I. Introduction
The significance of studying the kinetics of drug partitioning into red blood cells (RBCsb) in animals and humans is not fully appreciated, although the importance of routine determination of rate and extent of partitioning of investigational drugs has been stressed (Lee et al., 1981b; Hinderling 1984). As will be demonstrated in this review, failure to determine the kinetics of drugs in RBCs may be a lost opportunity. Knowledge of RBC partitioning of compounds enables: (a) a rational choice of appropriate biological fluid, either whole blood, plasma, or serum, for assay; (b) physiologically meaningful referencing of pharmacokinetic parameters of drugs to concentrations in either whole blood, plasma, or serum; (c) in vitro prediction of drug distribution in vivo; (d) determination of plasma protein binding of drugs; and (e) effective screening of drugs whose biophase resides within the RBCs, thereby enabling the study of the effects of drugs on RBCs.
The goals of this review are to: (a) summarize the present knowledge regarding the partitioning of drugs into RBCs, and (b) demonstrate the relevance of knowing the kinetics of RBC uptake of drugs. RBCs of animals and humans are known to be different (Bowyer, 1957). This review deals almost exclusively with human RBCs and their interactions with drugs.
II. Salient Features of Red Blood Cells
Among the cellular constituents of blood, the RBCs represent, by far, the largest population both in number and cell size. The RBCs make up more than 99% of the total cellular space of blood in humans (Diem and Lentner, 1975c). RBCs occupy a volume of approximately 25 to 30 mL·kg−1, of which 71% constitute an aqueous phase (Diem and Lentner, 1975a). A total of approximately 760 g of hemoglobin is contained in the RBCs, representing approximately 10% of the total body proteins of an adult human (Spector, 1956; Diem and Lentner, 1975c; Kawai et al., 1994). Hemoglobin interacts with small diffusible ligands such as O2, CO2, and NO and may be involved in the control of blood pressure (Jia et al., 1996). The RBCs draw energy from glucose metabolism via direct glycolysis and the hexose monophosphate shunt (Beutler et al., 1995a). Individual RBCs are biconcave discs and have a cell diameter of 7 to 9 μm and a thickness of 2 μm (Diem and Lentner, 1975a; Beutler et al., 1995b). The volume of individual RBCs is 90 fL and the surface area is 163 μm2 (Diem and Lentner, 1975b; Beutler et al., 1995a). The plasma membrane consists of a lipid bilayer of 10-nm width and contains channels of 0.4-nm radius (Solomon et al., 1983; Delaunay, 1995). The membrane potential is inside negative (Hoffman and Laris, 1974). The main constituents of the membrane are phospholipids, cholesterol, and proteins (Beutler et al., 1995b). Linked to the inner surface of the plasma membrane is a supramolecular system of skeleton proteins (Delaunay, 1995). The plasma membrane proteins include receptors, carriers, and enzymes (Gratzer, 1981; Benjamin and Dunham, 1983;Carruthers and Melchior, 1988). The carriers include anion exchanger (Cl−/HCO3−), glucose transporter, cation transporters (Na+/K+ pump, Ca2+ pump, Ca2+ activated K+ channel, Na+/H+ exchanger, and Na+/Li+ cotransporter), anion/cation cotransporters (Na+/K+/2Cl−cotransporter and K+/Cl−cotransporter), nucleoside transporter, and others (Belz et al., 1972;Benjamin and Dunham, 1983; Carruthers and Melchior, 1988; Lauf et al., 1992; Van Belle, 1993; Garay et al., 1994; Delaunay, 1995). Additional enzymes and drug-binding proteins are located in the cytosol (Agarwal et al., 1986; Cossum, 1988; Hooks, 1994). The various enzymes in the RBCs can metabolize many drugs and an excellent review of this topic was presented by Cossum in 1988. An updated list of drugs metabolized by RBCs is presented in table 1. The cytosolic pH of the RBCs is 7.1 to 7.3 (Funder and Wieth, 1966) and smaller than the pH of plasma water (7.4), which is a consequence of the partitioning of electrolytic drugs. Drugs may bind to the membrane and/or to hemoglobin, carboanhydrase, and binding proteins in the cytosol of the RBCs (Kwant and Seeman, 1969; Wind et al., 1973; Beerman et al., 1975; Bickel, 1975; Wallace and Riegelman, 1977; Ehrnebo, 1980).
Drugs metabolized by blood cells in humans
The RBCs belong to the cellular space within the body and share many characteristics with other cells of the body. However, RBCs differ from other body cells and blood cells such as leukocytes in that they lack a nucleus. They are devoid of endoplasmic reticulum, cytochrome P450 isozymes, and mitochondria (Cossum, 1988). Also, individual RBCs are unlike tissue cells in that they are not connected with each other but are suspended in blood plasma. RBCs have a life span of 100 to 120 days, during which they travel 250 km throughout the cardiovascular system (Beutler et al., 1995a). RBCs are easily accessible and can be kept functional for prolonged times under in vitro conditions.
Investigations of the RBC partitioning of relatively small organic cationic, anionic, and nonelectrolytic molecules have shown that lipophilicity, molecular size, and chiral characteristics are important (Giebel and Passow, 1960; Sha’afi et al., 1971; Deuticke, 1977). Lipophilic organic compounds penetrate the RBCs by dissolving into the lipid bilayer membrane (Schanker et al., 1961, 1964; Holder and Heyes, 1965). Small size hydrophilic compounds (<150 d) enter the RBCs through aqueous channels. RBC partitioning by passive diffusion has been reported for organic cationic and anionic drugs, as well as for nonelectrolytes (Schanker et al., 1961, 1964; Holder and Hayes, 1965;Hinderling, 1984; Shirkey et al., 1985; Sweeney et al., 1988; Ferrari and Cutler, 1990; Lin et al., 1992; Reichel et al., 1994).
In theory, rate and extent of RBC partitioning of drugs can be determined ex vivo or in vitro. In the ex vivo procedure, drug is administered to humans, a series of blood samples is taken, and, following centrifugal separation, the drug concentrations are measured in the RBCs and plasma. In the in vitro procedure, drug is added to RBCs suspended in plasma or in plasma water, and after mixing, the drug’s concentration is measured in RBCs and plasma or plasma water following centrifugal separation (fig. 1;Hinderling, 1984). Apart from practical considerations that favor the latter method over the former, there exist additional reasons to prefer the in vitro procedure. With many drugs, the rate of partitioning is fast, and distribution equilibrium is reached within a few seconds to minutes; in these cases, only the in vitro procedure enables determination of the rate of partitioning to obtain meaningful results. The extent of drug partitioning into RBCs should be determined under steady-state equilibrium conditions. Because the in vitro method uses a closed system, steady-state equilibrium conditions can be easily established. With rapidly penetrating drugs, quasi steady state of equilibrium conditions may also be reached under in vivo conditions even though the human body represents an open system. However, with more slowly equilibrating drugs, this is not the case, and the erythrocyte uptake process is more difficult to separate from the multiplicity of other kinetic events, such as tissue distribution and elimination from the body, which are occurring simultaneously. This is not an issue with the in vitro procedure (Wallace and Riegelman, 1977).
Determination of rate and extent of red blood cell partitioning of drugs in whole blood or in red blood cells suspended in plasma or buffer. Cb, drug concentration in whole blood; Cb′, drug concentration in red blood cell suspension (red blood cells suspended in plasma water or buffer); Cp, drug concentration in plasma; Cp,u, unbound drug concentration in plasma water or buffer; Ce, drug concentration in red blood cells; Hc, hematocrit.
To date, only few data that allow a comparison of the results obtained by the in vitro and ex vivo methods have been reported (Fleuren and Van Rossum, 1977; Veronese et al., 1980; Kawai et al., 1982; Borgå and Lindberg, 1984; Brocks et al., 1984; Hinderling, 1984; Shirkey et al., 1985). The extent of RBC partitioning using the two procedures was reportedly similar for digoxin, terbutaline, amiodarone, and chlorthalidone (Fleuren and Van Rossum, 1977; Veronese et al., 1980;Borgå and Lindberg, 1984; Hinderling, 1984). However, discrepancies between the two methods were apparent for the estimated rate of partitioning for terbutaline and digoxin, as well as for the extent of partitioning for hydroxychloroquine (Kawai et al., 1982; Borgå and Lindberg, 1984; Brocks et al., 1984). With terbutaline, the in vitro experiment was conducted at room temperature (Borgå and Lindberg, 1984), and rates of drug partitioning into RBCs are known to be temperature-dependent in many cases (Hinderling, 1984; Reichel et al., 1994). There is clearly a need for more investigations comparing the results on the RBC partitioning of drugs measured under appropriate in vitro and ex vivo conditions. The influence of the composition of the suspension fluid and the impact of repeated washing of the RBCs on the results obtained by the in vitro method should also be carefully delineated.
III. Principles and Definitions
The in vitro method for determining extent and rate of RBC partitioning of drugs uses red cell suspensions in plasma or plasma water. Alternatively, serum instead of plasma or a pH 7.4 buffer instead of plasma water may be used. The experiments are conducted at pH 7.4 and 37°C. The rate of drug partitioning into RBCs is determined in spiked whole blood or in a suspension of RBCs in plasma water, which are gently shaken to mimic the in vivo situation, where drug distribution occurs by diffusion and convection. Timed samples are taken, which are immediately cooled and centrifuged (Hinderling, 1984;Reichel et al., 1994). Subsequently, the drug concentrations in the separated RBCs and plasma (or plasma water) are determined, and the times required to reach equilibration between drug concentrations in the RBCs and plasma (or plasma water) are calculated. The extent of RBC partitioning of drugs, K, is obtained by measuring, after equilibration and subsequent centrifugal separation, the respective drug concentrations in the RBCs (Ce) and in plasma (Cp), or in plasma water (Cp,u), in accordance with equations (1) and (2) (fig. 1):
The relationship between Ke/p and Ke/p,u is given by:
The whole blood-to-plasma concentration ratio (Kb/p) represents an additional drug distribution parameter of interest. The relationship between Ke/p and Kb/p is defined by:
Ke/p,u can be interpreted as a measure of the absolute affinity of drug to the binding sites in the RBCs, whereas Ke/p expresses a drug’s affinity to binding sites in the RBCs, relative to those in plasma (e.g., those associated with albumin and α1-acid glycoprotein). After addition of drug to suspensions of RBCs in plasma, the binding sites in plasma and in the RBCs compete for drug.
For drugs with saturable plasma protein binding, Ke/p increases with increasing whole blood concentration, even though Ke/p,u may be concentration-independent. Constancy of the RBC partitioning over a defined drug concentration range can be verified by using a system of RBCs suspended in plasma water or buffer.
A nonelectrolyte drug that is distributed exclusively into the aqueous phase of the RBCs (71% of total RBC volume) is expected to have a value of Ke/p,u = 0.71. A value of Ke/p,u > 0.71 for a nonelectrolyte drug indicates additional binding to erythrocytic binding sites. For cationic and anionic drugs, the respective critical Kp/e,u values indicative for erythrocytic drug binding were postulated to be different as a result of the Donnan equilibrium (Schanker et al., 1961, 1964).
Drugs may bind to constituents in the RBCs and plasma. It is assumed that one class of drug binding sites exists in both RBCs and plasma. The concentration in the RBCs for drugs lipophilic enough to pass the RBC membrane is given by:
For hydrophilic drugs that only bind to the outer part of the membrane without actually passing it, Ce is defined by equation (6):
Combination of equations (5) and (7) yields the following expression for Ke/p of minimally lipophilic compounds:
Equations (8) and (9) are nonlinear and predict that if Cp,u exceeds a critical value, binding of drug to the sites in the RBCs and/or in plasma becomes saturable. Hence, Ke/p will not be constant and either decreases or increases with increasing Cp,u.
Special cases for sufficiently lipophilic compounds can be differentiated as follows: (a) both binding sites in RBCs and in plasma are saturated:
Cp,u ≫ Kd,e and Cp,u ≫ Kd,p:
Cp,u ≪ Kd,e and Cp,u ≫ Kd,p:
IV. State of the Art
A. Methodological Aspects
Meaningful in vitro determinations of rate and extent of RBC partitioning of drugs must be performed under controlled physiological conditions (pH = 7.4; temperature = 37°C) and over the entire clinically relevant concentration range of drug. It is prudent to study type (linear or nonlinear) and reversibility of the RBC partitioning of drugs. With racemic drugs, the possible stereospecificity of the RBC uptake kinetics of the individual enantiomers should be investigated. True and apparent stereospecific RBC uptake of racemates must be differentiated. This requires determination of Ke/p and Ke/p,u for the individual enantiomers. Different Ke/p,u values for the enantiomers indicate true stereospecific RBC uptake. Identical Ke/p,u values and different Ke/p values suggest stereospecific plasma protein binding. Reversibility of drug binding to the RBCs can be studied in repartitioning experiments, in which drug-loaded RBCs are resuspended in drug-free plasma or plasma water. If the kinetics of drug in partitioning and repartitioning experiments are found to be identical, irreversible drug binding to RBC constituents can be excluded (Hinderling, 1984). It is prudent to first acquire knowledge on the rate of partitioning of drug to estimate when steady state of equilibrium is reached so that the extent of RBC partitioning can be appropriately estimated.
As mentioned above, RBCs contain numerous enzymes, which under in vitro and ex vivo conditions have been shown to metabolize drugs contained in whole blood. It is therefore prudent to use specific assay methodologies separating parent drug and possible metabolite concentrations in determining rate and extent of RBC partitioning of compounds.
With some drugs, uptake in white blood cells and platelets is also significant (Hebden et al., 1970; Piazza et al., 1981; Bergqvist and Domeij-Nyberg, 1983; Sartoris et al., 1984). When aliquots of whole blood from subjects taking chloroquine were collected into tubes that either did or did not contain anticoagulant, and subsequently were centrifuged, the drug concentrations in serum were 4 times larger than in plasma (Bergqvist and Domeij-Nyberg, 1983). The larger concentrations in serum were explained by a release of drug from the platelets during blood coagulation. Similar results were obtained with desmethylchloroquine (Bergqvist and Domeij-Nyberg, 1983). Therefore, the widely held assumption that serum and plasma concentrations of drug are identical may not be correct for some drugs.
Spuriously elevated Ke/p and Kb/p values for cationic drugs bound to α1-acid glycoprotein are obtained if blood collection tubes with stoppers containing tris(2-butoxyethyl)phosphate, a plasticizer, are used (Fremstad and Bergerud, 1976; Midha et al., 1979). Occurrence of hemolysis will also affect the results of RBC partitioning experiments and must be avoided.
B. Rate of Partitioning
Most of the studies investigating the RBC partitioning of drugs have used blood from healthy volunteers. Fewer investigations employed blood from patients. The kinetics of partitioning into RBCs, i.e., cytokinetics, have been systematically investigated in vitro only with relatively few drugs (Wallace and Riegelman, 1977; Hinderling, 1984;Lin et al., 1992; Minami and Cutler, 1992; Reichel et al., 1994). In some cases, evidence was obtained, suggesting that passive diffusion through the RBC membrane was the underlying mechanism (Schanker et al., 1961; Hinderling, 1984; Sweeney et al., 1988; Shirkey et al., 1985). For salicylic acid and other hydroxybenzoic acids, parallel transport by both the anion exchanger and passive diffusion was postulated (Minami and Cutler, 1992). However, for chloroquine, contradictory results indicative of passive diffusion or carrier-mediated transport were reported (Yayon and Ginsburg, 1982; Ferrari and Cutler, 1990). With passive diffusion, the driving force is the unbound drug concentration in plasma water. For drugs with sufficient lipophilicity to pass the RBC membrane, equilibration is reached when the ratio of the unbound concentrations of drug in the aqueous phases of plasma and RBC cytosol remains constant. Within the RBCs, different kinetic subcompartments have been identified with digoxin and derivatives, and a compartment-model–independent parameter such as the mean transit time has been proposed as a measure for the rate of drug partitioning into the RBCs (Hinderling, 1984). Large differences in the rate of RBC partitioning between structurally related and unrelated compounds have been observed (Schanker et al., 1961, 1964; Kornguth and Kunin, 1976;Wallace and Riegelman, 1977; Skalski et al., 1978; Jun and Lee, 1980;Hinderling, 1984; Matsumoto et al., 1989; Ferrari and Cutler, 1990;Reichel et al., 1994). The estimated time to reach partitioning equilibrium between RBCs and plasma or plasma water ranges between a few seconds to several hours for different drugs (Hinderling, 1984;Reichel et al., 1994).
Several drugs with primary amino groups such as gentamycin, furosemide, procainamide, bumetanide, methotrexate and vancomycin show delayed equilibration between red cells and plasma (Lee et al., 1981a,b, 1984,1986; Chen et al., 1983; Chang et al., 1988; Shin Wan et al., 1992). Schiff base formation with free fatty acid aldehyde groups in the membrane of RBCs was postulated to be the underlying mechanism.
C. Extent of Partitioning
Whereas information on the rate of partitioning is scarce, data on the extent of partitioning into RBCs have been reported for a larger number of drugs (Hinderling, 1988). With β-blockers and other compounds, lipophilicity was shown to be the single most important determinant for the extent of the RBC partitioning (Schanker et al., 1964; Holder and Hayes, 1965; Hinderling et al., 1984). Systematic investigations of the linearity of the extent of RBC partitioning were conducted only for a relatively small number of drugs (table2). In some cases, the results showed medium concentration proportionate uptake (Hinderling et al., 1974;Veronese et al., 1980; Eckert and Hinderling, 1981; Roos and Hinderling, 1981; Hinderling, 1984; Shirkey et al., 1985; Ferrari and Cutler, 1990), in other cases, saturable uptake (Collste et al., 1976;Fleuren and Van Rossum, 1977; Wallace and Riegelman, 1977; Garrett et al., 1978; Bayne et al., 1981; Altmayer and Garrett, 1983; Niederberger et al., 1983; Brocks et al., 1984; San George et al., 1984; Urien et al., 1988; Beysens et al., 1991; Lin et al., 1992; Yatscoff et al., 1993a,b; Jusko et al., 1995; Snoek et al., 1996), and in still other cases, temperature- or pH-dependent drug distribution into RBCs (Holder and Hayes, 1965; Roos and Hinderling, 1981; Niederberger et al., 1983;Hinderling, 1984; Yatscoff and Jeffrey, 1987; Rudy and Poynor, 1990;Beysens et al., 1991; Yatscoff et al., 1993a; Reichel et al., 1994). Evidence for stereoselective RBC partitioning of racemic drugs has also been published (Lin et al., 1992; Chu et al., 1995a,b)
Kinetics of red blood cell partitioning of drugs
Only limited data are available on the impact of disease and demographic factors on the RBC partitioning of drugs. Significantly larger values for Ke/p,u were found in vitro for plasmodium-infected RBCs compared with uninfected RBCs for mefloquine (Vidrequin et al., 1996) and chloroquine (Verdier et al., 1985). There was no difference in Ke/p,u for pethidine between young healthy subjects and old patients, suggesting that age had no influence on the RBC partitioning of the drug (Holmberg et al., 1982). For flucloxacillin and cloxacillin, Ke/p,u was similar in neonates, mothers, and controls (Herngren et al., 1982). Identical Ke/p,u values were also found for flucloxacillin in healthy subjects and in patients with pacemakers (Anderson et al., 1985). Similarly, Ke/p,u for phenobarbital was comparable in pregnant women at term and controls. However, Ke/p,u of phenobarbital was statistically significantly smaller in neonates than in the mothers (Walin and Herngren, 1985). Because phenobarbital is known to be bound to hemoglobin (Hilzenbecher, 1972), it may be speculated that the drug’s affinity to fetal hemoglobin is decreased. In healthy subjects as well as in patients with renal or hepatic diseases, Ke/p was found to increase in proportion to the free fraction of valproic acid in plasma (Shirkey et al., 1985). This result suggested that Ke/p,u of the drug in healthy subjects and in the patients is similar. Likewise, similar Ke/p,u values were found for amobarbital, pentobarbital, and phenytoin in healthy volunteers and uremic patients (Ehrnebo and Odar-Cederlöf, 1975). Also, the Ke/p,u values for phenytoin in healthy subjects and cirrhotic or uremic patients were not different. No gender difference was found for the RBC partitioning of phenytoin. In patients with psoriasis, RBC uptake of dehydroepiandrosterone was reported to be smaller than in healthy subjects (Morsches et al., 1981).
Ex vivo and in vitro results indicate that coadministered acetazolamide is able to displace chlorthalidone bound to carbonic anhydrase in RBCs (Beerman et al., 1975). With other drug combinations, no competition for RBC binding sites was found (Ehrnebo and Odar-Cederlöf, 1977).
Reversibility of the RBC partitioning was shown for digoxin and derivatives, disopyramide, tetracycline, phenytoin, penicillin G, arbutin, p-nitrophenol and tacrolimus (Hinderling et al., 1974;Kornguth and Kunin, 1976; Ehrnebo and Odar-Cederlöf, 1977; Jun and Lee, 1980; Hinderling, 1984; Matsumoto et al., 1989; Beysens et al., 1991).
D. Binding Sites
The primary binding sites of drugs in the RBCs are associated with hemoglobin, proteins, or plasma membrane (table3). The diuretics acetazolamide, methazolamide, and chlorthalidone and the ocular pressure reducing agent, dorzolamide, act as inhibitors of carbonic anhydrase and are bound extensively to this enzyme (Collste et al., 1976; Fleuren and Van Rossum, 1977; Wallace and Riegelman, 1977; Bayne et al., 1981; Lin et al., 1992; Biollaz et al., 1995), which is present in the RBC cytosol in seven isoforms and in larger concentrations than in the kidney (Maren, 1967). More than 90% of the amounts of this enzyme in the body are found in the RBCs. The immunosuppressive agents cyclosporin A and tacrolimus are strongly bound to the cytosolic proteins, cyclophilin and tacrolimus binding protein, in the RBCs (Agarwal et al., 1986;Hooks, 1994). The adenosine uptake inhibitor draflazine is bound extensively to the nucleoside transporters located on the RBCs (Snoek et al., 1996). Codeine, chlorpromazine, imipramine, mefloquine, and pyrimethamine have been shown to bind to the RBC plasma membrane (Bickel, 1975; San George et al., 1984; Rudy and Poynor, 1990; Mohammed et al., 1993). Digoxin and derivatives, sulfonamides, mefloquine, phenytoin, phenothiazines, barbiturates, phenylbutazone and derivative, salicylic acid and congeners, imipramine and derivatives, phenytoin, proquazone, pyrimethamine, as well as other drugs, are reportedly bound largely to hemoglobin (Hilzenbecher, 1972; Berneis and Boguth, 1976; Roos and Hinderling, 1981; Hinderling, 1984; San George et al., 1984; Rudy and Poynor, 1990).
Drug binding sites in red blood cells
Different types of interactions of drugs with proteins in RBCs have been observed. As mentioned above, many drugs bind to hemoglobin. Some of these drugs induce reversible allosteric changes in the hemoglobin molecule (Keidan et al., 1989; Venitz et al., 1996; Stone et al., 1992) or may, like acetylsalicylic acid and similar compounds, acetylate hemoglobin (Klotz and Tam, 1973; Zaugg et al., 1980; Massil et al., 1984). Inhibition of RBC carbonic anhydrase by acetazolamide and related compounds was mentioned above. Zifrosilone can reportedly inhibit cholinesterase in RBCs (Cutler et al., 1995).
V. Significance of Studying Red Blood Cell Partitioning of Drugs
A. Rational Choice of Biological Fluid (Whole Blood, Red Blood Cells, Plasma/Serum) for Assaying Drug Concentrations
The vast majority of drugs are assayed in plasma or serum. For drugs with Kb/p or Ke/p larger than 2.0, measuring concentrations in whole blood or erythrocytes rather than in plasma (or serum) increases the sensitivity of an assay with a given lower limit of quantitation (table 4; Fleuren and Van Rossum, 1977;Wallace and Riegelman, 1977; Bayne et al., 1981; Niederberger et al., 1983; Rambo et al., 1985; Winstanley et al., 1987; Hamberger et al., 1988; Nosál et al., 1988; Tett et al., 1988; Beysens et al., 1991; Jusko and D’Ambrosio, 1991; Yatscoff et al., 1993b; Chu et al., 1995b; Jusko et al., 1995; Kurokawa et al., 1996; Snoek et al., 1996). The increased sensitivity permits follow-up of the kinetics of drug for at least one additional half-life. Assaying concentrations in whole blood, rather than in plasma or serum, should also be considered with drugs showing significant temperature-dependent RBC partitioning (Wenk and Follath, 1983; Beysens et al., 1991; Yatscoff et al., 1993a). As a result of temperature-sensitive RBC partitioning, plasma concentrations of these drugs depend on the temperature maintained during centrifugation of the whole blood samples. Therefore, if these drugs are to be measured in plasma, the temperature during the centrifugation process must be kept at 37°C to reflect in vivo conditions. Alternatively, such drugs may be assayed in whole blood, obviating the need for temperature control of the samples. Indeed, most assays used today in the therapeutic monitoring of cyclosporin A and tacrolimus measure drug in whole blood (Wenk and Follath, 1983; Beysens et al., 1991).
Drugs with large blood-to-plasma concentration ratios, Kb/p, in humans4-a
With the neuroleptic drugs butaperazine, haloperidol, and thioridazine, the RBC concentrations reportedly correlate better with therapeutic effects or dose than plasma concentrations (Garver et al., 1977; Casper et al., 1980; Svensson et al., 1986). Based on these findings, measurement of RBC concentrations for therapeutic monitoring was recommended for butaperazine and haloperidol. The RBC concentrations of the metabolites of tricyclic antidepressants are reportedly the best toxicity markers for impaired intraventricular conduction (Amitai et al., 1992). Similarly, the measurement of RBC concentrations has been recommended for lithium in evaluating adverse reactions and toxicity of the drug (Albrecht and Mueller-Oerlinghausen, 1976). With digoxin, RBC concentrations were found to better distinguish between toxic and nontoxic drug levels than did plasma concentrations (Kawai et al., 1982).
As mentioned above, the RBC partitioning of drugs may vary as a result of either pH-dependent binding to plasma proteins and/or RBCs. In this case, it may be more practical to measure drug concentrations in whole blood. This is true whether the blood samples are from subjects with normal blood pH or patients with alkalosis or acidosis. If the concentrations of such drugs are measured in plasma, care must be taken that the pH of the blood samples is controlled during centrifugal separation of RBCs and plasma. However, to understand the pharmacokinetics of pH-sensitive drugs in patients with abnormal blood pH, it is necessary to measure concentrations in both plasma and RBCs.
When considering assaying concentrations of drugs in whole blood, possible degradation by enzymes located in the RBCs must be excluded. As mentioned above, RBCs have been shown to contain various enzymes that can effectively metabolize drugs. As a general rule, stability of drugs should be determined routinely in whole blood, plasma, or serum, independent of which matrix is used for assaying drug concentrations.
Choosing whole blood as the biological fluid for measuring the concentrations of drug does not obviate the need for studying rate and extent of partitioning of drug from plasma into RBCs. Knowing the type of kinetics of drug RBC partitioning (linear or nonlinear) and the time to equilibration is mandatory for a correct interpretation of kinetic parameters derived from concentrations measured in whole blood. The above examples indicate that knowledge of drug partitioning into the cellular constituents of blood and familiarity with the covariates impacting this process are critical in choosing the appropriate assay matrix for drug.
B. Significance of Rate and Extent of Red Blood Cell Partitioning for Physiological Interpretation of Organ Clearances and Volumes of Distribution
As mentioned in Section V.A., most often, drug concentrations are measured in plasma or serum and not in whole blood; hence, plasma or serum is the reference fluid for the derived pharmacokinetic parameters such as clearances and volumes of distribution. However, whole blood, not plasma or serum, is flowing through the vessels of the human body. Therefore, it would appear that whole blood rather than plasma is the more appropriate reference fluid for calculating and interpreting clearances and volumes of distribution. In agreement with this rationale, routine determination of the whole blood to plasma concentration ratio was recently proposed for drugs under development (Peck et al., 1992).
It is commonly assumed that only unbound drug molecules in the plasma water phase of blood can leave the capillary bed in the liver and kidney for elimination (fig. 2;Wilkinson, 1987). Most drugs are bound to some extent to plasma proteins and/or partitioned into RBCs. For such drugs, bound, partitioned, and unbound fractions in blood coexist and equilibria are maintained between the free and bound species. In contrast to the unbound molecules, the bound or partitioned molecules are not immediately available for elimination because they cannot leave the capillary bed in eliminatory organs.
Multiple equilibria between bound/partitioned and unbound concentrations of drug in whole blood and significance of forward and backward rate constants for physiological interpretation of organ clearances. Ce, drug concentration in red blood cells; Cp,u, unbound drug concentration in plasma water; Cp, drug concentration in plasma; (Cp-Cp,u), bound drug concentration in plasma; k1, rate constant of partitioning of free drug from plasma water into red blood cells; k2, rate constant of repartitioning of drug from the red blood cells into plasma water; k3, rate constant of binding of free drug in plasma water to plasma proteins; k4, rate constant of debinding of drug bound to plasma proteins; k5, rate constant of elimination of drug from plasma water; t̄, mean transit time of blood in capillaries of liver (t̄ = 2.5 seconds) or kidney (t̄ = 28 seconds).
A mean transit time of capillary blood in the liver of 10 seconds was estimated in animal experiments (Goresky, 1963). Corresponding values of 2.5 seconds and 28 seconds were found for cortex and medulla of the kidney in dogs (Kramer et al., 1960). Free drug molecules can leave the capillaries in liver and kidney during these times. As a consequence of the equilibria between the free and bound drug species in whole blood, unbound drug molecules that undergo elimination may be replaced by previously bound or partitioned drug molecules, which as unbound molecules can now leave the capillaries of the liver and kidney. The efficiency of resupply of free drug by previously bound or partitioned molecules depends on the magnitude of the binding/partitioning and debinding/repartitioning processes, relative to the above residence times of blood in the capillaries of liver and kidney (fig. 2). If repartitioning of drug with significant distribution into RBCs (Ke/p > 0.25) is fast compared with the residence time of blood in the capillaries, significant amounts of previously partitioned drug molecules may be eliminated during each passage of blood through the liver or kidney. However, if the repartitioning of drug is slow compared with the capillary blood’s residence time, then negligible amounts of drug previously bound to blood constituents will be eliminated during each passage of blood through the liver or kidney. Only in the former case is whole blood the appropriate reference fluid with physiological relevance. In choosing the appropriate reference fluid for computing renal clearance, consideration must also be given to the possibility of pretubular deviation of a portion of RBCs by a “skimming mechanism” (Pappenheimer and Kinter, 1956; Kinter and Pappenheimer, 1956a,b). The above discussion assumes that debinding of drug from plasma proteins is fast compared with the residence time of blood in the capillaries of eliminatory organs.
Evidence obtained from in vivo studies and in vitro experiments in humans, and additional investigations using isolated rat liver and kidney perfusion techniques, indicates that the time for equilibration between plasma and RBCs relative to the mean transit time of blood in the capillaries of liver or kidney is in the same range for some compounds (Reichel et al., 1994; Morgan et al., 1996), but is much slower for a distinct number of other drugs (table5; Schanker et al., 1961, 1964; Goresky et al., 1975; Kornguth and Kunin, 1976; Wallace and Riegelman, 1977;Skalski et al., 1978; Noel, 1979; Jun and Lee, 1980; Lee et al., 1981a,b, 1984, 1986; Tucker et al., 1981; Chen et al., 1983, 1992;Hinderling, 1984; Wiersma et al., 1984; Chang et al., 1988; Lee and Chiou, 1989a,b; Matsumoto et al., 1989; Beysens et al., 1991; Shin Wan et al., 1992; Rolan et al., 1993; Hasegawa et al., 1996). For these latter drugs, calculating clearance referenced to drug concentration in whole blood from
Drugs with slow rates of red blood cell partitioning in humans5-a
Equation (15) indicates the variables impacting organ clearances, which are referenced to drug concentrations in whole blood. Equation (15) is only valid if Ke/p is constant across the eliminatory organ (kidney or liver) (Stec et al., 1979; Lee et al., 1980). This means that reequilibration between the concentrations in the RBCs and plasma for drugs with significant RBC partitioning occurs quickly, relative to the transit time of blood in the capillaries. Equations (14) and (15) indicate that except if Ke/p = Kb/p = 1.0, the values for Clp and Clb differ. For drugs with Ke/p or Kb/p > 1.0, Clp exceeds Clb, and for compounds with Ke/p or Kb/p < 1.0, Clb exceeds Clp.
Experimental limitations in humans may preclude a definitive assignment of the appropriate biological fluid of reference in calculating hepatic or renal clearance for a drug. However, the correct biological fluid of reference may be identified for dialysis clearance. In this case, all the critical variables (Ce, Cp, and Hc) can be determined in the arterial and venous lines of the dialyzer. Provided Ke/p and Hc are found to be constant, whole blood is the appropriate biological fluid of reference. It is mandatory to measure Ce and Cp in the blood samples obtained from the predialyzer and postdialyzer lines within a few seconds after drawing to conclusively demonstrate constancy of Ke/p across the dialyzer.
The concept of referencing pharmacokinetic parameters of appropriate drugs to whole blood, as the physiologically meaningful body fluid, has been applied also to intercompartmental clearances and volumes of distribution (Stec and Atkinson, 1981; Odeh et al., 1993). Experimental evidence in support of fast equilibration of drug between RBCs and plasma must be available, like for the above parameters, to justifiably reference intercompartmental clearances and volumes of distribution to whole blood.
C. Red Blood Cells as Probes for in Vivo Drug Distribution
A highly significant log-log linear correlation between Ke/p,u and the steady-state volume of distribution referenced to the unbound drug concentration in plasma water, Vu,ss, for 38 cationic drugs in humans has been reported (fig. 3; Hinderling, 1988). The data base for these drugs consisted of pairs of Ke/p,u values determined in vitro or ex vivo and Vss values obtained in vivo. It was assumed that the values obtained for Ke/p,u in vitro and ex vivo were equivalent. Values for Vu,ss were computed from Vu,ss = Vss/fu, where fu corresponds to the fraction of drug unbound in plasma water determined in vitro or ex vivo. Using this database, predictions of Vu,ss from Ke/p,u values were estimated to have a mean error of 50% and a bias of 2%. This magnitude of prediction error is small enough to make these estimates of Vu,ss useful in practice. Reportedly, highly significant log-log correlations also exist between Vu,ss and the n-octanol/water partition coefficient, P, for the 38 cationic drugs tested, suggesting that lipophilicity of the compounds is most important for the extent of drug distribution in the body and that electronic properties were not relevant.
Correlation between in vivo steady-state volume of distribution referenced to the unbound drug fraction in plasma and in vitro red blood cells to plasma water partition coefficient for 38 cationic drugs. Ke/p,u, red blood cell to plasma water partition coefficient for drug determined in vitro; Vu,ss, steady-state volume of distribution referenced to the unbound drug concentration in plasma water determined in vivo. 1, acebutolol; 2, alprenolol; 3, amitryptiline; 4, atenolol; 5, atropine; 6, chlordiazepoxide; 7, cimetidine; 8, diazepam; 9, diltiazem; 10, disopyramide; 11, haloperidol; 12, imipramine; 13, lidocaine; 14, lorcainide; 15, maprotiline; 16, mepivacaine; 17, metoprolol; 18, midazolam; 19, morphine; 20, N-acetylprocainamide; 21, nortryptiline; 22, papaverine; 23, pentazocine; 24, pethidine; 25, pindolol; 26, prazosin; 27, procainamide; 28, propranolol; 29, proquazone; 30, quinidine; 31, quinine; 32, flumazenil; 33, SM 1213; 34, sotalol; 35, tiapamil; 36, timolol; 37, tolamolol; 38, verapamil. Reproduced with permission fromHinderling (1988).
With drugs having volumes of distribution significantly larger than the volume of blood or plasma, the largest fraction of an administered dose is likely to be distributed into skeletal muscle, which constitutes approximately 40% of the total body mass (Fichtl and Kurz, 1978;Clausen and Bickel, 1993). Among muscular tissue proteins, actin and myosin constitute the largest fraction and were found to bind drugs significantly (Kurz and Fichtl, 1983). RBC partitioning and hemoglobin binding as well as muscular tissue binding were found to primarily depend on the lipophilicity of a drug (Schanker et al., 1961, 1964;Kurz and Fichtl, 1983; Hinderling et al., 1984; Hinderling, 1988), and statistically significant correlations were reported to exist between the binding of drugs to hemoglobin and muscular tissue (Kurz and Fichtl, 1983). It can be speculated that successful predictions of in vivo drug distribution from in vitro RBC partitioning are due to the relevance of lipophilicity for both processes.
D. Use of Red Blood Cell Suspensions in Plasma and Buffer to Determine Plasma Protein Binding of Drugs
The RBC partitioning technique can also be used as an alternative method to determine plasma protein binding (Garrett and Hunt, 1974;Veronese et al., 1980; Cenni et al., 1995). The experiments should be conducted under physiological conditions, i.e., at 37°C and pH 7.4. With this method, separate RBC suspensions in plasma and plasma water (or buffer) are prepared from blood samples obtained from individual subjects. After spiking with drug, equilibration, and subsequent centrifugal separation of the RBCs, the respective drug concentrations in RBCs, plasma, and plasma water are determined (fig. 1). The values for Ke/p and Ke/p,u are then computed, and the fraction of drug unbound in plasma is obtained from fu = (Ke/p)/(Ke/p,u) after rearrangement of equation (3). Minimal drug binding to and/or partitioning into RBCs, as well as absence of drug metabolism by the RBCs, are preconditions for the successful application of this alternative methodology to determine plasma protein binding.
RBC suspensions in plasma or buffer spiked with drug can be considered as a biological equilibrium dialysis (BED) system (Hinderling, 1987). The advantage of the BED system over the classical equilibrium dialysis system, using two chambers separated by an artificial semipermeable membrane (AED), is that the RBC membranes have a much larger surface. As a result, equilibration of drug in the BED system is much more quickly established than with the AED system (Hinderling, 1987). A comparison of the results on the equilibration time required for 22 acidic, basic, or nonelectrolytic drugs showed that the estimated equilibration time ranged between 2 and 45 minutes for the BED system and between 120 and 960 minutes for the AED system. As a result, water shift, change in pH, and degradation of plasma proteins, phenomena known to occur during prolonged equilibrium dialyses, are less likely to occur with the BED system than with the AED system (Hinderling, 1987). Excellent agreement was found between the respective estimates for the unbound fraction of drug in plasma by the two methods for 22 drugs with fu values ranging from 0.01 to 1.0 (fig.4). Similar results were obtained in comparing the fu estimates by the BED method and ultrafiltration for five drugs (Ho Ncoc-Ta Truong et al., 1984).
Plot of the unbound fraction of drug in plasma by the red blood cell partitioning method, fu (BED), against the corresponding value obtained by the traditional equilibrium dialysis method, fu (AED). fu: fraction of drug unbound in plasma water, AED, Equilibrium dialysis method employing artificial membranes, BED, Equilibrium dialysis method using biological membranes. 1, amitriptyline; 2, atropine; 3, binedaline; 4, fentanyl; 5, imipramine; 6, lidocaine; 7, minaprine; 8, nicardipine; 9, nortriptyline; 10, propranolol; 11, proquazone; 12, quinidine; 13, tropine; 14, amobarbital; 15, pentobarbital; 16, phenytoin; 17, digitoxin; 18, dihydrodigoxin; 19, digoxin; 20, digoxigeninbisdigitoxoside; 21, digoxigeninmonodigitoxoside; 22, β-methyldigoxin. Reproduced with permission from Hinderling (1987).
A modified BED method employing RBC suspensions in variably diluted plasma has been shown to allow measurements of protein binding of very lipophilic drugs like tetrahydrocannabinol, amiodarone, and etretinate, which bind significantly to artificial membranes rendering conventional equilibrium dialysis (AED) and ultrafiltration methods impractical (Garrett and Hunt, 1974; Veronese et al., 1980; Urien et al., 1992a,b;Cenni et al., 1995).
E. Significance of Studying the Effects of Drugs on Red Blood Cells
Screening drugs for beneficial and/or toxic effects on RBCs should be done early in drug development. Some of these investigations may be conducted in vitro with human RBCs. Beneficial effects of drugs include schizontocidal activity in parasitized RBCs, enhanced dissociation of O2 from hemoglobin, increased RBC deformability, decreased sickling of RBCs containing hemoglobin S, and inhibition of adenosine uptake into RBCs (Orringer et al., 1986;Keidan et al., 1989; Ambrus et al., 1990; Stone et al., 1992; Charache et al., 1995; Cenni and Betschart, 1995; Snoek et al., 1996; Venitz et al., 1996). Toxic effects of drugs include increased formation of methemoglobin and hemolysis (Grossman and Jollow, 1988; Pirmohamed et al., 1991; Fasanmade and Jusko, 1995). RBCs have been employed as surrogates for endothelial cells to study the pharmacodynamics of draflazine, an adenosine uptake inhibitor with potentially cardioprotective properties (Snoek et al., 1996). Historically, RBCs were successfully employed as models for cardiac myocytes in elucidating the molecular mechanism of action of cardiac glycosides (Schatzmann, 1953), and RBCs were more recently used as surrogates for body cells in general in studying the effects of tolrestat and other antidiabetics (Kuusisto et al., 1994; Van Griensven et al., 1995). Modeling was successfully applied in correlating the pharmacodynamics measured in the RBCs with the pharmacokinetics of some of these drugs (Fasanmade and Jusko, 1995; Van Griensven et al., 1995). Toxicity of primaquine and other drugs inducing massive denaturation of hemoglobin can be predicted to occur in subjects with glucose-6-phosphate dehydrogenase deficiency (Luzzatto, 1995). Decrease in ocular pressure and diuretically active carbonic anhydrase inhibitors can be identified in vitro by using RBC partitioning methods.
It is self-evident that routine determination of RBC partitioning of drugs in early stages of development can effectively aid in screening for candidates with the above or other beneficial or toxic effects on RBCs. Hence, it is surprising to note that the study of the partitioning of drugs into RBCs and the other cellular constituents of blood is not a standard part of routine investigations performed in screening drugs for potential beneficial or toxic effects.
F. Red Blood Cells as Markers
Myelosuppressive toxicity of azathioprine and 6-mercaptopurine is reportedly negatively correlated with thiopurine methyltransferase activity in the RBCs in leukemic and immunosuppressed patients (Lennard et al., 1990; Lennard, 1992). The determination of the activity of thiomethyltransferases and other polymorphic methyltransferases and monomorphic acetylases in RBCs has been proposed as a means of characterizing the metabolizer status of patients after significant correlations of the respective activities of these enzymes in the RBCs and in the liver or kidneys were demonstrated (Woodson et al., 1982;Scott et al., 1988; Szumlanski et al., 1992; Hutabarat et al., 1994). Different types of essential hypertension can reportedly be differentiated based on the type of ion transport defects found in the RBC membranes of hypertensive subjects (Garay et al., 1994; Duhm and Engelmann, 1992).
VI. Conclusions
It is surprising that in studying the disposition kinetics of drugs in blood and its subcompartments over the last 30 years, the study of RBC partitioning of drugs has received much less attention than the plasma protein binding of drugs. As a consequence, there are many opportunities for challenging research in this underdeveloped area.
It is currently routine practice to develop assays for drug concentration measurements in plasma or serum. In only a few instances have assays been developed to measure drugs in whole blood. However, the choice of the appropriate assay matrix is not always rationally based, and factors such as magnitude, pH, and temperature dependency of RBC partitioning are rarely considered in the decision-making process.
The goal of any drug concentration measurement in a blood, plasma, or serum sample must be to have the measured value precisely reflect the drug concentration that existed under in vivo conditions at the time of sampling. It is clear that the RBC partitioning of drugs varies: with some drugs, it is a fast process, with others, equilibration between plasma and RBCs is significantly delayed. With most drugs, RBC partitioning is temperature-dependent. A pH dependency of RBC partitioning has been observed with a number of compounds. Furthermore, some drugs’ interaction with RBCs may result in Schiff base formation. RBCs contain various enzymes that can metabolize many drugs. Thus, the handling of blood samples after collection should be done in accord with a protocol that is based on the results of previous studies, which establish the interaction between RBCs and the drug under development. Sample protocols should specify time to centrifugation, temperature, and pH conditions, as well as need for addition of suitable enzyme inhibitors.
The evidence presented in this review warrants a change from current practice in assay development and validation. Currently, the systematic investigation of drug distribution into and the possible metabolism by the RBCs is neglected. Protocols for the handling and storage of blood samples after their withdrawal should be established. The choice of the appropriate assay matrix should be rationally based.
Additional benefits that would result from a systematic study of the interaction between drugs and RBCs include: (a) a rationally derived and physiologically correct reference concentration of the drug for computing organ clearances, i.e., whole blood, plasma or serum; (b) estimates for the in vivo distribution of new cationic drugs; (c) use of the RBC partitioning procedure as an alternative method to determine plasma protein binding of drugs, particularly highly lipophilic drugs; and (d) early determination of desirable and undesirable effects of drugs on RBCs.
Acknowledgements
The author expresses his gratitude to Dr. Dieter Hartmann, Clin-Pharma Research Ltd. Birsfelden, Switzerland, and the reviewer, Dr. Richard Streiff, University of Florida, Gainesville, Florida, for a critical review of the paper and valuable suggestions. The expert secretarial help of Judy Pierce and Christine Juhasz is gratefully acknowledged.
Footnotes
-
↵FNa Address for correspondence: Peter H. Hinderling, Department of Clinical Pharmacology, Berlex Laboratories Inc., P.O. Box 1000, Montville, NJ 07045-1000.
- Abbreviations:
- AED
- equilibrium dialysis through artificial semipermeable membrane
- BED
- biological equilibrium dialysis through biological semipermeable membrane
- Ce
- drug concentrations in the RBCs
- Clb
- clearance of drug referenced to the drug concentration in whole blood
- Clp
- clearance of drug referenced to the drug concentration in plasma
- Cp
- drug concentrations in plasma
- fu
- fraction of drug in plasma unbound
- Hc hematocrit
- relative volume of RBCs in whole blood, RBCs suspended in plasma/serum, or plasma water/buffer
- I
- constant relating the unbound drug concentration in the aqueous phase of the RBCs to the unbound drug concentration in plasma or serum
- Kb/p
- whole blood to plasma partition coefficient of drug obtained from the ratio of the concentrations in whole blood and plasma or serum
- Ke/p
- RBC-to-plasma partition coefficient of drug obtained from the ratio of the concentrations in RBCs and plasma or serum
- Ke/p,u
- RBC-to-plasma water partition coefficient of drug obtained from the ratio of the concentrations in RBCs and plasma water or buffer
- nEt
- total binding site concentration in the RBCs
- nPt
- total binding site concentration in plasma or serum
- RBC
- red blood cell
- Vu,ss
- steady-state volume of distribution referenced to the unbound drug concentration in plasma water determined in vivo
- The American Society for Pharmacology and Experimental Therapeutics
References
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
