- Email: [email protected]
Assessment of Biliary Clearance in Early Drug Discovery using Sandwich-cultured Hepatocyte Model
Assessment of Biliary Clearance in Early Drug Discovery Using Sandwich-Cultured Hepatocyte Model GUOYU PAN,1 CARRI BOISELLE,2 JIANLING WANG2 1
Center for Drug Safety Evaluation and Research, Shanghai Institute of Materia Medica, Chinese Academy of Sciences, Shanghai, China 2
Metabolism and Pharmacokinetics, Novartis Institute for Biomedical Research, Cambridge, Massachusetts 02139
Received 6 September 2011; revised 19 December 2011; accepted 12 January 2012 Published online 9 February 2012 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.23070 ABSTRACT: It is challenging to predict biliary clearance (CLb ) for new chemical entities (NCEs) in early drug discovery. Although sandwich-cultured hepatocyte (SCH) model has offered a valuable tool for characterizing hepatobiliary disposition and drug–drug interaction potential of NCEs, no comprehensive study was reported to project in vivo biliary clearance (in vivo CLb,observed ) potential using in vitro SCH model during the drug discovery stage. In this study, the CLb of 110 discovery compounds was evaluated using rat SCH model. Parallel artificial membrane permeability assay, Caco-2, and rat liver microsomes were employed in parallel to explore the interplay of biliary excretion with cellular permeability and liver metabolism. Selected compounds were further tested in bile-duct-cannulated rats, confirming the value of the SCH model for ranking and predicting in vivo CLb,observed during drug discovery. For compounds with extremely low passive permeability and metabolism, rat SCH may underestimate in vivo CLb,observed . The combination of passive permeability, metabolic intrinsic clearance, and the SCH model could serve as an initial screening platform for biliary excretion potential as well as a means for improving compound liabilities and properties. A preliminary evaluation strategy was proposed to highlight biliary excretion risk evaluation during the drug discovery process. © 2012 Wiley Periodicals, Inc. and the American Pharmacists Association J Pharm Sci 101:1898–1908, 2012 Keywords: biliary excretion; hepatobiliary disposition; drug interactions; in vitro–in vivo correlations (IVIVC); hepatic clearance; permeability; transporters; hepatocytes
INTRODUCTION Biliary clearance (CLb ) is one of the critical drug elimination pathways for new chemical entities (NCEs) in drug discovery. This is especially true for compounds that undergo little metabolism in vivo.1 Understanding the extent of drug biliary excretion and the mechanistic insights such as determining the transporters involved is valuable for the selection and optimization of drug candidates in early drug discovery. Not only can it offer an early assessment in the context of potential toxicological and drug–drug interaction (DDI) risks, but it can also help bridge the gap between the data derived from in vitro and in vivo experiments. Correspondence to: Guoyu Pan (Telephone: 86-21-202310004023; Fax: 86-21-20231967; E-mail: [email protected]); Jianling Wang (Telephone: + 01-617-871-7140; Fax: + 01-617-8717061; E-mail: [email protected]) Journal of Pharmaceutical Sciences, Vol. 101, 1898–1908 (2012) © 2012 Wiley Periodicals, Inc. and the American Pharmacists Association
1898
Although in vivo (bile duct cannulated, BDC) and in situ animal models (e.g., isolated perfused livers) have been widely employed to investigate CLb and provide mechanistic information, they are typically labor intensive, costly, and not amendable for mechanistic studies in early drug discovery. In addition, the high complexity and variability of in vivo studies somewhat limit their applications in diagnosing in vitro–in vivo correlation (IVIVC) and when performing an in-depth mechanistic investigation. A number of in silico approaches were reported for predicting compound biliary excretion.2–3 In reality, it is challenging to predict a compound’s CLb by simply using its physicochemical properties. This is not surprising, as the CLb of a drug is greatly governed by not only hepatic metabolism but also by uptake/efflux transporters. Recently, sandwich-cultured hepatocytes (SCH) were utilized for predicting in vivo biliary clearance (in vivo CLb,observed ) by evaluating drug uptake and
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
APPLICATION OF SANDWICH-CULTURED HEPATOCYTES IN DRUG DISCOVERY
1899
Figure 1. ChemMAP score plot projection of 110 Novartis discovery new chemical entities (NCEs). Principal component analysis (PCA) was used to illustrate the position of the 110 NCEs in physical property space. The PCA score plot defined by principle components (PC) 1 and 2 shows the position of the bioactive ChemMAP reference molecules used to define the property space (black solid circles).9 The loading plot helps to interpret the position of NCEs (red solid circles) in property space (n = 110).
metabolism, as well as the potential for cytochrome P450 (CYP) induction.4 The assay represents a novel in vitro method for characterizing the hepatobiliary disposition and DDI potential of drug candidates and their metabolites.5 From a small collection of marketed drugs, the data from sandwich-cultured rat and human hepatocytes appeared to correlate reasonably well with in vivo CLb,observed data.6–7 Although promising as a compound-ranking tool for the in vivo biliary clearance risk applicable in early drug discovery,4 neither a systematic validation of SCH using real drug discovery NCEs from diversified chemical space and with suboptimal physicochemical drug properties, nor a quantitative model to properly account for in vitro biliary excretion was reported. Furthermore, guidance/strategy on how to bridge IVIVC gaps by considering the interplay with other absorption, distribution, metabolism, elimination, and toxicity (ADMET) properties is currently not available. Although CLb is influenced by both transporters and metabolic enzymes,8 little has been reported on the use of an in vitro model to account for the cooperative effects of transporters and metabolic enzymes on the hepatobiliary disposition of drug candidates. In this study, 12 commercially available compounds and 110 structurally diverse Novartis discovery compounds were evaluated using sandwich-cultured rat hepatocyte (SCRH) model. Clearly, most of the 110 NCEs were widely scattered in the new drug discovDOI 10.1002/jps
ery space (quadrants I and IV) rather than the traditional drug space (quadrants II and III9 ; Fig. 1). Passive permeability in parallel artificial membrane permeability assay (PAMPA), absorption permeability in Caco-2, and metabolic activity in rat liver microsomes were evaluated in parallel in order to explore the relationships between the in vitro biliary excretion, permeability, and metabolism of test compounds. Fifteen Novartis discovery compounds with in vitro (SCRH) and in vivo CLb,observed (BDC rat) results were compared to identify potential correlations and to set preliminary criteria to be used as a guideline for selecting discovery compounds to be screened in the future. A preliminary evaluation strategy involving SCH was proposed to highlight biliary excretion risk evaluation during the drug discovery process.
MATERIALS AND METHODS Chemicals Novartis test compounds were obtained from the Novartis compound archive (Cambridge, Massachusetts). Commercially available compounds were obtained from various vendors.
In Vitro PAMPA, Caco-2 Cell Culture, Hepatocytes, and Microsomal Clearance Assays The permeability assays were performed in two in vitro systems: PAMPA and Caco-2 cells, as JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
1900
PAN, BOISELLE, AND WANG
described elsewhere.10 The in vitro rat microsomal cryopreserved hepatocyte clearance was performed according to the protocol published by Obach et al.11
SANDWICH-CULTURED RAT HEPATOCYTE Calculation of biliary excretion index (BEI) and in vitro intrinsic biliary clearance (in vitro CLb,int ) in this report are similar to that performed by Liu et al.7
Sandwich-Cultured Rat Hepatocyte Sandwich-cultured 24-well plates with plated rat hepatocytes were purchased from Qualyst (Raleigh, North Carolina) after 3 days in culture. Upon receipt of the plates, the culture media was changed and the plates were incubated overnight at 37◦ C. The uptake study was carried out according to the method reported previously,7,12 with small modifications. Briefly, after an overnight incubation, SCRH was preincubated in 400 :L regular buffer (Hank’s balanced salt solution with calcium) or calcium-free buffer at 37◦ C for 10 min. The incubation in buffer without calcium opens the tight junctions of the bile canaliculi. After removing the buffer at the end of preincubation, 400 :L of 1 :M test compound prepared in regular buffer (dosing solution) was added to initiate the uptake process (10 min at 37◦ C). An aliquot of 100 :L sample was collected at both the beginning (time = 0 min) and the end (time = 10 min) of the incubation period. After incubation, the dosing solution was aspirated from the cells and uptake was terminated by washing the cells three times with 4◦ C regular buffer and aspirating the final wash. Plates were analyzed by liquid chromatography– tandem mass spectrometry (LC–MS/MS). Nonspecific binding (NSB) of each compound was evaluated by performing the assay using blank collagen-coated wells.
In Vivo CLb,observed Study After Intravenous Administration A BDC rat model (Charles River Laboratories International, Inc., Wilmington, Massachusetts) was employed to measure the bile and plasma concentrations of the test compounds. Compounds were administered intravenously in the animal studies, and bile and blood samples were collected in parallel at various time points. Compound concentrations were measured using LC–MS/MS. Data Analysis
In Vitro PAMPA, Caco-2 Cell Culture, Hepatocytes, and Microsomal Clearance Assays Calculation of the apparent permeability (Papp ) in PAMPA and Caco-2 cells, and efflux ratio (ER) in Caco-2 cells were performed according to the publication of Skolnik et al.10 The calculation of intrinsic clearance of rat liver microsomes (CLRLM,int ) and extraction ratio followed the Obach et al.11 method. JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
BEI =
Aplus
Ca++
− Aminus
Aplus
Ca++
Ca++
× 100%
(1)
Where A is the test compound accumulation amount in the wells of SCRH 24-well kit (pmol/well), Aplus Ca++ represents the test compound accumulation amount in regular buffer-treatment wells, and Aminus Ca++ represents the test compound accumulation amount in the calcium-free buffer-treatment wells. The value calculated in Eq. 1 represents the percentage of the initial drug concentration excreted into the bile. In vitro CLb,int =
Aplus
− Aminus AUC0−10 min
Ca++
Ca++
(2)
In Eq. 2, in vitro CLb,int of SCRH is calculated to represent the biliary excretion of the tested compound. Here, total concentration of the test compound during the entire incubation time was quantified using area under curve (AUC0–10min ) = time × (Cm,0min + Cm,10min )/2. Cm,0min and Cm,10min represent the substrate concentration at the beginning (0 min) and the end of the incubation period (10 min), respectively. In vitro CLb,int [:L/(min mg protein)] can be scaled to account for the total bile excretion in whole liver [CLb,int,s expressed as mL/(min kg)] for the in vivo rat model after being normalized by an empirical scaling factor (SF, Eq. 4). The SF (Eq. 3) of 8000 in the clearance calculation is based on allometric scaling; rat liver weight and protein content in liver tissue were assumed to be 40 g/kg of body weight and 0.20 g/ g of liver weight,13 respectively, in all calculations.7 The predicted biliary clearance (CLb,predicted ) was estimated according to the equation below (Eq. 5). The equation is based on the well-stirred model of hepatic disposition, where Q represents rat hepatic plasma flow [40 mL/(min kg body weight)]7 : SF = =
40 g liver 200 mg protein × 1 g liver 1 kg body mass 8000 mg protein kg body mass
(3)
CLb, int, s (:L/ min /kg body mass) g × SF = CLb,int × 1000 mg
CLb,predicted =
Q × CLb,int,s Q + CLb,int,s
(4)
(5) DOI 10.1002/jps
APPLICATION OF SANDWICH-CULTURED HEPATOCYTES IN DRUG DISCOVERY
1901
Figure 2. Sandwich-cultured rat hepatocyte biliary excretion index (BEI) compared with in vivo rat biliary clearance (in vivo CLb,observed ). Compound BEI values were divided into two clusters: zone a, low-biliary-clearance-risk compounds (BEI < 10%); zone b, high-biliaryclearance-risk compounds (BEI > 10%; n = 15).
Mass Balance Mass balance is calculated to evaluate the impact of drug metabolism on the cumulative uptake. Mass balance less than 70% was noted in comments. Amedium,10 min + Acell,10min × 100% Amedium,0 min
MB =
(6) Here, MB is the mass balance percentage. The amount (A) of the test compound has been assessed in the incubation medium (0 and 10 min) and the cells.
In Vivo Biliary Clearance in vivo biliary clearance was calculated using the following equation: CLb,observed = Abile /AUCplasma
(7)
where AUCplasma is the area under the curve of plasma concentration and time, and Abile is the amount of drug excreted into bile during that period.
RESULTS To evaluate the predictivity of the current SCRH model, the data derived from in vitro rat sandwiched experiments were compared with those collected from the in vivo BDC rat model (Fig. 2). With the limited dataset, it appeared that compounds with BEI values less than 10% in SCRH showed in vivo CLb,observed values less than 2 mL/(min kg) (zone a). In contrast, the DOI 10.1002/jps
compounds with high in vivo CLb,observed [e.g., >2 mL/ (min kg)], all clustered in the “in vitro BEI > 10%” zone (b). Overall, with the limited dataset, no clear linear correlation has been derived between in vitro BEI and in vivo CLb,observed , although in vitro BEI proves to be a useful binning tool when prioritizing compounds for the in vivo BDC studies. In this case, a BEI value of 10% obtained from in vitro SCRH studies is recommended to be the cutoff value for flagging compounds with significant CLb potential. For marketed drugs, SCRH–CLb,predicted values collected in our laboratory agree well with the in vivo CLb,observed data reported in literature (Table 1). Moreover, the marketed drugs and most of Novartis discovery NCEs appeared to scatter along the unity correlation line (dashed line in Fig. 3). A cluster of five Novartis NCEs deviated from the above trend line (circled in Fig. 3). Interestingly, those “outliers” exhibited low BEI (<10%) in the current SCRH studies. Taking the binning results between BEI from in vitro SCRH studies and CLb from BDC rat experiments [e.g., “BEI < 10%” corresponded well to “CLb < 2mL/ (min kg)” in Fig. 2] into consideration, the above trend line between predicted and observed rat CLb may still hold true for marketed drugs and discovery NCEs. The investigation of the interplay between physicochemical/ADME properties and the SCRH parameters led to the observation that moderate to high BEI and CLb,int in SCRH mostly occurred with lower Papp in PAMPA and Caco-2 (Figs. 4a and 4b and Table 2), and the probability of high biliary clearance risk appeared to decline significantly with increased permeability. Specifically, CLb,int was negligible when JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
1902
PAN, BOISELLE, AND WANG
Table 1.
Kinetic Parameters for Commercially Available Compounds in SCRH
Name Cephradine # D8-TCA
Digoxin # DPDPE # ICG
Indomethacin Ouabain Pravastatin Rosuvastatin Methotrexate Vincristine Doxorubicin # # #
BEI (%)
In Vitro CLb ,int [:L/(min mg protein)]
Predicted In Vivo CLb [mL/(min kg)]
Observed In Vivo CLb [mL/(min kg)]
0 56.6 39.7 88 63.1 0 43.6 30.5 26.4 35.6 33.8 9.3
≤0.01 2.05 0.33 1.21 3.15 ≤0.01 0.39 0.26 3.80 0.10 0.13 0.31
≤0.5 28.9 11.7 24.3 32 ≤0.5 13.2 10 31 4.5 5.5 11.4
3.114 29.815 10.716 18.717 3218 0.0119 14.720 14.221 24.312 12.122 14.223 16.624
D8-TCA: d8-taurocholate acid. DPDPE: [D-Pen2,5]Enkephalin hydrate. ICG: Indocyanine green.
PAMPA Papp at neutral pH was greater than 40 × 10−6 cm/s and/or Caco-2 Papp was greater than 10 × 10−6 cm/s. For instance, the high biliary potential [e.g., CLb,int > 0.3 :L/(min kg)] decreases from 31.7% to 4.3% and from 32.1% to 3.4% when under the above permeability boundary values for PAMPA and Caco2, respectively. Interestingly, the in vitro observations were also confirmed by in vivo BDC rat data (insets of
Figs. 4a and 4b). Similar trends were also observed between BEI and the Papp from either PAMPA or Caco-2 experiments. In addition to passive permeability as observed from the PAMPA experiments (Fig. 4), the impact of active transport on CLb was assessed in the context of the ER derived from Caco-2 experiments. Although it is difficult to develop a clean correlation between
Figure 3. Sandwich-cultured rat hepatocyte predicted biliary clearance (CLb,predicted ) was compared with in vivo rat biliary clearance (CLb,oberved ). Red hollow squares represent tested market compounds (n = 12). Blue solid diamonds represent Novartis compounds (n = 15). For the compounds with biliary excretion index (BEI) < 10% (zone a), in vivo CLb,observed is less than 2 mL/(min kg) and is regarded as negligible. For the compounds with BEI > 10% (zone b), CLb is estimated. Compounds in the red circle also had CLb estimated, but their BEI was less than 10%. JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
DOI 10.1002/jps
APPLICATION OF SANDWICH-CULTURED HEPATOCYTES IN DRUG DISCOVERY
1903
Figure 4. The impact of hepatocyte permeability on biliary excretion. (a–d) Sandwich-cultured rat hepatocyte biliary excretion parameters [biliary excretion index (BEI) and intrinsic biliary clearance (CLb,int )] were compared with parallel artificial membrane permeability assay (PAMPA) and Caco-2 permeability. Selected compounds with in vivo CLb,observed are presented in the inset graphs in the panels a and b (n = 108).
BEI or in vitro CLb,int and the Caco-2 efflux ratio (ER) for discovery NCEs, a trend was observed (Fig. 5). For instance, for NCEs with ER < 1, only a small fraction of them showed high CLb based on data from the in vitro SCRH (e.g., 15% for BEI > 10% or 6% for CLb > 0.3, respectively). However, the above fraction for high CLb drastically increased for NCEs with ER > 5 (e.g., 50% for BEI > 10% or 41% for CLb > 0.3, respectively). Thus, for those compounds having an elevated ER in Caco-2 cells, a concern regarding active secretion into the bile may be warranted. The interplay between CYP-based metabolic clearance and CLb was explored using the SCRH model (Fig. 6). Apparently, data were scattered in the lower left zone below the dotted diagonal line, which was included to help define the threshold. Specifically, no data were in the upper right half of the graph, suggesting that compounds highly metabolized in liver microsomes (extraction ratio > 0.75) are unlikely to
DOI 10.1002/jps
be actively excreted into the bile in SCRH [CLb,int < 1 :L/(min mg protein]. In contrast, most of the NCEs with high CLb,int were clustered on the low extraction zone (<0.4). In order to accurately interpret the SCRH data in the context of interplay between CYP metabolism and biliary excretion, it is critical to confirm if the SCRH system still maintains part of CYP activities. After incubation in SCRH, the remaining concentration of the CYP3A substrate (testosterone) and CYP1A substrate (phenacetin) decreased appreciably (Fig. 6). The calculated hepatic extraction clearance in SCRH was comparable to cryopreserved rat hepatocytes [52.7 vs. 52 mL/(min kg) for testosterone and 54.2 vs. 31.6 mL/(min·kg) for phenacetin]. The calculated hepatic extraction ratio in SCRH was close to 1. The results suggest that the SCRH system maintains the enzymatic activity for selected CYPs to some extent.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
1904
1.0 47.0 0.1 8.1 0.1 0.2 1.3 93.4 10.2 0.4 0.3 0.5 17.9 132.0 1.6 0.5 11.2 2.8 21.9 0.5 6.4 0.5 14.3 30.8 9.7 24.0 0.5 29.0 26.0 18.4
c For
b For
a 10%
was used to represent low metabolic compounds extraction ratio in liver microsomes (RLM liver extraction ratio < 10%). compounds without biliary excretion (BEI = 0%), 0.1 was used to replace zero for the convenience of plotting (log scale). compounds without biliary excretion [CLb ,predicted = 0 mL/(min kg)], 0.5 was used to replace zero for the convenience of plotting (log scale). d NVP-8 had no liver microsomal clearance data. Hepatocytes clearance data was used as surrogates (low metabolic ability in hepatocytes). RLM, rat liver microsomes; NA, not available.
0.1 18.3 0.5 30.0 0.1 1.4 0.1 21.0 22.5 1.4 4.3 0.1 22.1 24.0 5.5 81.0 <3.4 133.0 321.0 18.1 35.0 17.0 9.5a 11.0 81.0 38.0 <3.4 14.0 15.0 125.0 72.5 10.0 81.0 91.0 10 53.0 35.5 10d 26.0 73.0 55.5 10 28.0 33.0 80.0 31.6 2.5 31.6 4.0 12.6 10.0 0.8 0.8 20.0 63.1 79.4 79.4 0.0 4.0 7.9 0.9 16.0 0.3 15.2 0.9 3.1 1.4 1.9 NA 1.0 NA NA 1.6 26.7 0.1 2.7 0.5 2.0 0.9 0.3 2.5 0.2 0.7 NA 16.6 NA NA 1.0 0.1 2.7 4.8 2.2 4.7 2.5 6.2 2.6 4.8 4.2 8.4 3.8 8.3 3.9 5.5 3.3 7.0 385.5 505.6 502.6 489.9 546.5 311.4 1173.4 433.5 928.6 450.5 387.6 468.5 609.8 515.4 697.6
RLM Hepatic Extraction %a PAMPA Papp (×10–6 cm/s) Caco-2 Papp (B-A)/ Papp (A-B) Caco-2 Papp (A-B) (×10–6 cm/s) Clog p Molecular Weight (MW,g/mol)
DISCUSSION
NVP-1 NVP-2 NVP-3 NVP-4 NVP-5 NVP-6 NVP-7 NVP-8 NVP-9 NVP-10 NVP-11 NVP-12 NVP-13 NVP-14 NVP-15
Table 2.
The In Vivo and In Vitro Biliary Clearance Data Observed from Novartis Discovery NCEs (n = 15)
RLM CLint [mL/(min mg protein)]
BEI (%)b
Predicted CLb [mL/(min kg)]c
CLb,observed [mL/(min kg)]
PAN, BOISELLE, AND WANG
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
The polarized transporter expression and bile canaliculi network, such as bile pocket structures make SCRH a suitable in vitro model for investigating the biliary excretion of NCEs. The integrity of the tight junctions responsible for forming the canalicular network (bile pocket structures) is calcium dependent and can be opened by removing calcium from the incubation medium.25 Clearance measurements in the presence and absence of calcium allow for the determination of BEI and CLb,int , making it possible to assess the relative contributions of transporters involved in biliary drug excretion.7 Although the two parameters were employed in SCRH calculation, no report suggests utilizing these in vitro parameters to flag high biliary clearance risk in drug discovery. Biliary excretion index represented the percentage of retained substrate in the bile canaliculi and was believed to reflect the contribution, via active transport mechanisms, to biliary excretion.25 Our present work (Fig. 2) clearly demonstrated that the CLb data from BDC rat could be properly differentiated by the BEI data from the in vitro SCRH model (cutoff value = 10%), suggesting that SCRH may serve as a viable in vitro binning model for high biliary clearance risk for NCEs. Here, SCRH is valuable not only to help establish IVIVC of CLb , but also to prioritize costly in vivo BDC studies using the in vitro model. This is particularly useful for NCEs in which biliary elimination is the dominant elimination pathway. In this case, unnecessary in vivo studies can thus be avoided. It is worthwhile to note that BEI itself could not predict in vivo CLb,observed . This could be a result of differences between the in vitro and in vivo metabolic clearances (Fig. 6), or due to the contributions of other competitive elimination processes such as NCEs sinusoidal uptake activities.7 This could partially explain why some compounds (especially high-CLb compounds) with similar in vitro BEI values have fluctuation of over 10-fold for in vivo CLb,observed (Table 2, Fig. 2). Intrinsic biliary clearance, employed to describe the ability of compounds to be taken up into hepatocytes and excreted into bile, can be evaluated in the context of overall clearance involving multiple mechanisms. Indeed, the CLb,predicted of marketed drugs and most of NCEs from the current SCRH model had a reasonable correlation with the in vivo BDC data reported in literature sources (Table 1 and dashed line in Fig. 3). It should be noted that five Novartis NCEs deviating from the above trend line exhibited low BEI (10%). In other words, the absolute values of CLb,predicted may need to be evaluated with caution when BEI < 10%. In this case, the simplest approach may be to “qualify” CLb,predicted as “less than 2 mL/(min kg),” as suggested by Figure 2. From a mechanistic perspective, all of DOI 10.1002/jps
APPLICATION OF SANDWICH-CULTURED HEPATOCYTES IN DRUG DISCOVERY
1905
Figure 5. Sandwich-cultured rat hepatocyte biliary excretion [biliary excretion index (BEI) and intrinsic biliary clearance (CLb,int )] was compared with Caco-2 efflux ratio (ER) (n = 90).
the five NCE “outliers” with overpredicted CLb exhibited moderate to high CYP-related metabolic ability (extraction ratio = 0.53–0.89) and moderate to high permeability in PAMPA and Caco-2. Employing additional mechanistic studies may be valuable to pursue the impact of key physicochemical and ADME properties on the prediction of CLb . It is noteworthy that some marketed drugs (cephradine and indomethacin, both low-CLb drugs) also showed moderate variability against the above trend line, possibly because of variability among animal samples and/or analytical challenge when reaching the limit of detection. On the basis of our limited dataset, the IVIVC for CLb may not be very straightforward and depends greatly on interplay with other physicochemical and ADME properties.
It is also worthwhile to note that for a limited number of reported drugs, especially statins, improved linear correlation between SCRH CLb and in vivo CLb,observed was observed after protein binding correction.1226 However, for compounds with diverse structures and physicochemical properties, the IVIVC was not improved after protein binding correction, neither for market drugs nor Novartis compounds (based on the calculation from our dataset; data not shown). The detail mechanisms are still under investigation. It was speculated that the kinetic effects of plasma protein binding and the interplay between various competing and collaborating processes may contribute to this disconnection. For instance, it was implied that plasma protein binding is not
Figure 6. Sandwich-cultured rat hepatocyte intrinsic biliary clearance (CLb,int ) compared with the rat liver microsomes extraction ratio. Compounds with high intrinsic metabolic clearances were less likely to be actively excreted into bile. The extraction ratio of testosterone and phenacetin were comparable to the values in cryopreserved hepatocytes suspension (data not shown; n = 92). DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
1906
PAN, BOISELLE, AND WANG
the rate-limiting step for hepatic clearances in all circumstances.27 The equilibrium between the free drug in the aqueous phase and the drug in the membrane attenuates the impact on the free drug hypothesis.28 How compounds access binding sites (in vitro vs. in vivo systems) of the proteins also influence drug clearance.28 Overall, based on the existing evidence, it is not convincing that protein binding correction improves CLb prediction in SCRH studies. Therefore, no protein binding correction was employed to calibrate Eq. 5 in the present work. Although CLb is believed to account for the contributions of both efflux (into bile) and influx (uptake into hepatocyte) transporters in the biliary clearance process,25 BEI, by principle, only reflects that of efflux transporters to bile. Similar inverse trends between in vitro permeability data and CLb,int (Figs. 4a and 4b) or BEI (Figs. 4c and 4d), however, support the postulation that the biliary clearance process may primarily be governed by the transporter-mediated efflux process. In this case, compound uptake as a result of passive permeation into hepatocytes will be less critical to the overall level of CLb .29 Meanwhile, considering the volume and flow rate of bile and the limited canalicular membrane surface area, the direct contribution of passive permeation to the overall bile clearance may be negligible. Indeed, a similar observation was also reported between renal clearance and passive permeability.30 This prompts the need to utilize proper in vitro transporter-screening tools to flag the biliary clearance risk in early discovery. Efflux transporters located in hepatocyte canalicular membranes, such as P-glycoprotein (or multidrug resistance protein 1), MRP2, and breast cancer resistance protein, also exist in the human intestine and Caco-2 cells. BEI is an indicator of efflux transporter contribution in biliary excretion,25 and a clear trend between high ER in Caco-2 and high BEI and CLb in SCRH (Figs. 4 and 5) demonstrated that Caco-2 is among the viable in vitro tools for not only flagging CLb , but also helping explore the transport mechanisms. The above semiquantitative correlation between BEI or CLb and Caco-2 ER, albeit nonlinear, is rational, as it reflects the different levels of transporter expression between liver and gut.31 It should be noted, however, that Caco-2 ER itself is rather a semiquantitative flag for efflux substrate(s) than a quantitative parameter for its expression and activity.32 Compounds that are actively metabolized by CYP enzymes may not show high biliary excretion as a result of the location of CYP enzymes in hepatocytes.3 This is most likely because of the competing nature of efflux transporters and metabolic enzymes in hepatic drug elimination.29,33–34 However, most of the previous reports tested only a limited number of marketed drugs. The present work has successfully expanded the study by focusing on the interplay of CLb and JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
CYP-related metabolic activity using a collection of discovery compounds from diverse chemical scaffolds (Fig. 1). Although it has been reported that long-term cultured hepatocytes lose both phase I and phase II enzyme activity,35 and phase I metabolic activities deteriorate significantly in SCH,7 the data presented here suggests that some CYP activity remains in SCH (Fig. 6). For example, SCRH clearance values for the CYP3A substrate testosterone and the CYP1A substrate phenacetin are comparable to the clearance values in cryopreserved rat hepatocyte suspension. This is consistent with the previous reports that P450 activities (e.g., CYP3A) may be kept or even “recovered” after 3–4 days of culture in SCRH.36–37 Most of the compounds with high CLb,int in our dataset had medium to low clearance in rat liver microsomes. The results indicated that SCRH could be used to explore the interplay between transporters and metabolic enzymes in compound optimization and selection. Despite the competitive interplay between transportermediated CLb and CYP-based metabolic elimination, the “inverse relationship” between the two was neither perfectly “linear” nor “clean”, as depicted in Figure 6. Instead, a collection of low-CLb NCEs observed in SCRH showed varying CYP metabolic extraction ratios. In other words, the low CYP clearance does not always translate to high biliary clearance risk here. It is important to note that the microsomal data represents only phase I metabolic activities, whereas a clearance study in hepatocytes could better characterize the interplay between CLb and multiple metabolic mechanisms. The SCRH model could be more valuable in this process if CYP enzymes were induced.4 A preliminary biliary clearance risk evaluation strategy for drug discovery was proposed based on the SCRH results and related permeability/metabolism information obtained in this study (Fig. 7). For lead optimization and candidate selection, it is important to understand if CLb is the primary clearance pathway and/or limits the exposure of NCEs, and what may be potential drug interaction risk. In that sense, compounds with clearance IVIVC gaps would be required to undergo permeability screening (PAMPA/ Caco-2) before assessing the biliary excretion risk. A BEI value greater than 10% would be the first cutoff value for the SCRH model. Only when CLb,int is greater than 0.3 :L/(min mg protein) [CLb,predicted > 10 mL/(min kg)] would the BDC rat study be recommended to confirm the in vitro results. It should be noted that the in vitro model is not designed to replace the in vivo mechanistic studies that can offer comprehensive mechanistic insights. Rather, it could identify compounds with high CLb . However, in order to optimize NCEs ADMET properties, the contribution of each clearance route (biliary excretion, urinary excretion, metabolism, etc.) to the overall total clearance have to be determined together. In addition, the DOI 10.1002/jps
APPLICATION OF SANDWICH-CULTURED HEPATOCYTES IN DRUG DISCOVERY
1907
Figure 7. Proposed biliary clearance risk evaluation strategy for discovery stage compounds.
application of this strategy would greatly accelerate the biliary clearance risk evaluation process and reduce the burden of animal studies in discovery. This is especially true for compounds with similar scaffolds. Noticeably, the above strategy is to identify the clearance gap and flag high biliary clearance risk in rat in support for chemistry optimization in early drug discovery. At present, however, this strategy is not feasible to predict human biliary clearance risk or establish IVIVC,38 owing to limited clinic CLb data. The situation may be improved once additional human biliary excretion data become available.
CONCLUSION The SCRH model proved to be a valuable tool for ranking and predicting in vivo CLb,observed during drug discovery. Biliary excretion involves three processes: uptake across the sinusoidal membrane, metabolism, and excretion across the canalicular membrane.7 The in vivo and in vitro “intrinsic” clearance values calculated should be rate limited by the slowest step in these processes. Because CLb,int was determined on the basis of intracellular substrate amount, it should be considered as “apparent” CLb,int values. Therefore, it is difficult to evaluate uptake/efflux transporters alone without considering the contribution of passive diffusion and CYP metabolism in SCH, especially after the observation of P450 activities in cultured hepatocytes (Fig. 6). DOI 10.1002/jps
For compounds with extremely low passive permeability and low metabolism, the probability of CLb as the dominate elimination pathway may drastically increase. The combination of passive permeability, metabolic intrinsic clearance, and the SCRH model could be used as an initial screening platform for bile excretion potential as well as a means for improving compound liabilities and properties. Future studies using an extended list of discovery NCEs with varying biliary excretion ability, chemical diversity, and a suitable kinetic model (including passive permeability, transporters, and metabolism) will offer additional understanding for improved prediction of CLb .
ACKNOWLEDGMENTS The authors appreciate the critical review and constructive feedback from Dr. Leslie Bell. The authors also thank Suzanne Skolnik for offering diversity analysis of Novartis discovery NCEs, as well as Drs. Kenneth Brouwer and Claudia McGinnis for offering assistance in the initial setup of the Sandwich Cultured Rat Hepatocyte model in the Novartis lab.
REFERENCES 1. Bi YA, Kazolias D, Duignan DB. 2006. Use of cryopreserved human hepatocytes in sandwich culture to measure hepatobiliary transport. Drug Metab Dispos 34(9):1658–1665. 2. Luo G, Johnson S, Hsueh MM, Zheng J, Cai H, Xin B, Chong S, He K, Harper TW. 2010. In silico prediction of biliary excretion of drugs in rats based on physicochemical properties. Drug Metab Dispos 38(3):422–430. JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
1908
PAN, BOISELLE, AND WANG
3. Yang X, Gandhi YA, Duignan DB, Morris ME. 2009. Prediction of biliary excretion in rats and humans using molecular weight and quantitative structure–pharmacokinetic relationships. AAPS J 11(3):511–525. 4. Swift B, Pfeifer ND, Brouwer KL. 2010. Sandwich-cultured hepatocytes: An in vitro model to evaluate hepatobiliary transporter-based drug interactions and hepatotoxicity. Drug Metab Rev 42(3):446–471. 5. Hoffmaster KA, Turncliff RZ, LeCluyse EL, Kim RB, Meier PJ, Brouwer KL. 2004. P-glycoprotein expression, localization, and function in sandwich-cultured primary rat and human hepatocytes: Relevance to the hepatobiliary disposition of a model opioid peptide. Pharm Res 21(7):1294–1302. 6. Giacomini KM, Huang SM, Tweedie DJ, Benet LZ, Brouwer KL, Chu X, Dahlin A, Evers R, Fischer V, Hillgren KM, Hoffmaster KA, Ishikawa T, Keppler D, Kim RB, Lee CA, Niemi M, Polli JW, Sugiyama Y, Swaan PW, Ware JA, Wright SH, Yee SW, Zamek-Gliszczynski MJ, Zhang L. 2010. Membrane transporters in drug development. Nat Rev Drug Discov 9(3):215–236. 7. Liu X, Chism JP, LeCluyse EL, Brouwer KR, Brouwer KL. 1999. Correlation of biliary excretion in sandwich-cultured rat hepatocytes and in vivo in rats. Drug Metab Dispos 27(6):637–644. 8. Pang KS, Maeng HJ, Fan J. 2009. Interplay of transporters and enzymes in drug and metabolite processing. Mol Pharm 6(6):1734–1755. 9. Faller B, Ottaviani G, Ertl P, Berellini G, Collis A. 2011. Evolution of the physicochemical properties of marketed drugs: Can history foretell the future? Drug Discov Today 16(21–22):976–984. 10. Skolnik S, Lin X, Wang J, Chen XH, He T, Zhang B. 2010. Towards prediction of in vivo intestinal absorption using a 96well Caco-2 assay. J Pharm Sci 99(7):3246–3265. 11. Obach RS, Baxter JG, Liston TE, Silber BM, Jones BC, MacIntyre F, Rance DJ, Wastall P. 1997. The prediction of human pharmacokinetic parameters from preclinical and in vitro metabolism data. J Pharmacol Exp Ther 283(1):46–58. 12. Fukuda H, Ohashi R, Tsuda-Tsukimoto M, Tamai I. 2008. Effect of plasma protein binding on in vitro – in vivo correlation of biliary excretion of drugs evaluated by sandwich-cultured rat hepatocytes. Drug Metab Dispos 36(7):1275–1282. 13. Seglen PO. 1976. Preparation of isolated rat liver cells. Methods Cell Biol 13:29–83. 14. Nakashima, E., R. Matsushita, et al. Comparative pharmacokinetics of cefoperazone and cephradine in untreated streptozotocin diabetic rats. 1992, Drug Metab Dispos 20(5):730–735. 15. Inoue, M., Y. Morino, et al., Transhepatic transport of taurocholic acid in normal and mutant analbuminemic rats. 1985. Biochim Biophys Acta 833(2):211–216. 16. Watkins, J. B., 3rd and S. E. Sherman,Long-term diabetes alters the hepatobiliary clearance of acetaminophen, bilirubin and digoxin. 1992. J Pharmacol Exp Ther260(3), 1337–1343. 17. Qualyst Technical Application Bulletin. 2008, TAB Biol. 003v1, 18. Pollack, G. M. and K. L. Brouwer. Thermal injury decreases hepatic blood flow and the intrinsic clearance of indocyanine green in the rat, Pharm Res 1991, 8(1), 106–111. 19. Kouzuki, H., H. Suzuki, et al. Pharmacokinetic study of the hepatobiliary transport of indomethacin., Pharm Res 2000.174)432–438. 20. Russell, J. Q. and C. D. Klaassen Species variation in the biliary excretion of ouabain., J Pharmacol Exp Ther 1972, 183(3)513–519. 21. Yamazaki, M., S. Akiyama, et al., Biliary excretion of pravastatin in rats: contribution of the excretion pathway mediated
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33. 34.
35.
36.
37.
38.
by canalicular multispecific organic anion transporter., Drug Metab Dispos 1997, et al. 25(10), 1123–1129. Masuda, M., Y. I’Izuka, et al., Methotrexate is excreted into the bile by canalicular multispecific organic anion transporter in rats., Cancer Res1997 57(16), 3506–3510. Mellows, G. et al., The effect of XR9576 (tariquidar), a specific P-glycoprotein inhibitor, on the pharmacokinetics of doxorubicin, paclitaxel and vincristine in the rat., Proc Am Assoc Cancer Res (AACR) 2002, 43, 1363. Achira, M., R. Totsuka, et al, Tissue-specific regulation of expression and activity of P-glycoprotein in adjuvant arthritis rats., Eur J Pharm Sci 2002, 16(1–2)29–36. Liu X, LeCluyse EL, Brouwer KR, Gan LS, Lemasters JJ, Stieger B, Meier PJ, Brouwer KL. 1999. Biliary excretion in primary rat hepatocytes cultured in a collagen-sandwich configuration. Am J Physiol 277(1 Pt 1):G12–G21. Abe K, Bridges AS, Yue W, Brouwer KL. 2008. in vitro biliary clearance of angiotensin II receptor blockers and 3-hydroxy-3-methylglutaryl-coenzyme A reductase inhibitors in sandwich-cultured rat hepatocytes: Comparison with in vivo biliary clearance. J Pharmacol Exp Ther 326(3):983– 990. Baker M, Parton T. 2007. Kinetic determinants of hepatic clearance: Plasma protein binding and hepatic uptake. Xenobiotica 37(10–11):1110–1134. Smith DA, Di L, Kerns EH. 2010. The effect of plasma protein binding on in vivo efficacy: Misconceptions in drug discovery. Nat Rev Drug Discov 9(12):929–939. Watanabe T, Kusuhara H, Sugiyama Y. 2010. Application of physiologically based pharmacokinetic modeling and clearance concept to drugs showing transporter-mediated distribution and clearance in humans. J Pharmacokinet Pharmacodyn 37(6):575–590. Fagerholm U. 2008. Prediction of human pharmacokinetics— Biliary and intestinal clearance and enterohepatic circulation. J Pharm Pharmacol 60(5):535–542. Hilgendorf C, Ahlin G, Seithel A, Artursson P, Ungell AL, Karlsson J. 2007. Expression of thirty-six drug transporter genes in human intestine, liver, kidney, and organotypic cell lines. Drug Metab Dispos 35(8):1333–1340. Lin X, Skolnik S, Chen XH, Wang J. 2010. Attenuation of intestinal absorption by major efflux transporters: Quantitative tools and strategies using Caco-2 model. Drug Metab Dispos 39(2):265–274. Liu L, Pang KS. 2005. The roles of transporters and enzymes in hepatic drug processing. Drug Metab Dispos 33(1):1–9. Wu CY, Benet LZ. 2005. Predicting drug disposition via application of BCS: Transport/absorption/elimination interplay and development of a biopharmaceutics drug disposition classification system. Pharm Res 22(1):11–23. Niemann C, Gauthier JC, Richert L, Ivanov MA, Melcion C, Cordier A. 1991. Rat adult hepatocytes in primary pure and mixed monolayer culture. Comparison of the maintenance of mixed function oxidase and conjugation pathways of drug metabolism. Biochem Pharmacol 42(2):373–379. LeCluyse EL. 2001. Human hepatocyte culture systems for the in vitro evaluation of cytochrome P450 expression and regulation. Eur J Pharm Sci 13(4):343–368. Tuschl G, Hrach J, Walter Y, Hewitt PG, Mueller SO. 2009. Serum-free collagen sandwich cultures of adult rat hepatocytes maintain liver-like properties long term: A valuable model for in vitro toxicity and drug–drug interaction studies. Chem Biol Interact 181(1):124–137. Ghibellini G, Vasist LS, Leslie EM, Heizer WD, Kowalsky RJ, Calvo BF, Brouwer KL. 2007. in vitro – in vivo correlation of hepatobiliary drug clearance in humans. Clin Pharmacol Ther 81(3):406–413.
DOI 10.1002/jps
Center for Drug Safety Evaluation and Research, Shanghai Institute of Materia Medica, Chinese Academy of Sciences, Shanghai, China 2
Metabolism and Pharmacokinetics, Novartis Institute for Biomedical Research, Cambridge, Massachusetts 02139
Received 6 September 2011; revised 19 December 2011; accepted 12 January 2012 Published online 9 February 2012 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.23070 ABSTRACT: It is challenging to predict biliary clearance (CLb ) for new chemical entities (NCEs) in early drug discovery. Although sandwich-cultured hepatocyte (SCH) model has offered a valuable tool for characterizing hepatobiliary disposition and drug–drug interaction potential of NCEs, no comprehensive study was reported to project in vivo biliary clearance (in vivo CLb,observed ) potential using in vitro SCH model during the drug discovery stage. In this study, the CLb of 110 discovery compounds was evaluated using rat SCH model. Parallel artificial membrane permeability assay, Caco-2, and rat liver microsomes were employed in parallel to explore the interplay of biliary excretion with cellular permeability and liver metabolism. Selected compounds were further tested in bile-duct-cannulated rats, confirming the value of the SCH model for ranking and predicting in vivo CLb,observed during drug discovery. For compounds with extremely low passive permeability and metabolism, rat SCH may underestimate in vivo CLb,observed . The combination of passive permeability, metabolic intrinsic clearance, and the SCH model could serve as an initial screening platform for biliary excretion potential as well as a means for improving compound liabilities and properties. A preliminary evaluation strategy was proposed to highlight biliary excretion risk evaluation during the drug discovery process. © 2012 Wiley Periodicals, Inc. and the American Pharmacists Association J Pharm Sci 101:1898–1908, 2012 Keywords: biliary excretion; hepatobiliary disposition; drug interactions; in vitro–in vivo correlations (IVIVC); hepatic clearance; permeability; transporters; hepatocytes
INTRODUCTION Biliary clearance (CLb ) is one of the critical drug elimination pathways for new chemical entities (NCEs) in drug discovery. This is especially true for compounds that undergo little metabolism in vivo.1 Understanding the extent of drug biliary excretion and the mechanistic insights such as determining the transporters involved is valuable for the selection and optimization of drug candidates in early drug discovery. Not only can it offer an early assessment in the context of potential toxicological and drug–drug interaction (DDI) risks, but it can also help bridge the gap between the data derived from in vitro and in vivo experiments. Correspondence to: Guoyu Pan (Telephone: 86-21-202310004023; Fax: 86-21-20231967; E-mail: [email protected]); Jianling Wang (Telephone: + 01-617-871-7140; Fax: + 01-617-8717061; E-mail: [email protected]) Journal of Pharmaceutical Sciences, Vol. 101, 1898–1908 (2012) © 2012 Wiley Periodicals, Inc. and the American Pharmacists Association
1898
Although in vivo (bile duct cannulated, BDC) and in situ animal models (e.g., isolated perfused livers) have been widely employed to investigate CLb and provide mechanistic information, they are typically labor intensive, costly, and not amendable for mechanistic studies in early drug discovery. In addition, the high complexity and variability of in vivo studies somewhat limit their applications in diagnosing in vitro–in vivo correlation (IVIVC) and when performing an in-depth mechanistic investigation. A number of in silico approaches were reported for predicting compound biliary excretion.2–3 In reality, it is challenging to predict a compound’s CLb by simply using its physicochemical properties. This is not surprising, as the CLb of a drug is greatly governed by not only hepatic metabolism but also by uptake/efflux transporters. Recently, sandwich-cultured hepatocytes (SCH) were utilized for predicting in vivo biliary clearance (in vivo CLb,observed ) by evaluating drug uptake and
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
APPLICATION OF SANDWICH-CULTURED HEPATOCYTES IN DRUG DISCOVERY
1899
Figure 1. ChemMAP score plot projection of 110 Novartis discovery new chemical entities (NCEs). Principal component analysis (PCA) was used to illustrate the position of the 110 NCEs in physical property space. The PCA score plot defined by principle components (PC) 1 and 2 shows the position of the bioactive ChemMAP reference molecules used to define the property space (black solid circles).9 The loading plot helps to interpret the position of NCEs (red solid circles) in property space (n = 110).
metabolism, as well as the potential for cytochrome P450 (CYP) induction.4 The assay represents a novel in vitro method for characterizing the hepatobiliary disposition and DDI potential of drug candidates and their metabolites.5 From a small collection of marketed drugs, the data from sandwich-cultured rat and human hepatocytes appeared to correlate reasonably well with in vivo CLb,observed data.6–7 Although promising as a compound-ranking tool for the in vivo biliary clearance risk applicable in early drug discovery,4 neither a systematic validation of SCH using real drug discovery NCEs from diversified chemical space and with suboptimal physicochemical drug properties, nor a quantitative model to properly account for in vitro biliary excretion was reported. Furthermore, guidance/strategy on how to bridge IVIVC gaps by considering the interplay with other absorption, distribution, metabolism, elimination, and toxicity (ADMET) properties is currently not available. Although CLb is influenced by both transporters and metabolic enzymes,8 little has been reported on the use of an in vitro model to account for the cooperative effects of transporters and metabolic enzymes on the hepatobiliary disposition of drug candidates. In this study, 12 commercially available compounds and 110 structurally diverse Novartis discovery compounds were evaluated using sandwich-cultured rat hepatocyte (SCRH) model. Clearly, most of the 110 NCEs were widely scattered in the new drug discovDOI 10.1002/jps
ery space (quadrants I and IV) rather than the traditional drug space (quadrants II and III9 ; Fig. 1). Passive permeability in parallel artificial membrane permeability assay (PAMPA), absorption permeability in Caco-2, and metabolic activity in rat liver microsomes were evaluated in parallel in order to explore the relationships between the in vitro biliary excretion, permeability, and metabolism of test compounds. Fifteen Novartis discovery compounds with in vitro (SCRH) and in vivo CLb,observed (BDC rat) results were compared to identify potential correlations and to set preliminary criteria to be used as a guideline for selecting discovery compounds to be screened in the future. A preliminary evaluation strategy involving SCH was proposed to highlight biliary excretion risk evaluation during the drug discovery process.
MATERIALS AND METHODS Chemicals Novartis test compounds were obtained from the Novartis compound archive (Cambridge, Massachusetts). Commercially available compounds were obtained from various vendors.
In Vitro PAMPA, Caco-2 Cell Culture, Hepatocytes, and Microsomal Clearance Assays The permeability assays were performed in two in vitro systems: PAMPA and Caco-2 cells, as JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
1900
PAN, BOISELLE, AND WANG
described elsewhere.10 The in vitro rat microsomal cryopreserved hepatocyte clearance was performed according to the protocol published by Obach et al.11
SANDWICH-CULTURED RAT HEPATOCYTE Calculation of biliary excretion index (BEI) and in vitro intrinsic biliary clearance (in vitro CLb,int ) in this report are similar to that performed by Liu et al.7
Sandwich-Cultured Rat Hepatocyte Sandwich-cultured 24-well plates with plated rat hepatocytes were purchased from Qualyst (Raleigh, North Carolina) after 3 days in culture. Upon receipt of the plates, the culture media was changed and the plates were incubated overnight at 37◦ C. The uptake study was carried out according to the method reported previously,7,12 with small modifications. Briefly, after an overnight incubation, SCRH was preincubated in 400 :L regular buffer (Hank’s balanced salt solution with calcium) or calcium-free buffer at 37◦ C for 10 min. The incubation in buffer without calcium opens the tight junctions of the bile canaliculi. After removing the buffer at the end of preincubation, 400 :L of 1 :M test compound prepared in regular buffer (dosing solution) was added to initiate the uptake process (10 min at 37◦ C). An aliquot of 100 :L sample was collected at both the beginning (time = 0 min) and the end (time = 10 min) of the incubation period. After incubation, the dosing solution was aspirated from the cells and uptake was terminated by washing the cells three times with 4◦ C regular buffer and aspirating the final wash. Plates were analyzed by liquid chromatography– tandem mass spectrometry (LC–MS/MS). Nonspecific binding (NSB) of each compound was evaluated by performing the assay using blank collagen-coated wells.
In Vivo CLb,observed Study After Intravenous Administration A BDC rat model (Charles River Laboratories International, Inc., Wilmington, Massachusetts) was employed to measure the bile and plasma concentrations of the test compounds. Compounds were administered intravenously in the animal studies, and bile and blood samples were collected in parallel at various time points. Compound concentrations were measured using LC–MS/MS. Data Analysis
In Vitro PAMPA, Caco-2 Cell Culture, Hepatocytes, and Microsomal Clearance Assays Calculation of the apparent permeability (Papp ) in PAMPA and Caco-2 cells, and efflux ratio (ER) in Caco-2 cells were performed according to the publication of Skolnik et al.10 The calculation of intrinsic clearance of rat liver microsomes (CLRLM,int ) and extraction ratio followed the Obach et al.11 method. JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
BEI =
Aplus
Ca++
− Aminus
Aplus
Ca++
Ca++
× 100%
(1)
Where A is the test compound accumulation amount in the wells of SCRH 24-well kit (pmol/well), Aplus Ca++ represents the test compound accumulation amount in regular buffer-treatment wells, and Aminus Ca++ represents the test compound accumulation amount in the calcium-free buffer-treatment wells. The value calculated in Eq. 1 represents the percentage of the initial drug concentration excreted into the bile. In vitro CLb,int =
Aplus
− Aminus AUC0−10 min
Ca++
Ca++
(2)
In Eq. 2, in vitro CLb,int of SCRH is calculated to represent the biliary excretion of the tested compound. Here, total concentration of the test compound during the entire incubation time was quantified using area under curve (AUC0–10min ) = time × (Cm,0min + Cm,10min )/2. Cm,0min and Cm,10min represent the substrate concentration at the beginning (0 min) and the end of the incubation period (10 min), respectively. In vitro CLb,int [:L/(min mg protein)] can be scaled to account for the total bile excretion in whole liver [CLb,int,s expressed as mL/(min kg)] for the in vivo rat model after being normalized by an empirical scaling factor (SF, Eq. 4). The SF (Eq. 3) of 8000 in the clearance calculation is based on allometric scaling; rat liver weight and protein content in liver tissue were assumed to be 40 g/kg of body weight and 0.20 g/ g of liver weight,13 respectively, in all calculations.7 The predicted biliary clearance (CLb,predicted ) was estimated according to the equation below (Eq. 5). The equation is based on the well-stirred model of hepatic disposition, where Q represents rat hepatic plasma flow [40 mL/(min kg body weight)]7 : SF = =
40 g liver 200 mg protein × 1 g liver 1 kg body mass 8000 mg protein kg body mass
(3)
CLb, int, s (:L/ min /kg body mass) g × SF = CLb,int × 1000 mg
CLb,predicted =
Q × CLb,int,s Q + CLb,int,s
(4)
(5) DOI 10.1002/jps
APPLICATION OF SANDWICH-CULTURED HEPATOCYTES IN DRUG DISCOVERY
1901
Figure 2. Sandwich-cultured rat hepatocyte biliary excretion index (BEI) compared with in vivo rat biliary clearance (in vivo CLb,observed ). Compound BEI values were divided into two clusters: zone a, low-biliary-clearance-risk compounds (BEI < 10%); zone b, high-biliaryclearance-risk compounds (BEI > 10%; n = 15).
Mass Balance Mass balance is calculated to evaluate the impact of drug metabolism on the cumulative uptake. Mass balance less than 70% was noted in comments. Amedium,10 min + Acell,10min × 100% Amedium,0 min
MB =
(6) Here, MB is the mass balance percentage. The amount (A) of the test compound has been assessed in the incubation medium (0 and 10 min) and the cells.
In Vivo Biliary Clearance in vivo biliary clearance was calculated using the following equation: CLb,observed = Abile /AUCplasma
(7)
where AUCplasma is the area under the curve of plasma concentration and time, and Abile is the amount of drug excreted into bile during that period.
RESULTS To evaluate the predictivity of the current SCRH model, the data derived from in vitro rat sandwiched experiments were compared with those collected from the in vivo BDC rat model (Fig. 2). With the limited dataset, it appeared that compounds with BEI values less than 10% in SCRH showed in vivo CLb,observed values less than 2 mL/(min kg) (zone a). In contrast, the DOI 10.1002/jps
compounds with high in vivo CLb,observed [e.g., >2 mL/ (min kg)], all clustered in the “in vitro BEI > 10%” zone (b). Overall, with the limited dataset, no clear linear correlation has been derived between in vitro BEI and in vivo CLb,observed , although in vitro BEI proves to be a useful binning tool when prioritizing compounds for the in vivo BDC studies. In this case, a BEI value of 10% obtained from in vitro SCRH studies is recommended to be the cutoff value for flagging compounds with significant CLb potential. For marketed drugs, SCRH–CLb,predicted values collected in our laboratory agree well with the in vivo CLb,observed data reported in literature (Table 1). Moreover, the marketed drugs and most of Novartis discovery NCEs appeared to scatter along the unity correlation line (dashed line in Fig. 3). A cluster of five Novartis NCEs deviated from the above trend line (circled in Fig. 3). Interestingly, those “outliers” exhibited low BEI (<10%) in the current SCRH studies. Taking the binning results between BEI from in vitro SCRH studies and CLb from BDC rat experiments [e.g., “BEI < 10%” corresponded well to “CLb < 2mL/ (min kg)” in Fig. 2] into consideration, the above trend line between predicted and observed rat CLb may still hold true for marketed drugs and discovery NCEs. The investigation of the interplay between physicochemical/ADME properties and the SCRH parameters led to the observation that moderate to high BEI and CLb,int in SCRH mostly occurred with lower Papp in PAMPA and Caco-2 (Figs. 4a and 4b and Table 2), and the probability of high biliary clearance risk appeared to decline significantly with increased permeability. Specifically, CLb,int was negligible when JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
1902
PAN, BOISELLE, AND WANG
Table 1.
Kinetic Parameters for Commercially Available Compounds in SCRH
Name Cephradine # D8-TCA
Digoxin # DPDPE # ICG
Indomethacin Ouabain Pravastatin Rosuvastatin Methotrexate Vincristine Doxorubicin # # #
BEI (%)
In Vitro CLb ,int [:L/(min mg protein)]
Predicted In Vivo CLb [mL/(min kg)]
Observed In Vivo CLb [mL/(min kg)]
0 56.6 39.7 88 63.1 0 43.6 30.5 26.4 35.6 33.8 9.3
≤0.01 2.05 0.33 1.21 3.15 ≤0.01 0.39 0.26 3.80 0.10 0.13 0.31
≤0.5 28.9 11.7 24.3 32 ≤0.5 13.2 10 31 4.5 5.5 11.4
3.114 29.815 10.716 18.717 3218 0.0119 14.720 14.221 24.312 12.122 14.223 16.624
D8-TCA: d8-taurocholate acid. DPDPE: [D-Pen2,5]Enkephalin hydrate. ICG: Indocyanine green.
PAMPA Papp at neutral pH was greater than 40 × 10−6 cm/s and/or Caco-2 Papp was greater than 10 × 10−6 cm/s. For instance, the high biliary potential [e.g., CLb,int > 0.3 :L/(min kg)] decreases from 31.7% to 4.3% and from 32.1% to 3.4% when under the above permeability boundary values for PAMPA and Caco2, respectively. Interestingly, the in vitro observations were also confirmed by in vivo BDC rat data (insets of
Figs. 4a and 4b). Similar trends were also observed between BEI and the Papp from either PAMPA or Caco-2 experiments. In addition to passive permeability as observed from the PAMPA experiments (Fig. 4), the impact of active transport on CLb was assessed in the context of the ER derived from Caco-2 experiments. Although it is difficult to develop a clean correlation between
Figure 3. Sandwich-cultured rat hepatocyte predicted biliary clearance (CLb,predicted ) was compared with in vivo rat biliary clearance (CLb,oberved ). Red hollow squares represent tested market compounds (n = 12). Blue solid diamonds represent Novartis compounds (n = 15). For the compounds with biliary excretion index (BEI) < 10% (zone a), in vivo CLb,observed is less than 2 mL/(min kg) and is regarded as negligible. For the compounds with BEI > 10% (zone b), CLb is estimated. Compounds in the red circle also had CLb estimated, but their BEI was less than 10%. JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
DOI 10.1002/jps
APPLICATION OF SANDWICH-CULTURED HEPATOCYTES IN DRUG DISCOVERY
1903
Figure 4. The impact of hepatocyte permeability on biliary excretion. (a–d) Sandwich-cultured rat hepatocyte biliary excretion parameters [biliary excretion index (BEI) and intrinsic biliary clearance (CLb,int )] were compared with parallel artificial membrane permeability assay (PAMPA) and Caco-2 permeability. Selected compounds with in vivo CLb,observed are presented in the inset graphs in the panels a and b (n = 108).
BEI or in vitro CLb,int and the Caco-2 efflux ratio (ER) for discovery NCEs, a trend was observed (Fig. 5). For instance, for NCEs with ER < 1, only a small fraction of them showed high CLb based on data from the in vitro SCRH (e.g., 15% for BEI > 10% or 6% for CLb > 0.3, respectively). However, the above fraction for high CLb drastically increased for NCEs with ER > 5 (e.g., 50% for BEI > 10% or 41% for CLb > 0.3, respectively). Thus, for those compounds having an elevated ER in Caco-2 cells, a concern regarding active secretion into the bile may be warranted. The interplay between CYP-based metabolic clearance and CLb was explored using the SCRH model (Fig. 6). Apparently, data were scattered in the lower left zone below the dotted diagonal line, which was included to help define the threshold. Specifically, no data were in the upper right half of the graph, suggesting that compounds highly metabolized in liver microsomes (extraction ratio > 0.75) are unlikely to
DOI 10.1002/jps
be actively excreted into the bile in SCRH [CLb,int < 1 :L/(min mg protein]. In contrast, most of the NCEs with high CLb,int were clustered on the low extraction zone (<0.4). In order to accurately interpret the SCRH data in the context of interplay between CYP metabolism and biliary excretion, it is critical to confirm if the SCRH system still maintains part of CYP activities. After incubation in SCRH, the remaining concentration of the CYP3A substrate (testosterone) and CYP1A substrate (phenacetin) decreased appreciably (Fig. 6). The calculated hepatic extraction clearance in SCRH was comparable to cryopreserved rat hepatocytes [52.7 vs. 52 mL/(min kg) for testosterone and 54.2 vs. 31.6 mL/(min·kg) for phenacetin]. The calculated hepatic extraction ratio in SCRH was close to 1. The results suggest that the SCRH system maintains the enzymatic activity for selected CYPs to some extent.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
1904
1.0 47.0 0.1 8.1 0.1 0.2 1.3 93.4 10.2 0.4 0.3 0.5 17.9 132.0 1.6 0.5 11.2 2.8 21.9 0.5 6.4 0.5 14.3 30.8 9.7 24.0 0.5 29.0 26.0 18.4
c For
b For
a 10%
was used to represent low metabolic compounds extraction ratio in liver microsomes (RLM liver extraction ratio < 10%). compounds without biliary excretion (BEI = 0%), 0.1 was used to replace zero for the convenience of plotting (log scale). compounds without biliary excretion [CLb ,predicted = 0 mL/(min kg)], 0.5 was used to replace zero for the convenience of plotting (log scale). d NVP-8 had no liver microsomal clearance data. Hepatocytes clearance data was used as surrogates (low metabolic ability in hepatocytes). RLM, rat liver microsomes; NA, not available.
0.1 18.3 0.5 30.0 0.1 1.4 0.1 21.0 22.5 1.4 4.3 0.1 22.1 24.0 5.5 81.0 <3.4 133.0 321.0 18.1 35.0 17.0 9.5a 11.0 81.0 38.0 <3.4 14.0 15.0 125.0 72.5 10.0 81.0 91.0 10 53.0 35.5 10d 26.0 73.0 55.5 10 28.0 33.0 80.0 31.6 2.5 31.6 4.0 12.6 10.0 0.8 0.8 20.0 63.1 79.4 79.4 0.0 4.0 7.9 0.9 16.0 0.3 15.2 0.9 3.1 1.4 1.9 NA 1.0 NA NA 1.6 26.7 0.1 2.7 0.5 2.0 0.9 0.3 2.5 0.2 0.7 NA 16.6 NA NA 1.0 0.1 2.7 4.8 2.2 4.7 2.5 6.2 2.6 4.8 4.2 8.4 3.8 8.3 3.9 5.5 3.3 7.0 385.5 505.6 502.6 489.9 546.5 311.4 1173.4 433.5 928.6 450.5 387.6 468.5 609.8 515.4 697.6
RLM Hepatic Extraction %a PAMPA Papp (×10–6 cm/s) Caco-2 Papp (B-A)/ Papp (A-B) Caco-2 Papp (A-B) (×10–6 cm/s) Clog p Molecular Weight (MW,g/mol)
DISCUSSION
NVP-1 NVP-2 NVP-3 NVP-4 NVP-5 NVP-6 NVP-7 NVP-8 NVP-9 NVP-10 NVP-11 NVP-12 NVP-13 NVP-14 NVP-15
Table 2.
The In Vivo and In Vitro Biliary Clearance Data Observed from Novartis Discovery NCEs (n = 15)
RLM CLint [mL/(min mg protein)]
BEI (%)b
Predicted CLb [mL/(min kg)]c
CLb,observed [mL/(min kg)]
PAN, BOISELLE, AND WANG
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
The polarized transporter expression and bile canaliculi network, such as bile pocket structures make SCRH a suitable in vitro model for investigating the biliary excretion of NCEs. The integrity of the tight junctions responsible for forming the canalicular network (bile pocket structures) is calcium dependent and can be opened by removing calcium from the incubation medium.25 Clearance measurements in the presence and absence of calcium allow for the determination of BEI and CLb,int , making it possible to assess the relative contributions of transporters involved in biliary drug excretion.7 Although the two parameters were employed in SCRH calculation, no report suggests utilizing these in vitro parameters to flag high biliary clearance risk in drug discovery. Biliary excretion index represented the percentage of retained substrate in the bile canaliculi and was believed to reflect the contribution, via active transport mechanisms, to biliary excretion.25 Our present work (Fig. 2) clearly demonstrated that the CLb data from BDC rat could be properly differentiated by the BEI data from the in vitro SCRH model (cutoff value = 10%), suggesting that SCRH may serve as a viable in vitro binning model for high biliary clearance risk for NCEs. Here, SCRH is valuable not only to help establish IVIVC of CLb , but also to prioritize costly in vivo BDC studies using the in vitro model. This is particularly useful for NCEs in which biliary elimination is the dominant elimination pathway. In this case, unnecessary in vivo studies can thus be avoided. It is worthwhile to note that BEI itself could not predict in vivo CLb,observed . This could be a result of differences between the in vitro and in vivo metabolic clearances (Fig. 6), or due to the contributions of other competitive elimination processes such as NCEs sinusoidal uptake activities.7 This could partially explain why some compounds (especially high-CLb compounds) with similar in vitro BEI values have fluctuation of over 10-fold for in vivo CLb,observed (Table 2, Fig. 2). Intrinsic biliary clearance, employed to describe the ability of compounds to be taken up into hepatocytes and excreted into bile, can be evaluated in the context of overall clearance involving multiple mechanisms. Indeed, the CLb,predicted of marketed drugs and most of NCEs from the current SCRH model had a reasonable correlation with the in vivo BDC data reported in literature sources (Table 1 and dashed line in Fig. 3). It should be noted that five Novartis NCEs deviating from the above trend line exhibited low BEI (10%). In other words, the absolute values of CLb,predicted may need to be evaluated with caution when BEI < 10%. In this case, the simplest approach may be to “qualify” CLb,predicted as “less than 2 mL/(min kg),” as suggested by Figure 2. From a mechanistic perspective, all of DOI 10.1002/jps
APPLICATION OF SANDWICH-CULTURED HEPATOCYTES IN DRUG DISCOVERY
1905
Figure 5. Sandwich-cultured rat hepatocyte biliary excretion [biliary excretion index (BEI) and intrinsic biliary clearance (CLb,int )] was compared with Caco-2 efflux ratio (ER) (n = 90).
the five NCE “outliers” with overpredicted CLb exhibited moderate to high CYP-related metabolic ability (extraction ratio = 0.53–0.89) and moderate to high permeability in PAMPA and Caco-2. Employing additional mechanistic studies may be valuable to pursue the impact of key physicochemical and ADME properties on the prediction of CLb . It is noteworthy that some marketed drugs (cephradine and indomethacin, both low-CLb drugs) also showed moderate variability against the above trend line, possibly because of variability among animal samples and/or analytical challenge when reaching the limit of detection. On the basis of our limited dataset, the IVIVC for CLb may not be very straightforward and depends greatly on interplay with other physicochemical and ADME properties.
It is also worthwhile to note that for a limited number of reported drugs, especially statins, improved linear correlation between SCRH CLb and in vivo CLb,observed was observed after protein binding correction.1226 However, for compounds with diverse structures and physicochemical properties, the IVIVC was not improved after protein binding correction, neither for market drugs nor Novartis compounds (based on the calculation from our dataset; data not shown). The detail mechanisms are still under investigation. It was speculated that the kinetic effects of plasma protein binding and the interplay between various competing and collaborating processes may contribute to this disconnection. For instance, it was implied that plasma protein binding is not
Figure 6. Sandwich-cultured rat hepatocyte intrinsic biliary clearance (CLb,int ) compared with the rat liver microsomes extraction ratio. Compounds with high intrinsic metabolic clearances were less likely to be actively excreted into bile. The extraction ratio of testosterone and phenacetin were comparable to the values in cryopreserved hepatocytes suspension (data not shown; n = 92). DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
1906
PAN, BOISELLE, AND WANG
the rate-limiting step for hepatic clearances in all circumstances.27 The equilibrium between the free drug in the aqueous phase and the drug in the membrane attenuates the impact on the free drug hypothesis.28 How compounds access binding sites (in vitro vs. in vivo systems) of the proteins also influence drug clearance.28 Overall, based on the existing evidence, it is not convincing that protein binding correction improves CLb prediction in SCRH studies. Therefore, no protein binding correction was employed to calibrate Eq. 5 in the present work. Although CLb is believed to account for the contributions of both efflux (into bile) and influx (uptake into hepatocyte) transporters in the biliary clearance process,25 BEI, by principle, only reflects that of efflux transporters to bile. Similar inverse trends between in vitro permeability data and CLb,int (Figs. 4a and 4b) or BEI (Figs. 4c and 4d), however, support the postulation that the biliary clearance process may primarily be governed by the transporter-mediated efflux process. In this case, compound uptake as a result of passive permeation into hepatocytes will be less critical to the overall level of CLb .29 Meanwhile, considering the volume and flow rate of bile and the limited canalicular membrane surface area, the direct contribution of passive permeation to the overall bile clearance may be negligible. Indeed, a similar observation was also reported between renal clearance and passive permeability.30 This prompts the need to utilize proper in vitro transporter-screening tools to flag the biliary clearance risk in early discovery. Efflux transporters located in hepatocyte canalicular membranes, such as P-glycoprotein (or multidrug resistance protein 1), MRP2, and breast cancer resistance protein, also exist in the human intestine and Caco-2 cells. BEI is an indicator of efflux transporter contribution in biliary excretion,25 and a clear trend between high ER in Caco-2 and high BEI and CLb in SCRH (Figs. 4 and 5) demonstrated that Caco-2 is among the viable in vitro tools for not only flagging CLb , but also helping explore the transport mechanisms. The above semiquantitative correlation between BEI or CLb and Caco-2 ER, albeit nonlinear, is rational, as it reflects the different levels of transporter expression between liver and gut.31 It should be noted, however, that Caco-2 ER itself is rather a semiquantitative flag for efflux substrate(s) than a quantitative parameter for its expression and activity.32 Compounds that are actively metabolized by CYP enzymes may not show high biliary excretion as a result of the location of CYP enzymes in hepatocytes.3 This is most likely because of the competing nature of efflux transporters and metabolic enzymes in hepatic drug elimination.29,33–34 However, most of the previous reports tested only a limited number of marketed drugs. The present work has successfully expanded the study by focusing on the interplay of CLb and JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
CYP-related metabolic activity using a collection of discovery compounds from diverse chemical scaffolds (Fig. 1). Although it has been reported that long-term cultured hepatocytes lose both phase I and phase II enzyme activity,35 and phase I metabolic activities deteriorate significantly in SCH,7 the data presented here suggests that some CYP activity remains in SCH (Fig. 6). For example, SCRH clearance values for the CYP3A substrate testosterone and the CYP1A substrate phenacetin are comparable to the clearance values in cryopreserved rat hepatocyte suspension. This is consistent with the previous reports that P450 activities (e.g., CYP3A) may be kept or even “recovered” after 3–4 days of culture in SCRH.36–37 Most of the compounds with high CLb,int in our dataset had medium to low clearance in rat liver microsomes. The results indicated that SCRH could be used to explore the interplay between transporters and metabolic enzymes in compound optimization and selection. Despite the competitive interplay between transportermediated CLb and CYP-based metabolic elimination, the “inverse relationship” between the two was neither perfectly “linear” nor “clean”, as depicted in Figure 6. Instead, a collection of low-CLb NCEs observed in SCRH showed varying CYP metabolic extraction ratios. In other words, the low CYP clearance does not always translate to high biliary clearance risk here. It is important to note that the microsomal data represents only phase I metabolic activities, whereas a clearance study in hepatocytes could better characterize the interplay between CLb and multiple metabolic mechanisms. The SCRH model could be more valuable in this process if CYP enzymes were induced.4 A preliminary biliary clearance risk evaluation strategy for drug discovery was proposed based on the SCRH results and related permeability/metabolism information obtained in this study (Fig. 7). For lead optimization and candidate selection, it is important to understand if CLb is the primary clearance pathway and/or limits the exposure of NCEs, and what may be potential drug interaction risk. In that sense, compounds with clearance IVIVC gaps would be required to undergo permeability screening (PAMPA/ Caco-2) before assessing the biliary excretion risk. A BEI value greater than 10% would be the first cutoff value for the SCRH model. Only when CLb,int is greater than 0.3 :L/(min mg protein) [CLb,predicted > 10 mL/(min kg)] would the BDC rat study be recommended to confirm the in vitro results. It should be noted that the in vitro model is not designed to replace the in vivo mechanistic studies that can offer comprehensive mechanistic insights. Rather, it could identify compounds with high CLb . However, in order to optimize NCEs ADMET properties, the contribution of each clearance route (biliary excretion, urinary excretion, metabolism, etc.) to the overall total clearance have to be determined together. In addition, the DOI 10.1002/jps
APPLICATION OF SANDWICH-CULTURED HEPATOCYTES IN DRUG DISCOVERY
1907
Figure 7. Proposed biliary clearance risk evaluation strategy for discovery stage compounds.
application of this strategy would greatly accelerate the biliary clearance risk evaluation process and reduce the burden of animal studies in discovery. This is especially true for compounds with similar scaffolds. Noticeably, the above strategy is to identify the clearance gap and flag high biliary clearance risk in rat in support for chemistry optimization in early drug discovery. At present, however, this strategy is not feasible to predict human biliary clearance risk or establish IVIVC,38 owing to limited clinic CLb data. The situation may be improved once additional human biliary excretion data become available.
CONCLUSION The SCRH model proved to be a valuable tool for ranking and predicting in vivo CLb,observed during drug discovery. Biliary excretion involves three processes: uptake across the sinusoidal membrane, metabolism, and excretion across the canalicular membrane.7 The in vivo and in vitro “intrinsic” clearance values calculated should be rate limited by the slowest step in these processes. Because CLb,int was determined on the basis of intracellular substrate amount, it should be considered as “apparent” CLb,int values. Therefore, it is difficult to evaluate uptake/efflux transporters alone without considering the contribution of passive diffusion and CYP metabolism in SCH, especially after the observation of P450 activities in cultured hepatocytes (Fig. 6). DOI 10.1002/jps
For compounds with extremely low passive permeability and low metabolism, the probability of CLb as the dominate elimination pathway may drastically increase. The combination of passive permeability, metabolic intrinsic clearance, and the SCRH model could be used as an initial screening platform for bile excretion potential as well as a means for improving compound liabilities and properties. Future studies using an extended list of discovery NCEs with varying biliary excretion ability, chemical diversity, and a suitable kinetic model (including passive permeability, transporters, and metabolism) will offer additional understanding for improved prediction of CLb .
ACKNOWLEDGMENTS The authors appreciate the critical review and constructive feedback from Dr. Leslie Bell. The authors also thank Suzanne Skolnik for offering diversity analysis of Novartis discovery NCEs, as well as Drs. Kenneth Brouwer and Claudia McGinnis for offering assistance in the initial setup of the Sandwich Cultured Rat Hepatocyte model in the Novartis lab.
REFERENCES 1. Bi YA, Kazolias D, Duignan DB. 2006. Use of cryopreserved human hepatocytes in sandwich culture to measure hepatobiliary transport. Drug Metab Dispos 34(9):1658–1665. 2. Luo G, Johnson S, Hsueh MM, Zheng J, Cai H, Xin B, Chong S, He K, Harper TW. 2010. In silico prediction of biliary excretion of drugs in rats based on physicochemical properties. Drug Metab Dispos 38(3):422–430. JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
1908
PAN, BOISELLE, AND WANG
3. Yang X, Gandhi YA, Duignan DB, Morris ME. 2009. Prediction of biliary excretion in rats and humans using molecular weight and quantitative structure–pharmacokinetic relationships. AAPS J 11(3):511–525. 4. Swift B, Pfeifer ND, Brouwer KL. 2010. Sandwich-cultured hepatocytes: An in vitro model to evaluate hepatobiliary transporter-based drug interactions and hepatotoxicity. Drug Metab Rev 42(3):446–471. 5. Hoffmaster KA, Turncliff RZ, LeCluyse EL, Kim RB, Meier PJ, Brouwer KL. 2004. P-glycoprotein expression, localization, and function in sandwich-cultured primary rat and human hepatocytes: Relevance to the hepatobiliary disposition of a model opioid peptide. Pharm Res 21(7):1294–1302. 6. Giacomini KM, Huang SM, Tweedie DJ, Benet LZ, Brouwer KL, Chu X, Dahlin A, Evers R, Fischer V, Hillgren KM, Hoffmaster KA, Ishikawa T, Keppler D, Kim RB, Lee CA, Niemi M, Polli JW, Sugiyama Y, Swaan PW, Ware JA, Wright SH, Yee SW, Zamek-Gliszczynski MJ, Zhang L. 2010. Membrane transporters in drug development. Nat Rev Drug Discov 9(3):215–236. 7. Liu X, Chism JP, LeCluyse EL, Brouwer KR, Brouwer KL. 1999. Correlation of biliary excretion in sandwich-cultured rat hepatocytes and in vivo in rats. Drug Metab Dispos 27(6):637–644. 8. Pang KS, Maeng HJ, Fan J. 2009. Interplay of transporters and enzymes in drug and metabolite processing. Mol Pharm 6(6):1734–1755. 9. Faller B, Ottaviani G, Ertl P, Berellini G, Collis A. 2011. Evolution of the physicochemical properties of marketed drugs: Can history foretell the future? Drug Discov Today 16(21–22):976–984. 10. Skolnik S, Lin X, Wang J, Chen XH, He T, Zhang B. 2010. Towards prediction of in vivo intestinal absorption using a 96well Caco-2 assay. J Pharm Sci 99(7):3246–3265. 11. Obach RS, Baxter JG, Liston TE, Silber BM, Jones BC, MacIntyre F, Rance DJ, Wastall P. 1997. The prediction of human pharmacokinetic parameters from preclinical and in vitro metabolism data. J Pharmacol Exp Ther 283(1):46–58. 12. Fukuda H, Ohashi R, Tsuda-Tsukimoto M, Tamai I. 2008. Effect of plasma protein binding on in vitro – in vivo correlation of biliary excretion of drugs evaluated by sandwich-cultured rat hepatocytes. Drug Metab Dispos 36(7):1275–1282. 13. Seglen PO. 1976. Preparation of isolated rat liver cells. Methods Cell Biol 13:29–83. 14. Nakashima, E., R. Matsushita, et al. Comparative pharmacokinetics of cefoperazone and cephradine in untreated streptozotocin diabetic rats. 1992, Drug Metab Dispos 20(5):730–735. 15. Inoue, M., Y. Morino, et al., Transhepatic transport of taurocholic acid in normal and mutant analbuminemic rats. 1985. Biochim Biophys Acta 833(2):211–216. 16. Watkins, J. B., 3rd and S. E. Sherman,Long-term diabetes alters the hepatobiliary clearance of acetaminophen, bilirubin and digoxin. 1992. J Pharmacol Exp Ther260(3), 1337–1343. 17. Qualyst Technical Application Bulletin. 2008, TAB Biol. 003v1, 18. Pollack, G. M. and K. L. Brouwer. Thermal injury decreases hepatic blood flow and the intrinsic clearance of indocyanine green in the rat, Pharm Res 1991, 8(1), 106–111. 19. Kouzuki, H., H. Suzuki, et al. Pharmacokinetic study of the hepatobiliary transport of indomethacin., Pharm Res 2000.174)432–438. 20. Russell, J. Q. and C. D. Klaassen Species variation in the biliary excretion of ouabain., J Pharmacol Exp Ther 1972, 183(3)513–519. 21. Yamazaki, M., S. Akiyama, et al., Biliary excretion of pravastatin in rats: contribution of the excretion pathway mediated
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33. 34.
35.
36.
37.
38.
by canalicular multispecific organic anion transporter., Drug Metab Dispos 1997, et al. 25(10), 1123–1129. Masuda, M., Y. I’Izuka, et al., Methotrexate is excreted into the bile by canalicular multispecific organic anion transporter in rats., Cancer Res1997 57(16), 3506–3510. Mellows, G. et al., The effect of XR9576 (tariquidar), a specific P-glycoprotein inhibitor, on the pharmacokinetics of doxorubicin, paclitaxel and vincristine in the rat., Proc Am Assoc Cancer Res (AACR) 2002, 43, 1363. Achira, M., R. Totsuka, et al, Tissue-specific regulation of expression and activity of P-glycoprotein in adjuvant arthritis rats., Eur J Pharm Sci 2002, 16(1–2)29–36. Liu X, LeCluyse EL, Brouwer KR, Gan LS, Lemasters JJ, Stieger B, Meier PJ, Brouwer KL. 1999. Biliary excretion in primary rat hepatocytes cultured in a collagen-sandwich configuration. Am J Physiol 277(1 Pt 1):G12–G21. Abe K, Bridges AS, Yue W, Brouwer KL. 2008. in vitro biliary clearance of angiotensin II receptor blockers and 3-hydroxy-3-methylglutaryl-coenzyme A reductase inhibitors in sandwich-cultured rat hepatocytes: Comparison with in vivo biliary clearance. J Pharmacol Exp Ther 326(3):983– 990. Baker M, Parton T. 2007. Kinetic determinants of hepatic clearance: Plasma protein binding and hepatic uptake. Xenobiotica 37(10–11):1110–1134. Smith DA, Di L, Kerns EH. 2010. The effect of plasma protein binding on in vivo efficacy: Misconceptions in drug discovery. Nat Rev Drug Discov 9(12):929–939. Watanabe T, Kusuhara H, Sugiyama Y. 2010. Application of physiologically based pharmacokinetic modeling and clearance concept to drugs showing transporter-mediated distribution and clearance in humans. J Pharmacokinet Pharmacodyn 37(6):575–590. Fagerholm U. 2008. Prediction of human pharmacokinetics— Biliary and intestinal clearance and enterohepatic circulation. J Pharm Pharmacol 60(5):535–542. Hilgendorf C, Ahlin G, Seithel A, Artursson P, Ungell AL, Karlsson J. 2007. Expression of thirty-six drug transporter genes in human intestine, liver, kidney, and organotypic cell lines. Drug Metab Dispos 35(8):1333–1340. Lin X, Skolnik S, Chen XH, Wang J. 2010. Attenuation of intestinal absorption by major efflux transporters: Quantitative tools and strategies using Caco-2 model. Drug Metab Dispos 39(2):265–274. Liu L, Pang KS. 2005. The roles of transporters and enzymes in hepatic drug processing. Drug Metab Dispos 33(1):1–9. Wu CY, Benet LZ. 2005. Predicting drug disposition via application of BCS: Transport/absorption/elimination interplay and development of a biopharmaceutics drug disposition classification system. Pharm Res 22(1):11–23. Niemann C, Gauthier JC, Richert L, Ivanov MA, Melcion C, Cordier A. 1991. Rat adult hepatocytes in primary pure and mixed monolayer culture. Comparison of the maintenance of mixed function oxidase and conjugation pathways of drug metabolism. Biochem Pharmacol 42(2):373–379. LeCluyse EL. 2001. Human hepatocyte culture systems for the in vitro evaluation of cytochrome P450 expression and regulation. Eur J Pharm Sci 13(4):343–368. Tuschl G, Hrach J, Walter Y, Hewitt PG, Mueller SO. 2009. Serum-free collagen sandwich cultures of adult rat hepatocytes maintain liver-like properties long term: A valuable model for in vitro toxicity and drug–drug interaction studies. Chem Biol Interact 181(1):124–137. Ghibellini G, Vasist LS, Leslie EM, Heizer WD, Kowalsky RJ, Calvo BF, Brouwer KL. 2007. in vitro – in vivo correlation of hepatobiliary drug clearance in humans. Clin Pharmacol Ther 81(3):406–413.
DOI 10.1002/jps