Revolving Door Action of BCRP Facilitates or Controls the Efflux of Flavone Glucuronides from UGT1A9-Overexpressing HeLa Cells

Abstract

Cellular production of flavonoid glucuronides requires the action of both UDP-glucuronosyltransferases (UGT) and efflux transporters since glucuronides are too hydrophilic to diffuse across the cellular membrane. We determined the kinetics of efflux of 13 flavonoid glucuronides using the newly developed HeLa-UGT1A9 cells and correlated them with kinetic parameters derived using expressed UGT1A9. The results indicated that among the seven monohydroxyflavones (HFs), there was moderately good correlation (r²≥0.65) between fraction metabolized (f_met) derived from HeLa-UGT1A9 cells and CL_int derived from the UGT1A9-mediated metabolism. However, there was weak or no correlation between these two parameters for six dihyroxylflavones (DHFs). Furthermore, there was weak no correlation between various kinetic parameters (K_m, V_max or CL_int) for the efflux and the metabolism regardless if we were using 7 HFs, 6 DHFs or a combination thereof. Instead, cellular excretion of many flavonoids glucuronides appears to be controlled by the efflux transporter, and poor affinity of glucuronide to the efflux transporter resulted in major intracellular accumulation of glucuronides to a level that is above the dosing concentration of its aglycone. Hence, the efflux transporters appear to act as the “Revolving Door” to control the cellular excretion of glucuronides. In conclusion, the determination of a flavonoid's susceptibility to glucuronidation must be based on both its susceptibility to glucuronidation by the enzyme and resulting glucuronide's affinity to the relevant efflux transporters, which act as the “Revolving Door(s)” to facilitate or control its removal from the cells.

INTRODUCTION

Glucuronidation is a significant phase II metabolic pathway responsible for the elimination of a variety of endogenous and exogenous chemicals including drugs, hormones, and dietary phytochemicals such as polyphenols and flavonoids¹. Cellular glucuronide production is a two-step process, the formation of glucuronides catalyzed by various UGT isoforms, and excretion of glucuronides enabled by various anion transporting efflux transporters such as BCRP and MRPs¹. Efflux transporters were thought to act as a “Revolving Door” for facilitating and/or controlling the cellular glucuronide excretion. Therefore, a thorough understanding of the cellular glucuronide production process usually involves delineation of the glucuronidation steps that involve the relevant UDP-glucuronosyltransferases (UGTs) and subsequent glucuronide efflux steps by various efflux transporters. This understanding is important for the development of predictive algorithm useful for selecting drug candidates based on their metabolic susceptibility, because their bioavailability and drug interaction potentials are dependent on their metabolic susceptibility.

Towards this predictive goal, investigators including ourselves have spent significant effort in understanding and determining the structure-metabolic activity relationship (SAR) between chemical structures and glucuronidation rates by a specific UGT isoform^2-8. In addition, a few reports have shown that glucuronidation rates in tissue microsomes (e.g., liver microsomes) is directly correlated with the expression patterns of various UGT isoforms and activity of each isoform^9-11. Therefore, investigators have made significant strides in predicting glucuronidation rates of compounds by one or several major UGT isoforms.

Unlike the structural activity relationships demonstrated using various UGT isoforms such as UGT1A1⁵ and UGT1A9⁶, very little is known about the structural activity relationship of glucuronide efflux by an efflux transporter. Although several types of membrane vesicles overexpressing a particular type of responsible efflux transporter (e.g., BCRP or MRPs) have been used to shown that these efflux transporters are capable of mediating the efflux of various organic anions, an actual structural activity relationship has not been established for glucuronides. This is largely because of the lack of commercially available glucuronide standards needed to conduct the aforementioned studies.

Recently, we have developed a new tool: HeLa-UGT1A9 cells, which should allow us to determine the structure activity relationship without purified glucuronide standards¹². In these cells, we found that intracellular concentrations of glucuronides could rapidly reach steady state (usually within 30 min), which in turn allows the determination of the steady state efflux rates¹². The latter will allow us to determine the kinetic parameters associated with the dominating efflux transporters, a necessary step to delineate the structure-activity relationship (i.e., glucuronide efflux SAR). We also found that HeLa-UGT1A9 cells mainly used BCRP as its efflux transporter for flavonoid glucuronides using both molecular and kinetic characterization (using Ko143 as inhibitor). Specifically, complete inhibition of this transporter decreases flavonoid glucuronide clearance by more than 95%. We also found that MRP2 and MRP3 did not make significant contribution to the efflux of flavonoid glucuronides, using siRNAs and a chemical inhibitor of MRPs (LTC4)¹³. Therefore, these cells are used here to determine the SAR for BCRP mediated efflux of flavonoid glucuronides.

We have chosen flavone glucuronides as the model compounds for the present studies because flavonoids appear to possess a wide variety of biological activities in the areas of anticancer, anti-ageing, and anti-inflammation, all of which may be related to their antioxidant effects and glucuronides retain antioxidant activities. However, their development into therapeutic agents is severely challenged by the lack of oral bioavailability, mainly due to extensive phase II metabolism via glucuronidation and sulfonation. Therefore, we hypothesized that a better understanding of flavonoid glucuronidation process will reveal novel means of improving their oral bioavailability. Here, we studied BCRP-mediated efflux of 13 flavone glucuronides (Fig. 1), seven derived from mono-hydroxyl-flavones, and six di-hydroxyl-flavones in order to understand how changes in flavone glucuronide structures affect BCRP-mediated glucuronide efflux.

Fig. 1. Structures of seven mono-hydroxyflavones and six dihydroxyflavones.

Shown in the scheme are structures of aglycone forms of hydroxyflavone analogs.

Glucuronidated metabolites may be formed at the 5, 6, or 7 phenolic group on the A ring, or 2′-OH, 3′-OH or 4′-OH phenolic group on the B ring, or 3-OH group on the C ring.

2′-Hydroxyflavone (2′HF), 3′-Hydroxyflavone (3′HF),

4′-Hydroxyflavone (4′HF), 3-Hydroxyflavone (3HF),

5-Hydroxyflavone (5HF), 6-Hydroxyflavone (6HF),

7-Hydroxyflavone (7HF), 3,2′-Dihydroxyflavone (3,2′DHF),

3,3′-Dihydroxyflavone (3,3′DHF), 3,4′-Dihydroxyflavone (3,4′DHF),

3,5-Dihydroxyflavone (3,5DHF), 3,6-Dihydroxyflavone (3,6DHF),

3,7-Dihydroxyflavone (3,7DHF).

MATERIALS AND METHODS

Materials

Human UGT1A9-overexpressing HeLa cells (hereon referred to HeLa-UGT1A9 cells in this paper) were described previously¹². Human BCRP (Arg482) membrane vesicles and expressed human UGT1A9 were purchased from BD Biosciences (Woburn, MA). 2'-Hydroxyflavone (2'HF), 3'-Hydroxyflavone (3'HF), 4'-Hydroxyflavone (4'HF), 3-Hydroxyflavone (3HF), 5-Hydroxyflavone (5HF), 6-Hydroxyflavone (6HF), 7-Hydroxyflavone (7HF), 3,5-dihydroxyflavone (3,5DHF), 3,6-dihydroxyflavone (3,6DHF), 3,7-dihydroxyflavone (3,7DHF), 3,2′-dihydroxyflavone (3,2′DHF), 3,3′-dihydroxyflavone (3,3′DHF), and 3,4′-dihydroxyflavone (3,4′DHF) were purchased from Indofine Chemicals (Hillsborough, NJ). All other chemicals and solvents (analytical grade or better) were used as received.

Cell Culture

Conditions for culture and maintenance of HeLa-UGT1A9 cells were described previously and used without alteration¹². Cells from 3 days post-seeding were used for the glucuronide excretion experiments described below. Detailed molecular characterization of our UGT1A9 stably expressed Hela cells was performed previously¹². The numbers of cells were correlated to protein concentrations of harvested cells, which is determined using a protein assay kit (Bio-Rad, Hercules, CA) using bovine serum albumin as the standard. To determine the intracellular concentration of glucuronides based on amount of glucuronides present inside cells, HeLa cell volume was estimated to be 4μl/mg protein¹².

Excretion Experiments

Before experiments, the HeLa-UGT1A9 cells were washed twice with 37°C HBSS buffer (Hank's balanced salt solution, pH=7.4). The cells were then incubated in HBSS buffer containing one of the seven mono-hydroxyflavones (2'-, 3'-, 4'-, 3-, 5-, 6- and 7HF) or one of the six dihydroxyflavones (3,2′-, 3,3′-, 3,4′-, 3,5-, 3,6- and 3,7DHF) (defined as the “loading solution”) for a predetermined time interval (shown in Table S1, Supplemental Data) at 37°C. In the loading solution, the final flavone concentration was derived by the dilution of a (100X) concentrated stock solution (DMSO/Methanol=1:4) to ensure constant organic solvent content in all transport experiments. The sampling times were selected to ensure that the amounts excreted vs. time plots stay in the linear range. At each time point, 200 μl of incubating media from each well were collected as samples and an equal volume of loading solutions was used to replenish each well. The collected samples were each mixed with a 50μl “Stop Solution,” which is consisted of 94% acetonitrile and 6% acetic acid. Supernatants were ready for UPLC analysis after centrifugation (15min at 15,000 rpm).

Determination of Concentration of Intracellular Glucuronides in HeLa-UGT1A9 Cells

At the end of an excretion experiment, HeLa –UGT1A9 cells were washed twice with an ice-cold HBSS buffer to remove the extracellular aglycone and conjugates. The cells were then removed and collected in 120 or 200 μl HBSS buffer. In an ice-cold water bath (4°C) collected cells were lyzed in an Aquasonic 150D sonicator (VWR Scientific, Bristol, CT) operating at the maximum power (135 average watts) for 30min to ensure the release of aglycone and conjugates for measurement¹⁴. In the lysate, no further metabolism is expected since UGT1A9 in cellular lysate will not function without the co-factor, UDP-glucuronic acid. After centrifugation at 15,500 rpm for 20min, supernatants (100 μl) were collected and mixed with the Stop Solution (25 μl). The treated samples are ready for UPLC analysis after centrifugation (15min at 15,000 rpm) to remove larger particles and cellular debris.

Vesicular Transport Studies for BCRP

Vesicular transport studies were performed with rapid filtration techniques as described previously^15-17. In brief, membrane fractions containing inside out efflux transporters were incubated in the presence or absence of 5 mM ATP in a buffer containing 10 mM MgCl₂ and 10 mM Tris/HCl (pH 7.4). The transport was stopped by addition of 1ml of cold transport buffer to the membrane suspensions and then rapidly filtered through class F glass fiber filters (pore size, 0.45 μm). Filters were washed with 2× 5 ml of ice-cold wash buffer, cut and transferred to a 1.5-ml Eppendorf tube containing 400 μl of 50% methanol, followed by sonicating for 15min. After centrifugation at 15,000 rpm for 15 min, the supernatant was subjected into UPLC quantitative analysis. ATP-dependent transport was calculated by subtracting the values obtained in the presence of AMP from those in the presence of ATP.

Sample Analysis by UPLC

The conditions for UPLC analysis of HFs, DHFs and their respective glucuronides were modified based on a previously published method¹⁰. The conditions were: system, Waters Acquity™ UPLC with a binary pump and a 2996 DPA diode array detector (DAD, Waters, Milford, MA); column, BEH C18 column (50×2.1mm I.D.1.7 μm, Waters, Milford, MA); mobile phase A, 2.5mM ammonium acetate (pH 7.4); mobile phase B, 100% acetonitrile; gradient, 0-2 min (10-20%B), 2-3 min (20-40%B), 3-3.5 min (40-50%B), 3.5-4 min (50-90%B), 4-4.5 min (90%B), 4.5-5 min (90-10%B); injection volume, 10ul; detection wavelength for HFs, DHFs and their glucuronides were also shown in Table S1. The precision and accuracy were typically within acceptable range (<15%). The detection limits were at least 0.2 μM for both flavones and their glucuronides. The structures of glucuronides were further confirmed by the “UV spectrum maxima (λ_max) shift method” and UPLC-MS/MS as previously shown^18,12,10. Quantitation of the glucuronide was based on the standard curve of the parent compound and further calibrated using the conversion factor as described earlier^12,19,7. Conversion factors of 13 flavone glucuronides were also showed in Table S1 (Supplemental Data).

Biosynthesis of Glucuronides

The flavonoid glucuronides used in this study were biosynthesized using Sprague Dawley rat liver microsomes. The rat microsomes were obtained as described earlier (Chen et al., 2005b). The prepared glucuronide solution underwent liquid-liquid extraction using methylene chloride to remove the remaining aglycone. The resultant aqueous phase was then subjected to C-18 solid phase extraction using J.T. Baker Speedisk®48 Pressure Processors (Phillipsburg, NT). The column was conditioned with 4 ml of methanol followed by 2 ml of water. After loading the sample, polar impurities were removed with 2 ml water and the final elution was performed with 4 ml methanol. The methanol was evaporated and the remaining aqueous sample dissolved in ~4 ml of water. Glucuronide was isolated by collecting effluent from a preparative column (Phenomenex, Luna C18, 5 μm particle size, 250 × 10 mm) using an HPLC system (1050 series, Hewlett Packard). The mobile phase A was 100% water, mobile phase B was 100% acetonitrile. Gradient: 0 min: 10%B; 0-1hr: 10%-30%B. Multiple injections were made until the entire sample was isolated. The glucuronide solution was then lyophilized to remove the mobile phase.

Data Analysis

Glucuronide efflux by BCRP in HeLa-UGT1A9 cells was verified to be linear with respect to incubation time (within the planned experimental period, see Table S1, Supplemental Data) and protein concentration (excretion rates were expressed as amounts of glucuronides effluxed per min per mg protein or pmol/min/mg). The intracellular concentrations (at the end of experiment) of glucuronides were determined at the end of each efflux experiment for each loading concentration. Correlation plots between efflux rates and intracellular concentrations would allow the determination of the kinetic parameters of BCRP-mediated efflux using a variety of kinetic equations. The actual equation used is based on the shape of the Eadie-Hofstee plots. The most common equation is the Michaelis-Menten equation.

$$V=\frac{V_{\max }C}{K_m+C}$$ equation (1)

where K_m (substrate concentration at which half maximal velocity is achieved) represents the Michaelis constant, and V_max is the maximum excretion rate. If Eadie-Hofstee plots showed characteristic profiles of atypical kinetics such as autoactivation or biphasic kinetics^20,21, the data were fit to the corresponding equations as described previously²². Kinetics parameters for glucuronide excretion were performed with Microsoft Excel add-in program for modeling steady-state enzyme kinetics¹², the goodness of fit was evaluated on the basis of coefficient of determination (R²), Akaike's information criterion (AIC)²³, weighted sum of squared residual (WSS), and the rule of parsimony was applied. After the kinetic parameters were determine, the intrinsic clearance (CL_int=V_max/K_m) was calculated, and used to estimate the efficiency of the efflux transporter. Previously, intrinsic clearance is often used to predict the in-vivo pharmacokinetics of a drug using in vitro data. To facilitate the comparison, we used K_m, V_max and CL_int for glucuronide efflux and K_m’, V_max’ and CL_int’ for UGT-1A9-mediated glucuronidation.

Calculation of f_met Value

Fraction metabolized or f_met value was defined as the fraction of dose metabolized (equation (2) at the end of the experiment. The f_met value was considered as the more appropriate parameter to reflect the extent of metabolism in intact cells in the presence of a transporter-enzyme interplay¹⁶:

$$f_{met}=\frac{\sum Metabolite}{\sum Metabolites+\sum Parent}\times 100\%$$ equation (2)

Statistical Analysis

One-Way ANOVA was used to determine if the difference was statistically significant. The significance value was set at 0.05 or p<0.05.

RESULTS

Kinetics of HF Glucuronide Efflux in HeLa-UGT1A9 Cells

Efflux of glucuronides usually increased with intracellular concentration of glucuronides (Fig.2). Among the seven HF (mono-hydroxyflavones) glucuronides, all but one (3HF- glucuronide) displayed classical Michaelis-Menten kinetics, which is signified by a linear increase in the rate of efflux at low intracellular concentration, and plateauing rate of efflux at high intracellular concentrations (Fig.2). Interestingly, the intracellular concentration of HF glucuronides always increased with loading concentration at lower concentration (≤5 μM) but at higher concentrations, the intracellular concentrations either increased or fluctuated (Fig.3). The increase in intracellular concentration was expected since saturation of the efflux transporter leads to enhanced intracellular accumulation of glucuronides.

Fig. 2. Kinetics profiles of BCRP-mediated efflux of seven HFs glucuronides (A, 7HF-G; B, 2′HF-G; C, 3′HF-G; D, 3HF-G; E, 4′HF-G; F, 5HF-G; G, 6HF-G) (left) and corresponding Eadie-Hofstee plots for each kinetic profile (right).

For Fig.2 and 6, solid lines (left) denote excretion rate of flavone glucuronides. Each data point represents the average of three replicates, and error bars are the standard deviations of the mean (n=3). “Conc.” is the abbreviation for concentration.

Fig. 3. Intracellular HFs-G concentrations as a function of loading HFs concentration (A, 7HF; B, 2′HF; C, 3′HF; D, 3HF; E, 4′HF; F, 5HF; G, 6HF).

Each bar represents the average of three replicates, and error bars are the standard deviations of the mean (n=3). The symbol “*” means the difference between concentrations for each compound was statistically significant according to a one-way ANOVA test.

We then determined efflux of 7 HF-glucuronides in HeLa-UGT1A9 cells, and found that three HF glucuronides (i.e., glucuronides of 6HF, 5HF and 4’HF) displayed K_m value greater than 17 μM whereas the other 4 displayed K_m values that were less than 9.8 μM (Table 1). The smallest K_m value 5.4 μM (for 2’HF-glucuronide) was 9 times smaller than the highest K_m value (5HF glucuronide) (Table 1). In contrast, the V_max values differed by only 2.7 fold, with the largest value of 303 pmol/min/mg protein (for 7HF-glucuronide) and smallest value of 111 pmol/min/mg protein (for 4’HF-glucuronide). The fastest intrinsic clearance value for efflux was 33.6 μl/min/mg protein (for 7HF-glucuronide) and the smallest was 5.6 μl/min/mg protein (for 6HF-glucuronide), a 6 fold difference (Table 1).

Table 1.

The efflux kinetic parameters for glucuronides of seven mono-hydroxyflavones by BCRP, and the comparison between BCRP-mediated efflux of HFs-G and UGT1A9-mediated glucuronidation of HFs.

BCRP	Km (μM)	Vmax (pmol/mg/min)	CLint (Vmax/ Km) (μl/mg/min)	Log(CLint)	Kinetic type
6HF-G	26.8±4.8	151.0±8.6	5.6	0.75	MM
5HF-G	48.5±9.7	283.4±20.0	5.8	0.77	MM
4′HF-G	17.5±4.7	111.2±6.9	6.4	0.80	MM
3HF-G	9.7±1.9	165.6±10.1	17.1	1.23	BL
3′HF-G	5.6±2.8	128.1±10.0	22.9	1.36	MM
2′HF-G	5.4±1.1	167.7±4.4	31.2	1.49	MM
7HF-G	9.0±1.5	302.9±8.6	33.6	1.53	MM
Max/Min	9.0	2.7	6.0	0.8

UGT1A9	Km′ (μM)	Vmax′ (pmol/mg/min)	CLint′(Vmax′/ Km′) (μl/mg/min)	Log(CLint)′	Kinetic type
6HF*	2.27	76.1	34	1.53	MM
5HF*	0.96	118	123	2.09	MM
4′HF*	1.46	98.7	67.6	1.83	MM
3HF*	0.30	2100	7000	3.84	MM
3′HF*	2.02	926	458	2.66	MM
2′HF*	0.40	74.4	186	2.27	MM
7HF*	3.59	4895	1364	3.13	MM
Max/Min	12.0	65.8	208.8	2.3

For table 1 and table 2: MM: Michaelis-Menten equation; BL: Biphasic linear portion model; SI: Substrate inhibition equation. And * present results cited from our Lab recently report of Wu and Hu ³⁷.

We also compared the efflux kinetic parameters with kinetic parameters for UGT1A9-mediated glucuronidation (Table 1), which were previously published⁶, and the results showed that kinetic profiles were very similar in that they all shared classical Michaelis-Menten kinetics with the exception of 3-HF, whose efflux obeyed biphasic kinetics (Table 1) but whose glucuronidation followed the classical Michaelis-Menten kinetics. Using f_met as an indicator of cellular glucuronidation capability, we were able to show that f_met values decreased with increasing substrate concentration (Fig.4A), consistent with saturation of UGT1A9 and/or efflux transporter BCRP. We also plotted the f_met values at various concentrations against the CL_int’ values of glucuronidation of various substrates by UGT1A9 , and found that there was weak correlation (r² values in the range of 0.66-0.80) between f_met values of HF in intact HeLa-UGT1A9 cells and CL_int’ values of glucuronidation for HF using UGT1A9 isoform.

Fig. 4. Correlation between loading concentrations and metabolic efficiency or f_met of HeLa-UGT1A9 cells or CL_int of UGT1A9.

Fig.4A plots f_met values for seven HFs at six loading concentrations to contrast the effects of structural changes on f_met values. For Fig. 4A, each bar represents the average of three replicates, and error bars are the standard deviations of the mean (n=3). Fig. 4B plots correlation between f_met values derived from UGT1A9-overexpressing HeLa cell and Log(CL_int)′ derived from commercial UGT1A9 of seven HFs (2′HF, 3′HF, 4′HF, 3HF, 5HF, 6HF, 7HF). Correlation coefficients were calculated for each of the tested concentrations. Each data point represents the average of three replicates (n=3). The symbol “*” means the difference between compounds within each concentration was statistically significant according to a one-way ANOVA test.

Kinetics of HF Glucuronide Efflux in BCRP-overexpressing Membrane Vesicles

We determined kinetic parameters for efflux of 7 mono-hydroxyflavone glucuronides (HF-Gs) using BCRP-overexpressing membrane vesicles (Fig. 5). We found that the K_m values of most of the seven HF-Gs were similar (difference<3 fold) between membrane vesicles generated and HeLa-UGT1A9 cells generated. The smallest differences (by less than 50%) were for 6HF-G, 4’HF-G, 2’HF-G and 7HF-G, and the largest difference for 5HF-G (>10 fold) (Table 2). The two compounds in the middle (3HF-G and 3’HF-G) had a difference of approximately 3 fold.

Fig.5. Kinetic profiles of BCRP-mediated efflux of seven HF glucuronides.

The concentration versus rate plots are on the left and corresponding Eadie-Hofstee plots for each kinetic profile are on the right. Each data represents the average of three replicates, and error bars are the standard deviations of the mean (n=3).

Table 2.

Comparison of the efflux kinetic parameter K_m for glucuronides of seven mono-hydroxyflavones by BCRP membrane vesicles and HeLa-UGT1A9 cells.

Model/Flavonoid	6HF-G	5HF-G	4′HF-G	3HF-G	3′HF-G	2′HF-G	7HF-G
Vesicles	23.4	2.9	15.4	29	18.2	4.6	12.6
HeLa-UGT1A9	26.8	48.5	17.5	9.7	5.6	5.4	9.0

Kinetics of DHF Glucuronide Efflux in HeLa-UGT1A9 Cells

In our study of glucuronidation of DHFs, we found that likelihood of glucuronidation found using monohydroxyflavones or HFs could usually be extended to DHFs. For example, 3HF was glucuronidated at a much faster rate than 5HF (Table 1), and hence glucuronidation of 3,5-DHF usually occurred at the 3-OH position with no measurable glucuronidation at 5-OH position (Table 3). Furthermore, the addition of an OH group on the flavonoid backbone could increase or decrease a compound's susceptibility, depending on the site of substitution. In the current study, we investigated if tendencies in the efflux of O-glucuronide of HF could be extended to the efflux of O-glucuronide of DHF, as would be expected based on SAR of glucuronidation.

Table 3.

The efflux kinetic parameters for glucuronides of six dihydroxyflavones by BCRP, and the comparison between BCRP-mediated efflux of DHFs-3-O-G and UGT1A9-mediated glucuronidation of DHFs.

BCRP	Km (μM)	Vmax (pmol/mg/min)	CLint(Vmax/ Km) (μl/mg/min)	Log(CLint)	Kinetic type
3HF-G	9.7±1.9	165.6±10.1	17.1	1.23	BL
3,6DHF-3-O-G	31.9±7.1	34.7±1.3	1.1	0.04	MM
3,4′DHF-3-O-G	35.9±11.9	86.0±5.4	2.4	0.38	MM
3,2′DHF-3-O-G	35.9±14.9	99.4±25.4	2.8	0.44	SI
3,3′DHF-3-O-G	45.2±16.1	183.1±12.6	4.1	0.61	MM
3,5DHF-3-O-G	12.8±2.9	67.3±6.2	5.3	0.72	BL
3,7DHF-3-O-G	3.5±0.9	24.3±1.5	7.0	0.84	MM
Max/Min	13.0	7.5	6.4	0.8

UGT1A9	Km′ (μM)	Vmax′ (pmol/mg/min)	CLint′(Vmax′/ Km′) (μl/mg/min)	Log(CLint)′	Kinetic type
3HF*	0.30	2100	7000	3.84	MM
3,6DHF*	0.62	13000	20968	4.34	MM
3,4′DHF*	0.13	1800	13846	3.77	MM
3,2′DHF*	0.67	52	78	1.89	MM
3,3′ DHF*	0.11	3100	28182	4.44	MM
3,5 DHF*	0.25	1900	7600	3.88	MM
3,7DHF(3-OH)*	0.22	4600	20909	4.32	SI
Max/Min	6.1	250.0	363.1	2.6

Efflux of 6 DHF glucuronides was determined using the HeLa-UGT1A9 cells (Fig. 6). All six DHF has a 3-OH group and only one DHF (i.e., 3,7-DHF) formed two metabolites. Because our interest is to determine the SAR for BCRP-mediated efflux of 3-O-glucuronide, the efflux of 7-O-glucuronide of 3,7-DHF would not be discussed here. Among the six 3-O-glucuronides of DHF, only hydroxyl group at the 7 position increased the affinity to BCRP (smaller K_m value), when compared to 3HF-O-glucuronide. Substitution at 5-OH position had very little impact on K_m values, whereas substitution at other four positions substantially decreased the affinity to BCRP (K_m values higher by 100% or more, Table 3). On the other hand, all but one substitution (3’-OH) decreased V_max value, and as a consequence, CL_int values of glucuronide efflux were much smaller for 3–O-glucuronides of six DHF than 3-O-glucuronide of 3HF. In addition, the efflux kinetic profiles of DHF-3-O-glucuronides were mostly different from that of 3HF glucuronide, except for 3,5DHF, which shared the same profile (biphasic) as 3HF glucuronide. Hence, the efflux of DHF glucuronides was generally different from 3HF glucuronide, and cannot be predicted based on efflux of HF glucuronides.

Fig. 6. Kinetics profiles of BCRP-mediated efflux of six DHFs glucuronides (A,3,7DHF-3-O-G; B, 3,5DHF-3-O-G; C, 3,3′DHF-3-O-G; D, 3,2′DHF-3-O-G, E, 3,4′DHF-3-O-G and F, 3,6DHF-3-O-G) (left) and corresponding Eadie-Hofstee plots for each kinetic profile (right).

Each data represents the average of three replicates, and error bars are the standard deviations of the mean (n=3).

We also measured the intracellular concentrations of glucuronides as a function of loading concentration, and we were able to show that at concentration below or at 5 μM, intracellular concentrations always increased with loading concentration (Fig. 7), similar to our observations using seven HFs.

Fig. 7. Intracellular concentration of DHF glucuronides as a function of concentrations (A,3,7DHF; B, 3,5DHF; C, 3,3′DHF; D, 3,2′DHF; E, 3,4′DHF and F, 3,6DHF).

Each bar represents the average of three replicates, and error bars are the standard deviations of the mean (n=3). The symbol “*” means the difference between concentrations for each compound was statistically significant according to a one-way ANOVA test.

Lastly, we used f_met as a measurement of HeLa-1A9 cells’ ability to metabolize DHFs. As expected, f_met decreased with increasing concentration (Fig.8A). Substitution at 2’ and 7 position by a hydroxyl group did not increase f_met when comparing to 3HF. In fact, hydroxyl substation at these two positions led to the fastest decrease in f_met as a function of (rising) concentration. The f_met of 3,5-DHF was least sensitive to the rising concentration (Fig.8A). On the other hand, 3,6-DHF was the compound with the highest f_met value at all tested concentration, although it shared the top value with compounds such as 3,3’-DHF except at the lowest tested concentration (0.625 μM). In general, enzyme mediated CL_int’ value correlated weakly or not at all with f_met of six DHFs, with the best r² value of 0.679, suggesting that enzyme activities are usually not the rate-limiting step in the cellular production of glucuronides.

Fig. 8. Correlation between loading concentrations and metabolic efficiency or f_met derived from HeLa-UGT1A9 or CL_int of UGT1A9.

Fig.8A plots f_met values for six DHFs at five loading concentrations to contrast the effects of structural changes on f_met values. Fig. 8B plots correlation between f_met values derived from UGT1A9-overexpressing HeLa cell and Log(CL_int)′ derived from commercial UGT1A9 of six DHFs (3,2′-, 3,3′-, 3,4′-, 3,5-, 3,6- and 3,7DHF). Correlation coefficients were calculated for each of the tested concentrations. Each bar or point represents the average of three replicates, and error bars are the standard deviations of the mean (n=3). The symbol “*” means the difference between compounds within each concentration was statistically significant according to a one-way ANOVA test.

Correlation between UGT1A9-Mediated Glucuronidation and BCRP-mediated Glucuronide Efflux

It has been assumed routinely in medical literature that glucuronidation rates are primarily driven by activities of the enzyme(s) involved in the metabolism. Here, we plotted various kinetic parameters associated with flavonoid glucuronidation and glucuronide efflux. We plotted K_m’ values of UGT1A9-mediated metabolism versus K_m values of BCRP-mediated glucuronide efflux from HeLa-UGT1A9 cells and found no correlation (Fig.9A). We also plotted V_max/V_max’ values and CL_int/CL_int’ values, and found no correlation (Fig.9B, 9C). Therefore, UGT1A9 driven enzyme kinetic process did not dominate the cellular production of flavonoid glucuronide.

Fig. 9. Scatter plots of kinetic parameters derived from BCRP-mediated efflux (K_m, V_max and Log(CL_int)) of HFs-G and DHFs-3-O-G and those derived from UGT1A9-mediated glucuronidation (K_m′, V_max′ and Log(CL_int)′) of HFs and DHFs.

I, K_m-BCRP vs. K_m‘-UGT1A9; II, V_max-BCRP vs. V_max'-UGT1A9; III, Log(CL_int)-BCRP vs. Log(CL_int)'-UGT1A9.

Correlation between f_met and Rates of Metabolism or Efflux

An important question arising from the results above was: is there any relationship between f_met and rates of metabolism or efflux of a series of flavonoids. This is important because if there is a dominating process, then f_met is going to be more correlated with one of those rates. It is relatively straightforward to determining the relationship between f_met and UGT1A9 rates (Fig.4B), since we could compare the rates at the same concentration of flavonoids (this case the substrate). It is however much more difficult to do that for glucuronide efflux as the concentration of substrates (in this case various flavonoid glucuronides) are different inside the cells even though their extracellular concentrations were the same. For example, at 2.5 μM loading concentration, the intracellular concentrations of 7 different HF glucuronides differed by nearly 7 folds (Table 4). When intracellular concentrations of different HFs at the same loading concentration were plotted against corresponding efflux rates, there was no relationship, which was expected since different HF glucuronide had different affinity to the transporter. In other words, higher intracellular concentration did not necessarily lead to higher efflux rates.

Table 4.

The Intracellular concentration of HFs-G and DHFs-3-O-G at the end of the experiments for different compounds at different loading concentration of HFs and DHFs.

	Intracellular Con. of HFs-G or DHFs-3-O-G (μM)
Loading HFs (μM)	1.25	2.5	5	10	15	20
6HF-G	12.46±0.36	24.80±2.19	57.82±6.22	89.13±21.41	72.68±4.19	111.43±15.84
5HF-G	31.03±0.44	47.98±3.91	72.86±2.26	94.74±11.70	18.32±6.47	235.51±14.96
4′HF-G	8.02±0.8	32.53±4.77	41.61±8.17	64.05±10.82	146.46±34.83	71.96±25.09
3HF-G	7.14±3.02	38.56±8.36	206.98±12.39	335.57±41.55	356.82±94.04	280.60±28.78
3′HF-G	4.17±1.05	1.58±2.42	41.68±6.14	79.75±7.60	56.74±4.90	85.65±9.88
2′HF-G	4.63±0.43	29.58±3.75	150.96±24.87	152.62±12.47	58.09±11.61	92.74±11.26
7HF-G	2.88±1.09	7.32±2.85	50.90±2.82	90.51±17.80	142.37±22.20	178.64±17.54

Loading DHFs (μM)	0.3125	0.625	1.25	2.5	5	10
3,6DHF-3-O-G	53.21±5.63	114.59±5.20	172.05±23.62	287.25±17.13	484.66±44.86	542.10±11.52
3,4′DHF-3-O-G	2.67±0.09	14.27±2.00	71.15±6.11	241.60±26.10	645.32±5.81	715.62±31.7
3,2′DHF-3-O-G	8.95±1.09	21.35±2.07	42.35±3.71	76.11±0.52	133.0±7.41	232.28±8.60
3,3′DHF-3-O-G	3.65±0.09	15.36±0.42	96.59±7.30	175.30±16.06	338.58±15.67	345.71±27.23
3,5DHF-3-O-G	2.28±0.68	4.82±0.02	14.70±2.60	53.48±3.45	242.98±9.46	416.65±12.34
3,7DHF-3-O-G	0.9±0.04	1.93±0.24	4.51±0.62	11.04±1.19	40.53±1.51	7.55±0.49

To derive the efflux rates at the same intracellular concentration (50 μM), which was somewhat higher than the Km values of all the flavonoid glucuronides we studied, we used the kinetic parameters listed in Table 1 and Table 3. We then identified the extracellular concentration that was used to achieve the intracellular concentration of 50 μM (Table 4), and used the corresponding f_met value for plots. The calculated rates were then plotted against f_met. The results showed that there were very good correlation between calculated efflux rates (at an intracellular concentration of 50 μM) for 6 out of 7 HFs and 5 out of 6 DHFs (Fig.10AB). We did not do the same calculation for a lower concentration of glucuronides, because f_met estimated at the end of experiments were approximate values at best, as many substrates (especially HFs) were exhausted at the end of the experiments (2 hr time period). Lack of substrate would impact the amount of substrate available for metabolism, and therefore would decrease the f_met value.

Fig. 10. Correlation between Rates of Glucuronide Efflux and f_met at Similar Intracellular Concentration of Glucuronides.

The rates of efflux at 50 μM were calculated for the plot because each glucuronide tested had an intracellular concentration (measured experimentally) that was close to an intracellular concentration of 50 μM. Once the concentration was selected, we were then able to find the f_met value for that compound for use in the plot. At this condition, none of the compound was exhausted. Different loading concentrations were used for different compounds and a table that details the actual concentrations of each flavonoid used is provided in the Supplemental Materials. The outlier compound (in circle) was 3HF in Fig.10A, or 3,6DHF in Fig.10B.

DISCUSSION

We have demonstrated clearly that cellular production of flavonoid glucuronides is not governed by the SAR derived using the predominating UGT isoform UGT1A9, which is highly expressed in HeLa-UGT1A9 cells. In other words, cellular production of many but not all flavonoid glucuronides is mainly governed by the function of the efflux transporter BCRP in HeLa-UGT1A9 cells. We demonstrated clearly that BCRP, acting as the gate-keeper (or “Revolving Door”) for glucuronides formed within the cell membrane, could both facilitate the efflux and control the intracellular concentration of glucuronides. In other words, glucuronides with lower efflux rates via BCRP than UGT1A9-mediated metabolism rates tend to accumulate inside the cells. Using f_met as the indicator of cellular glucuronide production, there was a weak correlation between cellular f_met and CL_int’ of UGT1A9-mediated metabolism for seven HFs (mono-hydroxyflavones) (Fig. 4B) but very weak or no correlation for six DHFs (Fig. 8B). On the other hand, there were decent correlation between cellular f_met and rate of efflux values when intracellular concentration of were similar (Fig.10), although there were two notable outliers.

We first proposed that efflux transporter can act as a gate-keeper for the cellular glucuronide in 2005²⁴. We coined the term “Revolving Door” to explain the role of efflux transporter as the gate-keeper. We proposed that for many compounds, the lack of efficient or the presence of intrinsically inefficient “Revolving Door” would ultimately hinder the cellular production of glucuronides, especially flavonoid glucuronides, whose formation is rather rapid. Here, we provided convincing evidence to prove this hypothesis, albeit with a model system that is relatively simplistic compared to the majority of the metabolically active cells such as enterocytes, or hepatocytes. We showed that when a glucuronide is poor substrate of BCRP, faster metabolism of its aglycone (as measured by enzyme CL_int’ values) by the rate-determining enzyme (i.e., UGT1A9) will not determine how efficient the cells are able to produce the glucuronides (f_met) (Fig.8B), a phenomenon that has been observed but not well explained. Rather, the cells began to accumulate high concentration of the glucuronides (Fig.3 and Fig.7), which did not appear to cause any harms to the cells. We also showed that SAR governing the glucuronidation could be completely different from SAR governing the glucuronide efflux (Fig.9). In addition, we showed that kinetic of glucuronidation often did not follow kinetics of glucuronide efflux (Fig.2, Fig.5 and Fig.6). Taken together, these results demonstrated that the two processes (i.e., glucuronidation vs. glucuronide efflux) are independent of each other. Because they were reasonably good correlation between efflux rates and f_met values, we believe that the efflux transporter BCRP acts as a “Revolving Door” to facilitate and/or control the cellular production of glucuronides. Although it is commonly assumed that faster metabolite formation should be correlated with faster glucuronide efflux, the “Revolving Door” nature of the cellular glucuronide production process meant that it was not true for many of the thirteen model flavonoids (Fig. 9).

BCRP is a very important apically located efflux transporter for removing intracellular glucuronides of flavonoid²⁵, and recently we have shown that BCRP could control the distribution of genistein glucuronides in vivo²⁶. Results shown here support this assertion because slower efflux via the BCRP meant significant accumulation (i.e., higher intracellular concentration than extracellular concentration) of glucuronides within the cells (Fig.3 and Fig.7), and in a polarized cell such as enterocyte or hepatocyte, the glucuronide would have been excreted by an alternative efflux transporter, most likely located on a different membrane (e.g., MRP3 on the basolateral membrane). This could then result in increased plasma exposure to these metabolites as we have shown previously using genistein²⁶. Since glucuronides could be hydrolyzed back to aglycone via the action of glucuronidases²⁶, the ability to change the systemic exposure of glucuronides meant that it is possible to change the systemic exposure to aglycone, which are usually pharmacologically more active.

BCRP is, however, not the only efflux transporter that has a major impact on the cellular excretion of glucuronides from enterocytes and hepatocytes. On the apical side, another important transporter is MRP2, whereas on the basolateral side the important transporter appears to be MRP3. In Caco-2 cells, inhibitors of MRPs significantly reduced the efflux of flavonoid glucuronides from apical and basolateral sides^27,28. In MRP2-deficient rats, biliary excretion of phase II conjugates were severely impaired^29-32, whereas intestinal excretion of glucuronides was only moderately decreased³¹. In contrast, in MRP3 deficient mice, plasma exposure to phase II conjugates could be drastically reduced^{33,34,31,35,36}, whereas efflux via the apical efflux transporters could cause an increase in biliary excretion^33,31. Taken together, these results suggest that efflux transporters may determine the organ/tissue distribution of hydrophilic glucuronides.

In the current paper, we used f_met as an indicator of cell's ability to metabolize a flavonoid. By definition, f_met is the fraction of compound metabolized at the end of an experiment. We have expressed the f_met value as “% /mg/10min” to account for differences in incubation time and amount of cells used, because f_met will change with the length of incubation, and amount of cells used in each experiment. This is necessary since we were unable to conduct all experiments for the same length of time (e.g., shorter time for flavonoids that are rapidly metabolized and longer time for slowly metabolized flavonoids). Some variability in cell culture was unavoidable, although the difference was usually within 25%. Finally, some f_met values may be underestimated due to substrate exhaustion. This occurred rather frequently when we used a low concentration of flavonoids (<1.25 μM) and long sampling time (2 hr). Therefore, we cannot compare the f_met values between 7 HFs and six DHFs as the experiments were conducted for different length of time. This difference explains why higher rates of efflux for glucuronides of DHF did not translate into higher f_met for DHFs when compared to HFs in Fig.10, although an additional variable of different extracellular concentrations (see Supplemental Materials) may have also played a significant role.

In conclusion, we have demonstrated clearly that efflux transporters may act as gate-keeper for the cellular production of glucuronides, and the functions of these efflux transporters are similar to the “Revolving Door” with a limited capacity. Because of the presence of these efflux transporters as the “Revolving Door,” compounds that are rapidly metabolized by an UGT isoform are not necessarily rapidly glucuronidated in an intact cell system, where the action of an efflux transporter may determine the rate and extent of a substrate's metabolism. Therefore, determination of a compound's susceptibility to glucuronidation must come from both its susceptibility to glucuronidation by one or a few UGT isoforms and resulting glucuronide's affinity to the relevant efflux transporters. In other words, separate SARs for glucuronide formation and efflux are needed for accurate prediction of cellular glucuronide production.

ABBREVIATIONS

2’HF

2'-hydroxyflavone

3’HF

3'-hydroxyflavone

4’HF

4'-hydroxyflavone

3HF

3-hydroxyflavone

5HF

5-hydroxyflavone

6HF

6-hydroxyflavone

7HF

7-hydroxyflavone

3,2’DHF

3,2'-dihydroxyflavone

3,3’DHF

3,3'-dihydroxyflavone

3,4’DHF

3,4'-dihydroxyflavone

3,5DHF

3,5-dihydroxyflavone

3,6DHF

3,6-dihydroxyflavone

3,7DHF

3,7-dihydroxyflavone

DMSO

dimethyl sulfoxide

BCRP

breast cancer resistance protein

CL_int

Intrinsic Clearance

O-G

O-glucuronide

UPLC

ultra performance liquid chromatography

UGT

UDP-glucuronosyltransferases

UGPDA

uridine diphosphoglucuronic acid

AIC

Akaike's information criterion

WSS

Weighted Sum of Squared Residual

SAR

structure activity relationship

HBSS

Hank's balanced salt solution

DMEM

Dulbecco's modified Eagles medium

FBS

fetal bovine serum

MM

Michaelis-Menten equation

BL

Biphasic linear portion model

SI

Substrate inhibition equation