Abstract
This work examines the inter-relationship between the unbound drug fractions in blood and brain homogenate, passive membrane permeability, P-glycoprotein (Pgp) efflux ratio, and log octanol/water partition coefficients (cLogP) in determining the extent of central nervous system (CNS) penetration observed in vivo. The present results demonstrate that compounds often considered to be Pgp substrates in rodents (efflux ratio greater than 5 in multidrug resistant Madin-Darby canine kidney cells) with poor passive permeability may still exhibit reasonable CNS penetration in vivo; i.e., where the unbound fractions and nonspecific tissue binding act as a compensating force. In these instances, the efflux ratio and in vitro blood-brain partition ratio may be used to predict the in vivo blood-brain ratio. This relationship may be extended to account for the differences in CNS penetration observed in vivo between mdr1a/b wild type and knockout mice. In some instances, cross-species differences that might initially seem to be related to differing transporter expression can be rationalized from knowledge of unbound fractions alone. The results presented in this article suggest that the information exists to provide a coherent picture of the nature of CNS penetration in the drug discovery setting, allowing the focus to be shifted away from understanding CNS penetration toward the more important aspect of understanding CNS efficacy.
Within the modern drug discovery paradigm, drug metabolism and pharmacokinetics (DMPK) play an integral role in the process of compound selection and progression. Much of the impact of DMPK has been caused by its transformation from a largely descriptive discipline to that of a predictive science, fuelled by advances in bioanalysis and in vitro techniques. Hence discovery DMPK provides a powerful means for assessing the risks of taking potential assets into development.
Nevertheless, the development of molecules targeted at the central nervous system (CNS) remains a significant challenge caused by the increased regulation and protection afforded to the brain over other organs of the body. The major knowledge gaps are 1) understanding the physicochemical features that determine CNS penetration, 2) understanding the impact of the blood-brain barrier (BBB) on CNS uptake, and 3) providing a coherent measure of CNS penetration that can be related to drug efficacy. Regarding the latter point, although it is important to develop a link between the pharmacokinetics of a molecule and the biophase, arguably the critical issue is one of sufficient access of free drug to the requisite site of action.
Numerous models and measures of CNS uptake are available to assist in the search for centrally active agents. In situ brain perfusion techniques have highlighted the good correlation between increasing lipophilicity and CNS permeability. Polar drugs that are subject to paracellular absorption such as atenolol (logDoct,7.4 –2.1; Artursson, 1990) and sumatriptan (logDoct,7.4 –1.5; Pascual and Munoz, 2005) show minimal brain permeability due to the tight junction morphology at the BBB. Lipophilic compounds (i.e., propranolol, logDoct,7.4 1.2; Artursson, 1990), able to undergo passive transcellular diffusion in the unbound, unionized form, show improved BBB permeability relative to atenolol (Street et al., 1979). Outliers to this trend are often substrates for either influx of efflux transporters (Smith, 2003; Kuresh et al., 2004). Of particular note is P-glycoprotein (Pgp), whose promiscuous role in modulating the brain penetration of many therapeutic agents has been the subject of extensive research (Mahar Doan et al., 2002; Lin and Yamazaki, 2003; Lin, 2004).
Assessing the extent of Pgp efflux on CNS uptake may be obtained from either in vitro BBB models, such as Madin-Darby canine kidney (MDCK) cells overexpressing the human form of Pgp or in vivo models by comparing the CNS penetration in mdr1/a(+/+) mice to mdr1a/b(–/–) knockouts. From these in vitro methods, the efflux ratio (ER) is generally determined, defined as the -fold difference between the basolateral-to-apical (B–A) transfer rate of drug across a cell monolayer relative and its apical-to-basolateral (A–B) rate of transfer. In general molecules, exhibiting good passive membrane permeability and low efflux ratios (i.e., <2) tend to exhibit good CNS penetration in vivo. The cut-off for permeability is dependent upon the nature of the assay format, i.e., cell density, but usually these values are between 30 (Wang et al., 2005) and 200 nm/s (Mahar Doan et al., 2002).
Other measures of CNS penetration include brain/plasma or similarly brain/blood (Br:Bl) ratios from bulk tissue, cerebrospinal fluid (CSF) concentrations (Shen et al., 2004), or interstitial fluid concentrations measured via microdialysis (De Lange et al., 1997; Langer and Muller, 2004), with the former two techniques offering better throughput to the meet the demands in drug discovery. Recently, however, equilibrium dialysis has been reported as an alternative means of assessing unbound drug in brain tissue. Studies in rodents (Kalvass and Maurer, 2002; Maurer et al., 2005) have provided a reasonably good prediction of the brain/plasma ratios for compounds where the blood-brain barrier does not modulate CNS penetration (Doran et al., 2005). Despite the fact that the use of tissue homogenates results in the mixing of intracellular and extracellular elements, improved correlations have been reported between brain/plasma ratios and CSF concentrations with the incorporation of fu(brain).
In this study, we have extended the utility of equilibrium dialysis to include the effect of transporters. We report correlations of the unbound fraction in blood and brain and the in vitro Br:Bl partition ratio (Kbb) with several key parameters: efflux ratio, log octanol/water partition coefficients (cLogP), and in vivo Br:Bl ratios derived from the mdr1a/b knockout mouse model. Combining measurements of Kbb and efflux ratio markedly improves the in vitro prediction of in vivo CNS penetration for Pgp substrates in rats. The trend can be extended to account for differences in CNS penetration between wild type and knockout mice, highlighting the quantitative physiological relevance of the efflux ratio.
Materials and Methods
The following compounds were obtained from the Sigma Chemical Co. (Poole, Dorset, UK): paracetamol, ketorolac, ibuprofen, atenolol, flurbiprofen, indomethacin, naproxen, and diclofenac. All other compounds were available from in-house sources.
Calculated Physicochemical Properties. cLogP values were calculated using CLOGP software (Biobyte Corporation, Claremont, CA). Calculations of pKa values (denoted as ApKa1 and BpKa1 for acid and base dissociation) was performed with PKALC version 3.11 (CompuDrug Chemistry Ltd, Budapest, Hungary). Calculation of polar surface area was based on the work of Clark (1999).
Equilibrium Dialysis Measurements. The methodology employed in this study was a modification of that reported by (Kalvass and Maurer, 2002). In brief, a 96-well equilibrium dialysis apparatus was used to determine the free fraction in the blood and brain for each drug (HT Dialysis LLC, Gales Ferry, CT). Membranes (3-kDA cut-off) were conditioned in deionized water for 40 min, followed by conditioning in 80:20 deionized water:ethanol for 20 min, and then rinsed in deionized water before use. Rat blood and brain were obtained fresh on the day of the experiment. Blood was diluted 1:1 with PBS, whereas the brain tissue was homogenized with PBS to a final composition of 1:2 brain:PBS by means of ultrasonication (Tomtec Autogiser; Receptor Technologies, Adderbury, Oxon, UK) in an ice bath. Diluted blood and brain homogenate were spiked with the test compound (1000 ng/g), and 100-μl aliquots (n = 6 replicate determinations) were loaded into the 96-well equilibrium dialysis plate. Dialysis versus PBS (100 μl) was carried out for 5 h in a temperature-controlled incubator at 37°C (Stuart Scientific, Watford, UK) using an orbital microplate shaker at 125 revolutions/min (Stuart Scientific). At the end of the incubation period, aliquots of blood, brain homogenate, or PBS were transferred to ScreenMate tubes (Matrix Technologies, Hudson, NH), and the composition in each tube was balanced with control fluid, such that the volume of PBS to blood or brain was the same. Sample extraction was performed by the addition of 200 μl of acetonitrile containing an internal standard. Samples were allowed to mix for 15 min and then centrifuged at 2465g in 96-well blocks for 20 min (Eppendorf 5810R; VWR International, Poole, Dorset, UK). The unbound fraction was determined as the ratio of the peak area in buffer to that in blood or brain, with correction for dilution factor according to eq. 1 (Kalvass and Maurer, 2002), where D = dilution factor in blood or brain homogenate and fu(apparent) is the measured free fraction of diluted blood or brain tissue.
As described by Kalvass and Maurer (2002), if the unbound fractions in blood and brain are equal at steady state, then the in vivo Br:Bl ratio can be estimated from free fractions in blood and brain tissue as shown in eq. 2, where fu(blood) and fu(brain) denote the unbound fractions in blood and brain. The in vitro brain:blood partition ratio fu(blood)/fu(brain) is termed Kbb in this article.
Analysis of Test Compounds in Equilibrium Dialysis. All samples were analyzed by means of HPLC/MS/MS on a PE-Sciex API-4000 tandem quadrupole mass spectrometer (Applied Biosystems, Concord, ON, Canada), employing a Turbo V Ionspray operated at a source temperature of 700°C (80 psi of nitrogen). Samples (3–10 μl) were injected using a CTC Analytics HTS Pal autosampler (Presearch, Hitchin, UK) onto a Hypersil Aquastar 3.0 × 30 mm, 3-μm column (Thermo, Runcorn, Cheshire, UK) operated at 40°C and at an eluent flow rate of 1 ml/min. Analytes were eluted using a high-pressure linear gradient program by means of an HP1100 binary HPLC system (Agilent Technologies, Stockport, Cheshire, UK), using acetonitrile as solvent B. For HPLC/MS/MS analysis in positive ion mode, solvent A comprised 1 mM ammonium acetate containing 0.1% (v/v) formic acid, whereas in negative ion mode, solvent A comprised 1 mM ammonium acetate. The gradient was held at 5% solvent B for 0.2 min before increasing to 90% at 1.2 min and remaining at 90% until 1.6 min before returning to the starting conditions. The cycle time was 2.5 min/sample. Relative peak areas between the PBS and tissue half-wells were used to determine the respective free fractions.
Apparent Permeability and Pgp Efflux Measurements. All permeability and Pgp efflux measurements were performed as described previously (Wang et al., 2005). In brief, MDR-MDCK monolayers were grown to confluence on collagen-coated, microporous, and polycarbonate membranes in 12-well Costar Transwell plates (Corning Life Sciences, Acton, MA). The buffer used was Hanks' balanced salt solution containing 10 mM HEPES and 15 mM glucose at pH 7.4, and permeability was assessed in the presence and absence of 2 μM Pgp inhibitor GF120918A. All cells were preincubated for 30 min, and for the studies using the Pgp inhibitor, GF120918A (Lentz et al., 2000) was present on both sides of the membrane during the incubation and permeation period. Dosing solution concentrations were 3 μM in assay buffer. Cells were dosed on the apical side (A–B) or basolateral side (B–A) and incubated at 37°C in 5% (v/v) CO2 in a humidified incubator. Each determination was performed in triplicate. There was a single 60-min sampling time point. Permeability through a cell-free (blank) membrane was determined to assess nonspecific binding to the device and free diffusion of the test compounds through the filter. Membrane transfer of Lucifer yellow was also measured for each monolayer after being subjected to the test articles to assess the integrity of the cell monolayers. All samples were assayed by HPLC/MS/MS using electrospray ionization. Accompanying each set of test compounds, atenolol, pindolol, propranolol, and digoxin were also run as positive controls.
The apparent permeability, Papp, and percent recovery were calculated according to eqs. 3 and 4, where dCr/dt is the slope of the cumulative concentration in the receiver compartment versus time in micromole second–1, Vr is the volume of the receiver compartment in cm3, Vd is the volume of the donor compartment in cm3, A is the area of the cell monolayer (1.13 cm2 for 12-well Transwell), C0 is the concentration of the dosing solution in micromoles, Cfinalr is the cumulative receiver concentration in micromoles at the end of the incubation period, and Cfinald is the concentration of the donor in micromoles at the end of the incubation period.
Determination of Steady-State CNS Penetration in mdr1a/b(+/+) and mdr1a/b(–/–) Mice. Each mouse (at least 6 weeks old at the time of surgery) was cannulated via the jugular vein (for drug administration) while under isoflurane anesthesia, with the cannulae exteriorized at the back of the neck. Animals were placed in a jacket and tether counter-balance system and allowed to recover for at least 3 days before dosing. Three male control mdr1a/b(+/+) and three male mdr1a/b(–/–) knockout mice were dosed with each compound.
The dose solution for each test compound was prepared on the day of dosing using an appropriate formulation for intravenous administration. The infusion rate and infusion time required to reach steady-state blood concentrations (taken as at least 3.5 half-lives) were determined separately for each compound and were based on prior studies to establish the intravenous pharmacokinetics. Blood samples (25 μl) were taken from the tail vein at 0.5-h intervals during the last 2 h of the infusion to confirm that steady-state blood concentrations had been achieved. At the end of the infusion period, the mice were decapitated and the brains were removed and homogenized with an equal weight of purified water (ELGA Maxima; ELGA LabWater, High Wycombe, UK). Triplicate-weighed aliquots (50 μl) of brain homogenate were stored at ca. –80°C before HPLC/MS/MS analysis. Samples were extracted by a method based on protein precipitation using acetonitrile (250 μl containing a suitable internal standard). Brain concentrations were corrected for residual blood volume using 15 μl/g of brain tissue as the vascular space (Brown et al., 1986).
Determination of CNS Penetration in the Rat at Steady State or after Acute Dosing. The CNS penetration for most compounds (47 of 56) was assessed under steady-state conditions after intravenous administration. The time required reach to steady blood concentrations (taken as at least 3.5 half-lives) was determined from prior studies of the intravenous pharmacokinetics of individual compounds. In some cases, blood and brain concentrations were obtained from acute single time point studies, where sampling was obtained at Cmax.
For steady-state infusion studies, each rat was cannulated via the jugular vein (for drug administration) and femoral vein (for blood sampling) while under isoflurane anesthesia, with the cannulae being exteriorized at the back of the neck. Animals were placed in a jacket and tether and allowed to recover for at least 2 days before dosing.
The dose solution for each test compound was prepared on the day of dosing using an appropriate formulation for intravenous administration. The infusion rate and infusion time were determined separately for each compound. Blood samples were taken at 0.5-h intervals during the last 2 h of the infusion to confirm that steady-state blood concentrations had been achieved. At the end of the infusion period, the rats were exsanguinated and decapitated and the brains were removed. Blood samples (ca. 80 μl) were collected into tubes containing K2EDTA as anticoagulant and aliquoted immediately after collection. An aliquot (50 μl) of each blood sample was diluted with an equal volume of purified water and stored at ca. –80°C to await analysis. Each brain was homogenized with an equal weight of purified water (ELGA Maxima). Weighed aliquots (50 μl) of brain homogenate were taken and stored at ca. 80°C before HPLC/MS/MS analysis. Samples were extracted by a method based on protein precipitation using acetonitrile (250 μl containing a suitable internal standard). Brain concentrations were corrected for residual blood volume using 15 μl/g as the vascular space (Brown et al., 1986).
Bioanalysis of Blood and Brain Homogenate Extracts. In all cases, the sample extraction of blood and brain homogenate was performed by a method based on protein precipitation, using acetonitrile (250 μl) containing a structural analog of the analyte of interest. Extracts were vortex mixed for 20 min followed by centrifugation at 2465g for 20 min (Eppendorf 5810R). Blood and brain extracts were analyzed by means of HPLC/MS/MS, either PE-Sciex API-4000 or API-3000 tandem quadrupole mass spectrometers, employing a Turbo V Ionspray. For API-3000 methods, the interface temperature was set to 450°C, with an ionspray voltage of 1500 to 2000 V. For API-4000 methods, the interface temperature operated at 700°C, with an ionspray voltage of 1500 to 2000 V. Samples (3–10 μl) were injected using a CTC Analytics HTS Pal autosampler and chromatographed by means of an HP1100 binary HPLC system. Specific chromatographic conditions were developed for each analyte to complement their physicochemical properties. The following conditions were common to all assays: column temperature of 40°C, eluent flow rate of 1 ml/min with 250 μl/min of eluent flow directed to MS interface, and assay run times of 2.5 to 3 min/samples. Calibration curves were constructed in blood and brain homogenate to cover at least three orders of magnitude (i.e., 5–5000 ng/ml). Three calibration points per decade were included together with quality controls covering the concentration range of calibration.
Results
Table 1 details the compound set examined in the present analysis, which was selected to include both marketed and proprietary compounds covering a wide range of physicochemical properties and therapeutic targets within the neuroscience portfolio. The extent of CNS penetration varies across this compound set from those with poor brain uptake (Br:Bl ratio <0.05:1) to those showing good distribution into the brain (Br:Bl ratio ca. 4:1), after correction for the vascular volume. Lipophilicity ranges across 10 orders of magnitude (cLogP –1.9 to 7.9) with a molecular mass range spanning 150 to 550 Da, including neutral, acidic, and basic compounds as well as substrates for Pgp (comprising 12 compounds spanning six different therapeutic targets).
Results from Equilibrium Dialysis and Nonspecific Binding Measurements. The influence of free fraction on in vivo CNS penetration in the rat is highlighted in Fig. 1, A and B. As shown in Fig. 1A, the free drug fraction in blood exhibits a weak correlation with in vivo Br:Bl ratio (R2 = 0.05), and notably, the CNS penetration of highly protein bound drugs [i.e., fu(blood) < 0.1%] is not restricted by the fraction of unbound drug in the systemic circulation. In contrast, the correlation is improved by taking account of the free fractions in both blood and brain, i.e., Kbb (Fig. 1B; R2 = 0.54 for non-Pgp substrates). Analogous to the findings of Kalvass and Maurer (2002), the relative tissue affinities of blood and brain play an important role in determining the extent of CNS observed in vivo. As such, compounds with free fractions in blood as low as those of GSK-18 and GSK-33 (0.009 and 0.07% free, respectively) may still exhibit good CNS penetration due to the fact that the drug affinity for brain tissue is higher still (0.006 and 0.03% free, respectively).
Figure 2 shows a plot of fu(brain) versus fu(blood) highlighting three data series, namely Pgp substrates, acids (calculated pKa <7.4), and nonacids. Eleven of the 12 Pgp substrates are nonacids, and the relationship between fu(brain) and fu(blood) is similar to the other nonacids depicted in Fig. 2 (combined least-square linear relationship, y = 1.0685x + 0.0403, R2 = 0.8161). For the acidic compounds, however, some separation is noted whereby acidic compounds tend to exhibit a lower free fraction in blood relative to brain (least-square linear relationship, y = 0.0666x + 0.0011, R2 = 0.6664). Likewise, the acidic Pgp substrate GSK-11 (ApKa 4.3) shows comparable behavior to the other acidic compounds within the set. Consistent with the observation that the volume of distribution for acids is lower than that of nonacids (Benet et al. (1996), these results are consistent with the fact that acidic compounds exhibit a high affinity for blood proteins, such as albumin.
Of the nonacid compounds, five seem to exhibit fu(blood) values that more closely resemble those of the acid series. Two of these compounds are characterized by ApKa slightly greater than 7.4 and therefore would still be partially ionized in the acid form at physiological pH (i.e., valdecoxib (ApKa 8.8) and celecoxib (ApKa 9.5). The remaining three outliers are GSK-37, GSK-38, and GSK-39, where their respective fu(blood) values (0.0004 to 0.0011) also lead to unexpectedly low Br:Bl ratios in the rat. These compounds will be discussed further with respect to differences in CNS penetration across species.
Correlation between Efflux Ratio and in Vitro Brain: Blood Partition Ratio. For the Pgp substrates present in the compound set, Br:Bl ratios ranged from 0.03:1 to 4:1, indicating that, although this transporter limits CNS penetration, it does not always preclude it entirely. Table 2 shows that the efflux ratios were found to range from ca. 7 to 125, with apparent membrane permeabilities ranging from 2 to 115 nm/s (i.e., in the absence of Pgp inhibitor GF120918A) and passive membrane permeabilities from 15 and 350 nm/s (i.e., in the presence of the inhibitor). For seven of the 12 Pgp substrates shown in Fig. 1B, Kbb tends to overpredict the Br:Bl ratio observed in vivo. This is consistent with the fact that the determination of Kbb assumes that the unbound concentrations in the blood and brain compartments are the same at equilibrium, which would only be valid for compounds where the BBB does not influence drug uptake.
Table 2 also shows that the efflux ratio does not seem to be strong a predictor of in vivo CNS penetration. Two compounds (GSK-2 and GSK-3) have Papp values lower than 10 nm/s and efflux ratios of ca. 70, which is generally considered to be an indicator for poor CNS penetration. However, both compounds exhibited Br:Bl ratios of ca. 0.5:1. Furthermore, several other compounds identified as having high-efflux ratios were also found to have an in vivo Br:Bl ratio greater than 1.5:1, albeit with higher passive membrane permeabilities. Hence, although there are certainly many good examples where low-passive permeability and high-efflux ratio translate into poor CNS penetration, there are examples where these parameters do not adequately describe the extent of CNS penetration observed in vivo, indicating a role for other parameters in determining the CNS penetration of the particular compounds.
Figure 3 shows a graphical plot of the relationship between in vivo Br:Bl ratio and the combination of two in vitro parameters, namely the Kbb and the ER. A good linear correlation is observed for the 12 Pgp substrates (R2 = 0.93), with a slope of 10.71 and a small y-intercept (–0.19). Despite the efflux ratio being derived from an MDCK cell line expressing human Pgp, Fig. 3 shows that there is acceptable agreement with the in vivo rat data presented. The good correlation (R2 = 0.93) also indicates that two in vitro parameters may be used to adequately describe the extent of in vivo CNS penetration of these Pgp substrates; i.e., that the Kbb estimates the extent of partitioning between blood and brain tissue, whereas the efflux ratio acts as a correction for the blood-brain barrier component of Pgp efflux. The non-zero intercept may be a result of the fact that in vivo rat Br:Bl data are related to human Pgp efflux; Ohe et al. (2003) observed a similar y-intercept discrepancy, which improved on the use a cell line expressing rodent Pgp (mdr1a).
The correlation between Br:Bl ratio and Kbb/ER suggests that an improved in vitro estimate of in vivo CNS penetration may be derived by means of the linear equation shown in Fig. 3, in which both partitioning and efflux ratio are accounted for as shown in eq. 5.
Fig. 4 depicts a plot of rat Kbb values against in vivo Br:Bl ratio, where the Kbb for the Pgp substrates have been corrected for their efflux ratios. A comparison of Figs. 1B and 4 highlights the point that combining the information on both the efflux ratio and Kbb markedly improves the in vitro prediction of CNS penetration for Pgp substrates. This is true not only for the substrates where Pgp is seen to exert a large impact on the CNS penetration (i.e., GSK-4, GSK-5) but also for substrates where CNS penetration was not markedly affected by Pgp (i.e., GSK-1, GSK-9). Linear regression shows that the line of best fit for Kbb and Br:Bl ratio yields a slope less than unity (ca. 0.94) and a negative y-intercept of –0.13 (R2 = 0.75).
Correlations of Efflux Ratio with CNS Penetration in the mdr1a/b(+/+) and mdr1a/b(–/–) Mouse Model. The well accepted gold standard method for determining the Pgp status of a drug is the -fold difference in Br:Bl ratio between wild-type mice and their knockout counterparts, where Pgp is not present. The physiological relevance of the in vitro measurement of efflux ratio is further exemplified, comparing the in vivo CNS penetration between mdr1a/b(+/+) mice with mdr1a/b(–/–) knockouts, as shown in Table 3. The extent of CNS penetration in for mdr1a/b(+/+) mice may be determined from mdr1a/b(–/–) corrected by the in vitro efflux ratio. Figure 5 shows a plot of the Br:Bl ratio from mdr1a/b(–/–) mice corrected for efflux ratio against the Br:Bl ratio from mdr1a/b(+/+) mice, where a strong correlation (R2 = 0.96) is observed in analogy to the in vitro relationship depicted in Fig. 3. Hence, although the efflux ratio is derived from a relatively simple model of the blood-brain barrier, the present data suggest that the value can be used quantitatively to account for the CNS penetration observed in vivo relative to either Pgp knockout models or to in vitro measurements of brain:blood partitioning.
Species Differences Rationalized by in Vitro Brain: Blood Partitioning. As discussed above, although Pgp efflux is a major factor limiting CNS penetration in vivo, the physicochemical properties of a molecule may also limit brain uptake. In a recent drug discovery project, a species mismatch appeared for one compound class whereby appropriate efficacy was demonstrated in the guinea pig model while proving ineffective in the rat version of the same pharmacodynamics model. Prior knowledge of the cloned receptors suggested that this was not a consequence of target differences, because previous compounds had elicited pharmacological responses in both species. Analysis of blood and brain samples from these studies showed that there were significant differences between the Br:Bl ratios in the two species. Good CNS penetration was observed in the guinea pig and accompanied the efficacy observed in this species (Table 4), whereas poor CNS penetration in the rat was correlated with no efficacy in the rat model. One possible explanation highlighted for this mismatch was a species difference in active transporters, although in vitro screening (MDR1-MDCK assay) indicated that Pgp was unlikely to be the cause. Equilibrium dialysis provided a consistent explanation for this difference between the two species.
Although the unbound fractions in brain tissue are comparable across the two species, the unbound fractions in blood are significantly different (Table 4). Poor CNS distribution in the rat could be as a result of a markedly higher degree of drug binding to blood constituents leading to a lower free fraction. It must be noted, however, that this restriction in CNS penetration is not solely the result of high-affinity protein binding in blood; rather it is the result of the fact that the nonspecific binding in blood greatly exceeds that in brain tissue. This difference between guinea pig blood and rat blood leads to a 20- to 30-fold increase in the extent of protein binding in rat blood, thereby restricting the CNS penetration markedly in the rat.
Discussion
The results presented in Figs. 3, 4, 5 indicate that in vivo CNS penetration in the rat may be described adequately (R2 = 0.75) by the use of two in vitro parameters, Kbb accounting for the ability of the compound to distribute into brain tissue and the efflux ratio accounting for the influence of the BBB. Thus, in this current model, 75% of the variance is accounted for by incorporation of Kbb and ER (Fig. 4), but evidently other factors play a role. The term Kbb/ER is further expanded in eq. 6.
The MDCK cell system is polarized, such that Pgp acts to preferentially displace drug into the apical basin of the assay. Thus, Papp(A–B) represents transport of drug from the blood side of the BBB to the brain, whereas Papp(B–A) represents the converse transport from the brain back to the blood. Because the efflux experiment is performed by incubation of the drug in buffer containing no protein, the Papp values are analogous to the permeability results where fu = 1 (i.e., 100% unbound). Thus, the term Kbb/ER may be interpreted as a correction of the A-to-B and B-to-A transport by their respective free concentrations in blood and brain on either side of the MDCK monolayer. Related findings in the rat were reported recently by Ohe et al. (2003) comparing CSF-plasma ratio, plasma free fraction, and Pgp efflux ratio. A strong linear correlation was observed, highlighting that CSF seems to be a useful surrogate for the free brain concentration. In addition, the work of Ohe et al. (2003) suggests that the CSF concentration could be estimated from knowledge of the efflux ratio and fu(plasma). Likewise, the in vivo results presented in Fig. 5 suggest that once the relationship among wild type, knockout mouse, and efflux ratio has been established, the extent to which Pgp modulates in vivo CNS penetration could be estimated without the need for the knockout mouse leg of the experiment.
Optimal CNS penetration seems to be the product of the correct balance of permeability, a low propensity for active efflux, and the appropriate physicochemical properties to allow partitioning into brain tissue. Based on the present data, it seems that these factors are inter-related, such that poorly permeable compounds that are also Pgp substrates may still enter the CNS if brain tissue binding provides a sufficient enough driving force to overcome the effects of the efflux transporter. Supporting this hypothesis is a recent report by Liu et al. (2005), which modeled the time required for a drug to reach equilibrium brain concentrations. The analysis showed that uptake into the brain is a hybrid function of both permeability and fu(brain), such that poorly permeable molecules may still exhibit CNS, albeit requiring significantly longer for the compound to approach steady-state concentrations.
In terms of compound selection commonly operating in drug discovery, these results raise several questions on how one might apply the current permeability assays in routine screening. In a paradigm where MDR-MDCK data are generated to assess Pgp liability, would a compound be progressed into further studies if the efflux ratio were observed to be in excess of 70 and with correspondingly low-passive permeability? Of equal note, if in vivo screening is employed as primary filter, would a compound showing a brain:blood ratio of 4:1 be checked for Pgp liability before passing onto efficacy models? Traditionally, the answer to both questions would be negative. Therefore, employing in vitro Pgp results as a hard filter may preclude the progression of potentially CNS penetrant efficacious compounds, which may account for the fact that there are so few examples currently noted of CNS penetrant Pgp substrates. Current focus should be on the use of efficacy models to aid compound selection as early as feasibly possible, in concert with free drug exposure measurements in blood and brain.
Further questions are raised by Fig. 6, which highlights the trend between cLogP and log(fu(brain)). As lipophilicity rises, there is a general decline in the fraction of unbound drug in brain tissue, with a potential consequence that, although CNS penetration may be improved by increasing lipophilicity, much of the compound will be nonspecifically bound to brain tissue. Thus, the assumption that increasing lipophilicity will lead to increased efficacy for CNS compounds cannot be guaranteed and will be a balance between adequate penetration and adequate free drug in brain. A similar trend is observed between cLogP and log(fu(blood)) (data not shown). In situ brain perfusion studies have demonstrated that BBB permeability tends to increase with lipophilicity (Smith, 2003), and this observation may be related to the increased propensity for binding to brain tissue, which provides the driving force for drug uptake.
Taken as a whole, the results reported here support the hypothesis that BBB transporters, permeability, and free fraction in blood and brain all play a role in determining the degree of CNS penetration of a drug observed in vivo. Partitioning into brain tissue may act as a driving force to overcome at least some of the effects of efflux transporters at the blood-brain barrier by supplying a sufficiently large concentration gradient to promote CNS penetration. Thus, it cannot be assumed that because a compound is a substrate for Pgp, it cannot gain access to the brain. Of the same note, the use of equilibrium dialysis allows for a high-capacity means of assessing free drug fractions in blood and brain tissue and the opportunity to investigate species differences in CNS penetration. Further work is required to understand the role of other drug transporters in CNS disposition (i.e., breast cancer-resistant protein) and also in the applicability of estimating fu(brain) to understanding drug efficacy at the level of the brain, where local free concentrations may differ from that in the bulk milieu.
Acknowledgments
We thank David Begley for assistance and advice during preparation of this manuscript. In addition, the following people are acknowledged for contributions: Ann Lewis, Ian Owens, Jonathan Barford, Paul Goldsmith, Lee Abberley, Beverly Smith, Caroline Peet, Michael Briggs, Colin Howes, Lee Skinner, Teresa Heslop, Ann Metcalf, Joanne Schogger, Trevor White, Nigel Deeks, Tania Buck, and Evangelia Grigorakou.
Footnotes
- Received July 27, 2005.
- Accepted December 1, 2005.
-
doi:10.1124/jpet.105.092916.
-
ABBREVIATIONS: DMPK, drug metabolism and pharmacokinetics; CNS, central nervous system; BBB, blood-brain barrier; Br:Bl, brain/blood concentration ratio; Pgp, P-glycoprotein; MDCK, Madin-Darby canine kidney; Kbb, in vitro brain:blood partition ratio; A–B, apical-to-basolateral; B–A, basal-to-apical; ER, efflux ratio; MDR, multidrug resistance; CSF, cerebrospinal fluid; PBS, phosphate-buffered saline; HPLC, high-performance liquid chromatography; MS/MS, tandem mass spectrometry; GF120918A, N-(4-{2-[6,7-bis(methyloxy)-3,4-dihydro-2(1H)-isoquinolinyl]ethyl}phenyl)-5-(methyloxy)-9-oxo-9,10-dihydro-4-acridinecarboxamide hydrochloride.
- The American Society for Pharmacology and Experimental Therapeutics