Assessment of gadoxetate DCE-MRI as a biomarker of hepatobiliary transporter inhibition

Drug-induced liver injury (DILI) is a clinically important adverse drug reaction, which prevents the development of many otherwise safe and effective new drugs. Currently, there is a lack of sensitive and specific biomarkers that can be used to predict, assess and manage this toxicity. The aim of this work was to evaluate gadoxetate-enhanced MRI as a potential novel biomarker of hepatobiliary transporter inhibition in the rat. Initially, the volume fraction of extracellular space in the liver was determined using gadopentetate to enable an estimation of the gadoxetate concentration in hepatocytes. Using this information, a compartmental model was developed to characterise the pharmacokinetics of hepatic uptake and biliary excretion of gadoxetate. Subsequently, we explored the impact of an investigational hepatobiliary transporter inhibitor on the parameters of the model in vivo in rats. The investigational hepatobiliary transporter inhibitor reduced both the rate of uptake of gadoxetate into the hepatocyte, k1, and the Michaelis–Menten constant, Vmax, characterising its excretion into bile, whereas KM values for biliary efflux were increased. These effects were dose dependent and correlated with effects on plasma chemistry markers of liver dysfunction, in particular bilirubin and bile acids. These results indicate that gadoxetate-enhanced MRI provides a novel functional biomarker of inhibition of transporter-mediated hepatic uptake and clearance in the rat. Since gadoxetate is used clinically, the technology has the potential to provide a translatable biomarker of drug-induced perturbation of hepatic transporters that may also be useful in humans to explore deleterious functional alterations caused by transporter inhibition. Copyright © 2013 John Wiley & Sons, Ltd.


INTRODUCTION
Drug-induced liver injury (DILI) is responsible for up to 15% of cases of acute liver failure in Europe and the USA and is one of the leading causes of drug attrition and/or delayed progression of new drug candidates (1,2). Underlying mechanisms are still poorly understood and are difficult to identify (2,3). For many drugs that cause DILI in man, liver injury is not evident in preclinical safety studies undertaken in animals, or in early clinical trials, but is first detected only during large and expensive Phase III trials or after the drug has reached the market and a very large number of patients has been exposed (4). Typically, the first indications of liver dysfunction are asymptomatic raised plasma levels of hepatic enzymes such as aspartate aminotransferase (AST) and alanine aminotransferase (ALT) (5). Although these signals are considered useful for identification of drugs that have the potential to cause DILI in the human population, it is important to recognise that they provide no information on the mechanisms that underlie DILI. In addition, for the majority of patients who exhibit hepatic enzyme elevations the liver dysfunction is transient and is not followed by symptomatic DILI, even if drug treatment is continued (4,5). However, a small proportion of patients who exhibit hepatic enzyme elevations may subsequently develop DILI, which in some instances can result in life threatening liver failure (3). Currently, there remains an unmet need for novel biomarker approaches that can provide useful mechanistic insights and, for patients who exhibit hepatic enzyme elevations, can distinguish between individuals who are at risk of DILI and those for whom the liver dysfunction is transient or compensatory.
In man, the most frequent observed patterns of DILI are hepatocellular (i.e. affecting hepatocytes), cholestatic (affecting bile flow and the biliary system) and mixed hepatocellular/cholestatic (where features of both hepatocellular and cholestatic injury are evident) (4,5). Hepatocellular DILI poses particular clinical concern, since for some drugs this may result in substantial loss of hepatocyte mass and development of liver failure. In contrast, the major clinical consequences of cholestatic DILI are chronic liver dysfunction arising from sustained impairment of bile flow, which may result in hepatic or biliary fibrosis, cirrhosis or vanishing bile duct syndrome (6,7).
In recent years, an emerging body of evidence has highlighted the critical role played by hepatic transporters in uptake and biliary clearance of drugs and endogenous molecules (Fig. 1). Uptake of many drugs from blood plasma through the basolateral membrane into the hepatocyte is mediated by a family of multispecific organic anion transporting polypeptides (OATPs), which also mediate uptake of some bile salts and various metabolites (8,9). Many drugs have been found to inhibit the activities of hepatic OATP uptake transporters, and this process has the potential to cause drug-drug interactions that result in altered drug pharmacokinetics (10,11). In addition, drugs, their metabolites and numerous endogenous metabolites are secreted from hepatocytes into bile via specific ATP-dependent active transport processes such as the multidrug resistance-associated protein 2 (MRP2) (12,13). Genetically determined defects in MRP2 activity have been shown to result in elevated levels of conjugated bilirubin in plasma but not overt liver injury in humans (Dubin-Johnson syndrome) (14,15), and in some strains of rats (16,17). Bile salts are excreted into bile primarily by the bile salt export pump (BSEP), and genetically inherited defects in BSEP expression and BSEP activity result in cholestatic liver injury in humans (14,15). It has been observed that many drugs that cause cholestatic or mixed hepatocellular/cholestatic DILI inhibit BSEP activity (18)(19)(20). In view of this, it has been proposed that, in patients who develop liver injury caused by such drugs, inhibition of BSEP activity results in accumulation of bile salts within hepatocytes to cytotoxic levels, thereby leading to cholestatic liver injury (21).
In view of the observed association between hepatic biliary transport inhibition by drugs and increased risk of both drugdrug interactions and DILI, approaches that can be used to quantify transporter inhibition in vitro have been developed and are now starting to be utilised by drug developers. However, it is important to note that, at present, the effectiveness of this approach is limited by an incomplete scientific understanding of the relationship between biliary transporter inhibition in vitro, the functional consequences in vivo and the risk of developing DILI (19). This is due in large part to the restricted range of tools that are currently available to quantify hepatic transporter interactions in vivo, both in preclinical species and in man. The approaches used currently require highly invasive procedures (e.g. bile duct cannulation) or provide only indirect and mechanistically non-specific evidence of liver dysfunction (e.g. evaluation of plasma levels of bile salts, conjugated bilirubin, AST, ALT etc.). There is therefore an urgent need for improved in vivo approaches that can be used to study hepatobiliary transporter function. Non-invasive imaging techniques, such as MRI, are well placed to fill this gap.
Dynamic contrast-enhanced MRI (DCE-MRI) allows characterisation of functional aspects of physiology by adding a temporal dimension to the spatial information provided by MRI. DCE-MRI studies are normally performed using gadolinium-based contrast agents, which can be classified as extracellular (e.g. gadopentetate) or intracellular (e.g. gadoxetate), depending on their biodistribution in the body (22). Gadopentetate is a clinically approved ionic gadolinium chelate that after injection quickly distributes into the extracellular space and is completely cleared through the kidney (23,24). Gadoxetate is a clinically approved (25)(26)(27) hepatobiliary-specific MRI contrast agent (28)(29)(30). It is a lipophilic gadolinium chelate, which is excreted by both liver and kidney, in proportions that are species dependent (30). In rats, gadoxetate clearance is 70% biliary and 30% renal (28), while in humans 50% is excreted via each route (25). Gadoxetate utilises organic anion transport systems in vivo, since hepatic uptake is inhibited by bromosulfophthalein, bilirubin and rifampicin (29)(30)(31)(32). In vitro experiments have shown that gadoxetate uptake is mediated by the liver-specific cell transporters human OATP1B1, OATP1B3 and Na + taurocholate co-transporting polypeptide (NTCP) (20), and rat Oatp1a1 (previously called Oatp1) (33). In vivo experiments in TRrats have shown that biliary efflux of gadoxetate is mediated by Mrp2 (31,34).
In the present study we have developed a novel model that describes the uptake of gadoxetate from the extracellular space into the hepatocyte and excretion into bile. We first used gadopentetate DCE-MRI to calculate the volume fraction of extracellular space in rat liver and spleen. This information was required by the model to enable estimation of the intracellular concentration of gadoxetate in the liver. We then used the model to quantify the hepatic uptake and excretion of gadoxetate in the rat, in the presence and absence of an investigational chemokine receptor antagonist (CKA) compound that we have found to inhibit biliary transporter activity in vitro.
Our results indicate that this technique provides a novel approach for investigation of drug-induced perturbation of hepatic transporters in vivo.

Animal preparation and imaging procedure
All experiments were conducted in compliance with licences issued under the UK Animals (Scientific Procedures) Act 1986 after review by the local ethics committee.
Male Han Wistar rats (232-340 g, Harlan, Bicester, UK) were fasted overnight and anaesthetised using 3% isoflurane in air. Anaesthesia was maintained at 1.5-2.0% throughout the imaging procedure. The tail vein was cannulated with a 26G catheter and flushed with heparin solution to prevent clotting (100 IU/ml in saline). Throughout the imaging procedure, rectal temperature and respiration rate were monitored (SA Instruments, New York, USA) and maintained at 37 C and 50-60 breaths/min, respectively.
MRI was performed at 4.7 T (Oxford Instrument, Oxford, UK) incorporating a high power gradient set (200 mT/m, 121 mm inner diameter); Avance III, Bruker Biospin, Ettlingen, Germany, a 72 mm quadrature birdcage volume transceiver and Paravision 5.1 (Bruker Biospin). The imaging protocol included IntraGate FLASH (47) coronal images for anatomical reference of the abdominal cavity, an inversion recovery FISP (48,49) sequence for R1 mapping of the abdomen and IntraGate FLASH for DCE-MRI (see Table 1). Five minutes after the DCE-MRI acquisition commenced, the contrast agent (see following paragraphs) was administered intravenously (i.v.) via the tail vein with an automatic pump (5 ml/min, 305 piston pump, Gilson, Luton, UK).
We first estimated the volume fraction of extracellular space in the liver. Five healthy rats (282 AE 9 g) were imaged following the protocol described in the previous paragraph and dosed i.v. with 100 mmol/kg of gadopentetate (Magnevist, Bayer Schering Pharma, Berlin, Germany). DCE-MRI data were continuously acquired for 45 min. After the imaging session, animals were euthanised with an overdose of isoflurane.
Second, we characterised the hepatic clearance of gadoxetate when co-administered with an inhibitor of biliary transporter activity (50). Thirty rats (body weight 298 AE 26 g), with six rats per group, were dosed orally with either the chemokine antagonist (CKA) 1-(4-chloro-3-trifluoromethyl-benzyl)-5-hydroxy-1-Hindole-2-carboxylic acid (51) or vehicle. The CKA was synthesised by AstraZeneca and had a purity of more than 99%. The CKA was formulated as suspensions in the vehicle (water containing 0.5% (w/v) hydroxypropyl-methylcellulose/0.1% (w/v) polysorbate 80) and was dosed by oral gavage at 20, 200, 500 or 2000 mg/kg. Thirty minutes after administration of CKA or vehicle, animals were anaesthetised and imaged following the protocol described in Table 1. After a further 30 min (i.e. 60 min after the CKA or vehicle dose) DCE-MRI data acquisition commenced, and 5 min later 25 mmol/kg of gadoxetate (Primovist, Bayer Schering Pharma) was administered i.v. DCE-MRI data were acquired continuously for a total of 60 min. The start of DCE-MRI acquisition was defined as t = 0 min, injection of gadoxetate was t = 5 min and the end of DCE-MRI acquisition was t = 60 min. At the end of the imaging session, animals were euthanised with an overdose of isoflurane and blood and liver tissue samples were taken for analysis.

Plasma chemistry analysis
Blood samples were taken from the vena cava into lithiumheparin tubes and the plasma fraction was separated by centrifugation at 1200 g and 4 C for 10 min. Plasma samples were analysed on a Roche P Modular analyser (Roche Diagnostics, Burgess Hill, West Surrey, UK) using a total bilirubin assay, an alanine aminotransferase (ALAT/GPT) assay (both from Roche Diagnostics) and a total bile acid assay (Alere, Stockport, UK) according to the manufacturers' instructions.

Pathology analysis
At necropsy, 3-4 mm thick slices of the left lateral, right medial and caudate lobes of the liver were taken, immersed in 10% neutral buffered formalin and fixed for 48 h prior to standard tissue processing into paraffin wax. Sections (4-5 mm thick) were stained with haematoxylin and eosin (H&E) and examined by light microscopy.

Calculation of gadoxetate concentration
In all the experiments, regions of interest (ROIs) covering the liver and spleen were manually selected using ImageJ (53), from both IR-FISP and DCE-MRI acquisitions, to calculate the mean longitudinal relaxation rate before contrast injection, R 1 (0), and to extract the time course, respectively. It was assumed that the R 1 of each organ of interest was homogeneous within the whole organ, so R 1 (0) represents the mean pre-contrast longitudinal relaxation over the ROI. The mean post-contrast longitudinal relaxation over the ROI, R 1 (t), was calculated from the IntraGate FLASH images using the signal equation of a saturation recovery spoiled gradient echo (54): with T R and θ the repetition time and flip angle of the IntraGate FLASH acquisition, respectively, and a(t) defined by (54) with S pre and S post (t) the average signal intensity over the ROI of the baseline (i.e. pre-injection) and post-injection at a given dynamic time point, t, respectively. It has been shown that in tumours (55,56) and organs such as skeletal muscle (57) and brain (58), the intercompartmental water exchange is not negligible, and the system departs from the fast exchange limit regime (FXL), in which case the relationship between R 1 (t) and the concentration of the contrast agent becomes non-linear (59)(60)(61)(62). However, in the hepatocyte, the mean intracellular lifetime of a water molecule (t i = 40-50 ms (63,64)) and the water permeability coefficient (P = 66.4 Â 10 -4 cm/s (63)) are comparable to the corresponding values in an erythrocyte (t i =~10 ms, P = 50 Â 10 -4 cm/s), which is indeed in the FXL regime (57). Therefore the signal intensity can be converted to the concentration of extracellular contrast agent using the linear relationship (65) where C ES (t) represents the mean extracellular concentration of contrast agent across the ROI, v e the volume fraction of extracellular space (ES) accessible by the contrast agent and r 1 the relaxivity of the contrast agent. The relaxivities of gadoxetate and gadopentetate in liver and spleen at 4.7 T were assumed to be the same as in plasma and equal to 5.9 and 3.8 /s / mM, respectively (66). Multicompartmental models have been used to describe the kinetics of extracellular Gd-based MRI contrast agents (55,(59)(60)(61)(62)(67)(68)(69). Standard kinetic models usually consider three main compartments (69): plasma, extravascular extracellular space (EES) and intracellular compartments. However, in the case of a highly vascularised and perfused organ such as liver or spleen, where capillaries are highly permeable, the transfer exchange of contrast agent between plasma and EES occurs very fast (i.e. K trans ! 1 (68)), reducing it to two compartments: the extracellular space, comprising plasma and interstitial space (70,71), and the intracellular compartment, comprising hepatocytes and other liver cells, although we assume most of the intracellular space is occupied by hepatocytes (72). Following the reference region model approach (65), the extracellular concentration in the liver can be calculated by using another highly vascularised tissue, such as spleen, as the reference tissue.
For extracellular contrast agents the calculation of C ES from the MRI signal in the organ of interest is straightforward using Equation [3] if v e is known. v e can be estimated as the area under the curve (AUC) ratio between plasma and tissue concentrations: with C plasma (t) derived from the concentration of contrast agent in the whole blood, C blood (t), corrected for the haematocrit, Hct: C plasma (t) = C blood (t)/(1 À Hct). We have used this to measure the volume of extracellular space in liver and spleen with gadopentetate, which remains extracellular in these two tissues (71).
In later experiments examining gadoxetate uptake into hepatocytes we use the signal in the spleen to estimate C ES . Compared with measuring C ES from the blood signal in a major artery or vessel, the spleen offers the advantage of higher signal to noise ratio, larger region of interest and reduced need to include saturation bands in the MRI acquisition. Therefore we use liver v e;spleen C spleen t ð Þ [5] with C liver (t) and C spleen (t) the extracellular concentrations in liver and spleen, as derived from the MRI acquisition (i.e. C tissue (t) = v e C ES (t), where C ES (t) is the 'true' extracellular concentration in the tissue).
In the case of an intracellular tracer, the total concentration in the liver, C t (t), is the combination of the concentration in both the ES and intracellular (i.e. hepatocyte) compartments: If the system is in the FXL regime, the total concentration is determined directly from the signal intensity by using Equation [3], and the intracellular component is determined from Equation [6]. However, in this work we propose a more general approach and calculate the intracellular concentration from the equivalent signal intensity as follows.
The total MRI signal (normalised to baseline) in a voxel, S(t), is made up of the contributions from each compartment (55): with S e,i (t) and v e,i the signal intensity and measures of the water spaces in the extracellular (e) and intracellular (i) compartments, respectively (note that outside the FXL regime Equation [7] acquires a more complex form, with v e and v i (=1 À v e ) replaced by a S and a L defined in (60)). The equivalent concentration in the hepatocyte, C hep (t), is obtained by solving Equations [1]- [3] for the intracellular component in Equation [7]. Given that the equivalent extracellular concentration in the liver, C es (t), is known (Equation [5]), the extracellular signal intensity in the liver is determined by solving backwards Equations [1]- [3]. The fraction of the signal intensity in the intracellular compartment can easily be determined if we know the total signal intensity in the liver, its fraction in the extracellular space and the fractional water spaces on each compartment.

Non-linear compartmental model
Gadoxetate undergoes both renal (i.e. linear) and biliary (i.e. nonlinear, Michaelis-Menten) elimination (28,73). In rats, the pharmacokinetics of gadoxetate plasma concentration can be well described by a model with central and peripheral compartments (28)(29)(30)32,74). This model assumes that elimination from a central plasma compartment follows first order kinetics in the kidney and Michaelis-Menten kinetics in the hepatobiliary system (28,29,74). In humans, a two compartment model was proposed with first order kinetics to describe gadoxetate uptake from the liver ES compartment into a hepatobiliary compartment consisting of hepatocytes and biliary ductules (75) with no efflux.
Here, we propose a two compartment model that follows the Tofts and Kermode formulation (67) to characterise the kinetics of gadoxetate clearance in the liver in terms of the rate of unidirectional uptake from the liver ES compartment into hepatocytes and the Michaelis-Menten constants of its efflux into bile.
We can therefore isolate the intracellular compartment and characterise gadoxetate kinetics in the hepatocyte (Fig. 2) using the concentrations in each compartment derived in the previous section. The model defined in Figure 2 is expressed mathematically by a non-linear differential equation (73,74): with C hep (0) = 0, k 1 the uptake rate of gadoxetate from the extracellular space into the hepatocyte (i.e. Oatp1a1), and the Michaelis-Menten constants of biliary efflux V max (the maximum rate of efflux into bile) and K M (the concentration of gadoxetate when the rate of efflux is half the maximum rate, i.e. V max /2). These three parameters reflect the effect of the CKA on the kinetics of gadoxetate transport (76), and they are estimated for each dose group by solving the following minimisation problem: C hep (t) is determined by solving Equation [8] with the initial condition C hep (0) = 0, C hep t ð Þ represents the actual measurement from the MRI data and || Á || the second norm operator. Note the non-negativity constraint in all three parameters.
Parameter estimation was performed in MATLAB (R2011a, MathWorks, Natick, MA, USA) using the optimization and differential equation toolboxes.

Statistical analysis
The significance of the effect of the CKA on the kinetics of gadoxetate uptake (k 1 ) and clearance (V max and K M ) was investigated via ANOVA analysis (77) of the logarithm of each parameter, where variability was expressed in 95% confidence limits (CLs). Normality and homoscedasticity (i.e. variance homogeneity) of the data were confirmed via the Shapiro-Wilk normality test (78) and the Bartlett test of homogeneity of variances (79), respectively (95% CL). When comparing multiple sample groups in which marked changes were detected (95% CL), a multicomparison Tukey test (80) was performed to identify which of these differences were statistically significant. The directions of the changes were assessed with a one tailed t-test for the means between pairs of adjacent groups. All the statistical analyses were performed in R v2.13.0 (81), using the stats and agricolae (http://tarwi.lamolina.edu.pe/ fmendiburu/) packages.

RESULTS
Estimation of the volume fraction of extracellular space in liver Figure 3(a) shows the mean gadopentetate concentration data over time (five animals) in liver, spleen and blood following i.v. administration of the contrast agent. Using Equation [4] and a haematocrit value of 0.45, the estimated liver and spleen fractional extracellular spaces were equal to 0.23 AE 0.02 and 0.43 AE 0.02 (mean AE standard error of the mean (SEM)), respectively (p < 0.001 for the difference in extracellular spaces). Figure 3(b) shows that the liver gadopentetate time course can be represented by the spleen data scaled to an equivalent ES value. The blood data are included for reference. Evaluation of effects of CKA on hepatic clearance of gadoxetate in vivo . Compared with animals treated with vehicle, a marked and dose-dependent reduction in the rate of gadoxetate uptake into the liver and clearance from the liver into the GI tract was evident in animals treated with 200 and 500 mg/kg CKA. It is especially notable that enhancement of the small bowel lumen and also reflux of gadoxetate into the stomach was observed in the vehicle-treated animal. In contrast, no enhancement of the bowel was detected in the animal treated with 500 mg/kg CKA. Figure 5(b)-(d) shows the time course of the gadoxetate concentration in the liver, spleen and hepatocytes as determined by Equation [9] (time courses are the mean over the ROI within each group and error bars represent SEM).

Inhibition of transporter activity in vitro
As can be seen in Figure 5(c), after about 30 min the amplitude of the spleen signal was comparable to the noise level in the vehicle-and lower-dose-treated animals. In order to minimise the contribution of this noise to the parameter estimation, a biexponential function (67) was fitted to the spleen data and used as C spleen (t) in Equation [5]. The concentration of gadoxetate in the hepatocyte ( Figure 5(d)) was derived from its equivalent signal intensity using Equation [3]. The signal intensity in the hepatocyte was calculated from the signal intensity in liver and spleen (measured from the gadoxetate data) and the volume fraction of ES (calculated from the gadopentetate data) using Equation [7].  In Figure 6(a)-(c), representative plots of the fitting results obtained from rats treated with vehicle, or with CKA at 200 mg/ kg or 500 mg/kg, are shown. The calculated rate of uptake into hepatocytes, k 1 , and the Michaelis-Menten constants, V max and K M , for its biliary excretion in all dose groups are summarised in Figure 6(d)-(f) and are listed in Table 2. The data analysis showed no significant effect on the kinetic parameters caused by CKA administration at 20 mg/kg (p = 0.23, 0.46 and 0.44 for the differences compared with vehicle in k 1 , V max and K M , respectively), whereas all three parameters were altered in rats dosed with CKA at 200 mg/kg, 500 mg/kg or 2000 mg/kg. In particular, there was a marked dose-dependent reduction in k 1 and V max from 20 mg/kg to 200 mg/kg (p = 0.002 and 0.003 for k 1 and V max , respectively) and from 200 mg/kg to 500 mg/kg (p = 0.009 and 0.037 for k 1 and V max , respectively). No significant differences were observed between the changes in all three parameters in rats dosed with CKA at 500 mg/kg and 2000 mg/kg (p = 0.14, 0.56 and 0.33 for the differences in k 1 , V max and K M , respectively).

In vivo clinical chemistry and pathology findings
Evaluation of H&E-stained liver sections by light microscopy (Figure 7(a)-(c)) demonstrated that only one out of six animals dosed with 2000 mg/kg CKA exhibited significant centrilobular hepatocellular degeneration and necrosis, which was accompanied by neutrophil infiltration and associated sinusoidal congestion (Figure 7(b)). This was observed in the right median lobe section only and is illustrative of acute intoxication. Other liver lobe tissue sections from this animal resembled those from all other animals in this group (e.g. Figure 7(c)) and did not show any overt changes attributable to CKA treatment. Liver sections from rats given CKA at doses of 20, 200 or 500 mg/kg displayed no histological features indicative of drug-induced hepatotoxicity.
Increases in plasma markers indicative of functional hepatobiliary impairment, i.e. total bilirubin and bile acids, exhibited marked dose-dependent increases following administration of CKA (Figure 7(d),(e)). There was also a dose-related increase in plasma ALT levels in CKA-treated animals, indicating enzyme leakage from damaged liver cells, although data for individual animals were more variable than for other plasma chemistry parameters, and this was most evident in the groups treated with 500 and 2000 mg/kg CKA (Figure 7(f)).
A striking correlation was evident between CKA-induced elevations in plasma bile acids and total bilirubin and the effects of CKA administration on kinetic parameters of hepatic uptake and biliary excretion of gadoxetate (Fig. 8). In contrast, there was no correlation between the observed increases in plasma ALT levels and the hepatic uptake rate k 1 (r 2 = 0.44) or the biliary efflux K M (r 2 = 0.41).

DISCUSSION
Our investigations have shown that DCE-MRI with gadoxetate provides highly robust and reproducible data to evaluate hepatic transporter function in anesthetised rats in vivo. We have also presented a novel non-linear two compartment model that describes the kinetics of hepatic uptake and biliary excretion of gadoxetate in the hepatocyte.
Gadoxetate is formed by including the lipophilic ethoxybenzyl moiety (EOB) to the extracellular MRI contrast agent gadopentetate, which results in hepatocyte-specific uptake of gadoxetate (28,33). When constructing the model, we therefore assume that gadoxetate does not access the intracellular space of nonhepatocytes, and that its concentration in the extracellular space of the liver is equivalent to its concentration in the extracellular space of other organs (e.g. spleen) (33,34,75). The relative volume fraction of extracellular space between liver and spleen was estimated by using gadopentetate. We then used this information to estimate the intracellular, i.e. hepatocyte, concentration of gadoxetate and to determine the parameters of the non-linear model that enabled us to characterise its hepatic uptake and excretion kinetics.
The results we have obtained indicate that this model provides a simplified but effective description of hepatic transport of gadoxetate. The parameters provided by the model were its rate of hepatocyte uptake (k 1 ) and the Michaelis-Menten values for biliary efflux (K M and V max ). Using this approach, we have been able to explore the use of gadoxetate DCE-MRI to investigate functional effects on hepatobiliary transporter function in vivo. We have used a CKA test compound, which was found to inhibit the hepatobiliary transporters OATP1B1 (the human orthologue of rat Oatp1a1), Mrp2 and Bsep in vitro.
Although the DCE-MRI imaging data cannot in themselves provide definitive insight into the mechanism of transporter inhibition, it is clear that both a marked dose-dependent reduction of gadoxetate uptake into the liver and efflux from the hepatocyte are observed in the images in the presence of the CKA. The dose-dependent inhibition of hepatic uptake by the CKA, represented by k 1 , is indicative of inhibition of organic anion uptake transporter proteins, most likely Oatp1a1, since gadoxetate was previously identified as a substrate for this transporter (21). Although it is unknown which of the rodent Oatps is the counterpart of human OATP1B1, members of the OATP/Oatp family have similar substrate and inhibitor specificities, with some exceptions (82,83). It is therefore conceivable that the inhibitory  Table 2. Summary of the pharmacokinetic parameters that characterise the hepatic clearance of gadoxetate (mean AE SEM). Gadoxetate accumulates in the liver because the uptake rate, k 1 , is one to two orders of magnitude larger than the efflux rate, V max , producing liver enhancement in the MR images wileyonlinelibrary.com/journal/nbm effect of the CKA on hOATP1B1, which we observed in vitro, also applies to the rat ortholo.gue in vivo. The MRI signal enhancement observed in the liver is a consequence of accumulation of gadoxetate in hepatocytes due to reduced efflux into bile caused by inhibition of Mrp2 by the CKA. These in vivo data are supported by the in vitro Bsep and Mrp2 inhibition data obtained  and by a report that gadoxetate is a substrate of Mrp2 (34). The calculated increase in the K M of biliary efflux of gadoxetate is suggestive of competitive inhibition by the CKA (76). The observed effects of the CKA on the kinetics of hepatic clearance of gadoxetate are consistent with previous studies in rats, where administration of OATP inhibitors such as bromosulfophthalein, rifamycin, bilirubin, prednisolone, doxorubicin hydrochloride, cisplatin, propranolol hydrochloride and rifampicin resulted in a reduction of maximum hepatic MRI signal enhancement (28,29,(84)(85)(86)(87). This highlights the potential of gadoxetate MRI to be a useful tool for evaluation of the drug-drug interaction potential of pharmaceuticals. In particular, drugs that interact with members of the OATP family have been demonstrated to have clinically relevant interactions leading to altered pharmacokinetics and potentially toxicity (11,88). Although a recent study in healthy volunteers did not show a statistically significant difference in relative liver MRI signal enhancement in subjects receiving the OATP inhibitor erythromycin compared with control subjects (89), the concept has not yet been explored with other clinically relevant drugs in man. This approach has potential benefit in the early phases of clinical trials when a liver safety signal is seen that is consistent with hepatobiliary transport inhibition and subsequent cholestasis, but it is not clear whether the signal is transient and compensation will occur or whether the patient or volunteer will progress to liver damage. In this situation an assay of in vivo hepatobiliary transporter function could enable more effective decision making.
Given the short duration of the study, overt cellular damage is not expected at this time point, which is reflected in a relatively poor correlation observed between the effects of CKA administration on gadoxetate hepatic transport kinetics and plasma ALT levels. Elevated serum levels of ALT reflect leakage of the enzyme from damaged cells (primarily hepatocytes) and not hepatobiliary transporter function. In fact, only a single animal in the study (dosed with CKA at 2000 mg/kg) showed necrosis and centrilobular hepatocellular degeneration (i.e. cellular damage) together with a markedly elevated plasma ALT value (344 IU/l). In contrast, we observed a close correlation between the effects of CKA compound administration on gadoxetate hepatic transport kinetics and plasma markers of hepatobiliary clearance, i.e. levels of bile acids and bilirubin. This finding provides further confirmation that administration of CKA inhibits hepatobiliary transporter function in vivo and is consistent with the data demonstrating inhibition of Mrp2 and Bsep by the compound in vitro. It also demonstrates that gadoxetate DCE-MRI is a much more sensitive technique for exploration of functional hepatobiliary transport inhibition than liver histopathology, as is evaluation of plasma levels of bile acids and bilirubin. It must be noted that, although the acquisition of DCE-MRI data was started one hour before the blood samples were taken, they strongly correlated with a decrease in the hepatic uptake rate k 1 . This might suggest that gadoxetate DCE-MRI could be an earlier biomarker of impaired hepatic transporter activity than soluble plasma markers.
The plasma bile acid elevations caused by CKA administration can be attributed to the known inhibitory effect of the CKA on Bsep activity and have been observed with other drugs (90)(91)(92). These data also raise the intriguing possibility that inhibition of Bsep activity in vivo by the CKA administration might have contributed to the observed inhibition of biliary gadoxetate excretion, but possible transport of gadoxetate by Bsep has not been investigated. The elevated plasma bilirubin levels are likely to be a consequence of inhibition of hepatic uptake of bilirubin and its glucuronide form via OATPs (93,94) and of inhibition of biliary excretion of conjugated bilirubin by Mrp2 (95).
In recent years, evidence has been accumulated that suggests that impaired function of hepatobiliary transporters such as BSEP and MRP2, either through drug-induced inhibition or genetically determined reduction of activity, plays a role in drug-induced liver injury (19,96,97). Furthermore, inhibition of OATPs and MRP2 is associated with drug-induced hyperbilirubinaemia (98,99), and drug-drug interactions related to certain OATPs have clinical relevance (11). Consequently, candidate drugs are frequently assessed for transporter interactions during drug discovery using in vitro approaches. A major challenge when attempting to understand the functional significance of such in vitro data is that the relationship between inhibition of transporter activity by drugs in vitro and potential deleterious functional effects in vivo remains ill defined (100). This highlights the urgent need for new approaches that can be used to generate such understanding. Our results suggest that gadoxetate DCE-MRI may be a suitable technique for this purpose, especially where this is complemented by evaluation of bile acids and bilirubin as plasma biomarkers. In the future it will be important to generate data in animals treated with a broad range of different test compounds exhibiting a range of potency and selectivity of transport inhibition in order to develop a more complete understanding of the consequences that may ensue in vivo. In addition, gadoxetate MRI has been used extensively in the clinic (41)(42)(43)(44)(45)(46), which raises the possibility that the approach might also be useful during drug development, to aid in clinical hazard identification and risk assessment of novel candidate drugs.

CONCLUSION
This study has shown that gadoxetate DCE-MRI enables quantification of drug-induced alterations in hepatobiliary transporter activity in the rat in vivo. Administration to rats of a test compound (CKA) that inhibited the activity of the hepatic transporters OATP, Bsep and Mrp2 in vitro resulted in changes in the kinetics of gadoxetate clearance that could be explained in terms of effects on three parameters: the rate of gadoxetate uptake into hepatocytes, k 1 , and the Michaelis-Menten constants for efflux from hepatocytes into bile, V max and K M . These changes correlated closely with effects of CKA compound on serum biomarkers indicative of biliary transporter inhibition (total bile acids and bilirubin). We conclude that gadoxetate DCE-MRI can detect inhibition of hepatobiliary transporters and has the potential to serve as a novel biomarker of functional cholestasis. Furthermore, since gadoxetate is used in the clinic, it provides a novel approach that is potentially applicable both in laboratory animals and in patients in order to support safety evaluation of novel pharmaceuticals.