Blood cells are considered an important distributional compartment for metformin based on the high blood-to-plasma partition ratio (B/P) in humans (>10 at Cmin). However, literature reports of metformin's intrinsic in vitro B/P values are lacking. At present, the extent and rate of metformin cellular partitioning was determined in incubations of fresh human and rat blood with [14C]metformin for up to 1 week at concentrations spanning steady-state plasma Cmin, Cmax, and a concentration associated with lactic acidosis. The results showed that metformin's intrinsic equilibrium B/P was ∼0.8–1.4 in blood, which is <10% of the reported clinical value. Kinetics of metformin partitioning into human blood cells and repartitioning back into plasma were slow (repartitioning half-life ∼32–39 hours). These data, along with in vivo rapid and efficient renal clearance of plasma metformin (plasma renal extraction ratio ∼90%–100%), explain why the clinical terminal half-life of metformin in plasma (6 hours) is 3- to 4-fold shorter than the half-life in whole blood (18 hours) and erythrocytes (23 hours). The rate constant for metformin repartitioning from blood cells to plasma (∼0.02 h−1) is far slower than the clinical renal elimination rate constant (0.3 h−1). Blood distributional rate constants were incorporated into a metformin physiologically-based pharmacokinetic model, which predicted the differential elimination half-life in plasma and blood. The present study demonstrates that the extent of cellular drug partitioning in blood observed in a dynamic in vivo system may be very different from the static in vitro values when repartitioning from blood cells is far slower than clearance of drug in plasma.
The intrinsic rate and extent of blood cell partitioning of drugs can be reliably determined in a static in vitro system due to the absence of confounding processes present in vivo, including absorption, competing distribution to other tissues, and clearance (Hinderling, 1997). Although metformin is a first-line treatment of type 2 diabetes mellitus that has been commonly used for decades (Gong et al., 2012), its in vitro partitioning properties into the cellular component of blood have not been documented in the literature. Instead, erythrocytes are presumed to be an important distributional compartment for metformin based on the markedly longer blood versus plasma half-life in humans (Glucophage, 2009). Early clinical studies described metformin's in vivo blood-to-plasma partition ratio (B/P) value as high (B/P > 10 at Cmin), and partitioning as slow based on the apparent time-dependent B/P, which increased over the duration of the dosing interval (Sirtori et al., 1978; Pentikainen et al., 1979; Tucker et al., 1981). Following oral administration of a single 1.5-g dose in humans, metformin B/P gradually increased from <1 between 0 and 4 hours to >10 at 24 hours, with the B/P value becoming greater than unity at 4 hours (Tucker et al., 1981). More recent clinical studies demonstrated metformin elimination half-life in erythrocytes (23.4 ± 1.9 hours) to be considerably longer than that in plasma (2.7 ± 0.2 hours) (Robert et al., 2003). In one case study report, plasma metformin levels decreased 119-fold from Cmax after 3 days, while erythrocyte metformin declined only 3-fold from Cmax over the same period (Lalau and Lacroix, 2003).
Metformin plasma concentration–time profiles in humans are biexponential, with the alpha distributional phase occurring during the first 12 hours (Noel, 1979). These relatively slow distributional kinetics, combined with the high human volume of distribution (571 liters), empirically supported in vivo metformin distribution into a “deep” compartment (Noel, 1979). Based on the high clinical Cmin B/P values, erythrocytes were proposed to be an important contributor to this deep distributional compartment (Noel, 1979; Lalau and Lacroix, 2003). These observations are reflected in the product label, which states that “metformin partitions into erythrocytes, most likely as a function of time … erythrocyte mass may be a compartment of distribution” (Glucophage, 2009).
Due to the unusual pharmacokinetic properties of metformin, which include both slower distribution and “flip-flop” absorption kinetics versus plasma clearance (Stepensky et al., 2001), it is difficult to reliably determine the intrinsic rate and extent of blood cell partitioning in vivo. In contrast, intrinsic blood cell partitioning parameters can be accurately characterized in a static in vitro system (Wallace and Reigelman, 1977). The aim of the present study was to determine the intrinsic extent and rate of blood cell partitioning of metformin in an in vitro system to help contextualize previously reported in vivo pharmacokinetic observations.
Materials and Methods
[14C]Metformin hydrochloride was purchased from Moravek Biochemicals (Brea, CA) with a stated specific activity of 110.2 mCi/mmol and a stated radiochemical purity of 99.4%. Fresh human blood was collected by venipuncture at the GlaxoSmithKline employee clinic (Research Triangle Park, NC) from five consenting healthy volunteers (male and female), who abstained from drug use for 1 week and fasted overnight. The human biologic samples were sourced ethically, and their research use was in accord with the terms of the informed consent. Fresh blood from 10 nonfasted male Sprague-Dawley rats (Charles River Laboratories International, Wilmington, MA) was collected by cardiac puncture under terminal anesthesia (2%–5% isoflurane in an induction chamber). All studies were conducted in accordance with the GlaxoSmithKline Policy on the Care, Welfare and Treatment of Laboratory Animals and were reviewed either by the Institutional Animal Care and Use Committee at GlaxoSmithKline or by the ethical review process at the institution where the work was performed. All blood samples were collected into tubes containing K2-EDTA as anticoagulant and kept on ice before incubation. Hemoglobin Assay Kit was purchased from Sigma-Aldrich (St. Louis, MO). Reagents for liquid scintillation counting (LSC) were obtained from PerkinElmer (Waltham, MA).
Incubation of Metformin with Blood.
[14C]Metformin hydrochloride was dissolved in 50% acetonitrile to obtain a 10 mM metformin solution, which was further diluted in 50% acetonitrile to provide stock solution of [14C]metformin at 7.5, 1.5, and 0.075 mM (1000, 200, and 10 μg/ml). The [14C]metformin stock solution was spiked into aliquots of blood that was prewarmed at 37°C for 5 minutes to achieve a final concentration of 10, 2, and 0.1 μg/ml, respectively, and the final organic solvent concentration was 0.5%. Although the use of organic solvent is needless in the case of highly water-soluble metformin, <1% organic solvent concentration is commonly used in blood cell partitioning experiments, and there is no evidence to suggest that these low organic solvent concentrations compromise the physiologic relevance of the in vitro B/P value (Kalamaridis and DiLoreto, 2014). Each sample was mixed quickly by inversion and then placed on a rotating mixer (22 reversals/min) in a non-CO2 incubation chamber maintained at 37°C. In a preliminary study, samples were collected at 0, 1, 2, 4, 24, and 48 hours after incubation, and the results indicated that even after 48 hours of incubation, the partitioning of metformin into blood cells was not fully equilibrated (the B/P value increased from 0.6 at 0 hour to 0.8 at 48 hours with an initial blood concentration of 1 μg/ml). Therefore, in subsequent experiments, blood and plasma samples were collected at 0, 4, 24, 72, and 168 hours of incubation. The stability of [14C]metformin was confirmed by high-performance liquid chromatography with radiochemical detection (radiochemical purity was 99.2% at day 0, 98.8% at 5 months for 0.22 mM [14C]metformin in 90% ethanol); ∼100% recovery of intravenous metformin as unchanged drug in urine indicated stability in the human body and biologic matrices at 37°C over 48 hours (Pentikainen et al., 1979). Plasma was obtained by centrifuging blood at 1200g for 10 minutes. For human samples, blood from three volunteers was incubated individually. For rat samples, blood from 10 donors was pooled and incubated as triplicates. B/P of metformin was calculated at all time points. Plasma concentrations over time were used to calculate partitioning rate constant and metformin half-life in plasma.
To further investigate the repartitioning of metformin from human blood cells into plasma, human blood (4 ml, in triplicates) from a single male subject was spiked with [14C]metformin stock solution to obtain the final concentrations of 10, 2, and 0.1 μg/ml, respectively, and incubated at 37°C as described above for 72 hours (the longest time that human blood can be incubated and corrected for hemolysis as described below). Then the blood was centrifuged and plasma was removed. The remaining blood cells were washed with phosphate buffer (pH 7.4, room temperature) three times, and fresh blank human plasma was added to reconstitute to a hematocrit of ∼0.5 (Wallace and Reigelman, 1977). The reconstituted samples were incubated again at 37°C, and blood and plasma samples were collected at 0, 1, 2, 4, and 24 hours. Blood cellular concentrations after reincubation over time were used to calculate repartitioning rate constant and metformin half-life in blood cells.
Hematocrit was determined by centrifugation of filled hematocrit capillary tubes (Drummond Scientific, Broomall, PA) using an IEC MB Microhematocrit centrifuge (International Equipment Co., Chattanooga, TN) at full speed for 5 minutes and by reading using a Clay Adams microhematocrit reader (Becton Dickinson, Franklin Lakes, NJ).
LSC for Blood and Plasma.
Blood samples were combusted by using a Packard Tri-Carb 307 Automatic Sample Oxidizer (PerkinElmer), with Carbosorb E as the 14CO2-trapping agent and Permafluor E as the scintillant, prior to assay using LSC. Plasma samples were mixed with Ultima Gold LSC cocktail and counted by Tri-Carb 2910TR Liquid Scintillation Analyzer (PerkinElmer).
Correction of Plasma Concentration by Degree of Hemolysis.
The amount of hemoglobin in blood or plasma was determined by using Hemoglobin Assay Kit following the manufacturer’s instructions. The degree of hemolysis (DH) was calculated asHemolysis releases red blood cell–associated drugs into plasma. Therefore, to correct for the impact of hemolysis on the measured plasma concentrations (Cp) of metformin, the following equation was used to estimate plasma concentration (Cp′) of metformin in the absence of hemolysis based on values of blood concentration of metformin (Cb) and hematocrit (He) (Tan et al., 2012):
The blood and plasma concentrations of metformin over incubation time were analyzed by noncompartmental analysis with Phoenix WinNonlin (v.6.3; Certara, Princeton, NJ) to generate partitioning rate constant and half-life. Data are presented as mean ± S.D.
Blood cell partitioning and repartitioning rates determined in the present study were combined with literature data to simulate clinical plasma and blood concentration–time profiles. These bottom-up in vitro–to–in vivo simulations were carried out using the Simcyp Population-based Simulator (v.14.1; Certara). The default Simcyp metformin compound file was used with the following modifications. A minimal physiologically-based pharmacokinetic (PBPK) model was used to enable addition of a single adjusting compartment (SAC), which was used to describe the blood cell distribution of metformin (Fig. 1). The volume of the SAC was initially set to the Simcyp default physiologic blood volume (0.06 l/kg), and was further optimized (0.07 l/kg) to improve the recovery of metformin blood concentration–time profiles. The kin (0.006 h−1) and kout (0.02 h−1) of the SAC were set to the mean partitioning rate from plasma to blood cells and mean repartitioning rate from blood cells to plasma determined in vitro in the present study, respectively.
The in vitro B/P of metformin was determined in human and rat blood at initial blood concentrations of 0.1, 2, and 10 μg/ml, which correspond to the Cmin, Cmax, and a concentration exceeding the threshold associated with lactic acidosis (≥4 μg/ml) in humans, respectively (Tucker et al., 1981; Lalau and Lacroix, 2003). In human blood incubations, the B/P of metformin increased from 0.6–0.7 at the initial time point to 0.8–1.2 at 72 hours of incubation (Table 1). In rat blood, the B/P of metformin increased from the initial value of 0.6–0.7 to 0.8–1.4 at 72 and 168 hours of incubation (Table 1). Extensive hemolysis (>60%) in human blood after 1 week of incubation precluded quantification of the B/P at 168 hours (Supplemental Table 1), but less extensive rat blood hemolysis at 168 hours permitted calculation of the B/P value (Helman et al., 1974). High hemolysis in human blood at 168 hours may have been caused by compromised sterility, which was not directly assessed.
The first-order rate constants and associated half-life values for metformin distribution from plasma to blood cells, as well as redistribution from cells to plasma, are presented in Table 2.
The PBPK model (Fig. 1) was used to simulate plasma and blood concentration–time profiles following a single oral dose of 1.5 g (Fig. 2). The observed plasma and blood concentration–time profiles determined in a single subject (Tucker et al., 1981) were overlaid with the simulations. The simulated plasma and blood concentrations were in reasonable agreement with observed values: the predicted area under the curve and Cmax in plasma and blood were all within 1.5-fold of the reported values. Most importantly, the predicted elimination half-life in blood (35 hours) was within 1.5-fold of the observed elimination half-life of ∼24 hours. In summary, the developed model integrating blood cell partitioning rates determined in vitro predicted the differential elimination half-life in plasma and blood in humans.
In vitro rat blood metformin (1–20 μg/ml) B/P was previously reported to be 0.7–0.8 at 2 hours (Choi et al., 2006). Although the Choi et al. values are consistent with the initial B/P in the present study (Table 1), as well as with the in vivo value (B/P < 1) at 2 hours after oral dosing (Tucker et al., 1981), this 2-hour B/P value does not represent distributional equilibrium (Noel, 1979). Therefore, it had been assumed that the relatively low B/P value reported by Choi et al. was an artifact of metformin not having achieved distributional equilibrium in blood, and that the B/P value would continue increasing over a longer incubation period as observed in humans in vivo over 24 hours (Tucker et al., 1981). The novelty of the current study is that blood incubations were carried out to equilibrium (72–168 hours), and the metformin equilibrium intrinsic B/P value was demonstrated to be unity (0.8–1.4) at all tested concentrations in both rat and human blood.
The in vitro rate constant for metformin partitioning from human plasma to blood cells (Table 2) was ∼20% of the in vivo value for metformin entering the distributional compartment from the central compartment (0.025 ± 0.008 h−1) (Noel, 1979). This difference is consistent with the presence of other distributional compartments for metformin in vivo, including the small intestine, kidney, liver, and skeletal muscle (Wilcock and Bailey, 1994; Scheen, 1996). Distribution into these organs is a carrier-facilitated process, and thus distributional equilibrium is achieved faster in vivo (Gong et al., 2012; Zamek-Gliszczynski et al., 2013a). Although metformin tissue distribution does not stand out as particularly high individually [e.g., kidney/plasma ratio: 11.9 ± 1.1; and liver/plasma ratio: 4.5 ± 0.6 in mice (Zamek-Gliszczynski et al., 2013a)], collectively these additional sites of distribution contribute to the higher partitioning rate observed in vivo. This explanation of the in vitro versus in vivo discrepancy in rate constants is supported by the PBPK model (Figs. 1 and 2), which predicted the differential elimination half-life in plasma and blood in humans.
Likewise, the repartitioning rate constant from human blood cells into plasma (Table 2) was ∼30% of the rate constant for metformin exiting the distributional to the central compartment (0.067 ± 0.018 h−1) (Noel, 1979), which again reflected the existence of other distributional compartments in vivo. However, the in vitro blood cell repartitioning kinetics are consistent with those reported in vivo. The in vitro half-life for the transfer of metformin from blood cells to plasma (32.0–38.6 hours) determined in the present study is 1.3–1.6 times longer than the in vivo erythrocyte half-life value (23.4 ± 1.9 hours) (Robert et al., 2003), which is in reasonably good agreement considering that measurements were made in vitro versus in vivo.
Surprisingly, metformin repartitioning from blood cells into plasma was 2- to 5-fold faster than distribution from plasma into the cellular component of blood (Table 2). These in vitro findings are conceptually consistent with clinical distributional kinetics, where the rate constant for metformin redistribution from the distributional compartment back to the central compartment is 2- to 3-fold faster than the rate constant for transfer from the central compartment to the distributional compartment (Noel, 1979). The mechanistic explanation for these observations is unknown. Theoretically, plasma membrane permeability should be the same in both directions (Kalvass and Pollack, 2007; Zamek-Gliszczynski et al., 2013b). Furthermore, the electronegative membrane potential thermodynamically favors cellular partitioning of cationic metformin (Gong et al., 2012), and would support the opposite trend than the one observed. Finally, this observation cannot be attributed to more extensive nonspecific plasma protein binding, which is negligible for metformin (Glucophage, 2009). At a gross kinetic level, these observations can be reconciled by transport process(es), which are relatively more efficient in cellular efflux than uptake of metformin, as has been extensively demonstrated in central nervous system drug distribution (Kalvass et al., 2013). Transporters primarily involved in metformin's intestinal, hepatic, and skeletal muscle distribution and active tubular secretion are organic cation transporters, multidrug and toxin extrusion transporters, and plasma membrane monoamine transporter (Graham et al., 2011). ATP-binding cassette transporters are not known to transport metformin to an appreciable extent in vivo, although ex vivo studies suggested that P-glycoprotein and breast cancer resistance protein may efflux metformin at the blood-placental barrier (Hemauer et al., 2010). However, at the present time, insufficient data are available on the qualitative or quantitative protein expression of these transporters in blood cells to speculate about the mechanism(s) underlying the apparently more efficient efflux from than influx into blood cells.
Metformin's intrinsic B/P value of approximately unity is <10% of the value reported in vivo, where the B/P was >10 at 24 hours after oral dosing (Tucker et al., 1981). This >10-fold discrepancy can be reconciled by considering the distributional and plasma clearance kinetics. The rate constant for metformin repartitioning from blood cells to plasma (0.018–0.022 h−1) is an order of magnitude slower than the rate constant for plasma renal clearance in humans (0.293 ± 0.030 hour−1) (Noel, 1979). Metformin is eliminated rapidly and efficiently by the kidneys, the sole clearing organ for this drug (Glucophage, 2009). Metformin's renal extraction ratio is 90%–100% in mice, rats, and humans (plasma renal clearance approximates renal plasma flow rate) (Davies and Morris, 1993; Choi et al., 2006; Glucophage, 2009; Higgins et al., 2012; Zamek-Gliszczynski et al., 2013a). Relative to the 6.2-hour plasma half-life in humans, the half-life of repartitioning from the cellular component of blood to plasma is 6-fold longer (Table 2), resulting in a longer in vivo whole-blood half-life of 17.6 hours and an even longer in vivo erythrocyte half-life of 23.4 hours (Robert et al., 2003; Glucophage, 2009). This remarkable kinetic difference between blood cell redistribution and renal elimination of plasma metformin in vivo leads to the perceived high B/P when measured at Cmin (Tucker et al., 1981). To our knowledge, metformin exhibits the most drastic difference in the in vitro and in vivo blood cell partitioning due to very slow blood cell distribution relative to plasma clearance (Hinderling, 1997).
The putative high B/P of metformin displayed in vivo insinuates a potential mechanism of drug-drug interactions (DDIs) between metformin and other drugs, in addition to the broadly recognized inhibition of renal transporters (Tornio et al., 2012). If a DDI perpetrator drug were to decrease the B/P of metformin, then it could increase metformin systemic exposure measured in plasma. Results from the current study demonstrate that the equilibrium in vitro B/P value is not helpful in elucidating such a hypothetical distributional DDI, because the intrinsic B/P of unity is unlikely to be shifted. In addition, red blood cell volume accounts for <1% of metformin’s volume of distribution when plasma is used as the reference fluid for pharmacokinetics (Davies and Morris, 1993; Glucophage, 2009). Therefore, the small red blood cell contribution to metformin’s overall volume of distribution further rules out the possibility of such a distributional DDI. Finally, a theoretical shift in the B/P value may impact drug distribution, but not clearance, which ultimately determines exposure.
Based on the results of the present work, it may be tempting to speculate that metformin blood pharmacokinetics are a more relevant measure of systemic exposure than its plasma pharmacokinetics. Clinical blood metformin exposures far exceed those in plasma at time points beyond 4 hours (Tucker et al., 1981), which will lead to marked differences in basic pharmacokinetic parameters; namely, blood terminal half-life is longer due to lower apparent blood clearance. However, blood cell metformin is effectively not available for pharmacological activity, because following slow redistribution from blood cells to plasma, it undergoes rapid renal clearance. In contrast, plasma metformin is 100% unbound and available for pharmacological activity. As such, plasma metformin is the relevant measure of systemic drug exposure, and not total blood exposure. Finally, unlike for most drugs with rapid blood cell equilibration, metformin plasma and blood pharmacokinetic parameters cannot be converted based on the B/P value, because metformin does not rapidly attain equilibrium in blood cells, and its in vivo B/P value is time dependent.
In summary, the present study for the first time characterized the intrinsic extent and rate of metformin partitioning into blood cells in a static in vitro system. The results showed that the intrinsic B/P is around unity in both human and rat blood, and the rate of blood cellular redistribution is slow relative to the rate of renal clearance of plasma metformin. This study demonstrates that the extent of cellular drug partitioning in blood observed in a dynamic in vivo system may be very different from the static in vitro values, when the repartitioning from blood cells is far slower than clearance of drug in plasma.
The authors thank Katie L. Olson for obtaining fresh rat blood and oxidizing blood samples for liquid scintillation counting.
Participated in research design: Xie, Bowers, Zamek-Gliszczynski.
Conducted experiments: Xie.
Performed data analysis: Xie, Ke, Zamek-Gliszczynski.
Wrote or contributed to the writing of the manuscript: Xie, Ke, Bowers, Zamek-Gliszczynski.
- blood-to-plasma partition ratio
- drug-drug interaction
- liquid scintillation counting
- physiologically-based pharmacokinetic
- single adjusting compartment
- Copyright © 2015 by The American Society for Pharmacology and Experimental Therapeutics