During pregnancy, mean plasma levels of free and conjugated estrogens increase dramatically with gestational age (Levitz and Young, 1977). Additionally, 3-hydroxy glucuronidation of the steroid A ring appears to produce a unique terminal product in estriol metabolism, in contrast to the intermediary metabolism of other estriol conjugates (Levitz et al., 1984). We have postulated that increased levels of cholestatic estrogen glucuronides, such as β-estradiol 17-(β-d-glucuronide) (E₂17G) and estriol 16α-(β-d-glucuronide) (E₃16G), in pregnancy contribute to the decrease in bile secretory function observed in normal pregnancy. E₂17G is a prototype cholestatic estrogen D-ring glucuronide in the rat model, whereas β-estradiol 3-(β-d-glucuronide) (E₂3G) is a choleretic A-ring glucuronide (Meyers et al., 1981). The mechanistic basis for their differential effects on bile flow is unknown, but likely involves direct interactions with hepatic canalicular transporters. E₂17G is a well established Mrp2 probe substrate (Gerk and Vore, 2002), but it has not been determined whether E₂3G is an Mrp2 substrate. E₂3G inhibits 2,4-dinitrophenol-glutathione transport in rat canalicular membrane vesicles (Vore et al., 1996), and its biliary excretion in Mrp2-deficient Eisai hyperbilirubinemic rats is diminished (Takikawa et al., 1996), suggesting that E₂3G is an Mrp2 substrate.

Mrp2-mediated transport of E₂17G is required for its cholestatic activity, although the mechanism is not clear. E₂17G does not induce cholestasis in TR^- rats deficient in Mrp2, even though very high concentrations of E₂17G were used, sufficient to achieve equal biliary concentrations of E₂17G in wild-type Wistar and TR^- rats (Huang et al., 2000). These data imply that the interaction between Mrp2 and E₂17G is essential for cholestasis. We therefore questioned whether the inability of E₂3G to induce cholestasis is because it is not an Mrp2 substrate, or because its interaction with Mrp2 differs significantly from the interaction between Mrp2 and E₂17G.

E₂17G is transported by several ATP-binding cassette (ABC) transporters, including MDR1 (Huang et al., 1998), MRP1 (Keppler et al., 1997), MRP2 (Keppler et al., 1997), MRP3 (Hirohashi et al., 1999), MRP4 (Zelcer et al., 2003b), MRP7 (Chen et al., 2003), and ABCG2 (breast cancer resistance protein; or mitoxantrone resistance transporter) (Suzuki et al., 2003), but not MRP6 (Belinsky et al., 2002). Additionally, E₂17G is transported by the rat organic anion-transporting polypeptides 1 to 4 (Hagenbuch and Meier, 2003). By contrast, little is known of the transport mechanisms for E₂3G. E₂3G is a poor inhibitor of MRP1-mediated transport of leukotriene C₄ and E₂17G (Loe et al., 1996), and is a low-affinity inhibitor of both MRP4 and MRP7, with IC₅₀ values near 100 μM (Chen et al., 2003; Zelcer et al., 2003b). E₂3G (100 μM) did not inhibit MDR1-mediated transport of E₂17G, suggesting that E₂3G is not an MDR1 substrate (Huang et al., 1998). E₂3G may be a substrate for rat organic anion-transporting polypeptide 1, since it inhibits E₂17G transport with a K_i of 9.7 ± 0.7 μM (Kanai et al., 1996). Its substrate specificity for other transporters is unknown.

The purpose of this investigation was 1) to determine whether Mrp2 transports E₂3G, 2) to determine whether rat Mrp2-mediated E₂3G transport differs mechanistically from that of E₂17G, and 3) to obtain estimates of the kinetic parameters describing these interactions. The present research indicates that E₂3G, like E₂17G, is a high-affinity Mrp2 transport substrate. However, unlike E₂3G, E₂17G transport exhibits marked positive cooperativity, indicating that E₂17G, but not E₂3G, activates an allosteric site in Mrp2.

Materials and Methods

³H-E₂17G (40-45 Ci/mmol) and ³H-E₂3G (53-57 Ci/mmol) were obtained from PerkinElmer Life and Analytical Sciences (Boston, MA). Unlabeled estrogen conjugates (Fig. 1A) and all other reagents were obtained from Sigma-Aldrich (St. Louis, MO).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

A, structures of estrogen conjugates. B, Western blot of Mrp2 expressed in insect cells. Abbreviations are as indicated in the text.

The rat MRP2 plasmid was obtained from Dr. Peter Meier, and the full coding region was inserted into the pFastBac1 vector (Invitrogen, Carlsbad, CA) with a nine-amino acid carboxy-terminal hemagglutinin tag and recombined in DH10Bac Escherichia coli cells (Invitrogen) containing the baculovirus genome. A comparable carboxy-terminal 6×-histidine tag does not inhibit Mrp2 function (Hagmann et al., 2002). As a background control, the empty vector was also recombined as above, but without an insert. The recombinant bacmids were amplified, sequenced (Elim Biopharmaceuticals, Hayward, CA) to confirm the presence (or absence) of Mrp2, and transfected using CellFectin (Invitrogen) into Sf9 insect cells. The supernatant containing the recombinant baculovirus was harvested, amplified, and titered by a viral plaque assay. Sf9 cells (5 × 10⁸ cells) in suspension culture were infected (multiplicity of infection = 3), and 64 to 68 h later, cell membranes for transport experiments were harvested by layering on 38% sucrose and collecting the layer at the buffer-sucrose interface (Ito et al., 2001c). Membranes were vesiculated, snap frozen in liquid nitrogen, stored at -80°C, and designated as either Mrp2- or empty virus (EV)-infected cell membranes. Protein concentrations were determined by a modification of the method described using bovine serum albumin as a standard (Lowry et al., 1951).

Expression of rat Mrp2 was determined by Western blotting using 0.5 μg of sucrose-fractionated membrane protein. Proteins were denatured in the presence of sodium lauryl sulfate at 37°C for 30 min before loading onto an 8% Tris/glycine polyacrylamide Novex precast gel (Invitrogen), separated by standard electrophoresis, and transferred onto Polynitran nitrocellulose membranes (Schleicher & Schuell, Keene, NH). Membranes were blocked using 5% nonfat milk at room temperature for 1 h, and binding of primary antibody [mouse anti-human MRP2 (M₂III-6); Alexis Biochemicals, San Diego, CA) and secondary antibody (sheep anti-mouse, horseradish peroxidase-conjugated; Amersham Biosciences Inc., Piscataway, NJ) was performed in 5% nonfat milk at room temperature for 1 h. Chemiluminescence detection was done using ECL-Plus (Amersham Biosciences Inc.) and exposure to Biomax MR film (Kodak, Rochester, NY).

Transport experiments were performed in a Tris-sucrose buffer (Ito et al., 2001c), containing 5 mM ATP or AMP, 10 mM MgCl₂, 10 mM phosphocreatine, 100 μg/ml creatine phosphokinase, and unlabeled estrogen conjugates in either dimethyl sulfoxide (0.5%) or 10:4:1 Tris-sucrose buffer/propylene glycol/ethanol (2%) as vehicles. Preliminary studies showed that the choice of vehicle had no effect on transport. ATP-dependent transport of ³H-E₂17G or ³H-E₂3G into membrane vesicles (10 μg/20 μl) was measured in incubations at 37°C for 2 to 5 min, transport was stopped with 3.5 ml of ice-cold stop buffer (Ito et al., 2001c), and the mixture was quickly filtered onto Durapore 0.4-μm filters (Millipore Corporation, Bedford, MA). The filters were selected due to their minimal binding of E₂17G at low (90 nM) or high (100 μM) concentrations. The tubes and filters were rinsed as described (Boyer and Meier, 1990). ³H collected on the filters was detected by liquid scintillation counting using scintillation counting cocktail (Bio-Safe II; Research Products International Corp., Mt. Prospect, IL).

Transport corrected for that in the presence of AMP was termed ATP-dependent transport, whereas that corrected for background (EV) transport was termed Mrp2-mediated transport. Nonlinear regression was performed on saturation data by fitting the data to the Hill equation and weighting as indicated in figure legends. Other curves were unweighted for linear or nonlinear regression using Prism version 4 computer software (GraphPad Software Inc., San Diego, CA) for fitting as indicated in figure legends. Remaining data were analyzed by one-way analysis of variance (α = 0.05) followed by Dunnett's multiple comparison test (Prism).

Results

The expected underglycosylated rat Mrp2 protein (<190 kDa) was detected in Mrp2-expressing Sf9 cell membrane vesicles by means of Western analysis (Fig. 1B). Mrp2 was undetectable in the EV membrane vesicles.

The basic transport characteristics for E₂3G and E₂17G are shown in Fig. 2. To ensure that transport properties and kinetic values were attributable to Mrp2, transport studies were carried out in sucrose-fractionated Sf9 plasma membranes, and transport in Mrp2-expressing membranes was corrected for that in membranes from cells transfected with the EV. Transport of both compounds was linear to at least 5 min and was much greater in membrane vesicles expressing Mrp2 versus EV (Fig. 2, A and B). Similarly, transport of both compounds occurred into an osmotically sensitive space with a y-intercept close to zero, indicating that observed activity was mainly transport and binding was minimal (Fig. 2, C and D). Additionally, transport of both compounds was linear with respect to protein (data not shown). Mrp2-mediated ATP-dependent transport of E₂3G (Fig. 2E) was fit to a single apparent saturable transport site with S₅₀ = 55.7 (95% CI 44.2-70.1) μM and V_max = 326 (95% CI 286-366) pmol·mg^-1·min^-1, and a Hill coefficient of 0.88 (95% CI 0.85-0.91). Figure 2F shows the combined results of three independent saturation experiments with E₂17G. Fitting the data to simple linear (nonsaturable), one-site and two-site Michaelis-Menten equations and the Hill equations yielded the lowest Akaike information criteria (Prism 4.0) for the fit to the Hill equation, indicating that the fit to the Hill equation was best supported by the data. Due to the limited solubility of E₂17G, we were unable to fully saturate the transport to obtain unique estimates for V_max and K_m; however, for Mrp2-mediated transport, the data were best fit to an unweighted Hill equation with a Hill slope of 1.5 ± 0.2. These data suggest positive cooperativity for E₂17G transport. Notably, E₂17G association with either Mrp2 or EV membrane vesicles in the absence of ATP was not influenced by E₂17G concentration (0.02-160 μM) or by any other agent used here. Also, ATP-dependent transport of E₂3G and E₂17G in EV membranes followed simple Michaelis-Menten kinetics (Hill slope = 1), with K_m values of 221 (95% CI 100-490) μM and 56.2 (95% CI 33.2-79.3) μM, and V_max values of 132 (95% CI 40-225) and 135 (95% CI 104-166) pmol·mg^-1·min^-1, respectively. Finally, E₂3G was more soluble than E₂17G under identical experimental conditions, avoiding solubility problems that occur at high E₂17G concentrations.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Transport of E₂3G (A, C, and E) and E₂17G (B, D, and F) in sucrose-fractionated Sf9 membrane vesicles. Each data point represents mean ± S.D. from triplicate determinations. Filled or empty symbols represent data obtained from Mrp2-expressing or EV vesicles, respectively. A and B, E₂3G (30 nM) and E₂17G (77 nM) transport as a function of time. C and D, effect of sucrose (osmolarity) on E₂3G (30 nM) and E₂17G (77 nM) transport. Squares or circles represent data obtained in the presence of ATP or AMP, respectively. E, saturation of E₂3G transport. Nonlinear regression was performed fitting the data from two independent experiments (triplicate determinations) to the Hill equation, with a weighting factor of 1/y² (radiotracer concentrations 35-110 nM). F, data from three independent E₂17G saturation experiments, each determined in triplicate (radiotracer concentrations 43-96 nM). The data were fitted to the Hill equation (unweighted, Hill coefficient = 1.5).

To better understand the structure-activity relationships among relevant estrogen glucuronides (Fig. 1A) and Mrp2, we compared E₂17G and E₂3G transport, and the effects of E₃16G, a weak cholestatic glucuronide (Meyers et al., 1981) and β-estradiol 3-sulfate 17-(β-d-glucuronide) (E₂3SO₄17G), a naturally occurring choleretic biliary estradiol metabolite (Meyers et al., 1980; Takikawa et al., 1996), on their transport. The results for E₂17G transport are shown in Fig. 3. Again, unique fits were not obtained for the E₂17G saturation data due to its limited solubility (Fig. 2F), but the data were best fit to the Hill equation with a Hill coefficient of 1.5. E₂3G (15 μM, Fig. 3A) shifted the entire E₂17G transport versus concentration curve (Hill coefficient fixed at 1.5) to the right, consistent with competitive inhibition for transport. E₂3G (IC₅₀ = 14.2 μM, 95% CI 7.8-25.7 μM, Fig. 3B) was a complete and high-affinity inhibitor of E₂17G transport. Additionally, E₂3SO₄17G (5-50 μM) potently and completely (50 μM) inhibited transport of E₂17G (Fig. 3C). Association or diffusion of E₂17G into either Mrp2 or EV membranes in the absence of ATP did not change with the concentrations tested. Concentrations of ³H-E₂17G above 250 μM increased the radioactivity collected on the filters in the absence of ATP, suggesting precipitation from solution. Furthermore, both MDR1-mediated E₂17G transport in Sf9 membranes (Huang et al., 1998) and E₂17G transport by EV membranes in the present studies adhered to Michaelis-Menten kinetics. The data are therefore not consistent with nonspecific effects of E₂17G on the cell membrane.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

E₂17G transport and effects of estrogen glucuronides on E₂17G transport. Each data point represents mean ± S.D. of triplicate determinations from representative experiments. A, competitive displacement of E₂17G (111 nM) transport by unlabeled E₂3G (15 μM). B, inhibition of E₂17G (63 nM) transport by E₂3G. Data (unweighted) were fitted to a one-site competitive binding equation. C, inhibition of E₂17G (85 nM) transport by E₂3SO₄17G. Comparisons were made by Dunnett's test following one-way analysis of variance. ★ indicates p < 0.05 versus 0 μM E₂3SO₄17G.

Similar studies on E₂3G transport are shown in Fig. 4. E₂17G only partially inhibited (53%) E₂3G transport with an IC₅₀ = 33.4 μM (95% CI 26.4-45.2 μM; Fig. 4A). In contrast, E₃16G was a complete and high-affinity inhibitor of E₂3G transport (IC₅₀ = 2.23 μM, 95% CI 2.07-2.40 μM; Fig. 4B). Competitive inhibition studies were performed with E₂17G (5-125 μM) and E₂3SO₄17G (5 μM) (Fig. 4, C and D), and the nonlinear regression fit to the Hill equation (Table 1). Neither E₂17G nor E₂3SO₄17G had a significant effect on the V_max for E₂3G transport; as a result, V_max was shared for all the data sets. However, both E₂17G and E₂3SO₄17G significantly increased the apparent S₅₀ for E₂3G transport, indicative of competitive inhibition at the transport site. E₂3SO₄17G and the highest E₂17G concentration also slightly increased the Hill coefficients closer to unity.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

E₂3G transport and effects of estrogen glucuronides on E₂3G transport. Each data point represents mean ± S.D. of triplicate determinations from representative experiments. A and B, inhibition of E₂3G (49 nM) transport by E₂17G and E₃16G. Data for A and B were fitted to a one-site competitive binding equation (unweighted). C and D, competitive displacement of E₂3G (35 nM) transport (squares) by unlabeled E₂17G (C) and E₂3SO₄17G (5 μM; D). Nonlinear regression was performed fitting the data to the Hill equation, with a weighting factor of 1/y² as discussed in the text.

Discussion

The purpose of the present studies was to determine whether E₂3G is a substrate of Mrp2, to determine whether the noncholestatic E₂3G and the cholestatic E₂17G exhibited similar Mrp2-mediated transport properties, and to obtain estimates of the kinetic parameters for their transport. The present data clearly demonstrate that E₂3G is an excellent substrate for Mrp2, having an S₅₀ of 55.7 μM and a V_max of 326 pmol/mg/min. Thus, although several permissive multidrug resistance transporters such as P-glycoprotein and the MRPs transport E₂17G, E₂3G is most likely not a substrate for MDR1 and MRP1 (Loe et al., 1996; Huang et al., 1998), as mentioned previously. This suggests a potentially important difference in the substrate specificities of MRP1 and Mrp2/MRP2. Thus, if E₂3G is transported by Mrp2 but not other MRPs or MDR1, E₂3G would be a more selective probe than E₂17G for MRP2-mediated transport. The higher water solubility of E₂3G also increases its experimental utility. The data also establish Mrp2 as a hepatic efflux mechanism for E₂3G, and are consistent with data demonstrating the choleretic activity of E₂3G (Meyers et al., 1980) and its biliary excretion in normal rats (Takikawa et al., 1996). Although E₂3G is not overtly toxic, its biliary excretion is important due to the complex pattern of interconversion among the estrogen metabolites (Levitz and Young, 1977; Levitz et al., 1984). During the preparation of this article, ethinyl-estradiol-3-glucuronide, a structural analog of E₂3G, was also shown to be transported by human MRP2 but not MRP1 (Chu et al., 2004), consistent with the present findings.

The present data also indicate that the interactions between E₂3G and E₂17G and Mrp2 are very distinct. The Hill coefficient for Mrp2-mediated E₂3G transport was slightly, but significantly less than unity (0.88). Although the significance of this negative cooperativity is not known, it could contribute to the somewhat higher S₅₀ (56 μM) obtained for E₂3G transport versus its IC₅₀ to inhibit E₂17G transport (14.2 μM). In contrast, E₂17G transport occurred with significant positive cooperativity (Hill coefficient of 1.5). Our preliminary data also indicate a Hill coefficient of 2 for human MRP2-mediated E₂17G transport (Gerk et al., 2003). Studies reporting K_m values of 3 to 8 μM for rat Mrp2 expressed in Sf9 cells (Ito et al., 2001b,c) or in HEK-293 cells (Cui et al., 1999) did not examine the range of E₂17G concentrations used here. At higher concentrations, E₂17G exhibited nonclassical transport kinetics, consistent with allosteric activation of its transport, as recently proposed for human MRP2 (Bodo et al., 2003; Gerk et al., 2003; Zelcer et al., 2003a). Sf9 cells clearly possess endogenous transporter(s) for both E₂17G and E₂3G, with K_m estimates of 56.2 and 221 μM, respectively. It was necessary to correct for this endogenous transport activity, which became significant at high substrate concentrations, to obtain accurate estimates of the distinct kinetic parameters for Mrp2-mediated transport of E₂17G and E₂3G.

Evidence for two sites for E₂17G transport was also observed in the pattern of inhibition by other estrogen glucuronides. E₂3G and E₂3SO₄17G completely and potently inhibited E₂17G transport. Also, transport of E₂3G was potently inhibited by E₃16G (IC₅₀ 2.23 μM) and competitively inhibited by E₂3SO₄17G. Conversely, however, E₂17G did not completely inhibit E₂3G transport within the limits of E₂17G solubility. The data are consistent with a model in which E₂3G and E₂17G compete for binding to the Mrp2 transport site, as evidenced by the increased S₅₀ for E₂3G transport (Fig. 4C; Table 1) and the rightward shift of E₂17G transport in the presence of E₂3G (Fig. 3A). E₂3G transport clearly does not demonstrate positive cooperativity, and therefore, E₂3G does not activate the allosteric site. The slight negative cooperativity of E₂3G transport (Hill coefficient 0.88) may reflect antagonism rather than activation of the allosteric site. The data in Fig. 4, A and C, support a model with an allosteric site having low affinity for E₂17G (K_m >100 μM). Accordingly, low concentrations of unlabeled E₂17G competitively inhibit the transport of low concentrations of ³H-E₂3G or ³H-E₂17G, but as E₂17G concentrations increase, activation at the allosteric site would occur concurrently with competition for the transport site. Such a model with concurrent activation and competition explains the positively cooperative transport of E₂17G and the inability of E₂17G to completely inhibit E₂3G transport, as well as the lack of further increase in S₅₀ for E₂3G transport with increasing E₂17G concentrations (Table 1). The ability of E₃16G and E₂3SO₄17G to completely and potently inhibit transport also provides information on the nature of the transport site versus the allosteric site. The fact that E₂3SO₄17G inhibits completely indicates that sulfation of the A-ring of E₂17G abolished its ability to activate the allosteric site and implies that a phenolic A-ring is important for binding to this site. However, E₃16G, which like E₂17G is a glucuronide conjugate of the steroid D-ring with a phenolic A-ring, was also a potent and complete competitive inhibitor of E₂3G transport (IC₅₀ 2.2 μM); these data are also consistent with E₃16G having negligible activity at the allosteric site. These latter data indicate that the phenolic A-ring on a steroid glucuronide alone is not sufficient to activate the allosteric site; the greater hydrophilicity of E₃16G and/or the different stereochemistry of the glucuronic acid at the 16α-OH of estriol versus the 17β-OH of estradiol are also likely critical factors that decrease activation of the allosteric site. Finally, these data indicate that the transport site is much more permissive than is the allosteric activation site.

The presence of more than one binding site on Mrp2 is consistent with findings with other ABC transporters that mediate efflux of xenobiotics and their conjugates. P-Glycoprotein has three proposed binding sites, including two transport sites and an allosteric site (Shapiro et al., 1999). The existence of this MDR1 allosteric site was recently confirmed (Maki et al., 2003), demonstrating that this allosteric site influences substrate translocation and its subsequent dissociation from MDR1. Separate binding sites for E₂17G and sulfinpyrazone have been postulated for human MRP2, based on substitutions at W1254 (Ito et al., 2001a). The W1254F MRP2 mutant retained E₂17G transport, but not leukotriene C₄ or methotrexate transport. These authors concluded that each MRP2 substrate interacts with a unique, but overlapping set of contacts in a multipartite substrate-binding pocket. Also, Evers et al. (2000) postulated two MRP2 transport sites with positive cooperativity to explain the ability of sulfinpyrazone and indomethacin to stimulate glutathione transport. Whether the allosteric site described in the present studies represents another drug transport site or a third allosteric site is not clear. To date, E₂17G appears to be the only Mrp2 substrate that is able to activate its own transport.

The present data raise important questions regarding the physiological, pharmacological, and toxicological consequences of the ability of E₂17G to allosterically activate Mrp2 transport activity. The data suggest that binding of high, cholestatic concentrations of E₂17G to Mrp2 could lead to a marked stimulation of its transport into the canaliculus. Administration of E₂17G leads to endocytic retrieval of Mrp2 and Bsep from the canalicular membrane, thus decreasing their ability to transport osmotically active solutes into bile and decreasing bile flow (Mottino et al., 2002; Crocenzi et al., 2003). A key question is whether E₂17G-induced activation of Mrp2 is causally related to E₂17G-induced internalization of Mrp2 and Bsep and cholestasis. Cell shrinkage has been shown to cause internalization of transporters and cholestasis (Haussinger et al., 2000), but at present, there is no evidence for a mechanistic link between Mrp2 activation and decreased cell volume sufficient to trigger such internalization. In the presence of Mrp2, E₂17G enhances activation of chloride channels and, in cells swollen with hypotonic media, causes cell shrinkage (Li and Weinman, 2002). Further studies are needed to determine whether E₂17G-induced activation of Mrp2 sufficient to cause Mrp2 and Bsep retrieval is preceded by shrinkage of hepatocytes.

The ability of estrogen glucuronides to serve as Mrp2 substrates and competitively inhibit Mrp2 transport may contribute to the decreased bile secretory function observed in normal pregnancy, and to intrahepatic cholestasis of pregnancy observed in some women. Although E₃16G has a lower cholestatic potency (29%) compared with E₂17G in animal models (Meyers et al., 1980), the concentrations of E₃16G (∼100 nM) in human plasma at term of uncomplicated pregnancy are much greater than those of E₂17G (14 nM) (Levitz and Young, 1977; Numazawa et al., 1979). The present data show that E₃16G is a potent inhibitor of Mrp2, with an IC₅₀ value near 2 μM. Increased concentrations of E₃16G during normal pregnancy may reach levels in the hepatocyte sufficient to inhibit MRP2. Furthermore, both E₂17G and E₂3G inhibit glutathione biliary excretion (Mottino et al., 2003). In intrahepatic cholestasis of pregnancy, a vicious cycle may occur in which the accumulation of cholestatic estrogen glucuronides progressively inhibits MRP2 function by direct competitive inhibition as well as by triggering transporter retrieval, thus leading to further accumulation of cholestatic estrogen glucuronides and worsening of cholestasis.

In conclusion, these data provide direct evidence that E₂3G, like E₂17G, is transported by Mrp2, whereas E₂3SO₄17G and E₃16G are potent inhibitors of Mrp2 transport, consistent with their also being Mrp2 substrates. The interactions between Mrp2 and E₂17G are complex and require more than one distinct binding site. The data are consistent with a model in which E₂17G interacts with a transport site and an allosteric site that increases its own transport. E₂17G and E₂3G compete for binding to the transport site, but E₂3G does not activate the allosteric site. Mrp2 thus plays a major role in the biliary excretion of both cholestatic and noncholestatic estrogen glucuronides. Further studies are needed to identify the binding sites for E₂17G on Mrp2 and to determine whether there is a link between the binding of E₂17G to the allosteric site and its cholestatic activity.