The multidrug resistance-associated protein (MRP) family, currently consisting of nine members (MRP1-9, also named ABCC1-6 and ABCC10-12), is one branch of the ATP-binding cassette (ABC) superfamily of transmembrane proteins that use the energy of ATP-hydrolysis to translocate their substrates across biological membranes (Borst et al., 2000; Kruh and Belinsky, 2003; Schinkel and Jonker, 2003). ABCC2/MRP2 is expressed mainly in the apical membrane of liver canaliculi, renal proximal tubules, gut enterocytes, placenta, and brain-blood barrier. The substrates of ABCC2/MRP2 include nonconjugated amphipathic organic anions and glucuronide, glutathione, and sulfate conjugates (Suzuki and Sugiyama, 1998; Keppler and Konig, 2000). ABCC2/MRP2 also transports various unmodified drugs including vincristine (Keppler et al., 2000), doxorubicin (Koike et al., 1996; Cui et al., 1999), human immunodeficiency virus protease inhibitors (Gutmann et al., 1999), nucleoside phosphonates (Miller et al., 2001), p-aminohippuric acid (Leier et al., 2000), and fluoroquinolone antibiotics (Naruhashi et al., 2002). Therefore, ABCC2/MRP2, being a primary active efflux transporter, is functionally similar to P-glycoprotein as both are involved in the hepatobiliary elimination and intestinal absorption of many structurally diverse xenobiotics and their conjugates (Cui et al., 1999; Kusuhara and Sugiyama, 2002). The physiological, pharmacological, and clinical implications of ABCC2/MRP2 have been reported in reviews (Fardel et al., 2005; Robertson and Rankin, 2006).

It is of great value to establish structure-activity relationships (SARs) for ABCC2/MRP2 during the lead optimization stage of drug discovery. For example, researchers have been attempting to elucidate SARs for potent MRP1 inhibitors for cancer chemotherapy (Wang et al., 2004; van Zanden et al., 2005). Two types of molecules have been identified as MRP1 inhibitors or substrates based on their transport mechanism: 1) compounds cotransported with glutathione (GSH) and cytosolic glutathione S-transferases (GST) and 2) compounds interacting with MRP independently from GSH and GST. For compounds that cotransport with GSH and GST, transporter interactions are poorly understood. In contrast, for the GSH/GST-independent compounds, certain structural requirements have been identified (Boumendjel et al., 2005). For example, aromatic/heteroaromatic moieties, nitrogen atom, and carbonyl groups are frequently observed as MRP1 inhibitors/substrates. However, most of these reports are scattered observations rather than in-depth SAR investigations on MRP(s). In this study, we have carefully selected a group of biphenyl-substituted heterocycles and evaluated their transport and inhibition properties in Caco-2 cells and gene-transfected Madin-Darby canine kidney (MDCK) cells. Based on the analysis, we were able to derive a structure-activity relationship for this class of congeneric compounds and infer the structural requirements of ABCC2/MRP2 interactions.

Materials and Methods

Compounds and Reagents. Dulbecco's modified Eagle's medium, minimum essential medium, fetal bovine serum (FBS), nonessential amino acids, GlutaMAX, sodium pyruvate, gentamicin, l-glutamine, and Hanks' balanced salt solution (HBSS) were purchased from Invitrogen (Carlsbad, CA). MK571 was obtained from BioMol (Plymouth Meeting, PA). Calcein acetoxymethyl ester (calcein-AM) was obtained from Molecular Probes (Eugene, OR). Verapamil and novobiocin were purchased from Sigma-Aldrich (St. Louis, MO).

Cell Culture. Caco-2 cells (American Type Culture Collection, Manassas, VA) were maintained in Dulbecco's modified Eagle's medium with 10% FBS, 1% nonessential amino acids, 1% GlutaMAX, 1 mM sodium pyruvate, and 0.06 mg/ml gentamicin. Parental MDCK cells and MDCK cells stably expressing human ABCC2/MRP2 cDNA (MDCK-MRP2; Dr. P. Borst, The Netherlands Cancer Institute, Amsterdam, The Netherlands), were cultured in minimum essential medium with l-glutamine containing 10% FBS, 100 units of penicillin, and 100 μg/ml streptomycin. Both MDCK and Caco-2 cells were incubated at 37°C in 95% air-5% CO₂ with 95% humidity. Cell culture media were refreshed at 3-day intervals.

Transcellular Transport across Caco-2 Cell Monolayers and Michaelis-Menten Constants. Caco-2 cells were seeded at a density of 1 × 10⁵ cells/cm² onto Millipore 24-well insert plates (Millipore, Billerica, MA) and cultured for 21 to 25 days. Transepithelial electrical resistance values were measured to ensure that tight junctions are formed (≥600 ohms/cm², Millicell-ERS; Millipore). After the transwell filter was washed with HBSS (pH 7.4), each compound was applied to the donor side (either apical or basal chamber) at a concentration of 10 μM to initiate the transport assay. The transport incubation was maintained at 37°C for 2 h on a shaking incubator (Precision Scientific, Winchester, VA). Aliquots (200 μl) were collected from both the apical (A) and basal (B) sides, and compound concentrations were determined by liquid chromatography/MS/MS. The extent of permeation (P_app) was generated for both A→B and B→A transport. The efflux ratio was calculated from (P_app,B→_A)/(P_app,A→_B). An efflux ratio ≥2 indicates that transport is apically polarized (active efflux). The compounds actively effluxed by Caco-2 cells were further applied to the Caco-2 transport assay in the presence of 25 μM MK571 to assess transport under conditions of ABCC2/MRP2-specific inhibition. For measurement of the concentration-dependent transport in Caco-2 cells, the compounds (0.5-50 μM) were added to initiate transport, and an aliquot (200 μl) of the incubation buffer in the receiver chamber was collected at 120 min. The active efflux rate (B→A transport) was determined from the appearance of the compounds in the receiver compartment over the time period. Several models were fitted using nonlinear regression (WinNonlin; Pharsight, Mountain View, CA) to analyze the transport curves. The best-fit model was identified on the basis of the reduction in the sum of squared residuals and Akaike's information criterion.

Inhibition of Calcein Efflux Mediated by ABCC2/MRP2. MDCK-MRP2 cells were grown on a 96-well cell culture plate for 4 to 5 days and allowed to become confluent. After being washed twice with HBSS buffer, the cells were preincubated with 100 μl of test compound (80 μM) or MK571 (25 μM) in HBSS containing 1% dimethyl sulfoxide at 37°C for 15 min. Calcein-AM was then added into the well to yield a final concentration of 1 μM for 20 min. To minimize the calcein-AM efflux by P-gp or ABCC2/MRP2 during the loading phase, calcein-AM loading was conducted on ice. The cells were quickly washed twice with cold HBSS buffer and replaced with 100 μl of the test compound (80 μM) or MK571 (25 μM) for 1 h at 37°C. The cells were quickly washed twice with ice-cold HBSS buffer and then lysed by 1% Triton with 0.01% antifoam. The cell lysate was transferred to a 96-well clear bottom assay plate (Corning Glassworks, Corning, NY), and the fluorescent intensity of calcein was measured using a fluorescent spectrophotometer with an excitation wavelength of 485 nm and an emission wavelength of 530 nm. Quantification of ABCC2/MRP2 inhibition was completed using the following equation: % maximum = (FU_comp - FU_backgroup)/(FU_MK571 - FU_backgroup) ^* 100, where FU_comp is the fluorescence value in the presence of test compound and FU_MK571 is the fluorescence value in the presence of 25 μM MK571.

Structure and Physicochemical Parameters of Biphenyl Compounds. A series of biphenyl-substituted heterocyclic compounds was selected from archive storage at Pfizer Global Research and Development (St. Louis Laboratories, St. Louis, MO) based on similarity of substructure definitions. This 4-heterocyclic moiety contains a pyrazole with two branched polar functional groups: 3-carbamoyl and 4-urea (Table 1). The only structural variation among the selected compounds is located on biphenyl rings substituted with a variety of atoms or functional groups at 2-, 2′-, 3-, and 3′-positions. The substituting groups include halogens (fluorine, chlorine, and bromine) and methyl, trifluoromethyl, hydroxy, methoxy, and ethoxy groups, which differentiate the physicochemical properties and the three-dimensional conformations of each individual compound. The physicochemical properties characterizing the overall lipophilic and electronic properties of the test compounds were calculated using in silico methods from an internally produced computational package. These included ClogP, hydrogen bond donor and acceptor counts, and topological polar surface area (tPSA).

Measurement of Distribution Coefficients of Compounds. LogD values were measured by equilibrating compounds in octanol saturated with 50 mM sodium phosphate buffer (pH 7.4) of ionic strength (0.15 M) for 72 h at ambient temperature. The fraction of the drug in each layer was measured by HPLC using diode array detection. LogD at pH 7.4 was calculated by the equation:

Methods for Molecular Modeling. In recognition of ABCC2/MRP2 interactions, the three-dimensional structure of the ligand is considered to be an important parameter. Specifically, for the compounds under investigation, the torsion angles between the two phenyl groups appear to be extremely sensitive to the different ortho-substitutions (2,2- and 2′,2′-substitution). Initial three-dimensional conformers were generated via a two- to three-dimensional conversion algorithm, CONCORD (Tripos, St. Louis, MO). The conformers were then submitted to full energy minimizations by ab initio calculations using the Gaussian98 program (revision A.11, Gaussian Inc., Pittsburgh, PA). We used a basis set 6-31G^* and a density functional theory treatment with Becke potential, B3LYP/6-31G^*. To avoid convergence problems the structures were preminimized at the STO-3G level for 100 cycles before density functional theory calculations. A default convergence criterion was used.

To understand the importance of molecular fields around the molecules in their interaction with ABCC2/MRP2, comparative molecular field analysis (CoMFA) was applied to calculate the minimum energy conformers using Sybyl 7.0 (Tripos). The Gasteiger method was used to calculate the atomic partial charges. The molecules are aligned by their cogeneric core, specifically the amide urea pyrazole moiety. After the alignment, the molecules were placed in a three-dimensional cubic lattice with 2 Å spacing. The steric (van der Waals) and electrostatic (coulombic) fields were calculated for each molecule at each mesh point using a sp³ carbon probe with +1.0 charge. Any calculated steric and electrostatic energies greater than 30 kcal/mol were truncated to this value.

HPLC/MS/MS Analysis. The HPLC-MS system consisted of a Hewlett-Packard 1100 quaternary pump with membrane degasser (Hewlett Packard, Palo Alto, CA), a LEAP CTC Pal liquid handler (SpectraLab Scientific Inc., Toronto, ON, Canada), and a PE Sciex API 2000 mass spectrometer (Applied Biosystems, Foster City, CA). A 10-μl sample was injected onto a C₁₈ HPLC column (BetaMax Neutral 20 × 2.1 mm, Thermo Electron Corporation, Waltham, CA) and eluted by a mobile phase composed of 90% solvent A (95% H₂O-5% acetonitrile with 0.1% formic acid) and 10% solvent B (5% H₂O-95% acetonitrile with 0.1% formic acid). The peak areas of all the analytes and internal standard were quantitated using Analyst 1.4 (MDS Sciex, Concord, ON, Canada).

Data Analysis. Data are expressed as means ± S.D. of transport values obtained in three wells or filter inserts. Data are representative of a minimum of two experiments carried out on different days on different batches of cells.

Results and Discussion

ABCC2/MRP2-Mediated Transport in Caco-2 Monolayer and Structures of Test Compounds. Caco-2 cell monolayers grown on polycarbonate filters have been reported to express transporters that mimic human intestinal epithelial cells (Taipalensuu et al., 2001). This cell line has been widely used as an in vitro model to elucidate the pathways by which drugs permeate the intestinal mucosa, as well as the structure-transport relationship for efflux transporters (Walle et al., 1999; Artursson et al., 2001; Engman et al., 2001). In this study, a series of substituted biphenyl heterocyclic compounds were selected for investigation of their interactions with ABCC2/MRP2 transporter. As indicated in Table 1, the differences between the test compounds lie in their biphenyl portion, specifically the varying substitution groups at the 2-, 2′-, 3-, and 3′-positions of the biphenyl rings. In total, 26 compounds, which covered a broad spectrum of mono- and disubstitution patterns including halogens (fluorine, chlorine, and bromine) and methyl, trifluoromethyl, hydroxy, methoxy, and ethoxy groups, were included in our investigation. Their chemical structures and IUPAC names for the biphenyl substitution are provided in Table 1.

First, we evaluated the polarized transport of the compounds in Caco-2 cell monolayers to examine the involvement of efflux transporters. Twelve of the 26 compounds were effluxed by Caco-2 monolayers with the ratio of permeability B to A versus A to B from 8 to 50 (Table 1). The efflux by Caco-2 monolayers was offset by adding 25 μM MK571, an MRP antagonist and specific inhibitor (Table 1). However, neither verapamil (100 μM) nor novobiocin (100 μM), the P-gp or breast cancer resistance protein inhibitor (Gao et al., 2001; Yang et al., 2003; Shiozawa et al., 2004), was able to significantly alter the polarized transport by Caco-2 monolayer cells (data not shown). These results suggested that the polarized transport of this class of compounds in Caco-2 cells was mediated, at least in part, by MRPs. To minimize the potential effects of other efflux transporters, the transport experiments were also conducted in MDCK-MRP2 cells. However, the observed polarized transport in Caco-2 cells was not evident in MDCK-MRP2 monolayers, possibly because of the poor (∼3 × 10^-6 cm/s) apparent permeability for this set of compounds (data not shown). These results were not unexpected but rather were consistent with the published report that it was difficult to demonstrate the vectorial transport of ABCC2/MRP2 substrates due to their poor penetration across the basolateral membrane (Evers et al., 1998; Sasaki et al., 2002). Therefore, in consideration of the technical involvement of accessing the double-transfected MDCK cell lines, we regarded the Caco-2 cell line as a valuable tool for investigating the transport and inhibition properties of our test compounds. Similar application of Caco-2 cells has been reported by Letschert et al. (2005) and Matsushima et al. (2005). In fact, the mRNAs of MRP isoforms 1 through 6 (ABCC1-6) have been detected in the human intestine and Caco-2 cells (Prime-Chapman et al., 2004). More specifically, the ABCC2/MRP2 and ABCC3/MRP3 isoforms are expressed extensively in Caco-2 cells, whereas ABCC1/MRP1 and ABCC5/MRP5 expression is minimal (Hirohashi et al., 2000). Different isoforms are also differentiated by their selective expressions on disparate phases of the cell membrane. For example, ABCC2/MRP2 and ABCC3/MRP3 are, respectively, expressed on brush-border and basolateral membranes in Caco-2 cells (Rost et al., 2002). Therefore, the differential pharmacological sensitivity of apical efflux could provide a tool for dissecting the functional roles of the MRP isoforms.

Because of the specific localization of the MRP family, it is expected that ABCC2/MRP2 is the only MRP isoform involved in transport of the biphenyl compounds under study in Caco-2 cells. Unlike P-glycoprotein, which interacts with hydrophobic drugs, ABCC2/MRP2 protein transports hydrophilic molecules with or without acidic conjugates, such as GSH conjugates (Leier et al., 1996; Rappa et al., 1997). As an organic anion transporter, the ABCC/MRP family actively transports a broad range of substrates including glutathione-S-conjugates, glucuronide conjugates, bilirubin, organic anions, and conjugated drugs or their metabolites (Jedlitschky et al., 1997; Madon et al., 1997). However, for conjugates or compounds with GSH-dependent transport structure-activity relationships with ABCC/MRP are difficult to identify (Boumendjel et al., 2005). In the current study, Caco-2 monolayer assays were performed under such conditions to suppress formation of cotransport of glutathione or other conjugates by ABCC2/MRP2. SARs for ABCC2/MRP2 and biphenyl compound interaction were elucidated from the variation of the substituted group on the biphenyl ring with altered physicochemical properties and three-dimensional structure.

Correlation between Physicochemical Properties of Biphenyl-Substituted Heterocycles and Efflux or Inhibition of ABCC2/MRP2. As with many ABC transporters, ABCC2/MRP2 can be functionally inhibited by a wide range of structurally and pharmacologically unrelated compounds (Payen et al., 2000; Asakura et al., 2004). Substrate inhibition has also been observed for ABCC2/MRP2 (Sugie et al., 2004). To investigate the inhibitory profile of ABCC2/MRP2-mediated efflux, an inhibition study was performed in MDCK-MRP2 cells. The inhibition of ABCC2/MRP2 activity was achieved by using calcein as a model substrate. Nonfluorescent calcein-AM is a lipophilic, highly cell-permeable ester and a substrate for both P-gp and ABCC2/MRP2. The ester bond is rapidly cleaved by intracellular nonspecific esterases, generating highly fluorescent calcein, a specific ABCC2/MRP2 substrate with poor cell permeability (Legrand et al., 1998). The inhibition of calcein efflux was measured in the absence or presence of selected biphenyl compounds (80 μM) or MK571 (an ABCC2/MRP2 inhibitor, 25 μM). Compound inhibition of ABCC2/MRP2 was compared with and quantified by MK571 inhibition.

In general, for the compounds that were not effluxed by the Caco-2 monolayers, we observed no inhibition of the ABCC2/MRP2-mediated calcein efflux. In contrast, the compounds effluxed by the Caco-2 monolayers also demonstrated inhibition of the MRP2-mediated calcein efflux, ranging from weak to strong inhibition (10%-100%) compared with MK571 (Fig. 1). These results reveal that compounds being effluxed in Caco-2 cells were also capable of inhibiting ABCC2/MRP2-mediated efflux in MDCK-MRP2 cells, which suggests overlapping interactions between substrate and inhibitor binding to the ABCC2/MRP2 protein.

Several authors reported application of the CoMFA approach to understand the importance of molecular fields around the molecules in their interaction with ABCC2/MRP2. Likewise, Nakagome et al. (2003) reported that ligand recognition of canalicular multispecific organic anion transporter/MRP2 is achieved through interactions in two hydrophobic sites and two electrostatically positive and/or hydrogen bonding acceptor sites. Hirono et al. (2005) reported a similar model in which interaction between the ligands and MRP2 occurred in two hydrophobic and two electrostatically positive primary binding sites. CoMFA studies have been successfully applied to predict activity of compounds that inhibit the multidrug transporter P-glycoprotein for a series of anthranilamide derivatives (Labrie et al., 2006), 2,4,5- and 2,3,4,5-substituted imidazoles (Kim, 2001), natural and synthetic coumarins (Raad et al., 2006) and tariquadar analogs (Globisch et al., 2006). We attempted to apply the CoMFA approach to our dataset but were not able to build a statistically significant model (q² = 0.30 using 6 components for 26 compounds). Therefore, CoMFA, although it is a complex algorithm proven to be useful in elucidating protein-ligand interactions of many biological systems, was not able to reveal structural features sensitive to the transport properties we discovered for this set of compounds.

The physicochemical properties, including hydrophobicity, hydrogen bond donors, hydrogen bond acceptors, and polar groups, are common factors considered when one is evaluating the interactions between transporters and small molecules (Hirono et al., 2005; van Zanden et al., 2005). A simple model that relates the Wiener topological index and MRP1 inhibitory activity has been reported (Lather and Madan, 2005). The accuracy of prediction of such a model was found to be 88%. The Wiener topological index is proportional to the size of the molecular graph, a property that reflects the size of the molecule. In our analysis we used molecular weight as a parameter that describes molecular size but did not find any significant correlation with the MRP2-related activities. Furthermore, some structural elements for ABCC1/MRP1 have also been summarized. For example, aromatic or heteroaromatic moieties and carbonyl groups are frequently observed (Boumendjel et al., 2005). Because of the overlap of substrates and/or inhibitors between MRP1 and MRP2, it is assumed that the SAR for ABCC2/MRP2 could follow principles similar to those for ABCC1/MRP1 (Wang and Johnson, 2003).

In the current study, we attempted to account for the physiochemical properties and three-dimensional steric and electrostatic effects and investigated their relationship with the ABCC2/MRP2 interactions. For these purposes, the physicochemical parameters for this group of biphenyl substituted heterocycles, including ClogP, the number of hydrogen bond donors, the number of hydrogen bond acceptors. and the tPSA, were evaluated and correlated to ABCC2/MRP2 transport and inhibition. In addition to calculated properties, LogD values were measured to add one more dimension to the property pool to characterize the lipid-water distribution of the compounds under physiological pH. In general, for the biphenyl-substituted heterocycles with a variety of substituting groups on the biphenyl ring that were tested, ClogP ranged from 0.89 to 3.37, LogD (pH 7.4) ranged from 0.66 to 3.79, the number of hydrogen bond donors ranged from 5 to 7, the number of hydrogen bond acceptors ranged from 4 to 5, and the tPSA ranged from 125 to 156. Although most of the properties fell into the normal range for desirable adsorption, distribution, metabolism, and excretion properties (Lipinski et al., 2001), the tPSA appeared too high for passive transcellular absorption (Palm et al., 1997). Interestingly this finding is in keeping with the reduced transport in MDCK-MRP2 monolayers compared with Caco-2 cells we have observed, which also suggested poor permeability for this class of compounds.

To our surprise, no correlations were found between the physicochemical descriptors (ClogP, LogD, hydrogen bond donor/acceptor, and tPSA) and the ABCC2/MRP2 efflux in Caco-2 monolayers (substrate) (Fig. 1) or the inhibition of calcein efflux in MDCK-MRP2 cell (inhibitor) (Fig. 2). This finding implied that the nature and the strength of the interactions between ABCC2/MRP2 and this class of compounds were not determined by any overall molecular properties but possibly by certain other parameters that are more specific to the three-dimensional conformation of the molecules.

Correlation between Three-Dimensional Molecular Conformations and the Efflux/Inhibition Properties of ABCC2/MRP2. In the preliminary studies, it was observed that the inhibition of the ABCC2/MRP2 transporter by biphenyl-substituted heterocycles is sensitive to the substitution patterns at the 2- or 2′ (ortho)-positions. Further structure-activity analysis suggested that the inhibition or transport properties mediated by ABCC2/MRP2 is highly sensitive to the bulkiness of the ortho-substituents on the biphenyl ring system. It was suspected that the increase in bulkiness of the 2- or 2′-substitutions resulted in increased torsion angles between the two phenyl rings, and the conformational change was recognized by ABCC2/MRP2 in a favorable manner, leading to the strengthened three-dimensional interaction between the transporter and the test compounds. The three-dimensional molecular conformations, derived from electrostatic interactions and steric repulsions between the ortho-substitutions, seems to be the determining factor for ABCC2/MRP2 recognition and interaction.

To validate this hypothesis, we conducted high-level ab initio calculations during which the global energy minimum was sought and the geometries of the test compounds were fully optimized. As shown in Fig. 3, it was found that for the biphenyl ring system in the aforementioned scaffold, the torsion angles increased upon increasing bulkiness of the ortho-substituents in the following order: hydrogen < fluorine < hydroxy < chloride < methyl (Fig. 3). The change in torsional space is largely due to the steric repulsion between the two ortho-substituents from the opposing phenyl rings. Although the torsion of the plain biphenyl tends to be skewed to possibly maximize the π-conjugation, the bulky ortho-substituents force the biphenyl into more orthogonal conformations as we have seen here.

Both the efflux in Caco-2 monolayer and the inhibition of calcein efflux in the MDCK gene-transfected cells increased consistently with the increasing torsion angles (Fig. 4). On the basis of the range of torsion angles, the compounds were classified into three groups, with average torsions at 40, 56, and 84 degrees, respectively. Group I compounds are not effluxed in Caco-2 monolayers, and they are not ABCC2/MRP2 inhibitors. Group II and III compounds are ABCC2/MRP2 substrates with MK571-inhibitable efflux in Caco-2 cells. Furthermore, although a moderate inhibition (10-40%) was observed with group II compounds, strong inhibition of calcein efflux (100% compared with the maximal inhibition of MK571) was observed with group III compounds (Fig. 4). The results suggest that the torsion angles (i.e., steric effect) between the biphenyl rings of the test compounds is the key factor in determining the interaction with ABCC2/MRP2. The compounds with the smallest ortho-substitutions (hydrogen, fluorine, and oxygen) and thus the least torsion angle of the biphenyl (37-45°) are neither substrates nor inhibitors of ABCC2/MRP2 (Table 2). The ABCC2/MRP2 transporter interaction increases as the steric bulk of the ortho-substitution increases. When the 2-position is occupied by a methyl group, the torsion angle becomes moderate (54-65°), and the compound behaves as an ABCC2/MRP2 substrate, as well as a mild inhibitor (10-40% inhibition of transport when compared with MK571). For the 2,2′-dimethyl-substituted bi-phenyls, the torsion angles are further increased (78-87°), and the strongest inhibition of ABCC2/MRP2 transport was observed (Table 2). These results are in keeping with the report from van Zanden et al. (2005) in a study of a large number of flavonoids that the inhibition of MRP1 is related to the dihedral angle between the phenyl and the chromenone core, which suggests that the steric effect is a critical factor for ABCC1/MRP1 inhibition (van Zanden et al., 2005).

Several inhibition mechanisms for ABCC/MRPs have been proposed in the literature. The inhibitory compounds might interfere with ABCC/MRPs through drug binding, ATP binding, ATP hydrolysis, drug transport, or ADP release (Boumendjel, 2003). ABCC2/MRP2 inhibitors, such as probenecid and azithromycin, are also transported by ABCC2/MRP2, which suggests that their inhibition may be through competition for drug-binding sites on ABCC2/MRP2 (Chen et al., 2003; Sugie et al., 2004). The compounds from groups II and III possess both an inhibitory and a transport interaction with ABCC2/MRP2, suggesting that the interactions between ABCC2/MRP2 and the biphenyl compounds occur at a drug-binding site. Moreover, to further understand the transport kinetics for group II and III compounds, the concentration-dependent transport in Caco-2 cells was investigated. As shown in Fig. 5, the active efflux rate in Caco-2 cells increased with increasing concentrations of the test compound between 0.5 and 20 μM but plateaued when the concentration exceeded 20 μM. The active efflux rate followed Michaelis-Menten kinetics with the one-site binding model being the best fit:

where V is the apparent linear initial rate, [S] is the initial substrate concentration, V_max is the maximum transport rate, and K_m is the Michaelis-Menten constant. The estimated K_m values for group II and III compounds transported by Caco-2 were 7.63 ± 1.07 and 9.13 ± 2.73 μM, respectively. There were no statistically significant differences in the K_m and V_max values between the group II and group III compounds. These results suggest that the steric effect is more prominent on the inhibitory action than on the transport activity of ABCC2/MRP2. This is consistent with a recent report (Hirono et al., 2005) that the steric effect contributes more (63%) than the electrostatic (33%) and lipophilicity (4%) effects on the ligand binding affinity to rat Abcc2/Mrp2. It was also reported that there are two electrostatically positive binding regions and two primary hydrophobic binding regions on the rat Abcc2/Mrp2. Sequence identity between the human and rat MRP2 is 72%. If the human ABCC2/MRP2 has a binding pocket comparable to that of rat Abcc2/Mrp2, it is possible that the two branched polar groups, 3-carbamoyl and 4-urea on pyrazole, interact with the two primarily positive regions deep in the binding site, and the biphenyl ring interacts with the two hydrophobic binding regions. Further studies are needed to confirm whether the 4-(3-carbmoyl-4-urea)-pyrazole, which was kept constant in this study, is involved in the recognition of the ABCC2/MRP2 binding site.

In conclusion, for a series of biphenyl compounds we tested, the interactions between ABCC2/MRP2 and the small molecules do not correlate with a variety of physicochemical properties. However, the ABCC2/MPR2 interactions do correlate with an increase in the steric bulkiness of ortho-substitutions, indicating that the steric interaction is the predominant factor for understanding ABCC2/MRP2 interactions in this study. These results could help define the three-dimensional structural requirements of ABCC2/MRP2 interactions and provide SARs to support drug discovery.