Abstract
Although metformin hepatic distribution is critical to pharmacological activity, the drug is cleared by urinary excretion. Metformin hepatobiliary disposition was studied in rodents representative of clinical pharmacokinetics to elucidate why metformin is not appreciably eliminated in bile. On average, 1.0% ± 0.1% of the metformin oral dose was present in the liver (liver/plasma ratio = 4.5 ± 0.6) over a pharmacologically relevant dose and time range in mice (10–300 mg/kg; 1.5–2.5 hours; Tmax = 1.4 ± 0.5; bioavailability > 59%). Distribution to the kidneys was not markedly higher, which contained 0.87% ± 0.08% of the oral dose (kidney/plasma ratio = 11.9 ± 1.1). However, only 0.11% ± 0.02% of the intravenous and bioavailable oral dose was recovered in bile, suggesting that biliary excretion is not the only route of clearance for hepatic metformin. Consistent with negligible biliary excretion, pharmacokinetics were unaffected by bile duct cannulation, proving the effective absence of enterohepatic cycling. In single-pass liver perfusion studies, 2.4% ± 0.3% of the perfused metformin dose was distributed to the liver, which underwent >300-fold greater sinusoidal than biliary excretion during the subsequent drug-free washout perfusion (74.0% ± 39.3% versus 0.222% ± 0.003% recovery of hepatic metformin in perfusate versus bile, respectively). These studies demonstrate that despite similar magnitude of metformin liver and kidney distribution, metformin biliary excretion is negligible due to predominant sinusoidal efflux from the liver.
Introduction
Although the antidiabetic agent, metformin, is cleared by urinary excretion, its hepatic distribution is critical to pharmacological activity (Gong et al., 2012). Since metformin intravenous dose was completely recovered as parent in urine, and because drug-related material was not detected in feces, it has been broadly assumed that biliary excretion does not contribute to systemic clearance (Tucker et al., 1981; Glucophage, 2009; Graham et al., 2011). This assumption was confirmed in biliary excretion studies, in which ≤0.3% of the intravenous dose was recovered in bile (Maeda et al., 2007; Ito et al., 2010). Nonetheless, metformin distribution to the kidney is only approximately 3-fold higher than metformin distribution to the liver (kidney/plasma ratio = 6–25; liver/plasma ratio = 2–7) (Maeda et al., 2007; Ito et al., 2010; Higgins et al., 2012), which raises the question why is the drug wholly eliminated in urine with negligible biliary excretion?
Metformin has virtually no passive membrane permeability, such that cellular ingress and egress are transporter-governed processes (Graham et al., 2011). Hepatic uptake is mediated primarily by organic cation transporter (OCT) 1 and biliary excretion by multidrug and toxin extrusion protein (MATE) 1 (Gong et al., 2012). Various studies established that metformin hepatobiliary disposition is mechanistically and kinetically conserved between mice/rats and humans (Wang et al., 2002; Tsuda et al., 2009; Graham et al., 2011). Notably, OCT1 and MATE1 modulation have large effects on hepatic distribution and thus are pharmacodynamically consequential. Metformin hepatic concentrations were increased approximately 20-fold in Mate1-knockout mice, (Tsuda et al., 2009). Liver distribution was reduced up to 30-fold in mice lacking hepatic Oct1 (Wang et al., 2002). Clinically, OCT1 and MATE1 genetic polymorphisms resulted in altered metformin pharmacodynamics, consistent with the expected changes in hepatic drug exposure (Shu et al., 2007; Stocker et al., 2013). The active secretory component of metformin renal clearance, which accounts for approximately 80% of systemic clearance (glomerular filtration contributes approximately 20%), is mediated by OCT2 uptake into the renal proximal tubule followed by secretion into urine via MATE1 and MATE2-K (Gong et al., 2012). Renal clearance in rodents is kinetically analogous to that in humans; however, uptake is mechanistically mediated by both Oct1 and Oct2, whereas secretion into urine occurs solely via Mate1 in mice and rats (Tsuda et al., 2009; Higgins et al., 2012).
Why is metformin wholly eliminated in urine when kidney and liver distribution are similar in magnitude? One proposed answer was markedly greater renal uptake activity based on 10- to 100-fold greater in vitro transport efficiency by renal OCT2 than hepatic OCT1 (Kimura et al., 2005). This hypothesis essentially stipulates that metformin OCT/MATE vectorial transport across the liver is one to two orders of magnitude less efficient than transport across the renal proximal tubule, which is consistent with markedly lower hepatic than renal extraction ratio (approximately 2 and ≤100%, respectively), nearly complete urinary but negligible biliary excretion, and yet a comparable magnitude of tissue distribution (Chou, 2000; Ito et al., 2010; Higgins et al., 2012). Considerably less efficient metformin transport across the liver versus the renal proximal tubule is supported at a gross in vivo pharmacokinetic level (Ito et al., 2010). However, other mechanistic studies showed that renal OCT2 was only approximately 3-fold more efficient at transporting metformin than hepatic OCT1 (Choi et al., 2007), whereas renal and hepatic OCT expression was insufficiently different to explain the 50-fold difference in the extraction ratio (Nishimura and Naito, 2005; Tsuda et al., 2009). The hypothesis of less efficient metformin hepatobiliary transport appears to be correct overall, but its mechanistic justification may be more complex than merely less efficient hepatic OCT uptake activity.
Further complicating the understanding of metformin hepatobiliary disposition are the first-pass pharmacodynamic effect and absorption rate-limited oral pharmacokinetics (Stepensky et al., 2001, 2002). Specifically, metformin exhibits route of administration–dependent pharmacodynamics, in which the glucose-lowering effect increases in the following order of administration: intravenous < intraportal < oral (Stepensky et al., 2002). This first-pass pharmacodynamic effect, combined with absorption rate-limited oral pharmacokinetics (Stepensky et al., 2001) and >50% oral bioavailability (Tucker et al., 1981), has led to speculation of potentially high first-pass hepatic extraction, biliary clearance, and enterohepatic recycling during oral drug absorption. Although metformin biliary excretion is negligible after intravenous dosing (Tucker et al., 1981; Maeda et al., 2007; Ito et al., 2010), first-pass biliary clearance has not been characterized after oral administration.
This study demonstrates that metformin biliary excretion is negligible due to predominant sinusoidal efflux from the liver, which helps explain why the liver is an organ of distribution and not elimination. The data unequivocally support the absence of enterohepatic cycling after oral drug administration, disproving this potential contributing factor to metformin absorption rate-limited oral pharmacokinetics and first-pass pharmacodynamic effects.
Materials and Methods
Chemicals.
Metformin-HCl and D6-metformin-HCl were purchased from Toronto Research Chemicals (North York, ON, Canada).
Animals.
Wild-type FVB male mice (19–30 g) were purchased from Taconic (Hudson, NY). Male Sprague-Dawley rats (300–360 g) were purchased from Harlan (Indianapolis, IN). The vendor performed rat femoral artery and vein cannulations for in vivo blood sampling and intravenous dosing, respectively. In biliary excretion studies, bile ducts were also cannulated; bile flow was in the normal range for rats (approximately 10 μl/min) after intravenous (10 ± 1 μl/min) and oral (7 ± 3 μl/min) drug administration, as well as during liver perfusions (7 ± 3 μl/min). All animal procedures were approved by the Institutional Animal Care and Use Committee at Covance (Greenfield, IN), Harlan, and Eli Lilly and Company (Indianapolis, IN).
Tissue Drug Content.
Metformin (10, 30, 60, 100, and 300 mg/kg; 10 ml/kg in 1% hydroxyethyl cellulose/0.25% polysorbate-80/0.05% antifoam in water) was administered by oral gavage to mice. Mice were euthanized at 1.5, 2.0, and 2.5 hours postdose, when livers, kidneys, and plasma were collected for analysis of metformin concentrations.
Biliary Excretion and Pharmacokinetics.
The metformin plasma concentration time course was determined over 24 hours in an i.v./oral cross-over study in bile duct–cannulated rats and bile duct–intact rats. Metformin was administered by i.v. bolus injection (5 mg/kg; 2 ml/kg in isotonic phosphate-buffered saline, pH 7.4) and by oral gavage (10 mg/kg; 10 ml/kg in 1% hydroxyethyl cellulose/0.25% polysorbate-80/0.05% antifoam in water). Arterial blood samples were collected at the following times postdose: 0, 0.08, 0.25, 0.5, 1, 2, 4, 8, 12, and 24 hours. In addition, bile was collected in 6-hour intervals in bile duct–cannulated rats.
Rat Liver Perfusions.
Livers were prepared for perfusion by cannulation of the bile duct, portal vein (inflow), and inferior vena cava above the liver (outflow) as previously described (Zamek-Gliszczynski et al., 2011). After acclimation, livers were perfused in a single-pass manner with Krebs-Henseleit buffer containing 5 μM taurocholate (30 ml/min; 0–30 minutes with 10 μM metformin and drug-free 30–60 minutes). Bile and perfusate samples were collected every 5 minutes for up to 60 minutes; livers were snap-frozen at either 30 or 60 minutes.
Bioanalysis.
All samples were mixed with an organic internal standard solution to precipitate protein and centrifuged, and the resulting supernatants were directly analyzed. Metformin and its internal standard, D6-metformin, were eluted from a C18 column (Betasil Si 2.1 × 50 mm, 5 μm; Thermo Fisher Scientific, Inc., Waltham, MA) with a mobile phase gradient [1% trifluoroacetic acid/1% 1M ammonium bicarbonate in water (A) or acetonitrile (B)] optimized for each matrix. Analytes were detected in positive ion mode using multiple reaction monitoring (Sciex API 4000 triple quadrupole mass spectrometer equipped with a TurboIonSpray interface; Applied Biosystems/MDS, Foster City, CA): metformin 130.1 → 71.1 and D6-metformin 136.1 → 77.1 m/z. The dynamic range was 1–5000 ng/ml in all matrices. Samples with analyte concentrations above the upper limit of quantification were diluted with matrix to within the assay range.
Data Analysis.
Noncompartmental pharmacokinetic parameters were calculated using Watson version 7.4 (Thermo Fisher Scientific, Inc.). Excel 2007 (Microsoft, Redmond, WA) was used for all statistical analyses. The two-tailed t test was used to assess statistical significance between pharmacokinetic parameters estimated in bile duct–cannulated rats and bile duct–intact rats. The minimal criterion for significance was P < 0.05. Data are reported as the mean ± S.E.M. with the associated n reported in all cases.
Results
Table 1 summarizes the fraction of the metformin oral dose present in the liver and kidney over a subefficacious to maximally efficacious dose range (10–300 mg/kg) and time period (1.5–2.5 hours postdose) previously demonstrated to be pharmacologically relevant in mice (Higgins et al., 2012). This time frame spans the systemic Tmax to the time of maximal glucose-lowering effect after oral administration (Stepensky et al., 2002; Higgins et al., 2012). On average, 1.0% ± 0.1% of the metformin oral dose was present in the liver (liver/plasma ratio = 4.5 ± 0.6). For comparison, the kidneys on average contained 0.87% ± 0.08% of the oral dose (kidney/plasma ratio = 11.9 ± 1.1). Metformin oral bioavailability in mice was ≥59% (Higgins et al., 2012); thus, the low fraction of dose present in these tissues is not an artifact of poor absorption.
Metformin pharmacokinetics in bile duct–cannulated rats and bile duct–intact rats are presented in Fig. 1 and Table 2. On average, 0.11% ± 0.02% of intravenous and bioavailable oral dose was recovered in bile over 24 hours, a period accounting for >95% of systemic drug exposure. Consistent with negligible biliary excretion, metformin intravenous and oral pharmacokinetics were identical in bile duct–cannulated rats (Fig. 1; Table 2), supporting the effective absence of enterohepatic cycling.
Since 1.0% ± 0.1% of the metformin oral dose resided in the liver over a pharmacologically relevant period but only 0.11% ± 0.02% of the dose was ultimately recovered in bile, the potential for hepatic sinusoidal efflux was evaluated in perfusion studies (Fig. 2; Table 3). During liver perfusion with 10 μM metformin (0–30 minutes), hepatic bioavailability was nearly complete (98.8% ± 4.7%), although it was relatively lower at the first 5-minute time point (93.5% ± 5.6%). At the end of 30-minute drug perfusion, livers on average contained 2.4% ± 0.3% of total perfused dose (liver/perfusate = 2.1 ± 0.3), which decreased approximately 50-fold to 0.041% ± 0.004% of perfused metformin during the subsequent 30-minute drug-free perfusion. During the 30–60 minute washout perfusion period, metformin present in the liver at 30 minutes was >300-fold preferentially excreted into drug-free perfusate than bile (Fig. 2; Table 3).
Discussion
Metformin hepatic distribution was previously believed to involve unidirectional vectorial transport from blood to bile, with hepatic uptake by OCT1 and subsequent biliary excretion via MATE1 (Graham et al., 2011; Gong et al., 2012). For the first time, this study demonstrated that hepatic metformin is not exclusively excreted into bile but is primarily cleared from the liver via sinusoidal efflux. This finding is consistent with the liver being an organ of distribution but not elimination for metformin, and explains why biliary excretion is so low with no evident enterohepatic cycling.
It was previously proposed that markedly higher renal OCT uptake activity is responsible for metformin clearance by urinary and not biliary excretion (Kimura et al., 2005). Considerably less efficient vectorial transport across the liver than the renal proximal tubule would be consistent with a markedly lower hepatic than renal extraction ratio (approximately 2 and ≤100%, respectively), nearly complete urinary but negligible biliary excretion, and yet a fairly similar magnitude of tissue distribution (Chou, 2000; Ito et al., 2010; Higgins et al., 2012), and is supported at a gross in vivo pharmacokinetic level (Ito et al., 2010). However, the difference in renal versus hepatic OCT transport in other studies was not sufficiently large to fully explain the observed difference in the extent of urinary versus biliary excretion (Nishimura and Naito, 2005; Choi et al., 2007). Findings from this study conceptually support the inefficient transport of metformin from blood to bile; however, a major contributor to this inefficiency is extensive sinusoidal efflux, which counteracts OCT1 hepatic uptake, effectively reducing net uptake activity.
Although sinusoidal efflux is an appreciable clearance pathway for hepatic metformin, it merits noting that liver perfusions may have exaggerated its magnitude relative to biliary excretion. Known metformin transporters are classified as secondary active transport, which relies on various electrochemical gradients: electronegative membrane potential for OCTs and proton exchange for MATEs (Giacomini et al., 2010; Zamek-Gliszczynski et al., 2012). The drug-free washout phase of liver perfusion, when sinusoidal and biliary excretion of hepatic metformin were studied, represents a physiologically unrealistic condition in which the drug is present in the liver but not in the circulation. In this scenario, this infinite liver/perfusate concentration ratio may have enhanced the driving force for sinusoidal efflux. As such, the >300-fold higher hepatic excretion into drug-free perfusate than bile may not be physiologically relevant. Furthermore, an approximately 5-fold greater than physiologic hepatic perfusion/bile flow rate ratio [approximately 4300 in single-pass liver perfusion versus approximately 880 in vivo (Davies and Morris, 1993)] may have further contributed to the overestimation of metformin sinusoidal excretion.
A potentially elevated driving force for sinusoidal efflux during liver perfusion with drug-free buffer may have in turn resulted in underestimation of the fraction of hepatic metformin excreted in bile (0.2% of liver drug content). This point is supported by lower in vivo metformin bile/plasma concentration ratios (1.3 ± 0.3) and bile/perfusate ratios (0.8 ± 0.1) during liver perfusion with metformin than during the washout phase of perfusion, when the bile/perfusate concentration ratio was 10- to 16-fold higher (12.6 ± 3.2). However, even after correcting for the 10- to 16-fold concentration ratio difference, the fraction of hepatic metformin excreted in bile would be at most 3.2%. In vivo biliary recovery of the metformin dose was only 0.11% ± 0.02% but between the systemic Tmax and time of maximal glucose-lowering effect, on average 1.0% ± 0.1% of oral dose was present in the liver. Since biliary recovery was measured to infinity but hepatic content was determined at individual time points, biliary excretion accounts for at most <10% of total metformin clearance from the liver in vivo. In previous murine intravenous infusion studies, biliary dose recovery was ≤0.3% with approximately 2.5% of the dose in the liver at steady state (Maeda et al., 2007; Ito et al., 2010), independently confirming the fraction of hepatic metformin excreted in the bile to be at most <10% in vivo.
Notably, the current estimate of a 0.2 to <10% fraction of hepatic metformin cleared by biliary excretion contradicts the indirect estimate in Mate1-knockout mice. Metformin hepatic distribution was increased 5-fold after genetic ablation of Mate1 (20-fold increase in hepatic concentration = 4-fold increase in systemic concentration × 5-fold in increase in hepatic distribution) (Tsuda et al., 2009), which puts the metformin fraction excreted by Mate1 from the liver at a considerably higher 80% (Zamek-Gliszczynski et al., 2009). The reasons for this discrepancy remain to be elucidated, but Mate1-knockout findings indicate that the magnitude of increase in hepatic distribution is greater than can be accounted for by biliary excretion alone. Thus, sinusoidal efflux is likely to be additionally impaired due to either direct downregulation (e.g., Mate1 is also present on the basolateral membrane or its knockout results in compensatory downregulation of the relevant sinusoidal transporter) or indirect inhibition (e.g., absence of Mate1 results in competition for sinusoidal efflux with an accumulated endogenous Mate1 substrate or reduced driving force for sinusoidal efflux). Supporting the notion that impairment of hepatic Mate1 alters metformin hepatobiliary disposition in more ways than just biliary excretion are inhibition studies with pyrimethamine, in which hepatic metformin concentrations were increased 3-fold, while biliary excretion was unchanged and insufficiently extensive to explain the magnitude of increase in liver drug concentration (Ito et al., 2010). MATE1 is localized to the hepatic canalicular membrane (Otsuka et al., 2005), so the mechanism by which MATE1 affects extrabiliary clearance of hepatic metformin may be complex.
Metformin absorption rate-limited oral pharmacokinetics (Stepensky et al., 2001) and route of administration–dependent first-pass pharmacodynamics [intravenous < intraportal < oral (Stepensky et al., 2002)] have raised the possibility of high first-pass hepatic extraction, biliary excretion, and enterohepatic recycling during oral drug absorption. Metformin oral bioavailability is adequately high to support enterohepatic cycling (Tucker et al., 1981). Previous intravenous biliary/fecal mass balance studies contradict this first-pass hypothesis (Tucker et al., 1981; Maeda et al., 2007; Ito et al., 2010); however, the possibility was not directly tested after oral administration. This study unequivocally proved the absence of first-pass biliary excretion, which was similarly low after oral and intravenous dosing; consequently, enterohepatic cycling was not observed after drug administration by either route. Although the hepatic extraction ratio was initially 6.5 versus 1.2% at steady state in liver perfusions, biliary dose recovery was nonetheless low over the entire perfusion period.
In conclusion, this study demonstrates that metformin biliary excretion is negligible due to predominant sinusoidal efflux from the liver, which helps explain why the liver is an organ of distribution and not elimination. The data unequivocally support the absence of enterohepatic cycling after oral drug administration, disproving this potential contributing factor to metformin absorption rate-limited oral pharmacokinetics and first-pass pharmacodynamic effects.
Authorship Contributions
Participated in research design: Zamek-Gliszczynski, Bao, Day, Higgins.
Conducted experiments: Bao, Day.
Performed data analysis: Zamek-Gliszczynski, Bao, Day, Higgins.
Wrote or contributed to the writing of the manuscript: Zamek-Gliszczynski, Bao, Higgins.
Footnotes
- Received May 22, 2013.
- Accepted September 5, 2013.
Abbreviations
- MATE
- multidrug and toxin extrusion protein
- OCT
- organic cation transporter
- Copyright © 2013 by The American Society for Pharmacology and Experimental Therapeutics