Abstract
Species independence of brain tissue binding was assessed with a large number of structurally diverse compounds using equilibrium dialysis with brain homogenates of seven species and strains (Wistar Han rat, Sprague-Dawley rat, CD-1 mouse, Hartley guinea pig, beagle dog, cynomolgus monkey, and human). The results showed that the fractions unbound of the seven species and strains were strongly correlated with correlation coefficients ranging from 0.93 to 0.99. The cross-species/strain correlations were not significantly different from the interassay correlation with the same species. The linear correlation between Wistar Han and other species had a slope close to 1 and an intercept near 0. Based on orthogonal statistical analysis, no correction is needed for extrapolation of fraction unbound from Wistar Han rat to the other species or strains. Hence, brain tissue binding of Wistar Han rat can be used to obtain binding of other species and strains in drug discovery.
Introduction
The free drug hypothesis is a well accepted and widely applied concept in drug discovery and development (Liu et al., 2006; Hammarlund-Udenaes et al., 2008; Lin, 2008; Smith et al., 2010). This hypothesis states that the free (unbound) drug at the site of action exerts pharmacological activity, rather than total drug (bound and unbound), and that the free drug is able to distribute from the systemic circulation across membranes to tissues, rather than total drug. Provided that simple diffusion governs distribution, at equilibrium the free drug concentration in the systemic circulation and extravascular compartment will be equal, with total concentrations at a given time being governed by the respective binding constants for the compartments. Although these hypotheses are important for understanding the actions of all drugs, they are particularly important considerations for agents intended to exert their effect on the central nervous system (CNS). The CNS is an extravascular compartment and not directly accessed from the systemic circulation because of the blood-brain barrier. For a compound to reach a target in the CNS, the free drug in plasma must first cross this barrier. Subsequently, this free drug undergoes binding within the CNS to the brain tissues. Finally, the drug that is free in the CNS is available for pharmacological activity. It therefore follows that trying to understand CNS drug behavior using total drug concentration in the brain or plasma alone can be misleading (Smith et al., 2010). For neuroscience therapeutic targets, accurate understanding of total and free drug concentrations in plasma and brain, particularly brain interstitial fluid, is critical for developing PK/PD relationships, projecting doses, and designing clinical studies. Several techniques have been developed to obtain free drug concentrations in the brain directly or indirectly (Di et al., 2008; Liu et al., 2008; Di and Kerns, 2011), including in vivo microdialysis (direct) (Elmquist and Sawchuk, 1997; Hammarlund-Udenaes et al., 1997; de Lange et al., 1997; Hammarlund-Udenaes, 2000; Watson et al., 2006), CSF sampling (indirect) (Shen et al., 2004; Lin, 2008; Fridén et al., 2009b), and a combination of brain distribution through measuring plasma and brain concentration time courses (Kerns and Di, 2008) and brain tissue binding (quasi-direct) (Kalvass and Maurer, 2002; Mano et al., 2002; Maurer et al., 2005; Summerfield et al., 2006; Liu et al., 2009). The latter approach of measuring brain distribution and brain tissue binding is one of the most common strategies in the pharmaceutical industry to elucidate total and free drug PK relationships in brain and plasma. The approach shows advantages over other methods in that it is widely applicable to compounds with a wide range of physicochemical properties (microdialysis is limited to compounds for which recovery from apparatus is not prohibitive). Furthermore, it measures the compartment of interest versus sampling of CSF, which is technically not the compartment of interest. A drug concentration in CSF may not be in equilibrium with the biophase where the biological target resides in the brain parenchyma. Drugs can access CSF via both the cerebrovasculature and the choroidal epithelium. In addition, the approach is technically less challenging and more reproducible and has higher throughput than the other methods.
For brain tissue binding, both low and high throughput methods using brain homogenates (Kalvass and Maurer, 2002; Mano et al., 2002; Summerfield et al., 2006; Wan et al., 2007) and brain slices (Kakee et al., 1996; Becker and Liu, 2006; Fridén et al., 2007, 2009a) have been developed to determine the fraction unbound (fu) of drugs in brain tissues. Brain slices are considered more physiologically relevant than brain homogenates because the cellular structures (cell membrane, influx and efflux transporters, and intracellular fluids) are preserved in brain slices, whereas these are disrupted in brain homogenates. Nevertheless, data generated using the brain homogenate method have good correlation with brain slice data, especially when cytosolic pH partition was corrected for basic compounds (Becker and Liu, 2006; Lin, 2008; Fridén et al., 2011). This finding suggests that nonspecific binding to lipophilic components in the brain is the dominant mechanism for brain tissue binding and that the presence of intact structural elements plays a less significant role in determining brain binding. Furthermore, the free drug concentration determined from binding studies performed in brain homogenates and brain distribution data has good correlation with direct microdialysis (Fridén et al., 2007; Liu et al., 2009) and indirect CSF measurements (Maurer et al., 2005; Liu et al., 2006) of free in vivo brain interstitial drug concentration. The advantage of using brain homogenates is that they are readily available from vendors and can be stored frozen and thawed right before experiments, which is much easier than the process with brain slices. The good predictability and the ease of use make brain homogenate binding one of the most widely used methods for determining fraction unbound in brain tissues.
Species dependence of plasma protein binding is a well known phenomenon. Compounds can bind to specific binding sites of plasma proteins, leading to different fraction unbound in different species per given drug if this specific binding is species-unique. For example, the plasma protein binding of zamifenacin showed marked differences among the various species (fu = 0.0001 for human, 0.0010 for dog, and 0.0020 for rat) (Kratochwil et al., 2004). For this reason, plasma protein binding of drug candidates must be measured in multiple species to compare total exposures determined in those species at a given dose with what may be predicted to be seen in humans and to develop PK/PD relationships. Likewise, brain tissue binding has been routinely determined in multiple species to account for any potential species dependence. However, brain tissue has very different composition than plasma; brain has much higher lipid contents (11% lipid and 7.9% protein) than plasma (0.65% lipid and 18% protein) (Jeffrey and Summerfield, 2007; Di et al., 2008). Likewise, binding to brain tissue is different from binding to plasma tissue. Lipophilicity (logP) has been shown to be the dominant factor for brain tissue binding of compounds (Wan et al., 2007), suggesting that brain tissue binding might be less sensitive to species than plasma protein binding (nonspecific binding to lipids versus potential specific binding to proteins). Some initial studies with limited species or a limited number of test compounds indicated that brain tissue binding is species-independent. Wan et al. (2007) showed that rat and mouse brain fraction unbound had good correlation (R2 = 0.9887, n = 25). Summerfield et al. (2008) observed that brain fraction unbound among rat, Landrace pig, and human correlated well (R2 > 0.9, n = 21). In a review article, Read and Braggio (2010) reported that brain fraction unbound was conserved in eight species (dog, cynomolgus monkey, guinea pig, rat, man, marmoset, pig, and mouse) using a limited number of test compounds (seven compounds). Linear regression was applied in all the studies for data analysis, assuming that the independent variable X was observed accurately without any random experimental errors. Because all the studies reported either used a limited number of species or a limited number of test compounds, the question remains whether the species independence of brain tissue binding is a general phenomenon or whether it is only applicable to certain species, to certain classes of compounds, or to a certain nature of brain binding (e.g., tight versus mostly free). The goal of this study was to evaluate the degree and nature of potential species differences in brain tissue binding using a large number of test compounds that cover a wide range of physicochemical properties and brain binding characteristics in multiple animal species. From these results, we were able to draw conclusions as to whether brain binding is species-dependent or not for drug-like chemical space in animal species and strains commonly used for neuroscience pharmacology, PK, toxicity, and efficacy studies. Orthogonal regression was applied to provide more rigorous statistical analysis, and this more realistically reflects the fact that both X and Y contains experimental variations. The result of this work will determine whether determination of brain binding in a single representative species can replace multispecies determinations. This information will be very useful to guide experimental design, increase throughput, reduce cost, and minimize animal usage as it relates to understanding the nature of free and total brain and plasma PK and PK/PD relationships.
Materials and Methods
Materials.
Sprague-Dawley rat brain homogenates were purchased from Pel-Freez Biologicals (Rogers, AR). All other brain homogenates of nonhuman species (Wistar Han rat, CD-1 mouse, Hartley guinea pig, beagle dog, and cynomolgus monkey) were ordered as custom products through Bioreclamation Inc. (Hicksville, NY). Human brain tissue (occipital cortex) homogenates was obtained through Tissue Solutions Ltd. (Clydebank, UK) as a custom request. All brain tissues used in the study were mixed genders of male and female with the exception of human brain tissue, which was male. Test compounds were obtained from Pfizer Global Material Management (Groton, CT) or purchased from Sigma-Aldrich (St. Louis, MO). Pfizer research compound [(2E)-3-(4-{[(2S,3S,4S,5R)-5-{(1E)-N-[(3-chloro-2,6-difluorobenzyl)oxy]ethanimidoyl}-3,4-dihydroxytetrahydrofuran-2-yl]oxy}-3-hydroxyphenyl)-2-methyl-N-[(3aS,4R,5R,6S,7R, 7aR)-4,6,7-trihydroxyhexahydro-1,3-benzodioxol-5-yl]prop-2-enamide (CP-628374)] (Brighty et al., 1999) was used as an internal standard (IS) for LC-MS in both positive and negative ionization modes. The equilibrium dialysis device and cellulose membranes with molecular weight cutoff of 12,000 to 14,000 were obtained from HTDialysis, LLC (Gales Ferry, CT). Velocity V11 peelable seals were purchased from BD Falcon (Bedford, MA), 96-well deep well plates of 1.2- and 2.2-ml blocks were from Axygen Scientific Inc. (Union City, CA), and tips of 96 blocks were obtained from Apricot Designs (Monrovia, CA).
Preparation of Brain Homogenates.
The homogenates were all prepared by vendors using 1 g of brain in 4 ml of Dulbecco's phosphate buffer solution (DPBS) with high-speed tissue grinders. The homogenates were further processed using a glass Dounce homogenizer (Thermo Fisher Scientific, Waltham, MA) to reduce the size of the brain tissues. The homogenates were frozen at −20°C before use.
Equilibrium Dialysis for Brain Binding Study with Brain Homogenates.
The dialysis membranes were prepared before experiments. The cellulose membranes (molecular weight cutoff 12,000–14,000) were immersed in deionized water for 15 min, followed by 15 min in 30% ethanol-deionized water, 1 min in deionized water, and then at least 15 min or overnight in DPBS. The equilibrium dialysis device was assembled according to the manufacturer's instructions (http://htdialysis.com/page/1puq4/Operating_Instructions.html). An aliquot of compound dissolved in dimethyl sulfoxide stock solution (10 mM) was used to make 100 μM experimental stock solutions. The experimental stock solutions were diluted 1:100 in the brain homogenate and mixed well with a 96-well pipettor (Soken SigmaPet; Soken, Tokyo, Japan or PP550; Apricot Designs). The final compound concentration for the equilibrium dialysis experiment was 1 μM with 1% dimethyl sulfoxide. A 150-μl aliquot of brain homogenates spiked with a 1 μM concentration of compound was added to one side of the chamber (donor) and 150 μl of DPBS was added to the other side of the dialysis membrane (receiver). Before incubation, an aliquot of 20 μl of brain homogenates spiked with 1 μM concentrations of compounds was added to a 96-well deep plate containing 80 μl of DPBS and 200 μl of ice-cold acetonitrile (ACN) with MS IS CP-628374 (Brighty et al., 1999). These samples were used for recovery calculation and stability evaluation. The equilibrium dialysis device was covered with Breathe-Easy gas permeable membranes obtained from Diversified Biotech (Dedham, MA). Compounds were assessed in triplicate using three equilibrium dialysis devices for each experiment (replicates were between devices rather than within a given device). Equilibrium dialysis devices were placed on a shaking plate at 450 rpm and incubated for 6 h in a humidified incubator at 37°C with 5% CO2. The sampling procedure was designed such that the sample composition was consistent for all the samples to eliminate any potential confounding issues from varying background or ionization efficiency during analysis with LC-MS. At the end of the incubation, 20 μl of the brain homogenate samples from the donor wells were taken and added to a 96-well deep well plate containing 80 μl of DPBS and 200 μl of ice-cold ACN with IS (1.65 μg/ml). Aliquots of 80 μl of dialyzed DPBS were taken from the receiver wells and added to 20 μl of blank brain homogenate and 200 μl of ice-cold ACN with IS in a 96-well deep well plate. The plates were sealed and mixed with a vortex mixer (VWR, West Chester, PA) for 3 min and then centrifuged at 3000 rpm and 4°C (Eppendorf North America, New York, NY) for 5 min. The supernatant was transferred to a new deep well block, sealed, and subsequently analyzed using LC-tandem mass spectrometry as described in the following section.
Instrumentation and Software.
An API 3000 triple quadrupole mass spectrometer equipped with a TurboIonSpray source from Applied Biosystems (Foster City, CA), two Shimadzu LC10-AD pumps (Marlborough, MA), and a Gilson 215 injector (Middleton, WI) were used for sample analysis. Analyst 1.4.2 (Applied Biosystems), Galileo (Thermo Fisher Scientific), and custom software were applied for data collection, processing, and analysis (Janiszewski et al., 2001; Whalen et al., 2006). For each analyte, multiple reaction monitoring methods were generated using DiscoveryQuant 2.0 software (Applied Biosystems) with 3 μM analyte solutions. Samples (25 μl) were injected at 15-s intervals onto a DB11 1.5 × 5 mm column (Optimize Technologies, Oregon City, OR) at a flow rate of 1.2 ml/min with a step gradient: trap and desalt for 0.25 min with 2 mM NH4Ac-ACN-methanol (95:2.5:2.5) and then elute from 0.25 to 0.56 min with 2 mM NH4Ac-ACN-methanol (10:45:45).
Calculation of Fraction Unbound.
fu was calculated using eqs. 1 and 2 as described previously (Kalvass and Maurer, 2002). In brief, because unbound fractions determined from diluted brain homogenates were higher than those from undiluted brain tissue, corrections were made on the basis of the dilution factor (D). Masses in receiver and donor were determined from the area counts in these samples obtained from LC-tandem mass spectrometry analysis corrected to account for sampling volumes. A dilution factor of 5 (D = 5) was applied in the calculation that reflects the dilution of the brain tissue homogenates. where fu, d is the diluted fraction unbound, fu is the undiluted fraction unbound, and D is the dilution factor.
Physicochemical Property Calculations.
Physicochemical properties of the test compounds were calculated using various commercial and in-house software. LogD and pKa were obtained through ACD/PhysChem Batch (version 9.03; Advanced Chemistry Development, Inc., Toronto, ON, Canada). TPSA was computed according to published equations (Ertl et al., 2000) using in-house software.
Statistical Analysis.
For comparison between species, log transformation was applied to approximate normal distribution to avoid skewed distribution of fu in its original scale. The mean of the triplicate assessments was calculated after the log transformation, which was equal to the log of the geometric mean. Because the independent variable (X) in simple linear regression is assumed to be observed without random error, orthogonal regression instead of simple linear regression was applied to account for random errors included in all species for both X and Y. The variances of these random errors were assumed to be equal. Orthogonal regression is based on the least squares as in linear regression. However, the difference between observed dependent variable Y and the fitted line is replaced by the distance from data point (X, Y) to the fitted line. After the orthogonal regression, a composite hypothesis of intercept = 0 and slope = 1 was tested. For significance level 0.05, each parameter was tested with level 0.025 using the Bonferroni adjustment. The correlation coefficient between two random variables, which is a measure of the strength of the linear relationship between them, was calculated. SAS 9.2 (SAS Institute; Cary, NC) and customized programs were used for all the statistical analysis (Fuller, 1987).
Results and Discussion
Selection of Test Compounds and Species.
Forty-seven commercially available compounds covering a wide range of physicochemical properties representing acids, bases, neutrals, and zwitterions were selected for brain tissue binding studies with multiple species. The physicochemical properties of the test compounds are summarized in Table 1 and the diversity of the physicochemical properties is plotted in Fig. 1. The selection of the various physicochemical properties for diversity evaluation was mainly based on properties in Lipinski's rule of five (Lipinski et al., 1997). The compounds show diverse physicochemical properties with logD ranging from −1.43 to 6.01, molecular weight ranging from 151 to 823, and topological polar surface area ranging from 12 to 220. An addition criterion for compound selection was brain binding characteristics, and the set has been shown to cover a range of fraction unbound spanning 3 log units from 0.0005 to 0.5. The use of the large number of diverse compounds covering various chemical space with a range of properties should allow broader conclusions to be generated that apply across various compounds for brain tissue binding. Species included in the test are those that are commonly used for neuroscience drug discovery and in studies focused on understanding CNS drug distribution. The species included were Wistar Han rat, SD rat, CD-1 mouse, Hartley guinea pig, beagle dog, cynomolgus monkey, and human.
Comparison of Brain Fraction Unbound Generated in Multiple Species.
The brain fraction unbound for the 47-compound set was determined in the seven species and strains using equilibrium dialysis with diluted (5×) brain tissue homogenates, and the results are shown in Table 2. Concordance of the brain fraction unbound data between species, strain, and within strain (for Wistar Han, performed to gauge interexperimental correlation) were analyzed using rigorous statistical orthogonal regression. Orthogonal rather than linear regression was used for this analysis because both X and Y contain random experimental errors. The results showed that brain fraction unbound was highly correlated among the various species with the correlation coefficient ranging from 0.93 to 0.99. No significant species or strain differences in brain fraction unbound were observed. Of importance, the cross-species/strain correlations were not significantly different from the interassay correlation that was determined for Wistar Han rat. After the orthogonal regression, a composite hypothesis of intercept = 0 and slope = 1 was tested, and the results are shown in Table 3. For significant difference from intercept = 0 and slope = 1, the P value has to be <0.025. For the statistical tests between Wistar Han rat and other species, the intercepts were all very close to 0 (from 0.00788 to 0.166), the slopes were near 1 (from 0.968 to 1.05), and all the P values for both the intercepts and the slopes were >0.025, indicating that no corrections were needed when fraction unbound of species other than Wistar Han rat were extrapolated. The fraction unbound data for all the species are plotted in Fig. 2.
The results of the brain binding studies with the 47 diverse drug compounds measured in seven species and strains suggest that there is no significant species or strain dependence in brain fraction unbound and further suggests that brain tissue binding is governed predominantly by nonspecific binding. The lack of species differences (or the ubiquitous species-independent nature of brain binding) could potentially be explained by the higher lipid content in brain (Jeffrey and Summerfield, 2007; Di et al., 2008), which is similar across the species tested, leading to higher nonspecific binding than in plasma, or lack of brain proteins in sufficient concentration that selectively bind the compounds of interest. In either case, it seems reasonable to conclude that nonspecific factors govern brain tissue binding and that the key elements determining this nonspecific binding (at least in those studies using brain homogenates to determine binding) are highly similar across mammalian species of interest including human.
In conclusion, brain tissue binding information coupled with other absorption, distribution, metabolism, and excretion properties is critical for accurate assessment of dose, plasma and brain pharmacokinetics, and exposures, and in the development of PK/PD relationships. As such, measuring this parameter remains a critical activity necessary for the study of drugs intended to be used as CNS pharmacological agents. Based on our findings for brain fraction unbound determined in multiple species for a diverse set of drug compounds, we conclude that brain tissue binding is species-independent. A determination of brain fraction unbound in a single species (e.g., Wistar Han rat) can be used as a predictor for brain tissue binding of any preclinical species and strains as well as humans. This finding will greatly reduce the cost and resources needed for brain tissue binding measurements performed to help understand CNS behaviors of drugs. In addition, these findings have great value in helping to eliminate brain tissue binding as a possible cause for any observed (or predicted) differences in the behavior of CNS drugs between species.
Authorship Contributions
Participated in research design: Di, Umland, Chang, Lin, Scott, Troutman, and Liston.
Conducted experiments: Umland and Lin.
Performed data analysis: Di, Umland, Chang, Huang, and Lin.
Wrote or contributed to the writing of the manuscript: Di, Umland, Chang, Huang, Scott, Troutman, and Liston.
Footnotes
Article, publication date, and citation information can be found at http://dmd.aspetjournals.org.
doi:10.1124/dmd.111.038778.
-
ABBREVIATIONS:
- CNS
- central nervous system
- PK
- pharmacokinetic
- PD
- pharmacodynamic
- CSF
- cerebrospinal fluid
- IS
- internal standard
- LC
- liquid chromatography
- MS
- mass spectrometry
- DPBS
- Dulbecco's phosphate buffer solution
- ACN
- acetonitrile
- TPSA
- topological polar surface area
- CP-628374
- (2E)-3-(4-{[(2S,3S,4S,5R)-5-{(1E)-N-[(3-chloro-2,6-difluorobenzyl)oxy]ethanimidoyl}-3,4-dihydroxytetrahydrofuran-2-yl]oxy}-3-hydroxyphenyl)-2-methyl-N-[(3aS,4R,5R,6S,7R,7aR)-4,6,7-trihydroxyhexahydro-1,3-benzodioxol-5-yl]prop-2-enamide
- UK-240455
- N-(6,7-dichloro-2,3-dioxo-1,2,3,4-tetrahydroquinoxalin-5-yl)-N-(2-hydroxyethyl)methanesulfonamide.
- Received February 15, 2011.
- Accepted April 7, 2011.
- Copyright © 2011 by The American Society for Pharmacology and Experimental Therapeutics