- Email: [email protected]
fumic) for refined prediction of phase I mediated metabolic hepatic clearance
Journal of Pharmacological and Toxicological Methods 63 (2011) 35–39
Contents lists available at ScienceDirect
Journal of Pharmacological and Toxicological Methods j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j p h a r m t ox
Original article
Direct determination of the ratio of unbound fraction in plasma to unbound fraction in microsomal system (fup/fumic) for refined prediction of phase I mediated metabolic hepatic clearance Sujal V. Deshmukh ⁎, Andreas Harsch Drug Metabolism and Pharmacokinetics, Merck Research Laboratories, Boston, MA, United States
a r t i c l e
i n f o
Article history: Received 5 February 2010 Accepted 20 April 2010 Keywords: Microsomes Plasma Prediction Hepatic Metabolic Binding
a b s t r a c t At the drug discovery stage, in vivo metabolic hepatic clearance (CLhep) is commonly predicted using in vitro parent compound disappearance data generated in liver microsomes or hepatocytes. Correction for the unbound fraction of a compound in the in vitro system and in plasma/serum is known to be critical for the accuracy of metabolic clearance predictions. Discrete generation of these required experimental parameters can be laborious. Herein, we describe a straightforward and direct approach to obtain the ratio of unbound fraction in plasma (fup) to unbound fraction in the microsomal system (fumic) of a small molecule compound using equilibrium dialysis. Experimental conditions were optimized with respect to incubation time, temperature, and plate shaking speed. Results obtained from this system were validated for a set of test compounds by comparison to individually measured fup and fumic data using ultracentrifugation. The correlation for fup/fumic between the two methods for a set of 23 data points was very good with R2 of 0.94, slope of 1.05 and an intercept of 0.007. The impact of microsomal binding on predicted CLhep was illustrated for a tightly bound compound using a series of incubations with increasing concentration of monkey liver microsomal protein. Alteration of this experimental parameter profoundly affected calculated CLhep using the well-stirred model. Significant differences were observed in the prediction when the model was corrected for fup only; in contrast, the model corrected for plasma protein and microsomal protein binding predicted clearance values independent of the microsomal protein concentration. © 2010 Elsevier Inc. All rights reserved.
1. Introduction Prediction of in vivo clearance in humans is an integral part of new chemical entity profiling in drug discovery and early development. Since hepatic metabolic contributions often play a major role in overall xenobiotic clearance, intrinsic hepatic metabolic clearance (CLint) is determined fairly early in drug discovery programs. These data are commonly generated by measuring drug disappearance in subcellular hepatic fractions, such as S9 and microsomes (Obach, 1999). This experimental approach is broadly implemented due to its simplicity, ease of use, and amenability to high throughput formats. Although at an early drug discovery stage the exact mechanism of in vivo clearance of compounds is often unknown, it is important to accurately predict in vivo metabolic hepatic clearance (CLhep) to allow rank-ordering
Abbreviations: CLhep, in vivo metabolic hepatic clearance; fup, unbound fraction of a compound in plasma; fumic, unbound fraction of a compound in microsomal system; CLint, intrinsic hepatic metabolic clearance; LSC, liquid scintillation counter; dpm, disintegrations per minute; CLb, blood clearance. ⁎ Corresponding author. 33 Avenue Louis Pasteur, BMB 4-112, Boston, MA 02115, United States. Tel.: +1 617 734 2076; fax: +1 617 734 2084. E-mail address: [email protected] (S.V. Deshmukh). 1056-8719/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.vascn.2010.04.003
compounds based on this parameter. Several physiological models are available to scale in vitro intrinsic metabolic clearance data to in vivo organ clearance (Pang and Rowland, 1977; Rane, Wilkinson, & Shand, 1977; Wilkinson and Shand, 1975). Hepatic blood flow, hepatic intrinsic clearance and plasma/serum protein binding are minimal experimental inputs required to allow scalability of these models. Obach (1996, 1997, 1999) demonstrated the significance of including a measure of drug binding to microsomes as an additional important correction factor for accurate prediction of in vivo clearances. Several subsequent studies have evaluated nonspecific microsomal protein binding of xenobiotics and have confirmed its relevance in predicting in vivo clearance (Jones and Houston, 2004; Giuliano, Jairaj, Zafiu, & Laufer, 2005; Naritomi et al., 2001; Austin, Barton, Cockroft, Wenlock, & Riley, 2002). In all published reports, with the exception of Obach (1997) and Skaggs, Foti, and Fisher (2006) the unbound fraction in plasma and microsomes was separately determined. For proper correction, these independently determined factors were entered in the scaling equations as a ratio of fraction unbound in plasma (fup) to fraction unbound in microsomal system (fumic). Herein, a straightforward method for the direct and accurate experimental determination of this correction ratio (fup/fumic) is reported. Equilibrium dialysis was used to measure drug partitioning
36
S.V. Deshmukh, A. Harsch / Journal of Pharmacological and Toxicological Methods 63 (2011) 35–39
between plasma and microsomes, allowing direct calculation of fup/ fumic by determining the ratio of total drug concentration in plasma to total drug concentration in microsomes (Eq. (1)). At equilibrium, the unbound concentration in plasma, Cp × fup is equal to the unbound concentration in microsomal system, Cmic × fumic so that Cp × fup = Cmic × fumic. By rearranging this we get: fup = fumic = Cmic = Cp
ð1Þ
Where, Cmic Cp
is the total concentration of compound in microsomal system is the total concentration of compound in plasma
The experimental procedure was optimized using a focused set of compounds, and subsequently used to measure fup/fumic for compounds spanning a wide range of plasma and microsomal binding. Results were validated by comparison to calculated fup/fumic using individually determined measurements of fup and fumic by ultracentrifugation. Finally, the impact of microsomal and protein binding on CLhep was illustrated using a proprietary compound. 2. Materials and methods 2.1. Reagents 3
3
14
Radiolabeled verapamil ( H), digoxin ( H), and caffeine ( C) were purchased from Perkin Elmer Life and Analytical Sciences, Boston MA, 3 amitriptyline ( H) was obtained from American Radiolabeled Chemicals Inc. St. Louis, MO, and midazolam was purchased from Sigma Aldrich, St. Louis, MO. Proprietary radiolabeled compounds and diclofenac were obtained from the Labeled Compound Synthesis group, Merck Research Laboratories, Rahway, NJ. All other investigational compounds were provided by the Chemistry group, Merck Research Laboratories, Boston. Human liver microsomes were purchased from BD Biosciences, San Jose, CA. Mouse, rat, dog, and rhesus monkey liver microsomes were sourced from CellzDirect, Pittsboro, NC. Human plasma was purchased from Biomeda, Foster City, CA. Plasma from all other species was obtained from in-house sources.
In the case of midazolam, LC–MS/MS was used as described below. Radioactivity counts were replaced with area ratio to calculate unbound fraction. Each determination of unbound fraction was done as a singlet. 2.2.2. Protein binding ratio using equilibrium dialysis The ratio of unbound fraction across plasma and microsomes was directly determined using a 24 multi-well equilibrium dialysis system (BD Biosciences, Bedford, MA). The experiment was performed according to the guidelines provided by the vendor. Briefly, the membranes were soaked for 20 min in deionized water followed by 20 min soaking in 30% absolute alcohol:deionized water. The membranes were washed with water and further soaked in 50 mM potassium phosphate buffer for 20 min. 750 µL of plasma containing 1 µM of compound was loaded in the bottom chamber and 250 µL of the microsomal sample containing the same drug concentration was dispensed into the top chamber. Each sample was incubated in duplicate. For fup determination of M002 the microsomal chamber was replaced with buffer. In both matrices the final concentration of acetonitrile was maintained at 5% v/v as used in the ultracentrifugation method. The assembly was sealed using parafilm and subsequently placed on a shaker (Thermomixer, Eppendorf, New York, NY) at a speed of 300 rpm and allowed to incubate for 45 h at 25 °C. Samples set aside at room temperature served as controls. Plasma and microsomal samples were analyzed by radiometric detection or LC–MS/MS. For radiolabel compounds, 10 µL of the samples was mixed with 5 mL of scintillation cocktail and dpm were measured using a LSC, Perkin Elmer, Waltham, MA. For non-radiolabelled compounds, 50 µL aliquots of each matrix were precipitated with acetonitrile containing an appropriate internal standard. Samples were filtered, evaporated, and reconstituted in 100 µL acetonitrile:water (30:70 v/v). 10 µL sample was injected on a C18 column (2.0 mm× 30 mm, 3 µm particle size) and eluted using a gradient LC method with water containing 0.1% formic acid as the aqueous mobile phase, and acetonitrile containing 0.1% formic acid as the organic phase. Electrospray ionization with multiple reaction monitoring was used for MS/MS detection. Area ratios for each sample were determined by dividing the integrated analyte peak area to internal standard peak area. fup/fumic was calculated as follows: fup = fumic =
ðdpm or area ratio in microsomal incubateÞ ðdpm or area ratio in microsomal controlÞ ×
2.2. Experimental procedure 2.2.1. Protein binding using ultracentrifugation Plasma protein binding and microsomal protein binding using ultracentrifugation were determined by adding 100 µL of a 10 µM solution containing 2.27 µCi/mL radioactivity of a test compound in a 1:1 mixture of acetonitrile:water to 900 µL of either plasma or microsomes (0.25 mg/mL protein). Subsequently, 800 µL aliquots were transferred to 1 mL polycarbonate centrifuge tubes, Beckman Instruments Inc., Palo Alto, CA. The remaining fraction was stored at room temperature and was later used as control sample. Samples were centrifuged at ∼ 90,000 × g for 18 h at 25 °C using a Beckman Optima L-100 XP Ultracentrifuge, Beckman Coulter Inc., Fullerton, CA. Following centrifugation, six aliquots of 100 µL each, starting from the top layer, were gently pipetted out and placed in separate scintillation vials, and 5 mL of scintillation cocktail was added to each sample. The amount of radioactivity (disintegration per minute, dpm) in each fraction was measured using a liquid scintillation counter (LSC). The unbound fraction was calculated using the following equation:
Unbound fraction =
Radioactivity counts in aliquot with lowest counts : Radioactivity counts in control sample ð2Þ
ð3Þ
ðdpm or area ratio in plasma controlÞ : ðdpm or area ratio in plasma incubateÞ
Normalizing data to the control samples should account for any recovery differences in the assayed matrices, especially when conducting LC–MS/MS analysis. Standard curves were not generated to determine compound concentration, rather area ratios were used for fup/fumic calculation. It was ensured that MS responses for all samples were within the linear dynamic range of the instrument. High organic content may affect the protein binding for some compounds; it might thus be preferable to use lower organic content (0.5% v/v) for future data generation. We determined fup/fumic at 5% and 0.5% v/v of acetonitrile using 4 compounds with fup/fumic ranging from 0.04 to 0.28. The data suggests that fup/fumic was not affected by the amount of acetonitrile used, the values obtained with 5% and 0.5% acetonitrile were not statistically different. 2.2.3. Microsomal stability assay To determine CLint of M002 in monkey liver microsomes, 1 µM of the compound was added to varying concentration of monkey liver microsomes in the presence of 1 mM NADPH. The microsomal system was prepared using 50 mM potassium phosphate buffer, pH 7.4. The samples were incubated in duplicate at 37 °C for 0, 5, 10, 15, 20, and 30 min and concentration of M002 was determined using LC–MS/MS. CLint was calculated by the parent disappearance method according to
S.V. Deshmukh, A. Harsch / Journal of Pharmacological and Toxicological Methods 63 (2011) 35–39
37
Eq. (4). For calculation purposes the blood flow for rhesus monkey was considered to be 40 mL/min/kg (Davies and Morris, 1993) and the monkey liver microsomal recovery was estimated at 50 mg/g liver with a liver weight of 30 g/kg body weight. CLhep was calculated without any correction for protein binding, binding in plasma only, or binding in plasma and microsomal protein using Eqs. (5)–(7), respectively. CLint =
Dose ðμM = mg microsomal protein = mLÞ
ð4Þ
Area under parent disappearance curveðμM⋅minÞ ×
mg microsomal protein × g of liver weight g liver kg body weight
Predicted CLhep uncorrected for binding = Q hepatic ·
Predicted CLhep corrected for fup =
CLint Q hepatic + CLint
fup ðB = P Þ ⋅CLint Q hepatic · fu Q hepatic + ðB =pP Þ ·CLint
ðfup = fumic Þ Predicted CLhep corrected for fup and fumic = Q hepatic ·
ðB = P Þ
Q hepatic +
ð5Þ
ð6Þ
⋅CLint
ðfup = fumic Þ ðB = P Þ
·CLint
ð7Þ Where, Q is the hepatic blood flow; B/P is the blood to plasma ratio. 2.2.4. Blood to plasma ratio The blood to plasma ratio for M002 was determined in rhesus 3 monkey using radiolabeled [ H] M002. To 995 µL of blood, 5 µL of a 200 µM solution containing 2.27 µCi/mL of M002 was added and gently mixed. The blood was incubated at 37 °C for 30 min in a water bath and then centrifuged for 15 min at 3220 ×g. The above procedure was also performed using plasma instead of blood, this sample served as the control. Following centrifugation, 100 µL of plasma was transferred to a scintillation vial and 5 mL of scintillation cocktail was added to each sample and dpm measured using LSC Perkin Elmer, Waltham, MA. The blood to plasma ratio was calculated by dividing the dpm from control sample with dpm from plasma that was prepared following the addition of M002 to blood. The stability of M002 in plasma was tested under the assay conditions to ensure that the radioactivity at the end of the study represented M002.
Fig. 1. Effect of incubation time on fup/fumic for M001 (▲), verapamil (○) and caffeine (♦) using equilibrium dialysis. Data for each time point represents mean from a set of duplicates, conducted at 25 °C with a shaking speed of 300 rpm. The experiment was conducted at a compound concentration of 1 µM in human plasma and 0.25 mg/mL human liver microsomes.
compounds was incubated over 3 days at the optimized temperature and shaking speed to determine appropriate incubation times. Samples were taken at 19 h, 45 h, and 68 h after the start of incubation (Fig. 1). Caffeine maintained a fup/fumic ratio close to 1 throughout the sampled time interval, whereas M001, which exhibited a very low fup/fumic did not achieve equilibrium even after overnight incubation. A similar observation was also made for verapamil, its measured fup/fumic ratio continued to decrease up to 45 h and then remained constant. 3.2. Comparison of methods for the determination of fup/fumic The equilibrium dialysis assay was validated by comparing the direct determination of fup/fumic by equilibrium dialysis to data from ultracentrifugation obtained by individually measuring the unbound fraction in plasma and in microsomes and then calculating fup/fumic. This correlation analysis was done using 15 compounds with broad
2.2.5. In vivo clearance In vivo blood clearance (CLb) for M002 was determined in a rhesus monkey pharmacokinetic study. An intravenous dose of 0.25 mg/kg of M002 was administered to each subject (n = 2). Plasma samples were collected at 5 min, 15 min, 30 min, 1 h, 2 h, 4 h, 6 h, 8 h, and 24 h. Samples were analyzed by LC–MS/MS as described previously. Noncompartmental analysis (Watson 7.2) was used to calculate plasma clearance. CLb was calculated by dividing plasma clearance with the blood to plasma ratio. 3. Results 3.1. Assay optimization Equilibrium dialysis assay conditions were initially optimized for direct determination of fup/fumic in human plasma and liver microsomes using a set of 3 compounds (verapamil, caffeine, and M001 — a proprietary compound). These compounds were selected based on their broad range of binding ratio (fup/fumic) as determined by the ultracentrifugation method (Eq. (2)). The fup values for verapamil, caffeine, and M001 were 0.28, 0.91, and 0.05, respectively and the determined fumic were 0.85, 1, and 0.76, respectively. All analyses were performed by radiometric scintillation counting. The test set of 3
Fig. 2. Correlation for fup/fumic using ultracentrifugation (calculated by dividing the individually measured unbound fraction in plasma and microsomes) versus equilibrium dialysis (direct determination). Data represents mean fup/fumic ratio from a duplicate incubation for the equilibrium dialysis method and single measurement for the ultracentrifugation. The assays were conducted at a compound concentration of 1 µM in plasma and 0.25 mg/mL liver microsomes. The 23 data points represent fup/fumic ratio for 13 compounds using human plasma and liver microsomes and two compounds, M001 and M002 using mouse, rat, dog, monkey, and human plasma and liver microsomes.
38
S.V. Deshmukh, A. Harsch / Journal of Pharmacological and Toxicological Methods 63 (2011) 35–39
Table 1 Binding parameters for generic compounds. fupa
fumica fup/fumicb fup/fumicc
fup/fumic
fup/fumic
Calculated Experimental Calculated Experimental (Skaggs (Skaggs et al., 2006) et al., 2006) Amitriptyline Caffeine Diclofenac Digoxin Midazolam Verapamil a b c
0.097 0.91 0.012 0.81 0.01 0.28
0.67 1.00 0.89 0.98 0.7 0.85
0.145 0.910 0.013 0.830 0.015 0.330
0.094 0.930 0.026 0.930 0.022 0.230
0.274 – 0.011 – 0.104 0.248
1.14 – 0.178 – 0.456 0.689
Determined by ultracentrifugation. Ratio of fup and fumic for respective compound. Determined by equilibrium dialysis.
structural diversity and covering a wide range of binding ratios. The fup and fumic values ranged from 0.01–0.91 and 0.66–1, respectively with fup/fumic in the range of 0.01–0.9. The assay was predominantly performed using plasma and liver microsomes from humans, but matrices from other species (mouse, rat, dog, and monkey) were also used for comparison for the two test compounds, M001 and M002. A correlation plot of 23 data points for fup/fumic determined using ultracentrifugation and equilibrium dialysis techniques is shown in Fig. 2. The coefficient of determination, R2, was 0.94 with a slope of 1.05 and an intercept of 0.007. Data on a subset of the test compounds used in the correlation is represented in Table 1. Selection of the subset was based on availability of published literature on fup/fumic determined using equilibrium dialysis (Skaggs et al., 2006). 3.3. Impact of increasing concentration of microsomal protein on predicted CLhep The impact of fup/fumic on the prediction of CLhep (Eq. (7)) was illustrated using a proprietary compound M002. M002 was incubated at 1 μM with increasing concentration of rhesus monkey liver microsomes. The fup for M002 in rhesus monkey plasma was 0.02, which was determined by replacing the microsomes with buffer in the equilibrium dialysis experiment. The blood to plasma ratio in rhesus monkey for M002 was 0.75. The effect of increasing concentration of liver microsomal protein on CLint (Eq. (4)) and CLhep (Eqs. (5)–(7)) is summarized in Table 2. The CLb for M002 in monkey was determined to be 11.3 mL/min/kg. 4. Discussion Assay conditions were optimized for incubation temperature, time, and plate shaking speed. At 37 °C, extensive evaporation was observed within a few hours of incubation, thus this temperature was impractical to test the concept of direct ratio determination using an assay requiring extended incubation time. Evaporation was minimal at 25 °C thus this temperature was chosen for all subsequent studies. A platform shaker was used for sample agitation. At speeds greater
than 300 rpm there was a risk of cross-well spilling causing sample contamination. A speed of 300 rpm was found to be optimal for effective mixing. To evaluate the impact of incubation time on fup/ fumic, the equilibrium dialysis assay was conducted at the optimized temperature and shaking speed over 3 days using verapamil, caffeine and M001 as test compounds. Caffeine, which had similar unbound fraction in plasma and microsomal incubate achieved rapid equilibrium such that there was no change in fup/fumic over time (Fig. 1). On the other hand, verapamil and M001, which are tightly bound in the plasma and relatively unbound in the microsomal incubate, fup/fumic changed significantly over time (Fig. 1). Data from verapamil and M001 indicates that for compounds with significant differences between unbound fractions in plasma and microsomes, appropriate selection of the incubation time is critical for achieving equilibrium, which in turn is essential for accurate measurement of fup/fumic (Eq. (1)). Based on these initial data, an incubation time of 2 days or ∼45 h was selected in order to achieve equilibrium for compounds with significantly different unbound fractions in both matrices. It should be noted that the time to equilibrium is dependent on a range of compound-specific properties (binding to the biomatrices, diffusibility), device-specific parameters (surface size of the membrane, surface to well volume ratio, plate geometry, etc.) (Sebille, Zini, Madjar, Thuaud, & Tillement, 1990) as well as incubation conditions (pH, temperature, etc.). Efforts are currently underway to identify plates which will allow accelerated mass transfer and thus reduce incubation times. To validate the concept of direct determination of fup/fumic, we compared the experimental data from the equilibrium dialysis method to data calculated using ultracentrifugation. 13 compounds were evaluated for fup/fumic using human plasma and liver microsomes and for two compounds, M001 and M002 fup/fumic was determined using mouse, rat, dog, monkey, and human plasma and liver microsomes. For all data points there was a less than 2 fold difference between fup/fumic determined by equilibrium dialysis versus calculated via individually determined fup and fumic measurements using ultracentrifugation (Table 1 and Fig. 2). Skaggs et al. (2006) published a similar comparison for amitriptyline, diclofenac, midazolam and verapamil. Under their experimental conditions, the difference between calculated and experimental ratio ranged from 3 to 16 fold (Table 1). In our assay, an excellent correlation (R2 = 0.94) was obtained for the 23 data points with a slope of 1.05 and an intercept of 0.007 (Fig. 2). One possible reason for the excellent correlation achieved herein may be attributed to the optimized incubation time used in the equilibrium dialysis technique to ensure complete equilibrium. In the publications from Skaggs et al. (2006) and Obach (1997), it appears that the incubation time was 4 h and 4–5 h respective, at 37 °C which may not have been enough to achieve equilibrium for all compounds. In conclusion, the equilibrium dialysis assay conditions described herein were well suited for direct determination of fup/fumic. In addition to the simplicity of this experimental approach, there are possible intrinsic advantages to using direct fup/fumic determination (protein in both compartments) vis-à-vis the traditional route of independent plasma or microsomal protein binding determinations
Table 2 Effect of increasing concentration of monkey liver microsomes on fup/fumic, CLint, and CLhep. Conc. of monkey liver microsomes (mg/mL)
fup/fumica
CLint (mL/min/kg)
Predicted CLhep uncorrected for protein binding (mL/min/kg)
Predicted CLhep corrected for plasma protein binding (mL/min/kg)
Predicted CLhep corrected for plasma and microsomal protein binding (mL/min/kg)
0.1 0.25 0.5 1 2
0.05 0.06 0.07 0.11 0.17
199 153 103 75 42
33 32 29 26 20
4.7 3.7 2.6 1.9 1.1
10 9.4 7.8 8.7 7.6
a
fup for M002 was determined to be = 0.02.
S.V. Deshmukh, A. Harsch / Journal of Pharmacological and Toxicological Methods 63 (2011) 35–39
by equilibrium dialysis (protein in one compartment, buffer in the other compartment). Presence of protein in both compartments may help to keep compounds with poor physico-chemical properties in solution, and thus enable generation of fup/fumic for compounds that might not yield meaningful data when fup and fumic are measured independently, for example, due to lack of aqueous solubility. To demonstrate the impact of plasma protein binding and microsomal protein binding on CLhep, we used a proprietary compound — M002. This compound was known to be eliminated predominantly via phase I mediated hepatic metabolic clearance with minimal contribution from direct renal elimination based on data from a rat bile duct cannulated study (data not shown). Monkey was selected for this illustration since M002 showed measurable turnover in the monkey liver microsomes across a wide range of microsomal protein concentrations. For this case study, the concentration of monkey liver microsomal protein was varied from 0.1 to 2 mg/mL, and CLint and fup/fumic were determined using the microsomal stability assay and the optimized equilibrium dialysis assay, respectively. Increasing the concentration of microsomal protein resulted in a decrease in the CLint even after normalization to the protein concentration used in the incubation. In contrast, fup/fumic increased with increasing concentration of microsmal protein (Table 2). Since plasma concentrations are unaltered, fup is constant. Consequently, the observed increase in the fup/fumic ratio can be readily attributed to a decrease of fumic. In other words, the drug becomes more extensively bound to microsomal protein with increasing microsomal protein concentration. Application of the well-stirred model without correction for any binding (fup/fumic set to 1) led to a substantial over-prediction of the CLhep of M002, as expected, and the predicted CLhep values decreased with increasing microsomal protein concentrations (Table 2). This concentration dependence was even more pronounced after correction for plasma protein binding only (fup set to 0.02). This can be readily rationalized via a simple boundary analysis of the equation for the well-stirred model. Blood to plasma was not considered herein for illustration purposes. In situations where fup/fumic ×CLint is much smaller than Qhep, the equation for the well-stirred model simplifies to CLhep = fup/fumic × CLint. Changes in the fup/fumic ×CLint term will thus directly translate into changes in calculated CLhep. Herein, this is the case when the determined CLint are corrected for fup only. In cases when fup/fumic × CLint NN Qhep, the calculated CLhep will vary less and asymptotically approach Qhep as fup/ fumic × CLint increases. In this study, this is observed when CLhep is predicted using CLint without any correction for binding effects. In contrast to the scenarios discussed above, correction for plasma protein binding and microsomal protein binding (fup/fumic) (Eq. (7)) rendered the calculated CLhep value independent of the microsomal concentration used in the incubations, as the concentration dependent decrease of CLint was offset by the concentration dependent increase of fup/fumic. In this case study, the obtained hepatic metabolic clearance of M002 was also within 1.5 fold of the observed Clb of 11.3 mL/min/kg. The good alignment between the predicted phase I driven metabolic clearance based on the in vitro tools used herein and the observed total in vivo clearance suggests that elimination for this compound was predominantly via CYP-mediated hepatic metabolism. While M002 showed good alignment between predicted CLhep and observed in vivo Clp, this will not be the case for compounds where clearance is driven via mechanisms other than phase I hepatic metabolism. However, for any given compound, the predicted CLhep using the methodology described herein will provide an accurate assessment of the contribution of CYP-mediated clearance to the total observed in vivo drug clearance. Data for M002 is in full agreement with several other studies that have reported the significance of including microsomal protein binding along with plasma protein binding for predicting in vivo clearance from in vitro data (Obach, 1997; Obach, 1996). Our data confirms that incorporation of fup/fumic in clearance prediction models improved the accuracy of the calculated CLhep at different concentrations of microsomal protein, and especially when microsomal stability studies need to be executed at higher concentration of
39
microsomal protein. Typically, in early drug discovery compounds are rank-ordered for metabolic liability based on their in vitro microsomal stability data. Although this approach may be appropriate for compounds with minimal binding to microsomal and plasma proteins, it could be misleading for compounds that bind to microsomal protein used in the metabolic stability assay. The described correction factor (fup/fumic) is also critical when microsomal stability data at different protein concentrations needs to be compared. Based on our observations, even closely related compounds within the same structural series can exhibit quite dramatic differences in fup/fumic, indicating that this correction ratio is indeed critical for proper quantitative assessment of phase I driven CLhep. Due to its simplicity and amenability to automation, the direct determination of fup/fumic can be readily incorporated into early drug discovery, and will provide a more refined tool for structure activity relationship and compound prioritization around phase I driven metabolic liabilities. The current equilibrium dialysis assay was performed in a 24 well plate; efforts are currently underway to adopt this assay to an automated, faster, and higher throughput format. In addition we are also investigating different plate designs and sealing options which might allow the assay to be conducted at physiological temperature of 37 °C with a shorter incubation time. In summary, we have optimized and validated a simple assay to directly and accurately measure the ratio of unbound fraction in plasma to unbound fraction in microsomes for a diverse set of compounds. As an example we have illustrated the impact of incorporating this binding ratio on the prediction of CLhep, specifically when microsomal protein concentrations are altered in stability experiments. Acknowledgement The authors would like to thank Gloria Kwei for her support in the preparation of this manuscript. References Austin, R. P., Barton, P., Cockroft, S. L., Wenlock, M. C., & Riley, R. J. (2002). The influence of nonspecific microsomal binding on apparent intrinsic clearance, and its prediction from physicochemical properties. Drug Metabolism and Disposition, 30, 1497−1503. Davies, B., & Morris, T. (1993). Physiological parameters in laboratory animals and humans. Pharmaceutical Research, 10, 1093−1095. Giuliano, C., Jairaj, M., Zafiu, C. M., & Laufer, R. (2005). Direct determination of unbound intrinsic drug clearance in the microsomal stability assay. Drug Metabolism and Disposition, 33, 1319−1324. Jones, H. M., & Houston, J. B. (2004). Substrate depletion approach for determining in vitro metabolic clearance: Time dependencies in hepatocyte and microsomal incubations. Drug Metabolism and Disposition, 32, 973−982. Naritomi, Y., Terashita, S., Kimura, S., Suzuki, A., Kagayama, A., & Sugiyama, Y. (2001). Prediction of human hepatic clearance from in vivo animal experiments and in vitro metabolic studies with liver microsomes from animals and humans. Drug Metabolism and Disposition, 29, 1316−1324. Obach, R. S. (1996). The importance of nonspecific binding in in vitro matrices, its impact on enzyme kinetic studies of drug metabolism reactions, and implications for in vitro–in vivo correlations. Drug Metabolism and Disposition, 24, 1047−1049. Obach, R. S. (1997). Nonspecific binding to microsomes: Impact on scale-up of in vitro intrinsic clearance to hepatic clearance as assessed through examination of warfarin, imipramine, and propranolol. Drug Metabolism and Disposition, 25, 1359−1369. Obach, R. S. (1999). Prediction of human clearance of twenty-nine drugs from hepatic microsomal intrinsic clearance data: An examination of in vitro half-life approach and nonspecific binding to microsomes. Drug Metabolism and Disposition, 27, 1350−1359. Pang, K. S., & Rowland, M. (1977). Hepatic clearance of drugs. I. Theoretical considerations of a “well-stirred” model and a “parallel tube” model. Influence of hepatic blood flow, plasma and blood cell binding, and the hepatocellular enzymatic activity on hepatic drug clearance. Journal of Pharmacokinetics and Biopharmaceutics, 5, 625−653. Rane, A., Wilkinson, G. R., & Shand, D. G. (1977). Prediction of hepatic extraction ratio from in vitro measurement of intrinsic clearance. Journal of Pharmacology and Experimental Therapeutics, 200, 420−424. Sebille, B., Zini, R., Madjar, C. V., Thuaud, N., & Tillement, J. P. (1990). Separation procedures used to reveal and follow drug–protein binding. Journal of Chromatography, 531, 51−77. Skaggs, S. M., Foti, R. S., & Fisher, M. B. (2006). A streamlined method to predict hepatic clearance using human liver microsomes in the presence of human plasma. Journal of Pharmacological and Toxicological Methods, 53, 284−290. Wilkinson, G. R., & Shand, D. G. (1975). Commentary: A physiological approach to hepatic drug clearance. Clinical Pharmacology and Therapeutics, 18, 377−390.
Contents lists available at ScienceDirect
Journal of Pharmacological and Toxicological Methods j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j p h a r m t ox
Original article
Direct determination of the ratio of unbound fraction in plasma to unbound fraction in microsomal system (fup/fumic) for refined prediction of phase I mediated metabolic hepatic clearance Sujal V. Deshmukh ⁎, Andreas Harsch Drug Metabolism and Pharmacokinetics, Merck Research Laboratories, Boston, MA, United States
a r t i c l e
i n f o
Article history: Received 5 February 2010 Accepted 20 April 2010 Keywords: Microsomes Plasma Prediction Hepatic Metabolic Binding
a b s t r a c t At the drug discovery stage, in vivo metabolic hepatic clearance (CLhep) is commonly predicted using in vitro parent compound disappearance data generated in liver microsomes or hepatocytes. Correction for the unbound fraction of a compound in the in vitro system and in plasma/serum is known to be critical for the accuracy of metabolic clearance predictions. Discrete generation of these required experimental parameters can be laborious. Herein, we describe a straightforward and direct approach to obtain the ratio of unbound fraction in plasma (fup) to unbound fraction in the microsomal system (fumic) of a small molecule compound using equilibrium dialysis. Experimental conditions were optimized with respect to incubation time, temperature, and plate shaking speed. Results obtained from this system were validated for a set of test compounds by comparison to individually measured fup and fumic data using ultracentrifugation. The correlation for fup/fumic between the two methods for a set of 23 data points was very good with R2 of 0.94, slope of 1.05 and an intercept of 0.007. The impact of microsomal binding on predicted CLhep was illustrated for a tightly bound compound using a series of incubations with increasing concentration of monkey liver microsomal protein. Alteration of this experimental parameter profoundly affected calculated CLhep using the well-stirred model. Significant differences were observed in the prediction when the model was corrected for fup only; in contrast, the model corrected for plasma protein and microsomal protein binding predicted clearance values independent of the microsomal protein concentration. © 2010 Elsevier Inc. All rights reserved.
1. Introduction Prediction of in vivo clearance in humans is an integral part of new chemical entity profiling in drug discovery and early development. Since hepatic metabolic contributions often play a major role in overall xenobiotic clearance, intrinsic hepatic metabolic clearance (CLint) is determined fairly early in drug discovery programs. These data are commonly generated by measuring drug disappearance in subcellular hepatic fractions, such as S9 and microsomes (Obach, 1999). This experimental approach is broadly implemented due to its simplicity, ease of use, and amenability to high throughput formats. Although at an early drug discovery stage the exact mechanism of in vivo clearance of compounds is often unknown, it is important to accurately predict in vivo metabolic hepatic clearance (CLhep) to allow rank-ordering
Abbreviations: CLhep, in vivo metabolic hepatic clearance; fup, unbound fraction of a compound in plasma; fumic, unbound fraction of a compound in microsomal system; CLint, intrinsic hepatic metabolic clearance; LSC, liquid scintillation counter; dpm, disintegrations per minute; CLb, blood clearance. ⁎ Corresponding author. 33 Avenue Louis Pasteur, BMB 4-112, Boston, MA 02115, United States. Tel.: +1 617 734 2076; fax: +1 617 734 2084. E-mail address: [email protected] (S.V. Deshmukh). 1056-8719/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.vascn.2010.04.003
compounds based on this parameter. Several physiological models are available to scale in vitro intrinsic metabolic clearance data to in vivo organ clearance (Pang and Rowland, 1977; Rane, Wilkinson, & Shand, 1977; Wilkinson and Shand, 1975). Hepatic blood flow, hepatic intrinsic clearance and plasma/serum protein binding are minimal experimental inputs required to allow scalability of these models. Obach (1996, 1997, 1999) demonstrated the significance of including a measure of drug binding to microsomes as an additional important correction factor for accurate prediction of in vivo clearances. Several subsequent studies have evaluated nonspecific microsomal protein binding of xenobiotics and have confirmed its relevance in predicting in vivo clearance (Jones and Houston, 2004; Giuliano, Jairaj, Zafiu, & Laufer, 2005; Naritomi et al., 2001; Austin, Barton, Cockroft, Wenlock, & Riley, 2002). In all published reports, with the exception of Obach (1997) and Skaggs, Foti, and Fisher (2006) the unbound fraction in plasma and microsomes was separately determined. For proper correction, these independently determined factors were entered in the scaling equations as a ratio of fraction unbound in plasma (fup) to fraction unbound in microsomal system (fumic). Herein, a straightforward method for the direct and accurate experimental determination of this correction ratio (fup/fumic) is reported. Equilibrium dialysis was used to measure drug partitioning
36
S.V. Deshmukh, A. Harsch / Journal of Pharmacological and Toxicological Methods 63 (2011) 35–39
between plasma and microsomes, allowing direct calculation of fup/ fumic by determining the ratio of total drug concentration in plasma to total drug concentration in microsomes (Eq. (1)). At equilibrium, the unbound concentration in plasma, Cp × fup is equal to the unbound concentration in microsomal system, Cmic × fumic so that Cp × fup = Cmic × fumic. By rearranging this we get: fup = fumic = Cmic = Cp
ð1Þ
Where, Cmic Cp
is the total concentration of compound in microsomal system is the total concentration of compound in plasma
The experimental procedure was optimized using a focused set of compounds, and subsequently used to measure fup/fumic for compounds spanning a wide range of plasma and microsomal binding. Results were validated by comparison to calculated fup/fumic using individually determined measurements of fup and fumic by ultracentrifugation. Finally, the impact of microsomal and protein binding on CLhep was illustrated using a proprietary compound. 2. Materials and methods 2.1. Reagents 3
3
14
Radiolabeled verapamil ( H), digoxin ( H), and caffeine ( C) were purchased from Perkin Elmer Life and Analytical Sciences, Boston MA, 3 amitriptyline ( H) was obtained from American Radiolabeled Chemicals Inc. St. Louis, MO, and midazolam was purchased from Sigma Aldrich, St. Louis, MO. Proprietary radiolabeled compounds and diclofenac were obtained from the Labeled Compound Synthesis group, Merck Research Laboratories, Rahway, NJ. All other investigational compounds were provided by the Chemistry group, Merck Research Laboratories, Boston. Human liver microsomes were purchased from BD Biosciences, San Jose, CA. Mouse, rat, dog, and rhesus monkey liver microsomes were sourced from CellzDirect, Pittsboro, NC. Human plasma was purchased from Biomeda, Foster City, CA. Plasma from all other species was obtained from in-house sources.
In the case of midazolam, LC–MS/MS was used as described below. Radioactivity counts were replaced with area ratio to calculate unbound fraction. Each determination of unbound fraction was done as a singlet. 2.2.2. Protein binding ratio using equilibrium dialysis The ratio of unbound fraction across plasma and microsomes was directly determined using a 24 multi-well equilibrium dialysis system (BD Biosciences, Bedford, MA). The experiment was performed according to the guidelines provided by the vendor. Briefly, the membranes were soaked for 20 min in deionized water followed by 20 min soaking in 30% absolute alcohol:deionized water. The membranes were washed with water and further soaked in 50 mM potassium phosphate buffer for 20 min. 750 µL of plasma containing 1 µM of compound was loaded in the bottom chamber and 250 µL of the microsomal sample containing the same drug concentration was dispensed into the top chamber. Each sample was incubated in duplicate. For fup determination of M002 the microsomal chamber was replaced with buffer. In both matrices the final concentration of acetonitrile was maintained at 5% v/v as used in the ultracentrifugation method. The assembly was sealed using parafilm and subsequently placed on a shaker (Thermomixer, Eppendorf, New York, NY) at a speed of 300 rpm and allowed to incubate for 45 h at 25 °C. Samples set aside at room temperature served as controls. Plasma and microsomal samples were analyzed by radiometric detection or LC–MS/MS. For radiolabel compounds, 10 µL of the samples was mixed with 5 mL of scintillation cocktail and dpm were measured using a LSC, Perkin Elmer, Waltham, MA. For non-radiolabelled compounds, 50 µL aliquots of each matrix were precipitated with acetonitrile containing an appropriate internal standard. Samples were filtered, evaporated, and reconstituted in 100 µL acetonitrile:water (30:70 v/v). 10 µL sample was injected on a C18 column (2.0 mm× 30 mm, 3 µm particle size) and eluted using a gradient LC method with water containing 0.1% formic acid as the aqueous mobile phase, and acetonitrile containing 0.1% formic acid as the organic phase. Electrospray ionization with multiple reaction monitoring was used for MS/MS detection. Area ratios for each sample were determined by dividing the integrated analyte peak area to internal standard peak area. fup/fumic was calculated as follows: fup = fumic =
ðdpm or area ratio in microsomal incubateÞ ðdpm or area ratio in microsomal controlÞ ×
2.2. Experimental procedure 2.2.1. Protein binding using ultracentrifugation Plasma protein binding and microsomal protein binding using ultracentrifugation were determined by adding 100 µL of a 10 µM solution containing 2.27 µCi/mL radioactivity of a test compound in a 1:1 mixture of acetonitrile:water to 900 µL of either plasma or microsomes (0.25 mg/mL protein). Subsequently, 800 µL aliquots were transferred to 1 mL polycarbonate centrifuge tubes, Beckman Instruments Inc., Palo Alto, CA. The remaining fraction was stored at room temperature and was later used as control sample. Samples were centrifuged at ∼ 90,000 × g for 18 h at 25 °C using a Beckman Optima L-100 XP Ultracentrifuge, Beckman Coulter Inc., Fullerton, CA. Following centrifugation, six aliquots of 100 µL each, starting from the top layer, were gently pipetted out and placed in separate scintillation vials, and 5 mL of scintillation cocktail was added to each sample. The amount of radioactivity (disintegration per minute, dpm) in each fraction was measured using a liquid scintillation counter (LSC). The unbound fraction was calculated using the following equation:
Unbound fraction =
Radioactivity counts in aliquot with lowest counts : Radioactivity counts in control sample ð2Þ
ð3Þ
ðdpm or area ratio in plasma controlÞ : ðdpm or area ratio in plasma incubateÞ
Normalizing data to the control samples should account for any recovery differences in the assayed matrices, especially when conducting LC–MS/MS analysis. Standard curves were not generated to determine compound concentration, rather area ratios were used for fup/fumic calculation. It was ensured that MS responses for all samples were within the linear dynamic range of the instrument. High organic content may affect the protein binding for some compounds; it might thus be preferable to use lower organic content (0.5% v/v) for future data generation. We determined fup/fumic at 5% and 0.5% v/v of acetonitrile using 4 compounds with fup/fumic ranging from 0.04 to 0.28. The data suggests that fup/fumic was not affected by the amount of acetonitrile used, the values obtained with 5% and 0.5% acetonitrile were not statistically different. 2.2.3. Microsomal stability assay To determine CLint of M002 in monkey liver microsomes, 1 µM of the compound was added to varying concentration of monkey liver microsomes in the presence of 1 mM NADPH. The microsomal system was prepared using 50 mM potassium phosphate buffer, pH 7.4. The samples were incubated in duplicate at 37 °C for 0, 5, 10, 15, 20, and 30 min and concentration of M002 was determined using LC–MS/MS. CLint was calculated by the parent disappearance method according to
S.V. Deshmukh, A. Harsch / Journal of Pharmacological and Toxicological Methods 63 (2011) 35–39
37
Eq. (4). For calculation purposes the blood flow for rhesus monkey was considered to be 40 mL/min/kg (Davies and Morris, 1993) and the monkey liver microsomal recovery was estimated at 50 mg/g liver with a liver weight of 30 g/kg body weight. CLhep was calculated without any correction for protein binding, binding in plasma only, or binding in plasma and microsomal protein using Eqs. (5)–(7), respectively. CLint =
Dose ðμM = mg microsomal protein = mLÞ
ð4Þ
Area under parent disappearance curveðμM⋅minÞ ×
mg microsomal protein × g of liver weight g liver kg body weight
Predicted CLhep uncorrected for binding = Q hepatic ·
Predicted CLhep corrected for fup =
CLint Q hepatic + CLint
fup ðB = P Þ ⋅CLint Q hepatic · fu Q hepatic + ðB =pP Þ ·CLint
ðfup = fumic Þ Predicted CLhep corrected for fup and fumic = Q hepatic ·
ðB = P Þ
Q hepatic +
ð5Þ
ð6Þ
⋅CLint
ðfup = fumic Þ ðB = P Þ
·CLint
ð7Þ Where, Q is the hepatic blood flow; B/P is the blood to plasma ratio. 2.2.4. Blood to plasma ratio The blood to plasma ratio for M002 was determined in rhesus 3 monkey using radiolabeled [ H] M002. To 995 µL of blood, 5 µL of a 200 µM solution containing 2.27 µCi/mL of M002 was added and gently mixed. The blood was incubated at 37 °C for 30 min in a water bath and then centrifuged for 15 min at 3220 ×g. The above procedure was also performed using plasma instead of blood, this sample served as the control. Following centrifugation, 100 µL of plasma was transferred to a scintillation vial and 5 mL of scintillation cocktail was added to each sample and dpm measured using LSC Perkin Elmer, Waltham, MA. The blood to plasma ratio was calculated by dividing the dpm from control sample with dpm from plasma that was prepared following the addition of M002 to blood. The stability of M002 in plasma was tested under the assay conditions to ensure that the radioactivity at the end of the study represented M002.
Fig. 1. Effect of incubation time on fup/fumic for M001 (▲), verapamil (○) and caffeine (♦) using equilibrium dialysis. Data for each time point represents mean from a set of duplicates, conducted at 25 °C with a shaking speed of 300 rpm. The experiment was conducted at a compound concentration of 1 µM in human plasma and 0.25 mg/mL human liver microsomes.
compounds was incubated over 3 days at the optimized temperature and shaking speed to determine appropriate incubation times. Samples were taken at 19 h, 45 h, and 68 h after the start of incubation (Fig. 1). Caffeine maintained a fup/fumic ratio close to 1 throughout the sampled time interval, whereas M001, which exhibited a very low fup/fumic did not achieve equilibrium even after overnight incubation. A similar observation was also made for verapamil, its measured fup/fumic ratio continued to decrease up to 45 h and then remained constant. 3.2. Comparison of methods for the determination of fup/fumic The equilibrium dialysis assay was validated by comparing the direct determination of fup/fumic by equilibrium dialysis to data from ultracentrifugation obtained by individually measuring the unbound fraction in plasma and in microsomes and then calculating fup/fumic. This correlation analysis was done using 15 compounds with broad
2.2.5. In vivo clearance In vivo blood clearance (CLb) for M002 was determined in a rhesus monkey pharmacokinetic study. An intravenous dose of 0.25 mg/kg of M002 was administered to each subject (n = 2). Plasma samples were collected at 5 min, 15 min, 30 min, 1 h, 2 h, 4 h, 6 h, 8 h, and 24 h. Samples were analyzed by LC–MS/MS as described previously. Noncompartmental analysis (Watson 7.2) was used to calculate plasma clearance. CLb was calculated by dividing plasma clearance with the blood to plasma ratio. 3. Results 3.1. Assay optimization Equilibrium dialysis assay conditions were initially optimized for direct determination of fup/fumic in human plasma and liver microsomes using a set of 3 compounds (verapamil, caffeine, and M001 — a proprietary compound). These compounds were selected based on their broad range of binding ratio (fup/fumic) as determined by the ultracentrifugation method (Eq. (2)). The fup values for verapamil, caffeine, and M001 were 0.28, 0.91, and 0.05, respectively and the determined fumic were 0.85, 1, and 0.76, respectively. All analyses were performed by radiometric scintillation counting. The test set of 3
Fig. 2. Correlation for fup/fumic using ultracentrifugation (calculated by dividing the individually measured unbound fraction in plasma and microsomes) versus equilibrium dialysis (direct determination). Data represents mean fup/fumic ratio from a duplicate incubation for the equilibrium dialysis method and single measurement for the ultracentrifugation. The assays were conducted at a compound concentration of 1 µM in plasma and 0.25 mg/mL liver microsomes. The 23 data points represent fup/fumic ratio for 13 compounds using human plasma and liver microsomes and two compounds, M001 and M002 using mouse, rat, dog, monkey, and human plasma and liver microsomes.
38
S.V. Deshmukh, A. Harsch / Journal of Pharmacological and Toxicological Methods 63 (2011) 35–39
Table 1 Binding parameters for generic compounds. fupa
fumica fup/fumicb fup/fumicc
fup/fumic
fup/fumic
Calculated Experimental Calculated Experimental (Skaggs (Skaggs et al., 2006) et al., 2006) Amitriptyline Caffeine Diclofenac Digoxin Midazolam Verapamil a b c
0.097 0.91 0.012 0.81 0.01 0.28
0.67 1.00 0.89 0.98 0.7 0.85
0.145 0.910 0.013 0.830 0.015 0.330
0.094 0.930 0.026 0.930 0.022 0.230
0.274 – 0.011 – 0.104 0.248
1.14 – 0.178 – 0.456 0.689
Determined by ultracentrifugation. Ratio of fup and fumic for respective compound. Determined by equilibrium dialysis.
structural diversity and covering a wide range of binding ratios. The fup and fumic values ranged from 0.01–0.91 and 0.66–1, respectively with fup/fumic in the range of 0.01–0.9. The assay was predominantly performed using plasma and liver microsomes from humans, but matrices from other species (mouse, rat, dog, and monkey) were also used for comparison for the two test compounds, M001 and M002. A correlation plot of 23 data points for fup/fumic determined using ultracentrifugation and equilibrium dialysis techniques is shown in Fig. 2. The coefficient of determination, R2, was 0.94 with a slope of 1.05 and an intercept of 0.007. Data on a subset of the test compounds used in the correlation is represented in Table 1. Selection of the subset was based on availability of published literature on fup/fumic determined using equilibrium dialysis (Skaggs et al., 2006). 3.3. Impact of increasing concentration of microsomal protein on predicted CLhep The impact of fup/fumic on the prediction of CLhep (Eq. (7)) was illustrated using a proprietary compound M002. M002 was incubated at 1 μM with increasing concentration of rhesus monkey liver microsomes. The fup for M002 in rhesus monkey plasma was 0.02, which was determined by replacing the microsomes with buffer in the equilibrium dialysis experiment. The blood to plasma ratio in rhesus monkey for M002 was 0.75. The effect of increasing concentration of liver microsomal protein on CLint (Eq. (4)) and CLhep (Eqs. (5)–(7)) is summarized in Table 2. The CLb for M002 in monkey was determined to be 11.3 mL/min/kg. 4. Discussion Assay conditions were optimized for incubation temperature, time, and plate shaking speed. At 37 °C, extensive evaporation was observed within a few hours of incubation, thus this temperature was impractical to test the concept of direct ratio determination using an assay requiring extended incubation time. Evaporation was minimal at 25 °C thus this temperature was chosen for all subsequent studies. A platform shaker was used for sample agitation. At speeds greater
than 300 rpm there was a risk of cross-well spilling causing sample contamination. A speed of 300 rpm was found to be optimal for effective mixing. To evaluate the impact of incubation time on fup/ fumic, the equilibrium dialysis assay was conducted at the optimized temperature and shaking speed over 3 days using verapamil, caffeine and M001 as test compounds. Caffeine, which had similar unbound fraction in plasma and microsomal incubate achieved rapid equilibrium such that there was no change in fup/fumic over time (Fig. 1). On the other hand, verapamil and M001, which are tightly bound in the plasma and relatively unbound in the microsomal incubate, fup/fumic changed significantly over time (Fig. 1). Data from verapamil and M001 indicates that for compounds with significant differences between unbound fractions in plasma and microsomes, appropriate selection of the incubation time is critical for achieving equilibrium, which in turn is essential for accurate measurement of fup/fumic (Eq. (1)). Based on these initial data, an incubation time of 2 days or ∼45 h was selected in order to achieve equilibrium for compounds with significantly different unbound fractions in both matrices. It should be noted that the time to equilibrium is dependent on a range of compound-specific properties (binding to the biomatrices, diffusibility), device-specific parameters (surface size of the membrane, surface to well volume ratio, plate geometry, etc.) (Sebille, Zini, Madjar, Thuaud, & Tillement, 1990) as well as incubation conditions (pH, temperature, etc.). Efforts are currently underway to identify plates which will allow accelerated mass transfer and thus reduce incubation times. To validate the concept of direct determination of fup/fumic, we compared the experimental data from the equilibrium dialysis method to data calculated using ultracentrifugation. 13 compounds were evaluated for fup/fumic using human plasma and liver microsomes and for two compounds, M001 and M002 fup/fumic was determined using mouse, rat, dog, monkey, and human plasma and liver microsomes. For all data points there was a less than 2 fold difference between fup/fumic determined by equilibrium dialysis versus calculated via individually determined fup and fumic measurements using ultracentrifugation (Table 1 and Fig. 2). Skaggs et al. (2006) published a similar comparison for amitriptyline, diclofenac, midazolam and verapamil. Under their experimental conditions, the difference between calculated and experimental ratio ranged from 3 to 16 fold (Table 1). In our assay, an excellent correlation (R2 = 0.94) was obtained for the 23 data points with a slope of 1.05 and an intercept of 0.007 (Fig. 2). One possible reason for the excellent correlation achieved herein may be attributed to the optimized incubation time used in the equilibrium dialysis technique to ensure complete equilibrium. In the publications from Skaggs et al. (2006) and Obach (1997), it appears that the incubation time was 4 h and 4–5 h respective, at 37 °C which may not have been enough to achieve equilibrium for all compounds. In conclusion, the equilibrium dialysis assay conditions described herein were well suited for direct determination of fup/fumic. In addition to the simplicity of this experimental approach, there are possible intrinsic advantages to using direct fup/fumic determination (protein in both compartments) vis-à-vis the traditional route of independent plasma or microsomal protein binding determinations
Table 2 Effect of increasing concentration of monkey liver microsomes on fup/fumic, CLint, and CLhep. Conc. of monkey liver microsomes (mg/mL)
fup/fumica
CLint (mL/min/kg)
Predicted CLhep uncorrected for protein binding (mL/min/kg)
Predicted CLhep corrected for plasma protein binding (mL/min/kg)
Predicted CLhep corrected for plasma and microsomal protein binding (mL/min/kg)
0.1 0.25 0.5 1 2
0.05 0.06 0.07 0.11 0.17
199 153 103 75 42
33 32 29 26 20
4.7 3.7 2.6 1.9 1.1
10 9.4 7.8 8.7 7.6
a
fup for M002 was determined to be = 0.02.
S.V. Deshmukh, A. Harsch / Journal of Pharmacological and Toxicological Methods 63 (2011) 35–39
by equilibrium dialysis (protein in one compartment, buffer in the other compartment). Presence of protein in both compartments may help to keep compounds with poor physico-chemical properties in solution, and thus enable generation of fup/fumic for compounds that might not yield meaningful data when fup and fumic are measured independently, for example, due to lack of aqueous solubility. To demonstrate the impact of plasma protein binding and microsomal protein binding on CLhep, we used a proprietary compound — M002. This compound was known to be eliminated predominantly via phase I mediated hepatic metabolic clearance with minimal contribution from direct renal elimination based on data from a rat bile duct cannulated study (data not shown). Monkey was selected for this illustration since M002 showed measurable turnover in the monkey liver microsomes across a wide range of microsomal protein concentrations. For this case study, the concentration of monkey liver microsomal protein was varied from 0.1 to 2 mg/mL, and CLint and fup/fumic were determined using the microsomal stability assay and the optimized equilibrium dialysis assay, respectively. Increasing the concentration of microsomal protein resulted in a decrease in the CLint even after normalization to the protein concentration used in the incubation. In contrast, fup/fumic increased with increasing concentration of microsmal protein (Table 2). Since plasma concentrations are unaltered, fup is constant. Consequently, the observed increase in the fup/fumic ratio can be readily attributed to a decrease of fumic. In other words, the drug becomes more extensively bound to microsomal protein with increasing microsomal protein concentration. Application of the well-stirred model without correction for any binding (fup/fumic set to 1) led to a substantial over-prediction of the CLhep of M002, as expected, and the predicted CLhep values decreased with increasing microsomal protein concentrations (Table 2). This concentration dependence was even more pronounced after correction for plasma protein binding only (fup set to 0.02). This can be readily rationalized via a simple boundary analysis of the equation for the well-stirred model. Blood to plasma was not considered herein for illustration purposes. In situations where fup/fumic ×CLint is much smaller than Qhep, the equation for the well-stirred model simplifies to CLhep = fup/fumic × CLint. Changes in the fup/fumic ×CLint term will thus directly translate into changes in calculated CLhep. Herein, this is the case when the determined CLint are corrected for fup only. In cases when fup/fumic × CLint NN Qhep, the calculated CLhep will vary less and asymptotically approach Qhep as fup/ fumic × CLint increases. In this study, this is observed when CLhep is predicted using CLint without any correction for binding effects. In contrast to the scenarios discussed above, correction for plasma protein binding and microsomal protein binding (fup/fumic) (Eq. (7)) rendered the calculated CLhep value independent of the microsomal concentration used in the incubations, as the concentration dependent decrease of CLint was offset by the concentration dependent increase of fup/fumic. In this case study, the obtained hepatic metabolic clearance of M002 was also within 1.5 fold of the observed Clb of 11.3 mL/min/kg. The good alignment between the predicted phase I driven metabolic clearance based on the in vitro tools used herein and the observed total in vivo clearance suggests that elimination for this compound was predominantly via CYP-mediated hepatic metabolism. While M002 showed good alignment between predicted CLhep and observed in vivo Clp, this will not be the case for compounds where clearance is driven via mechanisms other than phase I hepatic metabolism. However, for any given compound, the predicted CLhep using the methodology described herein will provide an accurate assessment of the contribution of CYP-mediated clearance to the total observed in vivo drug clearance. Data for M002 is in full agreement with several other studies that have reported the significance of including microsomal protein binding along with plasma protein binding for predicting in vivo clearance from in vitro data (Obach, 1997; Obach, 1996). Our data confirms that incorporation of fup/fumic in clearance prediction models improved the accuracy of the calculated CLhep at different concentrations of microsomal protein, and especially when microsomal stability studies need to be executed at higher concentration of
39
microsomal protein. Typically, in early drug discovery compounds are rank-ordered for metabolic liability based on their in vitro microsomal stability data. Although this approach may be appropriate for compounds with minimal binding to microsomal and plasma proteins, it could be misleading for compounds that bind to microsomal protein used in the metabolic stability assay. The described correction factor (fup/fumic) is also critical when microsomal stability data at different protein concentrations needs to be compared. Based on our observations, even closely related compounds within the same structural series can exhibit quite dramatic differences in fup/fumic, indicating that this correction ratio is indeed critical for proper quantitative assessment of phase I driven CLhep. Due to its simplicity and amenability to automation, the direct determination of fup/fumic can be readily incorporated into early drug discovery, and will provide a more refined tool for structure activity relationship and compound prioritization around phase I driven metabolic liabilities. The current equilibrium dialysis assay was performed in a 24 well plate; efforts are currently underway to adopt this assay to an automated, faster, and higher throughput format. In addition we are also investigating different plate designs and sealing options which might allow the assay to be conducted at physiological temperature of 37 °C with a shorter incubation time. In summary, we have optimized and validated a simple assay to directly and accurately measure the ratio of unbound fraction in plasma to unbound fraction in microsomes for a diverse set of compounds. As an example we have illustrated the impact of incorporating this binding ratio on the prediction of CLhep, specifically when microsomal protein concentrations are altered in stability experiments. Acknowledgement The authors would like to thank Gloria Kwei for her support in the preparation of this manuscript. References Austin, R. P., Barton, P., Cockroft, S. L., Wenlock, M. C., & Riley, R. J. (2002). The influence of nonspecific microsomal binding on apparent intrinsic clearance, and its prediction from physicochemical properties. Drug Metabolism and Disposition, 30, 1497−1503. Davies, B., & Morris, T. (1993). Physiological parameters in laboratory animals and humans. Pharmaceutical Research, 10, 1093−1095. Giuliano, C., Jairaj, M., Zafiu, C. M., & Laufer, R. (2005). Direct determination of unbound intrinsic drug clearance in the microsomal stability assay. Drug Metabolism and Disposition, 33, 1319−1324. Jones, H. M., & Houston, J. B. (2004). Substrate depletion approach for determining in vitro metabolic clearance: Time dependencies in hepatocyte and microsomal incubations. Drug Metabolism and Disposition, 32, 973−982. Naritomi, Y., Terashita, S., Kimura, S., Suzuki, A., Kagayama, A., & Sugiyama, Y. (2001). Prediction of human hepatic clearance from in vivo animal experiments and in vitro metabolic studies with liver microsomes from animals and humans. Drug Metabolism and Disposition, 29, 1316−1324. Obach, R. S. (1996). The importance of nonspecific binding in in vitro matrices, its impact on enzyme kinetic studies of drug metabolism reactions, and implications for in vitro–in vivo correlations. Drug Metabolism and Disposition, 24, 1047−1049. Obach, R. S. (1997). Nonspecific binding to microsomes: Impact on scale-up of in vitro intrinsic clearance to hepatic clearance as assessed through examination of warfarin, imipramine, and propranolol. Drug Metabolism and Disposition, 25, 1359−1369. Obach, R. S. (1999). Prediction of human clearance of twenty-nine drugs from hepatic microsomal intrinsic clearance data: An examination of in vitro half-life approach and nonspecific binding to microsomes. Drug Metabolism and Disposition, 27, 1350−1359. Pang, K. S., & Rowland, M. (1977). Hepatic clearance of drugs. I. Theoretical considerations of a “well-stirred” model and a “parallel tube” model. Influence of hepatic blood flow, plasma and blood cell binding, and the hepatocellular enzymatic activity on hepatic drug clearance. Journal of Pharmacokinetics and Biopharmaceutics, 5, 625−653. Rane, A., Wilkinson, G. R., & Shand, D. G. (1977). Prediction of hepatic extraction ratio from in vitro measurement of intrinsic clearance. Journal of Pharmacology and Experimental Therapeutics, 200, 420−424. Sebille, B., Zini, R., Madjar, C. V., Thuaud, N., & Tillement, J. P. (1990). Separation procedures used to reveal and follow drug–protein binding. Journal of Chromatography, 531, 51−77. Skaggs, S. M., Foti, R. S., & Fisher, M. B. (2006). A streamlined method to predict hepatic clearance using human liver microsomes in the presence of human plasma. Journal of Pharmacological and Toxicological Methods, 53, 284−290. Wilkinson, G. R., & Shand, D. G. (1975). Commentary: A physiological approach to hepatic drug clearance. Clinical Pharmacology and Therapeutics, 18, 377−390.