Abstract
This review discusses strategies to optimize brain penetration from the perspective of drug discovery and development. Brain penetration kinetics can be described by the extent and time to reach brain equilibrium. The extent is defined as the ratio of free brain concentration to free plasma concentration at steady state. For all central nervous system (CNS) drug discovery programs, optimization of the extent of brain penetration should focus on designing and selecting compounds having low efflux transport at the blood-brain barrier (BBB). The time to reach brain equilibrium is determined by both BBB permeability and brain tissue binding. Rapid brain penetration can be achieved by increasing passive permeability and reducing brain tissue binding. Although many drug transporters have been identified at the BBB, the available literature demonstrates only the in vivo functional importance of P-glycoprotein (P-gp) in limiting brain penetration of its substrates. Drug-drug interactions mediated by P-gp at the BBB are possible due to inhibition or induction of P-gp. For newly identified drug transporters at the BBB, more research is needed to reveal their in vivo significance. We propose the following strategies for addressing drug transporters at the BBB. 1) Drug discovery screens should be used to eliminate good P-gp substrates for CNS targets. Special consideration could be given to moderate P-gp substrates as potential CNS drugs based on a high unmet medical need and the presence of a large safety margin. 2) Selection of P-gp substrates as drug candidates for non-CNS targets can reduce their CNS-mediated side effects.
Brain is separated from the systemic circulation by two barriers: the blood-brain barrier (BBB) and the blood-cerebrospinal-fluid barrier (BCSFB). The BBB is composed of cerebral endothelial cells that differ from those in the rest of the body by the presence of extensive tight junctions, absence of fenestrations, and sparse pinocytotic vesicular transport. The BCSFB is formed by a continuous layer of polarized epithelial cells that line the choroid plexus. The BBB and BCSFB exhibit very low paracellular permeability and express multiple drug transporters. These characteristics restrict the entry of hydrophilic compounds or efflux transport substrates into brain (Davson and Segal, 1995). In this review, we will summarize recent published data relevant to assess drug brain penetration and present the authors' opinions on how to effectively address BBB issues in drug discovery and development.
What Parameters Should Be Used to Assess Brain Penetration?
In drug discovery, it is obvious that one should select compounds with “good” brain penetration as CNS drugs, but it is not so obvious what parameters should be used to define “good” brain penetration. Two parameters, the ratio of brain and plasma concentration (Kp) and BBB permeability (quantified as the permeability surface area product, PS), have been used to describe brain penetration. Kp has been the most widely used parameter to evaluate and optimize brain penetration in drug discovery, but its relevance has been questioned (Pardridge, 2004). Analogous to the concept of the extent and rate of oral absorption, brain penetration can be assessed with two parameters, the extent and the time to reach brain equilibrium (Liu and Chen, 2005). The extent can be defined as the ratio of free brain and free plasma concentration at equilibrium, Kp, free, also known as Kp, uu (Syvanen et al., 2006). The time to reach brain equilibrium can be defined as the half-life to reach the equilibrium of brain and plasma concentration.
What Determines the Extent of Brain Penetration?
According to a three-compartment model (Fig. 1), the following equation (eq. 1) can be derived at steady state. Cluptake and Clefflux are the active uptake and efflux transport clearance at the BBB, respectively. Clbulk is the clearance due to brain interstitial fluid bulk flow, and Clmetabolism is the brain metabolic clearance. According to eq. 1, to augment the extent of brain penetration, one needs to either increase PS and Cluptake or reduce Clefflux, Clbulk, and Clmetabolism.
It is desirable to design a compound as a substrate of brain uptake transporters to enhance Cluptake. For example, large neutral amino acid transporter 1 transports L-DOPA and gabapentin across the BBB. Although L-DOPA has been available for more than 30 years, the same success in increasing brain penetration of other drugs has rarely been replicated, with the exception of its close-in analogs. Effective in vitro approaches have yet to be developed to screen brain uptake transporter substrates to deliver drugs through the endogenous transporters at the BBB. It would be more feasible to design lipophilic compounds (high PS values) without significant efflux transport (low Clefflux) than to design compounds as uptake transporter substrates.
Clbulk can play an important role in decreasing Kp, free for low permeability compounds. It has been estimated that bulk flow clearance spans the range of 0.2 to 0.3 μl/min/g (Cserr and Patlak, 1993). Take the example of mannitol, a compound of low permeability with a PS value of less than 1 μl/min/g. Bulk flow becomes significant compared to its permeability, resulting in a low Kp, free (0.01). For a compound with moderate to high permeability, Clbulk is insignificant. This is illustrated by caffeine, a compound of moderate to high permeability with a PS value of 13 μl/min/g. In this case, bulk flow clearance is much lower than the permeability and has an insignificant effect on Kp, free (1.0) (Hansen et al., 2002).
Brain metabolism, Clmetabolism, could also play a significant role in reducing Kp, free. Metabolizing enzymes, such as monoamine oxidase, flavin-containing monooxygenase, cytochrome P450, and glucuronosyltransferases have been identified in brain endothelial cells and brain tissue (el-Bacha and Minn, 1999; Fang, 2000; Gervasini et al., 2004; Strazielle et al., 2004). Hence, the stability of a compound in brain tissue needs to be examined in early drug discovery.
How to Use Kp to Assess Brain Penetration?
Kp is the most commonly used parameter to evaluate brain penetration and has been used as the primary parameter to optimize brain drug delivery in CNS drug discovery. In a recent study, Kp was determined for a set of the 32 most prescribed CNS drugs and ranged from 0.1 to 24 in mice (Fig. 2A). Thus, a compound having a Kp value as low as 0.1, such as sulpiride, can still be a successful CNS drug, suggesting that it is difficult to assess brain penetration based upon Kp alone (Doran et al., 2005).
According to the definition of Kp (Kp = Cbrain/Cplasma), fu, plasma (fu, plasma = Cu, plasma/Cplasma), fu, brain (fu, brain = Cu, brain/Cbrain), and Kp, free (Kp, free = Cu, brain/Cu, plasma), the relationship between Kp and Kp, free can be derived as follows: where Cplasma and Cu, plasma are the total and unbound plasma concentration, respectively; Cbrain and Cu, brain are the total and unbound brain concentration, respectively; and fu, plasma and fu, brain are the plasma and brain unbound fraction, respectively. From eq. 2, it is evident that a low Kp can be due to high nonspecific binding in plasma, low binding in brain, or low Kp, free. There was a 240-fold difference for Kp among the 32 most prescribed CNS drugs in mice but only a 34-fold difference for Kp, free (Fig. 2, A and B), indicating that nonspecific binding (e.g., fu, plasma/fu, brain) is a significant component for Kp. Gupta et al. (2006) reported that Kp for S- and R-cetirizine is 0.22 and 0.04, respectively. These Kp values apparently indicate that S-cetirizine penetrates brain tissue better than R-cetirizine. However, the Kp, free is 0.17 and 0.14 for S- and R-cetirizine, respectively, indicating that there is no stereoselective brain penetration. Further investigation revealed that the protein binding for these enantiomers was different. The fraction unbound for S- and R-cetirizine is 0.5 and 0.15, respectively. The stereoselective Kp is caused by differential binding to plasma proteins rather than transport at the BBB. Thus, when Kp is used in drug discovery to optimize brain penetration, it is very important to understand the impact of the binding in plasma and brain.
There are several methods to estimate Kp, free. Brain microdialysis is a direct approach to determine free brain concentration. However, the utility of microdialysis in the drug discovery setting is limited because it requires extensive resources and is not easily applied to highly lipophilic compounds. According to eq. 2, Kp, free can be calculated from Kp, fu, plasma, and fu, brain, where Kp can be determined from in vivo studies and fu, plasma and fu, brain can be estimated using in vitro equilibrium dialysis approaches with plasma and brain tissue homogenate, respectively (Maurer et al., 2005). Recently, we examined the correlation between the Kp, free determined using microdialysis and the Kp, free estimated from eq. 2, and a good correlation was observed for four model compounds (Liu et al., 2007). A potential caveat in using brain tissue homogenate to estimate fu, brain is that homogenization may change binding properties by unmasking binding sites that are not accessible to a drug in vivo. These concerns may be addressed by using a brain slice approach in which the brain structure remains intact (Becker and Liu, 2006; Fridén et al., 2007).
Kp, free may also be estimated from studies utilizing genetically modified and wild-type (WT) animals, such as Mdr1a/1b knockout (KO) and WT mice. Under the assumption that plasma and brain binding between the two strains of animals are identical and the only difference is a specific drug transporter, Kp, free can be estimated from the ratio of Kp, WT/Kp, KO (eqs. 1 and 2). For 25 of a set of 31 compounds, the Kp, free values estimated from eq. 2 are within 3-fold of Kp, WT/Kp, KO (Fig. 3). As expected from eqs. 1 and 2, for P-gp substrates, such as risperidone and metoclopramide, their Kp, free values in WT mice, which are much lower than unity, are within 3-fold of Kp, WT/Kp, KO, but their Kp, free in P-gp KO mice are much greater than Kp, WT/Kp, KO.
Kp, free also may be estimated from CSF drug concentration, assuming CSF drug concentration represents the free brain concentration. Shen et al. (2004) observed that for moderate to high-permeability compounds, CSF concentration approximates free brain concentration but, for low permeability compounds, CSF concentration may not represent free brain concentration. Our results indicate that CSF is better than plasma free concentration to predict brain free concentration (Liu et al., 2006). Although the cellular location of P-gp suggests that it pumps its substrates from plasma into CSF (Rao et al., 1999), in vivo data do not support this view. The free brain/CSF concentration ratios of three typical P-gp substrates, loperamide, verapamil, and quinidine, in the P-gp KO and competent mice were 1.5, 1.9, and 3.6, which are much less than Kp, KO/Kp, WT ratios of 9.3, 17, and 36, respectively (Doran et al., 2005). However, other transporters, such as Mrp1, do not play a significant role at the BBB but do at the BCSFB (Wijnholds et al., 2000). Therefore, in drug discovery settings where in vitro or in silico data demonstrate that compounds are highly permeable and are not substrates for efflux transporters, plasma free or CSF concentrations provide a simple way to estimate Kp, free.
What Determines the Time to Reach Brain Equilibrium?
An empirical approach to identify compounds with quick brain penetration is to screen compounds with high BBB permeability. To understand the theoretical basis of this practice, we developed a concept of intrinsic brain equilibrium half-life (t1/2eq, in) to quantitate how quickly a compound can enter into the brain (Liu et al., 2005). t1/2eq, in is defined as the time required for free brain concentration to reach 50% of free plasma concentration (eq. 3). where Vb represents the physiological volume of brain tissue. This equation demonstrates that a combination of BBB permeability and brain tissue binding determines the time to reach brain equilibrium. This theoretical analysis was supported by experimental observations. Theobromine has a low to moderate PS (23 ml/h/kg) and a high fu, brain (0.61), resulting in a PS × fu, brain of 14 ml/h/kg. In contrast, fluoxetine has a high PS (619 ml/h/kg) and a low fu, brain (0.00094), resulting in a PS × fu, brain of 0.6 ml/h/kg. Consistent with a higher PS × fu, brain product, the observed t1/2eq, in for theobromine (∼0.1 h) was shorter than that of fluoxetine (∼1 h) (Liu et al., 2005). Similar conclusions were made by Syvanen et al. (2006).
It has been postulated that BBB permeability and brain tissue binding correlate (Liu et al., 2005). For example, more lipophilic compounds tend to have higher BBB permeability and bind to brain tissue more extensively, whereas hydrophilic compounds have lower permeability and less extensive tissue binding. For many CNS drug-like molecules, plasma concentrations can quickly equilibrate with brain concentrations, even though their BBB permeability can vary substantially. In a study reported by Liu et al. (2005), the brain concentration of six of seven model compounds equilibrated with plasma concentration within 2 h after subcutaneous dose. Similar conclusions were drawn from a brain microdialysis study where compounds having much different BBB permeability were able to reach brain equilibrium quickly (Hammarlund-Udenaes et al., 1997). Therefore, a lead compound should not be eliminated as a candidate compound only because it shows low BBB permeability.
What Are the Main Drug Transporters at the BBB?
P-glycoprotein (P-gp, gene symbol Abcb1), multidrug resistance-associated proteins (Mrp, Abcc), breast cancer resistance protein (Bcrp, Abcg2), several organic anion transport polypeptides (Oatp, Slco), as well as one organic anion transporter (Oat3, Slc22a8) have been identified at the BBB and/or the BCSFB (Schinkel and Jonker, 2003; Kusuhara and Sugiyama, 2005). Recently, Yousif et al. (2007) examined the gene profile and expression for Mdr1a, Mdr1b, Bcrp, Mrp1–5, and Oatp1a4 (Oatp2) in rat brain and found that only Mdr1a, Bcrp, Mrp4, and Oatp1a4 gene profiles were similar to those of endothelium markers, indicating their presence at the BBB. Although significant progress has been made in identifying drug transporters at the BBB and the BCSFB and various in vitro assays have been developed to screen the substrates of the transporters, the in vivo significance of many of the drug transporters remains to be evaluated.
What Are the Potential Risks and Benefits to Develop P-gp Substrates/Inhibitors as Drug Candidates ?
P-gp is considered the most important efflux drug transporter at the BBB. The risk to develop a P-gp substrate as a CNS drug is the reduction of its therapeutic window. The reduced therapeutic window occurs because higher plasma free concentrations are necessary to compensate for the ef-flux transport and drive the brain free concentrations to the desired level. In this situation, the higher plasma concentrations could increase the risk of peripheral toxicities. Therefore, an ideal CNS drug should not be a good substrate for P-gp efflux. However, scarce published data are available to define what P-gp-mediated transport is acceptable for drug candidates. In a comprehensive study, P-gp transport of the 32 most prescribed CNS drugs were examined in mdr1a/1b KO and WT mice. It was found that 22% of the compounds showed an efflux ratio of unity, 72% between 1 and 3 and 6% between 3 and 10 (Fig. 4) (Doran et al., 2005). These results indicate that the majority (92%) of CNS drugs tested showed no to weak P-gp-mediated transport. These data support the conclusion that one should avoid developing good P-gp substrates as CNS drugs. On the other hand, P-gp-mediated drug transport per se would not be the sole reason to terminate candidate development if there is a large projected therapeutic window in humans.
There are several examples to indicate the benefits of developing P-gp substrates as peripheral targeted drugs to reduce CNS side effects. First generation H1 antagonists, such as diphenhydramine, triprolidine, and hydroxyzine, produce histamine blockade at H1 receptors in the CNS and frequently cause somnolence or other CNS adverse effects. However, the second generation H1 antagonists, such as cetirizine, loratadine, fexofenadine, and desloratadine, produce relatively little somnolence or other CNS side effects at recommended doses. Chen et al. (2003) demonstrated that the first generation of H1 antagonists are non-P-gp substrates and that the second generation of H1 antagonists are P-gp substrates. This results in high brain concentrations for the first generation but low brain concentration of the second generation of H1 antagonists. Similar observations were made by several other groups (Polli et al., 2003).
What Is the Possibility for Drug-Drug Interactions Mediated by P-gp at the BBB?
Several clinical and animal studies indicate the possibilities of drug-drug interactions (DDI) due to P-gp inhibition at the BBB. Loperamide is a potent μ-opiate agonist that reduces gut motility by its action at opiate receptors in the gut. It does not produce μ-opioid-mediated central nervous system effects at clinical doses, because loperamide is a substrate for P-gp efflux that limits its brain penetration. Coadministration of 600 mg of quinidine, a P-gp inhibitor, and 16 mg of loperamide in humans reduced the respiratory response to CO2 by 20% compared with loperamide alone (Sadeque et al., 2000). The direct evidence for clinical DDI mediated by P-gp at the BBB in humans was from a cyclosporine A and verapamil interaction study (Sasongko et al., 2005). The Kp of verapamil increased 9.5-fold in P-gp KO mice and 10.5-fold in rats pretreated with 50 mg/kg cyclosporine, suggesting a high DDI potential between verapamil and cyclosporine in humans (Hendrikse et al., 1998). In a clinical study, the human Kp of [11 C]verapamil was increased by 88% (p < 0.001) in the presence of cyclosporine without affecting [11C]verapamil metabolism or plasma protein binding (Sasongko et al., 2005). Although the magnitude of the change caused by cyclosporine A was modest, this work demonstrates that P-gp inhibition at the BBB exists in humans. Because cyclosporine A is the most potent P-gp inhibitor on the market, this study suggests that the risk of DDI due to P-gp inhibition is low unless the P-gp inhibition activity of the drug in question is much greater than cyclosporine A.
A recent study suggests that the possibility of DDI resulting from P-gp induction at the BBB by pregnane X receptor (PXR) ligands is possible. There have been multiple reports on the up-regulation of intestinal P-gp by PXR ligands rifampicin and hyperforin (a constituent from St. Johns Wort), which contributes to the decrease of plasma levels of several drugs, such as cyclosporine A and digoxin in humans when coadministered with the herbal supplement (Marchetti et al., 2007). Bauer et al. (2006) demonstrated that P-gp at the BBB can also be induced by rifampicin in transgenic mice expressing human PXR. P-gp expression levels at the BBB were significantly higher in the rifampicin-treated mice than in the vehicle-treated mice. In the same study, the CNS effects of methadone, a P-gp substrate, were substantially lower in the rifampicin-treated mice than in the vehicle-treated mice, suggesting that P-gp efflux activity at the BBB was significantly induced by rifampicin. These data indicate that DDI caused by P-gp induction, although it remains to be proven, is possible in humans.
Zhang et al. (2006a) presented the position of the United States Food and Drug Administration on transporter-mediated DDI focusing particularly on P-gp. Subsequently, the United States Food and Drug Administration developed guidelines for drug transporter-mediated DDI. Although the proposed guidelines represent a proactive approach to address the concerns of transport-mediated DDI, the effectiveness of the approach remains to be assessed when more data become available.
Is P-gp Activity Species-Dependent?
Literature views on species difference for P-gp activity are not consistent. One study showed a difference in P-gp ATPase binding affinity among rhesus monkey, dog, and human (Xia et al., 2006). Furthermore, the Km values of diltiazem exhibited approximately 16.5-fold differences among human, monkey, canine, rat, and mouse P-gp-transfected cell lines (Katoh et al., 2006). Yamazaki et al. (2001) reported different efflux ratios between mouse and human P-gp-transfected cells, suggesting species differences for P-gp activity. In contrast, a more recent study using a set of 3300 compounds demonstrated a 93% overlap between mouse and human P-gp-mediated transport (Feng et al., 2008). Thus, significant mouse-human differences in P-gp activity may be a rare phenomenon.
Suzuyama et al. (2007) showed that P-gp inhibition was species-dependent in in vitro systems. Using daunorubicin, digoxin, and cyclosporine A as P-gp substrates, we observed that the IC50 of quinidine and verapamil in human, monkey, canine, rat, and mouse P-gp-transfected cells could vary by 6-fold, although the majority of the IC50 values are within 2-fold. In contrast, the Kp of verapamil increased by 78% in rats and 75% in humans where plasma concentration of cyclosporine was 2.8 μM, indicating no species difference for verapamil-cyclosporine DDI (Hsiao et al., 2006).
What Is the in Vivo Significance of Bcrp (Abcg2), Mrps (Abcc), Oatps (Slco), and Oat (Slc22a) at the BBB?
The expression of Bcrp in mouse brain capillaries was demonstrated by quantitative polymerase chain reaction, Western blot, and immunohistochemical analysis (Lee et al., 2005). The high expression of Bcrp in brain microvessels indicates that Bcrp may play an important role at the BBB as an efflux pump. Several studies have examined the in vivo functions of Bcrp at the BBB in rodents (Table 1). The available data were unable to demonstrate the importance of Bcrp at the BBB. For example, in mdr1a KO mice, the brain uptake of mitoxantrone increased 3-fold in the presence of 2 μM elacridar, a P-gp and Bcrp dual inhibitor, suggesting that Bcrp functions as an efflux pump at the BBB (Cisternino et al., 2004). However, no enhancement of brain penetration was observed for mitoxantrone in Bcrp KO mice compared with the WT mice, indicating that Bcrp may not be an important efflux pump at the BBB (Lee et al., 2005). Another example is the brain penetration of a Bcrp substrate, imatinib (Breedveld et al., 2005). Its brain penetration was increased by 2.5-fold in Bcrp KO mice compared with the WT mice. However, the ratio increased 3.5-fold in P-gp KO compared with the WT mice, indicating that the efflux transport of imatinib is mainly mediated by P-gp.
Several Mrp/MRP proteins have been suggested to be important at the BBB according to their expression levels. The literature reports that the membrane localization and function of Mrps are still controversial (Dallas et al., 2006; Hoffmann et al., 2006). Available data show the functional importance of Mrps at the BCSFB but not at the BBB. Etoposide is an Mrp substrate as determined by in vitro assays. Its Kp was not altered in Mrp1/mdr1a/mdr1b triple KO mice, but its CSF concentration changed substantially (Wijnholds et al., 2000). Brain ECF-plasma concentration ratio of phenytoin, an Mrp substrate, increased by less than 2-fold in a probenecid inhibition study (Potschka et al., 2003). In Mrp2-transport-deficient rats, the brain ECF-plasma ratios of phenytoin were enhanced only 1.3-fold compared with those of the normal rats. Leggas et al. (2004) measured the plasma, brain, and CSF concentrations of topotecan in Mrp4 KO and WT mice post-i.v. dose. Significant increases in plasma, brain, and CSF concentrations were observed in the Mrp4 KO; however, plasma and brain concentration-time profiles were not provided. Thus, the changes observed in their investigation raise the interest in the function of Mrp4 at the BBB and BCSFB, but more studies are needed to confirm the function of Mrp4 at the BBB and BCSFB. Imaoka et al. (2007) found that the Kp values of Mrp4 substrates adefovir and tenofovir were not statistically different between Mrp4 KO and WT mice after 90 min of i.v. infusion. Therefore, the importance of Mrps at the BBB has yet to be demonstrated.
The expression and cellular locations of Oatp family (Oatp1a4, Oatp1a5, Oatp1c1, OATP1A2) and Oat (Oat3/OAT3) have been demonstrated recently at the BBB and the BCSFB (Kusuhara and Sugiyama, 2005). Studies are needed to demonstrate the in vivo importance of these transporters.
Considerations of the Methods Used to Study Brain Penetration
What Are the Key Issues for in Vivo Methods?
In addition to Kp and brain microdialysis methods as discussed under How to Use Kp to Assess Brain Penetration? , the in situ brain perfusion method has been developed to directly determine BBB permeability (Takasato et al., 1984). Dagenais et al. (2000) developed a mouse brain in situ perfusion method and demonstrated that brain uptake clearance was lower in P-gp-competent mice than in P-gp KO mice. To increase the throughput of the mouse brain uptake assay, a modified i.v. administration method has been developed (Raub, 2006). According to eqs. 1 and 3, BBB permeability alone cannot be used to predict either the extent of brain penetration or the time to reach equilibrium. Therefore, these data alone cannot be used to select drug candidates. For example, it has been observed that the permeability can differ by more than 100-fold among compounds that can penetrate into the brain (Liu et al., 2004). No clear association was observed between the permeability and efflux transport in a recent study (Summerfield et al., 2007). A rational way to use the permeability data is to assess whether a compound is an efflux or uptake transporter substrate. One can do this by comparing the observed BBB permeability data with the data predicted from an in silico model, as discussed under What in Silico BBB Models Need to be Developed?, or in the context of lipophilicity of a historical dataset as described by Raub (2006).
Measurements of receptor occupancy have become a common approach in drug discovery and development to assess brain penetration and interaction between compounds and their targets. In the preclinical setting, this type of study requires radiolabeled compounds. It has been demonstrated that liquid chromatography/mass spectrometry can be used to determine the brain concentration of the tracer ligand (Barth et al., 2006). Using liquid chromatography/mass spectrometry, the occupancy measurement for dopamine D2, 5-hydroxytryptamine2A, and NK-1 receptors has been demonstrated. This approach will allow one to screen the best ligands for occupancy studies in animals and humans. In clinical trials, human receptor occupancy studies can be conducted using imaging technologies, such as positron emission tomography or single-photon emission tomography. These imaging techniques can reveal important relationships between dose, exposure, and target occupancy in humans. It can also help to determine whether the drug safety profiles observed in Phase I trials warrant more expensive Phase II proof-of-concept trials and to select the optimal doses for these trials.
What Are the Challenges for in Vitro BBB Models?
Caco-2 and Madin-Darby canine kidney cells are commonly used to predict the BBB permeability in drug discovery. However, these cells are derived from colon or kidney epithelial cells, and their permeability and transporter characteristics are different from the brain endothelial cells. To overcome the limitations of these cells, extensive studies are being conducted to develop better in vitro BBB models using either primary brain endothelial cells or immortalized brain endothelial cells. One of the main limitations for these in vitro BBB models is the high paracellular permeability due to less than optimal function of the tight junctions. Coculture with astrocytes has been used to mimic the in vivo environment. A recent work (Zhang et al., 2006b) indicates a transepithelial electrical resistance value of 400 ohm · cm2 can be achieved in primary cultured porcine brain microvessel endothelial cells. To further mimic the shear stress caused by blood flow, a dynamic coculture model in a hollow-fiber cartridge has been developed (Santaguida et al., 2006). Because of a complicated set up and low throughput of this dynamic model, it will be difficult to be used in drug discovery in its current format. An ideal in vitro model should have similar paracellular permeability and transporter characteristics as the in vivo BBB and should be easily set up for routine drug screening. More research is needed to develop such an in vitro BBB model.
What in Silico BBB Models Need to be Developed?
Most BBB in silico models were developed to predict logKp or logBB. The main limitation is that Kp is not a good parameter for characterizing the brain penetration as discussed above. To address the limitation of the logKp model, Gratton et al. (1997) and Liu et al. (2004) developed a BBB permeability (logPS) model using the data generated with the in situ brain perfusion method. The logPS model may be used in conjunction with in vivo logPS. If the observed permeability is substantially lower or higher than the predicted value, it indicates that efflux or uptake transporters modulate the brain penetration for the tested compound. For example, the PS values of uptake transporter substrates phenylalanine and levodopa were underpredicted, and the PS values of P-gp substrates, digoxin, CP-141938, and quinidine, were overpredicted (Liu et al., 2004). Models that are developed to predict drug transport will be more useful because Kp, free is largely governed by drug transporters at the BBB.
Conclusions
The optimization of brain penetration in drug discovery needs to select compounds with high Kp, free by screening out very poorly permeable compounds and, more importantly, efflux transporter substrates. For those CNS drugs whose indications require a quick onset of action, short time to reach brain equilibrium is essential. This can be achieved by screening compounds with a combination of high permeability and low brain tissue binding because high permeability alone cannot ensure rapid brain penetration. Many drug transporters are expressed at the BBB; however, the available data demonstrate that only P-gp is important in limiting the brain penetration of its substrates in vivo. DDI mediated by P-gp at the BBB are possible due to inhibition or induction of P-gp. For newly identified drug transporters at the BBB, more research is needed to reveal their in vivo significance. We propose the following strategies for addressing drug transporters at the BBB. 1) Drug discovery screens should be used to eliminate good P-gp substrates for CNS targets. Special consideration could be given to moderate P-gp substrates as potential CNS drugs based on a high unmet medical need and the presence of a large safety margin. 2) Selection of P-gp substrates as drug candidates for non-CNS targets can reduce their CNS-mediated side effects.
Footnotes
-
Article, publication date, and citation information can be found at http://jpet.aspetjournals.org.
-
doi:10.1124/jpet.107.130294.
-
ABBREVIATIONS: BBB, blood-brain barrier; CNS, central nervous system; P-gp, P-glycoprotein; BCSFB, blood-cerebrospinal fluid barrier; CSF, cerebrospinal fluid; PS, BBB surface area permeability product; Kp, brain plasma concentration ratio; Kp, free and Kp, uu, brain plasma free concentration ratio; Cluptake, uptake transporter clearance at the BBB; Clefflux, efflux transporter clearance at the BBB; Clbulk, the clearance due to brain interstitial fluid bulk flow; Clmetabolism, brain metabolic clearance; fu, plasma, plasma unbound fraction; fu, brain, brain unbound fraction; Cplasma, plasma concentration; Cbrain, brain concentration; Cu, plasma, plasma unbound concentration; Cu, brain, brain unbound concentration; t1/2eq, in, intrinsic brain equilibrium half-life; DDI, drug-drug interaction; KO, knockout; WT, wild type; Mrp, multidrug resistance-associated protein; Bcrp, breast cancer resistance protein; Oatp, organic anion transport polypeptide; Oat, organic anion transporter; PXR, pregnane X receptor; CP-141938, N-(4-methoxy-3-[(2-phenyl-piperadin-3-amino)-methyl]-phenyl)-N-methyl-methane-sulfonamide.
- Received August 17, 2007.
- Accepted January 17, 2008.
- The American Society for Pharmacology and Experimental Therapeutics