). In silico Kps led to overpredicted plasma concentrations during the infusion phase, presumably due

to underpredicted distribution to tissues. Plasma concentration was underpredicted in the elimination phase possibly

because overpredicted plasma concentrations in the infusion phase led to excessive clearance of plasma drug. Kp

predicted from another in silico method (Poulin and Theil, 2000), or estimated from experimental rat tissue data


at ASPE



Dow

nloaded from


DMD #58461

20

(U.S. Food and Drug Administration, 1998), were also tested, but led to similar misprediction in human PK (data not

shown). Only the Kp values derived from the PET data were associated with reasonable predictions. It is worth

noting that the PET Kp is essentially calculated as the ratio of tissue and plasma AUCs (Equations 8 and 11), a

method previously reported (Gallo et al., 1987). We used AUC0-1.5 h (the duration of the PET study) to approximate

equilibrium AUC, because the two simulated AUC values are close (data not shown). The predicted Kp values

largely reflect nonspecific binding in the tissues, but the values derived from the PET imaging reflect combinations

of nonspecific binding, passive permeability, and efflux and influx transport or clearance processes. Tissues like

muscle may have limited active uptake and largely would be expected to reflect signal from the parent compound

rather than glucuronide, so the PET-derived Kp more accurately estimates this than current in silico methods do. As

noted in the sensitivity analysis, the muscle Kp is an important determinant of the volume of distribution reflected in

the plasma concentrations, so the 5-fold higher value obtained from the PET imaging impacts the plasma predictions.

By contrast, the much higher value for the PET-derived Kp for liver likely reflects a combination of influx and

efflux transport and metabolism (see Figure 5B). Similarly, the observed radioactivity in the PET study in kidney

may reflect some active uptake and in the small intestine, release of bile, impacting those Kps. Thus, the PET-

derived Kp values provide the best fit to the data through better estimates of the Kp or through compensating for

activities that may not be in the model structure, e.g., active uptake in kidney for parent or glucuronide.

Although the model can accurately predict the 20 min infusion 40 mg human IV data, it still overpredicts

30 min infusion 40 mg human IV data reflecting inconsistencies between these two studies. The discrepancy

between the predictions and data may be due to experimental factors including differences in OATP or UGT

polymorphic forms in the study participants (Ieiri et al., 2011; Yamada et al., 2011) versus the forms in the

hepatocytes used in vitro, as well as culturing or other factors in the in vitro studies or analytical differences in the in

vivo studies.

To account for the slow rate of telmisartan loss during the elimination phase despite the high metabolic

clearance predicted from in vitro studies, deconjugation of glucuronide and enterohepatic circulation of parent were

added into the generic model. Contradictory information about telmisartan enterohepatic recirculation has been

reported. Stangier et al. claimed that there was no evidence of enterohepatic circulation (Stangier et al., 2000a).

However, direct evidence was observed in rats (U.S. Food and Drug Administration, 1998), and two groups claimed


at ASPE



Dow

nloaded from


DMD #58461

21

that double peaks observed in human plasma concentration versus time profiles suggested recirculation occurring

and contributing to the long elimination half-life (Miura et al., 2009; Ieiri et al., 2011). Telmisartan has an apparent

volume of distribution 30 times higher than that of most other acidic compounds (around 7 L/kg versus 0.22 L/kg as

the median value of 159 acidic compounds (Obach et al., 2008) ). The high apparent volume of distribution and low

clearance are characteristics of compounds with efficient and dominant enterohepatic recirculation (Roberts et al.,

2002). However, efficient recirculation requires efficient biliary excretion (Roberts et al., 2002), which is not

observed for telmisartan in SCHH in this study. On the contrary, a high biliary excretion rate is observed for

telmisartan glucuronide in SCHH, which is consistent with previous hypothesis that in man, as in other species,

telmisartan is excreted in bile as the glucuronide (Wienen et al., 2000). Only limited glucuronide has been detected

in plasma and feces, which is believed to be due to very efficient cleavage of glucuronide in the intestine by

bacterial glucuronidases (Stangier et al., 2000a; Wienen et al., 2000). With all this information, it appears likely that

telmisartan enterohepatic recirculation is achieved by efficient biliary excretion and deconjugation of glucuronide.

Such recirculation has been reported for other compounds with prolonged elimination and their conjugates, the

topoisomerase I inhibitor SN-38 and its glucuronide (Kuhn, 1998) and 2-fluoro-β-alanine and its bile acid

conjugates (Zhang et al., 1993). With ka and Fa estimated from either rat or human oral data, the enterohepatic

circulation of parent compound derived from deconjugation of glucuronide enables the model to accurately simulate

the plasma concentration-time course from 0 to 12 hours after either IV or oral dosing. To predict intestinal

deconjugation of glucuronide and resultant enterohepatic recirculation before a clinical study, assays that can

quantify the glucuronide cleavage rate by intestinal bacterial glucuronidases are required in future. The empirical

scaling factor for biliary clearance of glucuronide calibrated by PET data in this study was 5 to 13 times larger than

our published biliary scaling factor. This may be due to the fact that the molecular structure, solubility and other

properties of the glucuronide are very different from the OATP substrates from which the empirical scaling factor

was estimated.

In the sensitivity analysis, plasma concentration, AUCplasma, and liver amount are increasingly sensitive to

Fa as time increases, probably driven by the enterohepatic recirculation. The plasma concentration is sensitive to

Kpmuscle until about 2.5 hours, which indicates Kp values have a major impact during the distribution phase but not

in the elimination phase. The total amount of telmisartan and its glucuronide in the liver is more sensitive to CLact

(the max value of the normalized sensitivity coefficient is 0.22, hence this parameter is not shown in Figure 7) than


at ASPE



Dow

nloaded from


DMD #58461

22

Fa, fu, liver, glu, and CLbile, glu in the distribution phase, however less sensitive to CLact than the other three parameters

during the elimination phase. These results indicate that predictions have different sensitivity to parameters between

the distribution and elimination phases, supporting our two-step modeling scheme.

In summary, a PBPK model of telmisartan and telmisartan glucuronide was developed. Parameterized with

appropriate Kp values, the PBPK model was able to predict the early intravenous plasma concentration of

telmisartan without fitting. A prediction of the liver amount of telmisartan and telmisartan glucuronide was made

and calibrated with PET data. Enterohepatic circulation was proposed and tested using the model to explain the flat

and extended elimination phase for telmisartan. This study demonstrated usefulness of the previously proposed

PBPK modeling approach for IV PK prediction, and identified additional factors that should be considered in

predictions on future OATP substrates.


at ASPE



Dow

nloaded from


DMD #58461

23

Authorship Contributions

Participated in research design: Li, Ghosh, Maurer, and Barton.

Conducted experiments: Kimito.

Performed data analysis: Li and Ghosh.

Wrote or contributed to the writing of the manuscript: Li, Ghosh, Maurer, Kimito, and Barton.


at ASPE



Dow

nloaded from


DMD #58461

24

Reference

Bi YA, Kazolias D, and Duignan DB (2006) Use of cryopreserved human hepatocytes in sandwich culture to

measure hepatobiliary transport. Drug metabolism and disposition: the biological fate of chemicals

34:1658-1665.

Casali AM, Siringo S, Sofia S, Bolondi L, Di Febo G, and Cavalli G (1994) Quantitative analysis of intrahepatic bile

duct component in normal adult human liver and in primary biliary cirrhosis. Pathology, research and

practice 190:201-206.

Davies B and Morris T (1993) Physiological parameters in laboratory animals and humans. Pharmaceutical

research 10:1093-1095.

Ebner T, Heinzel G, Prox A, Beschke K, and Wachsmuth H (1999) Disposition and chemical stability of telmisartan

1-O-acylglucuronide. Drug metabolism and disposition: the biological fate of chemicals 27:1143-1149.

Furukawa T, Nakamori F, Tetsuka K, Naritomi Y, Moriguchi H, Yamano K, Terashita S, and Teramura T (2012)

Quantitative prediction of intestinal glucuronidation of drugs in rats using in vitro metabolic clearance data.

Drug metabolism and pharmacokinetics 27:171-180.

Gallo JM, Lam FC, and Perrier DG (1987) Area method for the estimation of partition coefficients for physiological

pharmacokinetic models. Journal of pharmacokinetics and biopharmaceutics 15:271-280.

Gan J, Chen W, Shen H, Gao L, Hong Y, Tian Y, Li W, Zhang Y, Tang Y, Zhang H, Humphreys WG, and

Rodrigues AD (2010) Repaglinide-gemfibrozil drug interaction: inhibition of repaglinide glucuronidation

as a potential additional contributing mechanism. British journal of clinical pharmacology 70:870-880.

Gill KL, Houston JB, and Galetin A (2012) Characterization of in vitro glucuronidation clearance of a range of

drugs in human kidney microsomes: comparison with liver and intestinal glucuronidation and impact of

albumin. Drug metabolism and disposition: the biological fate of chemicals 40:825-835.

Graham H, Walker M, Jones O, Yates J, Galetin A, and Aarons L (2012) Comparison of in-vivo and in-silico

methods used for prediction of tissue: plasma partition coefficients in rat. The Journal of pharmacy and

pharmacology 64:383-396.

Houston JB (1994) Utility of in vitro drug metabolism data in predicting in vivo metabolic clearance. Biochem

Pharmacol 47:1469-1479.


at ASPE



Dow

nloaded from


DMD #58461

25

Ieiri I, Nishimura C, Maeda K, Sasaki T, Kimura M, Chiyoda T, Hirota T, Irie S, Shimizu H, Noguchi T, Yoshida K,

and Sugiyama Y (2011) Pharmacokinetic and pharmacogenomic profiles of telmisartan after the oral

microdose and therapeutic dose. Pharmacogenet Genomics 21:495-505.

Ishiguro N, Maeda K, Kishimoto W, Saito A, Harada A, Ebner T, Roth W, Igarashi T, and Sugiyama Y (2006)

Predominant contribution of OATP1B3 to the hepatic uptake of telmisartan, an angiotensin II receptor

antagonist, in humans. Drug metabolism and disposition: the biological fate of chemicals 34:1109-1115.

Jamei M, Bajot F, Neuhoff S, Barter Z, Yang J, Rostami-Hodjegan A, and Rowland-Yeo K (2014) A mechanistic

framework for in vitro-in vivo extrapolation of liver membrane transporters: prediction of drug-drug

interaction between rosuvastatin and cyclosporine. Clinical pharmacokinetics 53:73-87.

Jones HM, Barton HA, Lai Y, Bi YA, Kimoto E, Kempshall S, Tate SC, El-Kattan A, Houston JB, Galetin A, and

Fenner KS (2012) Mechanistic pharmacokinetic modeling for the prediction of transporter-mediated

disposition in humans from sandwich culture human hepatocyte data. Drug metabolism and disposition: the

biological fate of chemicals 40:1007-1017.

Jones HM, Gardner IB, Collard WT, Stanley PJ, Oxley P, Hosea NA, Plowchalk D, Gernhardt S, Lin J, Dickins M,

Rahavendran SR, Jones BC, Watson KJ, Pertinez H, Kumar V, and Cole S (2011) Simulation of human

intravenous and oral pharmacokinetics of 21 diverse compounds using physiologically based

pharmacokinetic modelling. Clinical pharmacokinetics 50:331-347.

Jones HM, Parrott N, Jorga K, and Lave T (2006) A novel strategy for physiologically based predictions of human

pharmacokinetics. Clinical pharmacokinetics 45:511-542.

Kuhn JG (1998) Pharmacology of irinotecan. Oncology 12:39-42.

Li R, Barton HA, Yates PD, Ghosh A, Wolford AC, Riccardi KA, and Maurer TS (2014) A "middle-out" approach

to human pharmacokinetic predictions for OATP substrates using physiologically-based pharmacokinetic

modeling. Journal of pharmacokinetics and pharmacodynamics 41:197-209.

Lindahl A, Sandstrom R, Ungell AL, Abrahamsson B, Knutson TW, Knutson L, and Lennernas H (1996) Jejunal

permeability and hepatic extraction of fluvastatin in humans. Clinical pharmacology and therapeutics

60:493-503.


at ASPE



Dow

nloaded from


DMD #58461

26

Menochet K, Kenworthy KE, Houston JB, and Galetin A (2012) Use of mechanistic modeling to assess

interindividual variability and interspecies differences in active uptake in human and rat hepatocytes. Drug

metabolism and disposition: the biological fate of chemicals 40:1744-1756.

Miura M, Satoh S, Kagaya H, Saito M, Inoue T, Ohkubo T, Habuchi T, and Suzuki T (2009) Effect of telmisartan,

valsartan and candesartan on mycophenolate mofetil pharmacokinetics in Japanese renal transplant

recipients. Journal of Clinical Pharmacy and Therapeutics 34:683-692.

Muck W, Ritter W, Ochmann K, Unger S, Ahr G, Wingender W, and Kuhlmann J (1997) Absolute and relative

bioavailability of the HMG-CoA reductase inhibitor cerivastatin. International journal of clinical

pharmacology and therapeutics 35:255-260.

Niemi M, Backman JT, Kajosaari LI, Leathart JB, Neuvonen M, Daly AK, Eichelbaum M, Kivisto KT, and

Neuvonen PJ (2005) Polymorphic organic anion transporting polypeptide 1B1 is a major determinant of

repaglinide pharmacokinetics. Clinical pharmacology and therapeutics 77:468-478.

Obach RS, Lombardo F, and Waters NJ (2008) Trend analysis of a database of intravenous pharmacokinetic

parameters in humans for 670 drug compounds. Drug metabolism and disposition: the biological fate of

chemicals 36:1385-1405.

Poirier A, Cascais AC, Funk C, and Lave T (2009) Prediction of pharmacokinetic profile of valsartan in human

based on in vitro uptake transport data. Journal of pharmacokinetics and pharmacodynamics 36:585-611.

Poulin P and Theil FP (2000) A priori prediction of tissue:plasma partition coefficients of drugs to facilitate the use

of physiologically-based pharmacokinetic models in drug discovery. J Pharm Sci 89:16-35.

Roberts MS, Magnusson BM, Burczynski FJ, and Weiss M (2002) Enterohepatic circulation: physiological,

pharmacokinetic and clinical implications. Clinical pharmacokinetics 41:751-790.

Rodgers T and Rowland M (2006) Physiologically based pharmacokinetic modelling 2: predicting the tissue

distribution of acids, very weak bases, neutrals and zwitterions. J Pharm Sci 95:1238-1257.

Rostami-Hodjegan A and Tucker GT (2007) Simulation and prediction of in vivo drug metabolism in human

populations from in vitro data. Nature reviews Drug discovery 6:140-148.

Schlyer DJ (2004) PET tracers and radiochemistry. Annals of the Academy of Medicine, Singapore 33:146-154.


at ASPE



Dow

nloaded from


DMD #58461

27

Shah DK and Betts AM (2012) Towards a platform PBPK model to characterize the plasma and tissue disposition of

monoclonal antibodies in preclinical species and human. Journal of pharmacokinetics and

pharmacodynamics 39:67-86.

Shimizu K, Takashima T, Yamane T, Sasaki M, Kageyama H, Hashizume Y, Maeda K, Sugiyama Y, Watanabe Y,

and Senda M (2012) Whole-body distribution and radiation dosimetry of [11C]telmisartan as a biomarker

for hepatic organic anion transporting polypeptide (OATP) 1B3. Nucl Med Biol 39:847-853.

Stangier J, Schmid J, Turck D, Switek H, Verhagen A, Peeters PA, van Marle SP, Tamminga WJ, Sollie FA, and

Jonkman JH (2000a) Absorption, metabolism, and excretion of intravenously and orally administered

[14C]telmisartan in healthy volunteers. J Clin Pharmacol 40:1312-1322.

Stangier J, Su CA, and Roth W (2000b) Pharmacokinetics of orally and intravenously administered telmisartan in

healthy young and elderly volunteers and in hypertensive patients. J Int Med Res 28:149-167.

U.S. Food and Drug Administration (1998) Micardis (Telmisartan) NDA 20850 Pharmacology Reviews, in:

http://wwwaccessdatafdagov/drugsatfda_docs/nda/98/20850_MICARDIS_pharmr_P1pdf (Center for Drug

Evaluation and Research ed, pp 27 - 28.

Wallenstein GC, Stangier J, and Ludden TM (1999) Analyzing rich data using different methods provided by

NONMEM: pharmacokinetics of telmisartan following intravenous infusion to healthy volunteers.

Pharmaceutical research 16:772-776.

Wang M, Shan B, Li W, Wang M, and Su X (2009) Influence of indapamide on pharmacokinetics of telmisartan in

male and female rats. Chinese Journal of Pharmacology and Toxicology 23:1-5.

Watanabe T, Kusuhara H, Maeda K, Shitara Y, and Sugiyama Y (2009) Physiologically based pharmacokinetic

modeling to predict transporter-mediated clearance and distribution of pravastatin in humans. The Journal

of pharmacology and experimental therapeutics 328:652-662.

Weber C, Schmitt R, Birnboeck H, Hopfgartner G, van Marle SP, Peeters PA, Jonkman JH, and Jones CR (1996)

Pharmacokinetics and pharmacodynamics of the endothelin-receptor antagonist bosentan in healthy human

subjects. Clinical pharmacology and therapeutics 60:124-137.

Wienen W, Entzeroth M, van Meel JCA, Stangier J, Busch U, Ebner T, Schmid J, Lehmann H, Matzek K,

Kempthorne-Rawson J, Gladigau V, and Hauel NH (2000) A Review on Telmisartan: A Novel, Long-

Acting Angiotensin II-Receptor Antagonist. Cardiovascular Drug Reviews 18:127-154.


at ASPE



Dow

nloaded from


DMD #58461

28

Yamada A, Maeda K, Ishiguro N, Tsuda Y, Igarashi T, Ebner T, Roth W, Ikushiro S, and Sugiyama Y (2011) The

impact of pharmacogenetics of metabolic enzymes and transporters on the pharmacokinetics of telmisartan

in healthy volunteers. Pharmacogenet Genomics 21:523-530.

Yun YE and Edginton AN (2013) Correlation-based prediction of tissue-to-plasma partition coefficients using

readily available input parameters. Xenobiotica; the fate of foreign compounds in biological systems

43:839-852.

Zhang R, Liu T, Soong SJ, and Diasio RB (1993) A mathematical model of the kinetics and tissue distribution of 2-

fluoro-beta-alanine, the major catabolite of 5-fluorouracil. Biochem Pharmacol 45:2063-2069.


at ASPE



Dow

nloaded from


DMD #58461

29

Figure Legends

Fig. 1. Schematic diagram of the in vitro model. Condition A: Ca2+/Mg2+ containing media, where passive

diffusion, active uptake, and metabolic clearances are active. Compounds cleared by biliary clearance are stored in

bile canaliculi (Equation 14). Condition B: 0.1 mM rifamycin SV, where active uptake clearance is fully inhibited;

and metabolic clearance is partially inhibited (Equation 15). Condition C: Ca2+/Mg2+ free, 1 mM EGTA, where

compounds cleared by biliary clearance are released into media from the bile canaliculi (Equation 16).

Fig. 2. Schematic diagram of the enterohepatic circulation model. Black arrows indicate transport of telmisartan;

while white arrows indicate transport of telmisartan glucuronide; red arrows indicate conversion between the two

compounds; and purple arrow indicates fecal excretion.

Fig. 3. Observed and predicted SCHH data. Circles represent the observed data, which were mean values of two

replicates. Solid lines represent simulations with the fitted parameter values. Data from lot HU4241 (A - F,

telmisartan and glucuronide measured) and BD310 (G - I, only telmisartan measured) were fitted individually.

Fig. 4. Predicted and observed human intravenous plasma concentration-time profile of telmisartan with

different Kp values following 40 mg (A) 30 min and (B) 20 min IV infusion dosing. Black circles represent

observed data (Stangier et al., 2000a,b), red solid lines represent predictions using PET derived Kp values, and

black dashed lines represent predictions using Rodgers and Rowland Kp values.

Fig. 5. Observed and simulated amount of telmisartan and telmisartan glucuronide in the liver. (A) Summed

total amount of telmisartan and telmisartan glucuronide in the liver per dosed amount of telmisartan. Circles

represent the observations in the PET study (Shimizu et al., 2012), black dashed line represents simulation when

SFbile for telmisartan glucuronide is set to previously published SFbile (Li et al., 2014), and red solid line represents


at ASPE



Dow

nloaded from


DMD #58461

30

simulation with fitted SFbile for telmisartan glucuronide. (B) Simulated amount of telmisartan (green dashed line)

and amount of telmisartan glucuronide (blue solid line) in the liver with fitted SFbile for telmisartan glucuronide.

Fig. 6. Simulations with and without enterohepatic recirculation following (A) IV infusion or (B) oral dosing.

Circles represent the observations, lines represent simulations with (green solid line) and without (red dashed line)

enterohepatic recirculation.

Fig. 7. Time-dependent local sensitivity of compound specific parameters on (A) plasma concentration of

telmisartan, (B) AUCplasma of telmisartan, and (C) amount of telmisartan and glucuronide in the liver. The

plots only show the parameters that had normalized sensitivity coefficients greater than 0.3 or less than −0.3.


at ASPE



Dow

nloaded from


DMD #58461

31

Tables

Table 1. Summary of Input Parameters for Telmisartan PBPK Model

Parameters Values (CV) Sources

logD7.4 2.5 In-house data

pKa 3.8 In-house data

Blood/plasma ratio (RB/P) 0.78 (Stangier et al., 2000a)

Erythrocyte/plasma ratio (RE/P) 0.5 (Stangier et al., 2000a)

Unbound fraction in plasma (fu, p) 0.005 (Stangier et al., 2000b)

Unbound fraction in human liver

tissue (fu, liver) 0.013 (16%) In-house data

CLu, int, met (μL/min/mg microsomal

protein) 1210 (32%) In-house HLM with BSA data

CLu, int, act (μL/min/106 cells) 53 (53%) This study

CLu, int, pass (μL/min/106 cells) 72 (43%) This study

CLu, int, bile (μL/min/106 cells) 3 (84%) This study

Unbound fraction of glucuronide in

human liver tissue (fu, liver, glu) 0.0073 (15%) In-house data

CLu, int, bile, glu (μL/min/106 cells) 74 (16%) This study


at ASPE



Dow

nloaded from


DMD #58461

32

Table 2. Predicted partition coefficients and volume of distribution1

Tissues Rodger and Rowland Kp PET Kp

adipose 0.35 (0.35)2

bone 0.13 0.47

brain 0.11 0.25

gut 0.22 1.67

heart 0.18 1.54

kidney 0.16 4.40

lung 0.25 1.07

muscle 0.08 0.40

skin 0.37 0.26

spleen 0.11 0.41

testes 0.08 0.26

rest of body 0.08 0.53

Liver3 0.12 11.37

Vdss (L/kg) 0.20 0.67

Clinical Vdss (L/kg)4 8.9 (Stangier et al., 2000a)

6.6 to 7.3 (Stangier et al., 2000b)

1. The body weight is assumed to be 70 kg for all Vdss values.

2. Parentheses indicate the PET Kp is not available for adipose, hence Rodger and Rowland Kp is used for

plasma PK or Vdss prediction.

3. The Kp of liver was not used in PBPK modeling, but only to estimate Vdss here.

4. Clinical Vdss are published values.


at ASPE



Dow

nloaded from


DMD #58461

33

Table 3 AUC1 estimated by nonparametric analysis and PBPK model

Study type AUCplasma, 0 – 2.5 hrs

(h·μg/L)

AUCplasma, 0 – 12 hrs

(h·μg/L)

Nonparametric analysis1

IV (Stangier et al., 2000b) 449 558

IV (Stangier et al., 2000a) 664 852

PBPK model2

Rowland and Rodger Kp3 731 -

PET Kp3 709 -

Without enterohepatic

recirculation4 - 742

With enterohepatic

recirculation4 - 946

1. All studies were performed with 40 mg telmisartan IV infusion dosing;

2. AUC were estimated using trapezoid integration;

3. Calculated without enterohepatic recirculation;

4. Calculated with PET Kp.


at ASPE



Dow

nloaded from


Figure 1

This article has not been copyedited and form

atted. The final version m

ay differ from this version.

DM

D Fast Forw

ard. Published on August 4, 2014 as D

OI: 10.1124/dm

d.114.058461

at ASPET Journals on June 10, 2018 dmd.aspetjournals.org Downloaded from


Figure 2


at ASPE



Dow

nloaded from


Figure 3


at ASPE



Dow

nloaded from


Figure 4




DM

D Fast Forw


OI: 10.1124/dm

d.114.058461 at ASPET Journals on June 10, 2018 dmd.aspetjournals.org Downloaded from


Figure 5




DM

D Fast Forw


OI: 10.1124/dm

d.114.058461 at ASPET Journals on June 10, 2018 dmd.aspetjournals.org Downloaded from


Figure 6


at ASPE



Dow

nloaded from


Figure 7(A)

(B) (C)




DM

D Fast Forw


OI: 10.1124/dm

d.114.058461

at ASPET Journals on June 10, 2018 dmd.aspetjournals.org Downloaded from


Documents

Physiologically based pharmacokinetic prediction …dmd.aspetjournals.org/.../2014/08/04/dmd.114.058461.full.pdfDMD #58461 1 Physiologically based pharmacokinetic prediction of telmisartan