Secondary farnesoid X receptor (FXR) effects, in addition to vitamin D receptor (VDR) effects, were observed in the rat liver after treatment with 1α,25-dihydroxyvitamin D3 [1,25(OH)2D3], the natural ligand of VDR, caused by increased bile acid absorption as a consequence of apical sodium-dependent bile acid transporter induction. To investigate whether the increased multidrug resistance protein 1 (Mdr1)/P-glycoprotein (P-gp) expression in the rat liver and kidney was caused by the VDR and not the FXR, we examined changes in Mdr1/P-gp expression in fxr(+/+) and fxr(−/−) mice after intraperitoneal dosing of vehicle versus 1,25(OH)2D3 (0 or 2.5 μg/kg every other day for 8 days). Renal and brain levels of Mdr1 mRNA and P-gp protein were significantly increased in both fxr(+/+) and fxr(−/−) mice treated with 1,25(OH)2D3, confirming that Mdr1/P-gp induction occurred independently of the FXR. Increased P-gp function was evident in 1,25(OH)2D3-treated fxr(+/+) mice given intravenous bolus doses of the P-gp probe, [3H]digoxin (0.1 mg/kg). Decreased blood (24%) and brain (29%) exposure, estimated as reduced areas under the curve, caused by increased renal (74%) and total body (34%) clearances of digoxin, were observed in treated mice. These events were predicted by physiologically based pharmacokinetic modeling that showed increased renal secretory intrinsic clearance (3.45-fold) and brain efflux intrinsic clearance (1.47-fold) in the 1,25(OH)2D3-treated mouse, trends that correlated well with increases in P-gp protein expression in tissues. The clearance changes were less apparent because of the high degree of renal reabsorption of digoxin. The observations suggest an important role of the VDR in the regulation of P-gp in the renal and brain disposition of P-gp substrates.
P-glycoprotein (P-gp), the gene product of the multidrug resistance protein 1 (MDR1), is a 170-kDa membrane transporter and member of the ATP-binding cassette superfamily (Juliano and Ling, 1976). P-gp functions as an ATP-powered drug efflux pump that interacts with numerous large, nonpolar, and weakly amphipathic compounds and cations of no apparent structural similarity. P-gp is highly expressed in many major organs and tissues in the body, including the intestine, liver, brain, kidney, colon, testes, and placenta (Cordon-Cardo et al., 1990). For this reason, P-gp plays a critical role in drug absorption, distribution, and elimination and is recognized as an important target for drug-drug interactions (DDIs) (Yu, 1999).
Over the last few decades, multiple nuclear receptors and transcription factors were found to regulate MDR1/P-gp expression (Reschly and Krasowski, 2006). These include the pregnane X receptor (PXR) and the constitutive androstane receptor, commonly referred to as xenobiotic-sensing nuclear receptors. Transactivation of the mouse Mdr1 gene by the humanized PXR was observed to result in altered drug disposition in transgenic mice (Bauer et al., 2006). The rodent Mdr1 or human MDR1 gene is also regulated by the farnesoid X receptor (FXR) (Landrier et al., 2006; Martin et al., 2008), and the FXR ligand, chenodeoxycholic acid, increased MDR1 mRNA in HepG2 cells (Martin et al., 2008). The vitamin D receptor (VDR) displays similar homology with PXR and constitutive androstane receptor (Reschly and Krasowski, 2006) and is another nuclear receptor that transactivates MDR1. 1α,25-Dihydroxyvitamin D3 [1,25(OH)2D3], the natural, active ligand of VDR (Tanaka et al., 1973), resulted in up-regulation of MDR1/P-gp expression in human colon carcinoma cell lines such as Caco-2, LS180, and LS174T cells (Aiba et al., 2005; Fan et al., 2009; Tachibana et al., 2009) and elevated P-gp levels in human airway epithelium-derived Caclu-3 cells (Patel et al., 2002). These results are consistent with the existence of a vitamin D response element in the MDR1 gene (Saeki et al., 2008).
In rats treated with 1,25(OH)2D3 in vivo, VDR led to increases in both Mdr1a mRNA and P-gp protein in the liver and kidney (Chow et al., 2009, 2010). Because of low levels of VDR in the rat liver (Gascon-Barré et al., 2003), induction of Mdr1a was suspected to be a result of indirect FXR effects elicited by induction of the apical sodium-dependent bile acid transporter (ASBT) in the rat intestine, culminating in increased absorption of bile acids into the portal blood (Chen et al., 2006) rather than from direct VDR effects. Upon entering the liver, bile acids, ligands of FXR, elicited hepatic FXR effects (Chow et al., 2009). Inasmuch as MDR1/P-gp may also be induced by bile acids (Martin et al., 2008), the increase in hepatic P-gp could be the consequence of both the direct VDR effects and/or indirect FXR effects on 1,25(OH)2D3 treatment to the rat in vivo.
To determine whether regulation of Mdr1/P-gp was via the direct action of the VDR or indirectly via the FXR, we carried out the present study to examine Mdr1/P-gp changes in fxr(−/−) mice and compared the changes in gene expression with those for the wild-type fxr(+/+) mice after vehicle and 1,25(OH)2D3 treatment. FXR effects become obviated in fxr(−/−) knockout mice; despite that, combined VDR and FXR effects persisted in fxr(+/+) wild-type mice in vivo. A protracted dosing regimen of 50 ng (2.5 μg/kg) of 1,25(OH)2D3 by intraperitoneal injection every other day for 8 days was chosen to lessen hypercalcemia (Chow et al., 2011). Changes in protein and mRNA expression were first determined to rule out the contribution of FXR in the up-regulation of Mdr1/P-gp in both fxr(+/+) and fxr(−/−) mice. We further investigated the fate of intravenous administration of tritiated [3H]digoxin, a P-gp probe, that is eliminated via excretion only in the mouse, to demonstrate changes in digoxin disposition caused by elevated P-gp after 1,25(OH)2D3 treatment. In vivo tissue levels of [3H]digoxin, which underwent enterohepatic recirculation and renal tubular reabsorption, in treated and untreated fxr(+/+) mice were fit to a physiologically based pharmacokinetic (PBPK) model to appraise the role of the VDR in altering P-gp function. Special attention was given to brain P-gp, the site of action of analgesics, antiepileptics, anticancer, and antiretroviral drugs, and in the heart, the site of action of digoxin.
Materials and Methods
1α,25(OH)2D3 was purchased in powder form from Sigma-Aldrich Canada (Mississauga, ON, Canada). The primary antibodies for Western blotting were obtained from various sources. Anti-Pgp, anti-GAPDH, and anti-Lamin B were from Abcam Inc. (Cambridge, MA); villin was from Santa Cruz Biotechnology Inc. (Santa Cruz, CA); and anti-VDR was from Thermo Fisher Scientific (Waltham, MA). cDNA synthesis and real-time PCR reagents were obtained from Applied Biosystems (Foster City, CA). [3H]Digoxin (specific activity, 40 mCi/μmol) was purchased from PerkinElmer Life and Analytical Sciences (Walham, MA) and purified by HPLC to >99% radiochemical purity (Liu et al., 2006). All other consumable reagents, including unlabeled digoxin, were obtained from Sigma-Aldrich Canada and Thermo Fisher Scientific (Mississauga, ON, Canada).
Induction Studies with 1,25(OH)2D3 in fxr(+/+) and fxr(−/−) Mice In Vivo
Both the male and female fxr(−/−) mice were kind gifts from Dr. Frank J. Gonzalez (National Institutes of Health, Bethesda, MD). The fxr(−/−) mice contained only the last exon of the FXR ligand binding domain and all of the 3′ untranslated region of the FXR gene (Sinal et al., 2000). The C57BL/6 pure strain male fxr(−/−) mice were genotyped using the following fxr primers: forward (wild-type allele), 5′-TCTCTTTAAGTGATGACGGGAATCT-3′; forward2 (null allele, 5′-GCTCTAAGGAGAGTCACTTGTGCA-3′; and reverse, 5′-GCATGCTCTGTTCATAAACGCCAT-3′, as described by Sinal et al. (2000). Male wild-type [fxr(+/+)] and knockout [fxr(−/−)] mice (8–12 weeks), bred in the animal facility at the University of Toronto, were given water and food ad libitum and maintained under a 12:12-h light and dark cycle in accordance with approved protocols. Mice were injected with 0 or 2.5 μg/kg 1,25(OH)2D3 in sterile corn oil intraperitoneally every other day for 8 days. The concentration of 1,25(OH)2D3 in anhydrous ethanol was determined spectrophotometrically at 265 nm (UV-1700; Shimadzu Scientific Instruments, Columbia, MD), and the 1,25(OH)2D3 solution was diluted in sterile corn oil (Sigma-Aldrich Canada) for injection (Chow et al., 2009). The alternate- day regimen was chosen because of the lessened hypercalcemia observed in comparison with those given doses in consecutive days (Chow et al., 2011).
On the ninth day, mice were anesthetized with an intraperitoneal injection of ketamine and xylazine (150 and 10 mg/kg, respectively). After flushing of the mouse blood from the lower vena cava with 10 ml of ice-cold saline, the intestine, liver, brain, and kidneys were removed and placed on ice. The ileal segment, taken as 6 cm proximal to the ileocecal junction and known to consist of the highest abundance of P-gp (Stephens et al., 2001; Liu et al., 2006), was used for protein and mRNA analyses. The ileum was flushed with physiologic saline solution containing 1 mM phenylmethylsulfonyl fluoride (PMSF) to eliminate feces, everted, and placed in the same saline solution containing PMSF before being scraped with a tissue scraper for the collection of enterocytes (Chow et al., 2009). The liver, brain, kidneys, and heart were weighed and reduced to small pieces. The scraped enterocytes and tissue pieces were snap-frozen with liquid nitrogen and stored at −80°C until further analyses.
Preparation of Subcellular Fractions
Frozen mucosal scrapings (50–100 mg of tissue) were homogenized with 1 ml of Tris-HCl buffer (0.1 M, pH 7.4) containing 1% protease inhibitor cocktail (Sigma-Aldrich Canada) and sonicated, as described by Chow et el. (2009). After centrifugation at 1000g for 10 min at 4°C, the resulting supernatant was spun again at 21,000g for 1 h at 4°C to yield a pellet or crude membrane fraction. The pellet was placed in a resuspension buffer (50 mM mannitol, 20 mM HEPES, 20 mM Trizma base, pH 7.4), which was premixed with 1% protease inhibitor cocktail (Sigma-Aldrich Canada). The liver, brain, and kidney tissue samples were homogenized (1:5 w/v) in a homogenizing buffer (250 mM sucrose, 10 mM HEPES, and 10 mM Tris-HCl, pH 7.4) that was premixed with 1% protease inhibitor cocktail (Sigma-Aldrich Canada). The homogenate was centrifuged at 3000g for 10 min at 4°C, and the resulting supernatant was spun again at 33,000g for 60 min at 4°C. The crude membrane pellet was placed in the resuspension buffer. A unified procedure was used for preparation of the crude nuclear fraction. All tissues were homogenized in an identical fashion with the same homogenizing buffer (250 mM sucrose, 10 mM HEPES, and 10 mM Tris-HCl, pH 7.4) containing 1% protease inhibitor cocktail, and the homogenate was spun at 3000g wherein the resultant pellet was resuspended in a nuclear buffer (60 mM KCl, 15 mM NaCl, 5 mM MgCl2·6H2O, 0.1 mM EGTA, 300 mM sucrose, 0.5 mM dithiothreitol, 0.1 mM PMSF, 300 mM sucrose, and 15 mM Trizma HCl, pH 7.4) containing 1% protease inhibitor cocktail. Protein concentration was measured by the Lowry method (Lowry et al., 1951).
For determination of the changes in VDR and P-gp protein levels among tissues, 20 to 80 μg was used; linearity for the relative intensity was shown to exist for the different amounts of tissue used. The sample was loaded and separated by 7.5 and 10% SDS-polyacrylamide gels, respectively, for VDR and P-gp analyses and transferred onto nitrocellulose membranes (GE Healthcare, Chalfont St. Giles, Buckinghamshire, UK), as described by Chow et al. (2009). The membrane was blocked with 5% (w/v) skim milk in Tris-buffered saline, pH 7.4, with 0.1% Tween 20 (TBS-T) (Sigma-Aldrich Canada) and washed three times with TBS-T before incubating it with the primary antibody solution (2% skim milk) overnight at 4°C. Thereafter, the membrane was washed three times with TBS-T and incubated with a secondary antibody (2% skim milk) at room temperature for 2 h. The membrane was washed three times, and then incubated with chemiluminescence reagents from GE Healthcare for visualization of the band intensity, quantified by scanning densitometry (National Institutes of Health Image software; http://rsb.info.nih.gov/nih-image/). Protein loading error was corrected by normalizing the target protein band against the protein band of the housekeeping gene: villin for intestinal samples, Lamin B for the comparison of VDR among tissues, and GAPDH for the same tissue type.
Quantitative Real-Time Polymerase Chain Reaction
The detailed procedure of RNA extraction has been described previously (Chow et al., 2009). For total RNA isolation, scraped enterocytes and other organ tissues were homogenized with TRIzol solution (50–100 mg/ml) and extracted with the TRIzol extraction method (Sigma-Aldrich Canada) according to the manufacturer's protocol, with modifications. The RNA purity of each sample was checked by absorbance ratios of 260 nm/280 nm and 260 nm/230 nm (≥1.7). The 1.5 μg of total RNA was converted to cDNA. Real-time quantitative PCR was performed with the SYBR Green detection system (Applied Biosystems 7500 Real-Time PCR System; Applied Biosystems, Streetsville, ON, Canada). Information on the primer sequence is summarized in Table 1. Primer specificity was checked by BLAST analysis (http://www.ncbi.nlm.nih.gov/BLAST/). Critical threshold cycle (CT) values of the target genes were collected using Applied Biosystems Sequence Detection software version 1.4. The target gene mRNA data were normalized to the housekeeping gene: villin for intestinal samples and cyclophillin for other tissue samples. The difference in CT values (ΔCT) between target and housekeeping genes was compared with the corresponding ΔCT of the vehicle control (ΔΔCT) and expressed as fold expression, 2−(ΔΔCT), for relative mRNA quantification.
Pharmacokinetic Study of [3H]Digoxin in Vehicle- or 1,25(OH)2D3-Treated Mice
The digoxin study was performed with fxr(+/+) mice only (same as C57BL/6 pure) treated with vehicle (corn oil) or 2.5 μg/kg (or 50 ng/mouse) of 1,25(OH)2D3 given intraperitoneally every other day for 8 days. On the ninth day, each mouse received a bolus injection of 0.1 mg/kg digoxin and ∼1.6 million dpm of [3H]digoxin in ∼100 μl of filtered saline solution containing 1.5% propylene glycerol via the tail vein and was placed inside a glass beaker atop a piece of suspended aluminum wire mesh for separate collection of the feces and urine. Three to five mice were rendered unconscious in a carbon dioxide chamber and used for blood collection by cardiac puncture via a 1-ml syringe, which was prerinsed with heparin (1000 IU/ml). After flushing of blood with ice-cold saline, perfused into the lower vena cava, the liver, kidney, heart, brain, and small intestine were rapidly removed, weighed, snap-frozen in liquid nitrogen, and stored at −80°C for future analyses. The feces above the wire mesh were collected and pooled together with flushed luminal contents of the small intestine and colon with ice-cold saline into pretared 15-ml polyethylene tubes. The urine from the collection beaker was pooled, together with water rinses (twice with 1 ml of water).
Blood samples (0.25 ml) were deproteinized upon addition of methanol 1:4 (v/v). The sample was mixed for 1 min and centrifuged at 14,000g for 10 min at 4°C, and 0.9 ml of the supernatant was removed for liquid scintillation counting. Varying known counts of [3H]digoxin were added to blank blood (same volume as samples) and used as standards and processed under identical conditions for construction of a calibration curve in the determination of total radioactivity of [3H]digoxin in blood samples. Likewise, the liver, kidney, heart, brain, and small intestine tissue were homogenized 1:3 (w/v) in saline solution, and 0.25 ml (tissue) or 1.2 ml (brain) of the homogenate was deproteinized with MeOH, 1:4 (v/v) while 1.0 to 4.8 ml was removed for liquid scintillation counting. Blank tissue homogenate samples, spiked with the appropriate aliquots of dpm for construction of calibration curves for each tissue, were processed according to the same deproteinization procedure. Counts in the fecal mixtures were estimated to denote the extents of biliary and luminal excretion; the contents were homogenized, and the radioactivity was extracted with ethyl acetate, 1:4 (v/v), after mixing (vortex) vigorously for 5 min and spun at 14,000g for 10 min at 4°C. An aliquot (0.95 ml) of the upper layer was removed for liquid scintillation counting. Again, standards of known dpm in fecal material were processed in an identical manner and used for the construction of the calibration curve. The total radioactivity in urine samples was determined directly by liquid scintillation counting.
The HPLC assay of Liu et al. (2006) was used to separate digoxin from its di- and mono-digitosoxides and the aglycone. The dpm in the deproteinized samples from blood, urine, feces, and tissue were separated by HPLC to resolve [3H]digoxin from its metabolite (Liu et al., 2006). The supernatant was evaporated under nitrogen gas and then reconstituted with 100 μl of methanol. The reconstituted residue was centrifuged, and 75 μl of the supernatant was injected into the HPLC Shimadzu system as described by Liu et al. (2006). Separation was achieved by a C18 reverse-phase column (Altech Associates, Deerfield, IL; 4.6 × 250 mm, 10-μm particle size) and a binary gradient consisting of water and acetonitrile, with an initial condition of 18% acetonitrile maintained at a flow rate of 1 ml/min, then increased to acetonitrile to 28% (Liu et al., 2006). The eluted fractions were collected at 1-min intervals and counted, and fractions of digoxin and its metabolites recovered in the sample were multiplied to the total count of the sample to arrive at individual dpm of digoxin and the metabolites. Results from HPLC revealed that, on average, unchanged [3H]digoxin represented approximately 98% of the total radioactivity for all blood and tissue samples (data not shown). A previous report also showed that the metabolic clearance of digoxin in mice was only approximately 3% of total clearance (Kawahara et al., 1999). Thus, the total radioactivity of the sample was taken to represent unchanged [3H]digoxin.
Modeling and Fitting
Whole-Body Physiologically Based Pharmacokinetic Modeling.
The concept of intestinal segregated flow was incorporated in the PBPK model (Cong et al., 2000; Liu et al., 2006). In this model, a minor portion of the intestinal flow (5–30%) perfuses the enterocyte region, whereas the majority of flow perfuses the serosal region (remaining flow >70%) (Cong et al., 2000). The model consists of tissue compartments that described digoxin concentrations in the blood (CB), heart (Cheart), kidney (CK), liver (CL), enterocyte (Cen), and serosal (Cs) tissues of the small intestine, other tissues (Cother), brain tissue (CBr) and blood of the brain (CBr,B) (Fig. 1). The volume and tissue to blood partition coefficient (tissue concentration/blood concentration) of each tissue compartment, denoted as V and KTB, respectively, are further qualified with the appropriate subscript for that tissue. The tissue/blood partition coefficient (KTB) of digoxin for the small intestine, liver, kidney, and heart were obtained experimentally from the tissue/blood ratios toward the end of the study (close to 600 min). We recognize that the tissue/blood concentration ratio would underestimate the true partitioning ratio within eliminating organs (Khor et al., 1991; Chiba et al., 1998). For the intestine, KTB,I or CI/CB,I (intestine tissue/blood leaving intestine) is assumed to be identical for both the enterocyte and serosal tissue. This is an oversimplication because KTB,I would not equal Cen/Cenb (ratio of the enterocyte tissue concentration relative to that in blood leaving the enterocyte) in view of the known luminal secretion by the P-gp. The estimate of KTB,I should be better estimated by Cs/Csb (ratio of the serosal tissue concentration relative to that in blood leaving the serosal tissue because of lack of elimination), because the serosa represents the noneliminating tissue of the intestine. The distortion of KTB,I by luminal secretion should be low because the flow to the enterocyte region was low, rendering a flow-weighted average concentration in intestine that would be quite close to the true estimate, especially when luminal secretion is absent.
The intrinsic secretory clearances for the kidney, liver, small intestine, and brain efflux that are representative of P-gp activities are denoted as CLint,sec,K, CLint,sec,H, CLint,sec,I, and CLefBr, respectively; CLinBr represents the digoxin uptake intrinsic clearance into the brain tissue. Biliary secretion (CLint,sec,H) followed by reabsorption (rate constant, ka) allows for the enterohepatic circulation of digoxin. For the kidney, the filtrated and secreted digoxin is prone to reabsorption, and the net excretion is modified by the fraction reabsorbed (FR) within the renal tubule, as described by Levy (1980). The roles of urinary pH, the pKa of digoxin, and the urinary flow rate on digoxin reabsorption were not considered, and the extent of reabsorption was assumed to be the same for both the control and 1,25(OH)2D3-treated mice. The differential equations that relate to mass transfer across these organs/tissues are summarized in the Appendix.
Fitting of the PBPK model to data from the vehicle control and 1,25(OH)2D3-treated fxr(+/+) mice was performed with the program Scientist (version 2.0; Micromath, St. Louis, MO). Appropriate weighting schemes (unity, 1/observation, and 1/observation2) were used. Some parameters were fixed in the fitting procedure; these included the blood and tissue volumes (V) and organ flow rates (Q), unbound fraction of digoxin (fP and fB), and blood/plasma concentration ratio (B/P). The tissue/blood partition coefficients (KTB) of digoxin for the small intestine, liver, kidney, and heart were obtained experimentally from the tissue/blood ratios toward the end of the study (close to 600 min).
The first strategy was to estimate the parameters first with the control data. The blood, urine, feces, and tissue data for the small intestine, liver, kidney, brain, and heart in control mice were used in the fitting to obtain the tissue/blood partition coefficient of the lumped tissues KTB,other, ka, CLinBr, FR, the fractional intestinal flow entering the enterocyte region (fQ), and the apparent intrinsic clearances. Because the unbound fractions of digoxin in the intestine, liver, kidney, and brain (fI, fL, fK, and fBr, respectively) were unknown, the fitted intrinsic secretory clearances (CLint,sec) and efflux clearance (CLefBr) were expressed as the product of tissue unbound fraction in tissue and intrinsic secretory clearance (for example, the apparent renal intrinsic secretory clearance, CL′int,sec,K is expressed as fKCLint,sec,K). With the assumption that fQ, KTB,other, ka, CLinBr, and FR were identical between control and treated mice, estimates obtained in the first fit were used to estimate the apparent intrinsic clearances of the treated mice in a subsequent fit. The second strategy was to use both sets of control and treated data in the same (or forced) fit to arrive at parameter estimates. In the fit, fQ, KTB,other, ka, CLinBr, and FR were again considered as common and unchanged parameters for both control and treated mice, and the apparent intrinsic secretory clearances of the small intestine, liver, and kidney, and efflux clearances of the brain (fBrCLefBr) were allowed to alter because of 1,25(OH)2D3 treatment. The weighting of two or one/observation2 yielded the highest model selection criterion and lowest coefficient of variation (S.D./parameter value).
Protein and mRNA data were expressed as mean ± S.D. The two-tailed Student's t test was used to compare differences between the vehicle control and treatment groups. For mRNA and protein analyses, data for the vehicle-treated sample from the fxr(+/+) mouse were set as the control (value set as unity) and used for comparison with those of other control and treatment samples. A P value of less than 0.05 was viewed as significant.
VDR and Mdr1/P-gp mRNA and Protein Expression in the Ileum, Liver, Kidney, and Brain of fxr(+/+) and fxr(−/−) Mice
Distribution of VDR Protein Expression Among Tissues.
With the anticipation that the abundance of VDR would vary among tissues, samples containing 50 μg of total crude nuclear protein were analyzed to examine the distribution of VDR protein in the nuclear fractions in the ileum, liver, kidney, and brain of fxr(+/+) and fxr(−/−) mice (data not shown). Linearity was shown to exist within the protein concentration range. However, larger variations in GAPDH were found in comparison with those for Lamin B among different tissues after loading of the same amount of tissue protein (data not shown). Thus, VDR protein intensities among different tissues were normalized to Lamin B. VDR protein expression for the ileum and kidney in both fxr(+/+) and fxr(−/−) mice was high and similar; whereas VDR protein expression was significantly less in the liver (29–35% of ileum control) and even lower for the brain (13–27% of ileum control) of both fxr(+/+) and fxr(−/−) mice. These results generally showed a lack of difference in VDR protein expression between the fxr(+/+) and fxr(−/−) mice.
Effect of 1,25(OH)2D3 on VDR mRNA and Protein Expression on fxr(+/+) and fxr(−/−) Mice.
Basal levels of VDR mRNA in the ileum, kidney, and brain were similar in both fxr(+/+) and fxr(−/−) mice, whereas hepatic VDR mRNA in fxr(−/−) mice was three times higher that of fxr(+/+) mice (Fig. 2A). Upon 1,25(OH)2D3 treatment, a significant increase in VDR mRNA and protein (∼3-fold) was observed for the kidney with 1,25(OH)2D3 treatment for both the fxr(+/+) and fxr(−/−) mice (Fig. 2), although there was a slightly lower VDR protein in the 1,25(OH)2D3-treated fxr(−/−) mouse brains compared with that of control fxr(−/−) mouse (Fig. 2B). The reason for the latter was unknown.
Effects of 1,25(OH)2D3 on Mdr1 mRNA and P-gp Protein Expression in Both fxr(+/+) and fxr(−/−) Mice.
Levels of Mdr1 mRNA and P-gp protein expression in the ileum and liver of both the fxr(+/+) and fxr(−/−) mice remained unaltered with 1,25(OH)2D3 treatment (Fig. 3). In contrast, treatment of 1,25(OH)2D3 led to increased Mdr1 mRNA and P-gp protein expression in the kidney and brain of both fxr(+/+) and fxr(−/−) mice (Fig. 3), although the natural abundance of Mdr1 mRNA in the brain of fxr(−/−) mouse was considerably lower than that of the wild-type counterpart, the fxr(+/+) mouse (P < 0.05) (Fig. 3A).
Effects of 1,25(OH)2D3 Treatment on the Pharmacokinetics of [3H]Digoxin in fxr(+/+) Mice
Blood Decay Profiles and Excretion of [3H]Digoxin after Intravenous Administration in fxr(+/+) Mice.
Figure 4 shows the blood concentration-time profiles and cumulative amounts recovered in urine and fecal matter versus time after a single intravenous dose administration (0.1 mg/kg) of [3H]digoxin. A biexponential decay of [3H]digoxin was noted in the blood concentration (normalized to dose) versus time profile (Fig. 4A). The concentration of [3H]digoxin was significantly lower in the 1,25(OH)2D3-treated group only at 360 min (6 h). The individual data points were averaged, and the mean blood values were used to calculate the area under the curve [AUC(0→∞)] by the trapezoidal rule and extrapolation of the last concentration over the terminal half-life (β) (Table 2). This AUC(0→∞) value for digoxin in mice treated with 1,25(OH)2D3 was only 76% of that of control mice. The apparent total body and renal clearances were increased by 34 and 74%, respectively, and the terminal half-life, decreased by 30% in the mice treated with 1,25(OH)2D3 (Table 2). The Vdarea (CLtotal/β) for both the control and treatment group remained relatively unchanged.
The cumulative amounts of [3H]digoxin excreted in urine at 600 min and feces were comparable (Fig. 4, B and C). Amounts in urine in 1,25(OH)2D3-treated mice were higher than those in control mice at almost every sampling time subsequent to 30 min (P < 0.05), except at 360 min (Fig. 4B), and were significantly higher (42.6 ± 9.5 versus 27.0 ± 5.6% dose) at 600 min. The cumulative fecal amounts of [3H]digoxin in the 1,25(OH)2D3-treated group were increased only slightly compared with those of control mice (Fig. 4C). The apparent renal clearance (Fig. 5A), estimated as the slope upon plotting the cumulative amount of [3H]digoxin excreted into urine versus blood AUC of 1,25(OH)2D3-treated mice (0.074 ml/min) was 74% higher than that of the control mice (0.0426 ml/min). Our control values were slightly lower than those estimated by others for the renal clearance (0.069 ml/min) (Kawahara et al., 1999) and total clearance (0.083 ml/min) of digoxin (Griffiths et al., 1984). The filtration clearance, calculated as fPGFR [where fp is 0.78 (Davies and Morris, 1993; Kawahara et al., 1999)], was 0.22 ml/min, a value much higher in relation to observed renal clearance (0.0426 ml/min), suggesting that reabsorption played a significant role in the net renal clearance of digoxin in mice.
The fecal (sum of net intestinal and biliary) clearance (0.0969 ml/min), estimated as the slope upon plotting the cumulative amount of [3H]digoxin recovered into feces versus the blood AUC, was higher than the renal clearance of digoxin (Table 2). 1,25(OH)2D3 treatment increased the fecal clearance by 30% over the control group (0.0742 ml/min) (Fig. 5B). Reabsorption of the biliarily and intestinally excreted digoxin followed by reabsorption in the intestine (enterohepatic recirculation) must have occurred. The possibility was also mentioned by Kawahara et al. (1999), and this could have affected the estimate of the fecal secretary clearances, which were deviated from the slopes of Fig. 5B.
Estimation of Area Under the Curves.
The amounts of [3H]digoxin, normalized to per gram of tissue in the small intestine, liver, kidney, brain, and heart were plotted against time (Fig. 6). Levels of digoxin in the small intestine, liver, kidney, and heart were generally similar, except for the brain. In both control and treated groups, [3H]digoxin levels in the small intestine, liver, kidney, and heart displayed instantaneous accumulation within the first few minutes, followed by a rapid decay by 10 min, then gradually decayed thereafter (Fig. 6, A, B, C, and E). However, the peaks occurred slightly longer (∼100 min) in the brain. In the brain, levels of [3H]digoxin in the treatment group were markedly lower than those of controls (Fig. 6D). When the AUCs in the small intestine, liver, kidney, brain, heart, and blood, derived from observations on the amounts per gram of tissue, were estimated by the trapezoidal rule (Table 2), an apparently lower AUC(0–600 min) (23%) was found for the brain in the treatment group. The AUC(0–600 min) for the small intestine, liver, kidney, and heart remained relatively unchanged for both groups (Table 2).
Tissue to Blood AUC Versus Time Profile.
The tissue partitioning coefficient of digoxin was estimated as the ratio of AUC(0→t) of tissue normalized to that of blood AUC(0→t) (Fig. 7). In control mice, digoxin tissue/blood AUC ratios of 6.0, 2.5, 1.12, and 1.3 were reached at 600 min for the small intestine, liver, kidney, and heart, respectively (Fig. 7, A, B, C, and E), and treatment with 1,25(OH)2D3 exerted only minimal effects on the tissue/blood AUC ratio. In the brain, a gradual rise in AUC ratio over time was observed for both treatment and control groups (Fig. 7D). The ratio was consistently lower than unity in the brain and was reduced with 1,25(OH)2D3 treatment. A plateau level was not reached for the brain to blood AUC ratio at 600 min (Fig. 7D).
Whole-Body PBPK Modeling
For both types of fits, the blood, urine, feces, small intestine, liver, kidney, brain, and heart data were fit to the whole PBPK model (Fig. 1) with literature values of tissue volumes and blood flows (Table 3) and the observed tissue partition coefficients (KTB) (from Fig. 7). For the first of the sequential fits, fQ, KTB,other, CLinBr, ka, and FR were 0.185 ± 0.082, 2.50 ± 0.17, 0.00303 ± 0.000491 ml/min, 0.00293 ± 0.00076 min−1, and 0.830 ± 1.39, respectively, for the vehicle control group, and those values were assigned to estimate the apparent intrinsic clearances for the 1,25(OH)2D3 treatment group. The composite fits revealed a higher [3H]digoxin elimination from blood caused by an increased apparent renal intrinsic clearance fKCLint,sec,K (from 0.0323 ± 2.12 to 0.188 ± 0.027 ml/min) and a greater efflux from brain tissue, fBrCLefBr (from 0.00228 ± 0.00482 to 0.00329 ± 0.00409 ml/min) for the 1,25(OH)2D3 treatment group (Table 4). The apparent renal secretory intrinsic clearance (fKCLint,sec,K) after 1,25(OH)2D3 treatment was 3.65-fold that of control, a value that correlated well to the level of P-gp induction [2.65-fold for fxr(+/+) mice] and less so to the change in apparent renal clearance (1.74-fold) (Fig. 5A) caused by the presence of filtration and the high degree of reabsorption of digoxin. The parameter estimated for intestinal (fICLint,sec,I) and hepatic (fHCLint,sec,H) secretions via P-gp were similar between the control and treatment groups (0.0227 ± 0.0085 versus 0.0225 ± 0.0050 ml/min for the intestine and 0.0157 ± 0.00125 versus 0.0152 ± 0.0086 ml/min for the liver; Table 4). These estimates, even when summed (0.0384 to 0.0377 ml/min), were low in comparison to the fecal clearances (0.074 and 0.097 ml/min) estimated from the slopes of Fig. 5B, suggesting that the unbound fractions fI and fH must be very low. The estimates suggest that P-gp activities for both intestinal and biliary excretion were relatively constant (Table 4). Moreover, the fitted results showed that the weighting of 1/observation2 yielded the highest model selection criterion (MSC) (Table 4), and lowest coefficient of variation (S.D./parameter value).
Estimates from the simultaneous (force) fit were generally similar to those obtained from the sequential fits (Table 4). The common parameters, fQ, KTB,other, CLinBr, ka, and FR, were estimated to be 0.158 ± 0.347, 2.42 ± 0.82, 0.000301 ± 0.000389 ml/min, 0.00216 ± 0.00162 min−1, and 0.858 ± 2.53, respectively (Table 4). Increased apparent renal intrinsic clearance fKCLint,sec,K (from 0.074 ± 5.34 to 0.255 ± 0.029 ml/min) and greater efflux from the brain tissue, fBrCLefBr (from 0.00230 ± 0.00422 to 0.00337 ± 0.00363 ml/min) were observed for the 1,25(OH)2D3 treatment group (Table 4). The apparent renal secretory intrinsic clearance (fKCLint,sec,K) after 1,25(OH)2D3 treatment was 3.45-fold that of control, a value that correlated to well to the level of P-gp induction (2.65-fold; Table 5) but less so to the change in renal clearance (1.74-fold) (Fig. 5A). These changes were not exact matches because of the presence of filtration and high degree of reabsorption of digoxin. Simulations performed using the FR as 0.857 and fQ as 0.158 and parameters obtained with the simultaneous fit showed that the blood and kidney and brain concentrations were generally insensitive to a doubling in fKCLint,sec,K (varied from 0.074 ml/min to ½, 2, 3, and 5 times) (Fig. 8). In contrast, for drugs whose FR approaches 0, changes in fKCLint,sec,K would significantly affect tissue levels and renal excretion (simulation not shown). The ratio for brain efflux was 1.47 times higher for the 1,25(OH)2D3 treatment group, a value similar to that for Western blotting (1.8 times; Table 5). Values of the apparent fICLint,sec,I and fHCLint,set,H were of similar magnitude (0.0168 to 0.0189 ml/min; Table 4) and were unchanged with 1,25(OH)2D3 treatment, as also found from Western blotting (Table 5). Again, their sum was much less than the apparent fecal clearances (0.074 and 0.097 ml/min), supporting the view that fI and fH must be very small. Overall, the fitted results showed that the weighting of 1/observation2 yielded the highest MSC (Table 4) and lowest coefficient of variation (S.D./parameter value).
In both types of fits, values of fQ (Table 4) were similar (0.185 and 0.158) to that found for the rat intestinal preparation (Cong et al., 2000; Liu et al., 2006) and less than 20% of the total flow, favoring the concept of segregated flow to the enterocyte (Cong et al., 2000). Although the fitted fKCLint,sec,K for the force fit was almost twice that for the sequential fits, the change caused by 1,25(OH)2D3, denoted as the ratio of the apparent intrinsic clearances, was similar (Table 4). In comparison, the force fit yielded higher coefficient of variations but a higher MSC (3.58), showing that this strategy was superior over that of the sequential fits. Moreover, when the traditional PBPK intestine model (Cong et al., 2000) was used, a slightly inferior fit with higher coefficients of variation, lower MSC, and higher sum of squared residuals were observed (data not shown).
In this study, we have used both wild-type fxr(+/+) and knockout fxr(−/−) mice to discern whether regulation of Mdr1a/P-gp observed in the rat after 1,25(OH)2D3 dosing was caused by the direct actions of the VDR or through indirect actions by the FXR (Chow et al., 2009, 2010). We appraised whether the VDR played a direct role in the up-regulation of the Mdr1 gene in the mouse. We found that there existed different responses between rats and mice caused by species differences that could be the result of different regulatory responses from nuclear receptors. First, ASBT induction with 1,25(OH)2D3, although observable in the rat, is deemed nonexistent in the mouse because of the presence of liver receptor homolog 1 cis-acting element on the mouse ASBT promoter in the intestine that exerts a negative feedback on ASBT on FXR activation (Chen et al., 2003, 2006). For this reason, the FXR effects are lessened in the mouse, and both ASBT mRNA and protein expression and portal bile acid concentrations were unchanged (data not shown). By contrast, the low levels of VDR present in the rat liver (Gascon-Barré et al., 2003; Chow et al., 2009) are unlikely to forge direct links between transactivation of the Mdr1 gene and the VDR. Rather, FXR effects are suspected to be operative, especially in rat liver because of the low VDR levels and increased portal bile acids concentrations (Chow et al., 2009, 2011). Second, protein levels of VDR in the mouse liver are relatively higher than those in the rat liver (Chow et al., 2010) in comparison with the ileum, and, upon activation, VDR could potentially be exerting a greater direct effect on Mdr1/P-gp expression in the mouse liver in vivo. These perceived differences aptly explain the different VDR observations on changes in Mdr1/P-gp expression of the mouse and rat. There was no induction of ileal Mdr1/P-gp in our study, and a slight, but insignificant, increase in Mdr1 mRNA was observed in livers treated with 1,25(OH)2D3 in fxr(+/+) mice but not fxr(−/−) mice (Fig. 3); therefore, the involvement of FXR in the regulation of hepatic Mdr1 in the rat cannot be ruled out. Our previous investigation had shown that intraperitoneal administration of 1,25(OH)2D3 to rats did not elevate P-gp levels in the intestine (Chow et al., 2010), although P-gp was increased in the liver (Chow et al., 2009). The lack of induction of intestinal Mdr1 in both the mouse and rat despite the abundance of VDR could be caused by low levels of 1,25(OH)2D3 reaching the intestine or the amount of 1,25(OH)2D3 needed for activation (Chow et al., 2009).
Of significance is that VDR activation increased Mdr1 and P-gp expression in the kidneys and brains of both the fxr(+/+) and fxr(−/−) mice (Fig. 3), and the mechanism of induction is independent of FXR. The induction of renal Mdr1/P-gp in 1,25(OH)2D3-treated mice led to increases in renal and total body clearances and lowered blood AUC(0→∞), and these shortened the elimination half-life (Table 2). The up-regulation of P-gp expression in the kidney is thus expected to exert a significant impact on the disposition of digoxin, a P-gp substrate that is primarily renally excreted in humans. However, the lower sensitivity of the renal clearance to changes in efflux P-gp activity was caused by the high fraction reabsorbed for digoxin (Fig. 8 and Table 5). For other renally excreted compounds that are less reabsorbed, P-gp induction is expected to play a greater role in increasing the renal clearance. Increases in brain P-gp levels also led to lower brain accumulation (Fig. 6D and Table 2) and lower brain/blood partitioning (Fig. 7D) of digoxin even though VDR mRNA and protein expression in the murine brain was low. There was a clear decrease in the brain/blood AUC ratio of the treated group versus the control group between 60 to 600 min, and the decrease in ratio was the greatest at 600 min (29% decrease).
The induction caused by 1,25(OH)2D3 on the up-regulation of the Mdr1 gene via the VDR was explained using PBPK modeling, which predicted the changes in digoxin disposition in different tissues as a result of increased P-gp activity in the kidney and brain after1,25(OH)2D3 treatment in the mouse. The PBPK model accurately predicted the data pertaining to digoxin disposition in the blood, urine, feces, brain, liver, kidney, heart, and small intestine (Figs. 4 and 6) and provided an accurate description of the insignificant P-gp activity in the intestine and the less than expected effect on renal secretion caused by the high extent of digoxin reabsorption in the kidney (Fig. 8 and Tables 4 and 5). By contrast, the renal clearances of digoxin in humans (125 ml/min) and rats (1.25 ml/min) (Harrison and Gibaldi, 1977a,b) were slightly larger than, or comparable with, their corresponding filtration clearances [90 ml/min in humans and 0.80 ml/min in rats estimated from the literature (Steiness, 1974; Evans et al., 1990; Davies and Morris, 1993)], suggesting that secretion in these species plays a major role in renal excretion. These observations point to species differences in renal excretion and indicate that a higher reabsorption of digoxin occurs in the murine kidney.
This study demonstrates that 1,25(OH)2D3 treatment increased P-gp levels in both the brain and kidney in mice. This observation leads one to consider treatment with 1,25(OH)2D3 or the vitamin D analogs and a P-gp substrate drug would lead to increased renal clearance of drugs. The up-regulation of brain MDR1/P-gp by 1,25(OH)2D3 would have a significant impact on drug disposition, cell homeostasis, and altered pharmacological and toxicological outcomes. Indeed, high doses of 1,25(OH)2D3 and the vitamin D analogs have been used as a therapeutic class of drugs for the treatment of hyperparathyroidism, kidney diseases, and cancer (Masuda and Jones, 2006). These treatments may have the potential to change transporters and enzyme expressions and alter drug disposition (Chow et al., 2009, 2010, 2011). DDIs involving P-gp in the brain may be beneficial or detrimental. For example, vitamin D analogs given concomitantly with P-gp substrates targeting the brain will increase drug efflux from the tissue, rendering the drug ineffective. When 1,25(OH)2D3 is used in combination with anticancer agents such as paclitaxel, a known P-gp substrate, to treat cancer (Masuda and Jones, 2006), lower therapeutic effects may result in the brain. P-gp substrates targeting the brain include the antiretrovirals such as atazanavir, ritonavir, and saquinavir for HIV treatment, antipsychotics such as risperidone (Kim, 2002; Zastre et al., 2009), antiepileptic drugs such as topiramate (Luna-Tortós et al., 2009), or central nervous system drugs such as morphine (Groenendaal et al., 2007), and all of these probably will be affected. On the other hand, increased P-gp activity may decrease brain concentration of a drug such as oseltamivir (Tamiflu), a drug used for influenza treatment, which is a P-gp substrate (Morimoto et al., 2008). An up-regulation of P-gp may decrease brain concentration of oseltamivir in this case.
In summary, the VDR can up-regulate P-gp expression in the kidney and brain in mouse in vivo independently of the FXR. Unequivocally, the induction of Mdr1/P-gp in the rat and mouse by the VDR has been translated to observations in human cell lines, and the intestinal P-gp may be further involved in DDIs. VDR activation of MDR1 mRNA and P-gp had been observed in human colonic cell lines (Aiba et al., 2005; Fan et al., 2009; Tachibana et al., 2009), a notion consistent with the existence of vitamin D response elements (Saeki et al., 2008). As a result, it is highly likely that 1,25(OH)2D3 treatment in combination with other therapeutic agents would cause significant DDIs.
Participated in research design: Chow, Durk, Cummins, and Pang.
Conducted experiments: Chow and Durk.
Contributed new reagents or analytic tools: Pang.
Performed data analysis: Chow, Durk, and Pang.
Wrote or contributed to the writing of the manuscript: Chow, Durk, Cummins, and Pang.
In the equations below, B/P is the blood to plasma partition coefficient; V is the volume of the tissue; C is the concentration; Alumen and Aurine are the amounts of [3H]digoxin excreted in feces and urine compartment, respectively; Q is the blood flow rate; QHA and QPV are the blood flow rate of hepatic artery and portal blood; QHV is the sum of QHA and QPV; KTB is the tissue to blood concentration ratio, assessed as the AUC tissue/blood ratio; fB, fP,fBr, fK, fL, and fI are the unbound fractions in blood, plasma, brain, kidney, liver, and intestine, respectively; fQ is the proportion of intestinal blood flow perfusing the enterocyte region in the small intestine (Liu et al., 2006); CLinBr and CLefBr are the intrinsic influx and efflux in the brain, respectively; ka is the absorption rate constant of the intestine; GFR is the glomerular filtration rate; and FR is the fraction reabsorbed in the kidney. The mass balance equations are listed below.
In the blood (B) compartment:
In the brain tissue (Br) compartment:
In the brain blood compartment:
In the heart compartment:
In other tissue compartments:
In the kidney compartment:
In the above equation, the arterial unbound plasma concentration equals fPCB/(B/P). The filtered and excreted components are reabsorbed according to the fraction reabsorbed, FR.
In the liver compartment, with segregated flows returning from the intestine: where QHV is the sum of QHA and QPV.
In the intestinal compartments of the small intestine:
In the fecal compartment:
This work was supported by the Canadian Institutes for Health Research [Grant MOP89850]. E.C.Y.C. was supported by a University of Toronto Open Fellowship and a National Sciences and Engineering Research Council of Canada Alexander Graham Bell Canada Graduate Scholarship. M.R.D. was supported by a Canadian Institutes of Health Research Strategic Training Grant in Biological Therapeutics.
Article, publication date, and citation information can be found at http://jpet.aspetjournals.org.
- 1α,25-dihydroxyvitamin D3
- apical sodium-dependent bile acid transporter
- area under the curve
- drug-drug interaction
- glyceraldehyde 3-phosphate dehydrogenase
- farnesoid X receptor
- high-pressure liquid chromatography
- multidrug resistance protein 1
- physiologically based pharmacokinetic model
- phenylmethylsulfonyl fluoride
- pregnane X receptor
- Tris-buffered saline with 0.1% Tween 20
- vitamin D receptor
- polymerase chain reaction
- fraction reabsorbed
- blood/plasma concentration ratio
- glomerular filtration rate
- model selection criterion.
- Received January 11, 2011.
- Accepted March 16, 2011.
- Copyright © 2011 by The American Society for Pharmacology and Experimental Therapeutics