Utility of unbound plasma drug levels and P-glycoprotein transport data in prediction of central nervous system exposure

Transcription

1 Xenobiotica, 2009; 39(09): RESEARCH ARTICLE Utility of unbound drug levels and P-glycoprotein transport data in prediction of central nervous system exposure H. He 1, K.A. Lyons 1, X. Shen 2, Z. Yao 2, K. Bleasby 1, G. Chan 1, M. Hafey 1, X. Li 1, S. Xu 1, G. M. Salituro 1, L. H. Cohen 1, and W. Tang 1 1 Department of Drug Metabolism and Pharmacokinetics and 2 Laboratory Animal Resources, Merck Research Laboratories, Rahway, NJ, USA Abstract 1. Drug concentrations in cerebrospinal fluid have been assumed to be a natural surrogate for total drug exposures in the central nervous system. The present communication reports a data set from a study of 30 compounds in mice. An attempt was made to correlate cerebrospinal fluid and unbound drug concentrations via incorporation of in vitro P-glycoprotein (Pgp)-mediated transport data. 2. Pgp-deficient (Pgp / ) and wild-type mice were dosed with compounds of interest by oral gavage (orally) at 5 mg kg 1. Plasma and cerebrospinal fluid samples were collected at 1 h post-dosing, and analysed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) for drug concentrations. Mouse and human Pgp-mediated transport were evaluated in vitro by a bi-directional (B to A and A to B) transport assay using LLC-PK1 cells expressing mouse (mdr1a) and human (MDR1) forms of Pgp, respectively. 3. Compounds with B to A/A to B transport ratios < 2 were defined as non-substrates of Pgp, whereas those exhibiting B to A/A to B transport ratios 2 were considered Pgp substrates. Plasma protein binding was also determined in vitro via equilibrium dialysis. Of the 30 compounds, 13 were identified to be mouse Pgp substrates, all of which were also human Pgp substrates, demonstrating a good agreement between mouse and human data. 4. In Pgp wild-type mice, the unbound and cerebrospinal fluid concentrations of the non-pgp substrates correlated well, with a regression slope of approximately 1.0. A similar relationship existed for Pgp substrates in Pgp / mice. On the other hand, an improved correlation of cerebrospinal fluid and systemic exposures of the Pgp substrates in Pgp wild-type mice was observed when the unbound concentrations were normalized to the corresponding B to A/A to B transport ratios. 5. These results reinforce the premise that a combined use of unbound drug concentrations and in vitro Pgp transport data may be of value for the estimation of central nervous system exposures. Keywords: Cerebrospinal fluid (CSF); liquid chromatography-tandem mass spectrometry (LC-MS/MS); P-glycoprotein transporter (Pgp); protein binding Introduction Drug metabolism and pharmacokinetics play an important role in drug-discovery programmes. Understanding compound properties such as absorption, distribution, metabolism, and excretion can be valuable to an initial estimation of efficacious dose ranges and safety margins. For drug-discovery programmes targeting the central nervous system (CNS), one has to consider the blood brain barrier (BBB) and the blood cerebrospinal barrier (BCSFB) that separate the systemic circulation and brain (De Lange 2004; Lin 2008). Lipophilic compounds, in their proteinunbound and un-ionized forms, may undergo passive transcellular diffusion from blood to the CNS, exhibiting good BBB permeability (Street et al. 1979; Summerfield et al. 2006). The exceptions to this trend are often attributed to either influx or efflux via active transport of the compounds (Summerfield et al. 2006). In this regard, the multidrug resistance transporter P-glycoprotein (Pgp) has been identified as the primary efflux transporter responsible Address for Correspondence: L. H. Cohen, Merck Research Laboratories, Mailstop RY800B201, Rahway, NJ , USA. lucinda_cohen@merck.com (Received 06 April 2009; revised 01 May 2009; accepted 04 May 2009) ISSN print/issn online 2009 Informa UK Ltd DOI: /

2 688 H. He et al. for low CNS drug exposures in a number of instances (Lin 2004). Evidence available currently suggests that both Pgp and permeability (lipophilicity) could play important roles in brain penetration of drugs (Lin 2008). Cerebrospinal fluid (CSF) is a constituent of the CNS, and drug concentrations in the CSF are often taken as an index for their exposures in the CNS (Ohe et al. 2003; Lin 2008). CSF is known to contain only trace levels of proteins; therefore, drug concentrations in the CSF are likely to reflect unbound drug concentrations at equilibrium if active transport is not involved. However, this relationship collapses when Pgp-mediated efflux exists, based on data from several studies exploring the dependence of brain-to- (or CSF-to-) drug partition ratios on unbound drug concentrations in brain and (Kalvass & Maurer 2002; Ohe et al. 2003; Maurer et al. 2005; Kalvass et al. 2007). The present paper reports a data set from a study of 30 compounds for their exposures in Pgp-deficient (Pgp / ) and wild-type mice. An attempt was made to correlate CSF and unbound drug concentrations via incorporation of Pgp-mediated transport estimated in vitro. The results suggest that a combined use of unbound concentration and Pgp transport data may be valuable in support of CNS targeting drug-discovery programmes. Materials and methods Chemicals Thirty compounds of interest were synthesized in the Department of Medicinal Chemistry, Merck Research Laboratories. Dulbecco s phosphate buffer saline (GIBCO, Grand Island, NY, USA). C57BL/6J mice frozen was obtained from Bioreclamation, Inc. (East Meadow, NY, USA). P-glycoprotein transport studies Bi-directional transport was measured across the LLC- PK1 cell monolayer and monolayer stably expressing human MDR1 (LLC-MDR1) or mouse Mdr1a (LLC- Mdr1a) Pgp (kindly provided by Netherlands Cancer Institute under a licensing agreement). Compounds were dosed at 5 µm (except for titration compounds, which were at 1.0 or 0.1 µm) to the apical compartment to determine transport from apical (A) to basolateral (B) and to the basolateral compartment to determine transport from B to A. Samples from both sides of monolayer were taken at 3 h. The P app B to A/A to B ratio gives an indication whether compounds are Pgp transporter substrates. The following equations were used to calculate the P app and B to A/A to B ratios: P app [(Volume of receptor chamber (ml))/(area of membrane (cm 2 ) 3 initial concentration (µm)] 3 [ concentration in receiver (µm))/(incubation time (s)] B to A/A to B ratio = P app (B to A)/P app (A to B) Equilibrium dialysis experiments Plasma unbound fractions were determined in 96-well equilibrium dialyser (5000 Da molecular weight cut off membrane, Harvard Apparatus, Holliston, MA, USA). C57BL/6J mice frozen and Dulbecco s phosphate buffer saline was used for protein binding incubations. The compound of interest was added to to achieve a final concentration of 1 µm. Aliquots of 200 µl and Dulbecco s phosphate buffer saline were loaded into a 96-well equilibrium dialyser on opposite sides of the membrane and incubated at 37 C for 22 h with rotation. A total of 50 µl aliquots of and buffer were taken for liquid chromatography-tandem mass spectrometry (LC/MS/MS) analysis. The unbound fraction (f u ) of the PPB assay were calculated by following equation. Buffer fu = [ ] Plasma [ ] (1) The unbound fraction value was then used to calculate unbound concentrations. [ Plasma] Plasma f u =[ ] 3 total u (2) Animal experiments All experiments were performed according to procedures approved by the Merck Institutional Animal Care and Use Committee. C57BL/6J 7 9-week-old male mice (The Jackson Lab, Bar Harbor, ME, USA) and 7 9-weekold male, Pgp-deficient (mdr1a / ) and wild-type CF-1 mice (Charles River Laboratories, Raleigh, NC, USA) were used. Following overnight fasting, mice were administered compounds by oral gavage at 5 mg kg 1. The dosing solutions (ph approximately 7) were formulated at 0. 5 mg ml 1 in ethanol:water (5:95). Plasma, brain, and cerebrospinal fluid (CSF) samples were collected at 1 h post-dosing, the presumed C max. Blood was collected via cardiocentesis and treated in ethylenediamine tetraacetic acid (EDTA) tubes to separate the. The whole brain was harvested and blotted dry using a cotton swab. A 26-gauge butterfly needle was inserted into the cistema magna proximal to the occipital bone and the atlas, avoiding visible capillary vessels under the membrane. A 0.45 µm filter (Costar, Coming, NY, USA) was attached to the butterfly needle for protection and CSF

3 Does unbound drug and Pgp transport predict CNS exposure? 689 was gently aspirated using mouth suction. CSF volume quantitation was determined through pre- weighting and post-weighting of micro-centrifuge tubes. For in vivo exposure studies, (25 µl), CSF (5 10 µl, based on sample weights) and 100 µl brain homogenate (1 g brain homogenized in 3 ml water) were analysed by LC/MS/MS. LC-MS/MS analysis The compounds, CSF, brain homogenate and in vitro protein binding incubation samples were extracted by protein precipitation using 25 µl sample and 800 µl acetonitrile. CSF samples were spiked with an equivalent volume of, extracted in the same manner, and quantitated against the same calibration curve. Brain tissue and protein-binding buffer samples were quantitated against calibration curves prepared from identical matrix standards (tissue homogenate standards for tissues, for example). A 700 µl aliquot of acetonitrile supernatants was dried under nitrogen and reconstituted with 400 µl mobile phase. A 5 µl aliquot was loaded on the LC column. The compounds were separated on Fluophase (PFP) (Thermo Scientific) mm, 5 m, with mobile phase of 78% acetonitrile in water containing 5 mm ammonium formate and 0.1% formic acid at a flow rate of 0. 3 ml/min 1 operated on a Shimadzu LC-10ADvp pump, SIL auto sampler, and SCL 10A controller. All compounds and internal standards were detected by tandem mass spectrometry using API 4000 triple quadrupole mass spectrometer (Applied Biosystems/MDS Sciex, Ontario, Canada) at positive ionization model. The Turbo-Ion-Spray source was operated at turbo gas 45, ion-spray voltage 4500 V and the temperature 450 C. Each calibration curve consisted of eleven calibration points (0.5, 1, 2, 5, 10, 20, 50, 150, 500, 1000, and 2000 ng ml 1 ) prepared by a liquid handling robot (Beckman, Biomek 2000) and three concentrations of quality control samples (5, 50, and 500 ng ml 1 ) prepared manually. The calibration curves were determined using least-square regression analysis (weighting 1/x²) and Analyst 1.4 software (Applied Biosystems/MDS Sciex, Ontario, Canada). Results Physicochemical properties Thirty compounds from a drug-discovery programme that requires CNS exposure for efficacy were employed in this study. The molecular weights of these compounds ranged from 318 to 449 Da, log D values from 1.43 to 1.24 at ph 7.4, and log P from 0.04 to 3.2. These data, summarized in Table 1, suggest that the majority of the compounds can be characterized as low molecular weight and relatively hydrophilic molecules. In addition, all compounds exhibited adequate aqueous solubility and, therefore, were formulated in an aqueous solution containing a small amount of ethanol (5%) for dosing to mice. Transport by Pgp In vitro transport data, including permeability across the parental LLC-PK1 cell monolayer, ratios of the B to A/A to B transport across the cell monolayer expressing human MDR1 or mouse mdr1a, are presented in Table 1. Among the compounds tested, 27 out of the 30 exhibited good permeability, with their P app values greater than cm s 1 (Table 1). Compounds with the B to A/A to B ratios 2 were defined as Pgp substrates. On that basis, 13 compounds were identified as the substrates of mouse mdr1a, all of which also were the substrates of human MDR1. Compound number 15 was categorized as a non-pgp substrate because it did not show characteristic transport across the cell monolayer expressing human MDR1 and its B to A/A to B ratio of 2.2 for transport across the monolayer expressing mouse mdr1a was on the borderline of the arbitrary cut-off value of 2 (Table 1). Three Pgp substrates, namely compound numbers 20, 26 and 30, showed poor permeability, with P app values at cm s 1 (Table 1). On this basis, these compounds were excluded from further regression analysis. Plasma protein binding Reversible binding of the compounds to proteins was determined using an equilibrium dialysis method. The experimental conditions were set based on the information provided by the manufacturer and our own pilot test (data not shown). The majority of the compounds showed low to moderate protein binding in mouse, with unbound fractions (f u, %) greater than 20%. The data are summarized in Table 2. Brain-to- concentration ratios Following oral dosing to mice, the compounds were measured for their concentrations in mouse brain and samples, and the data are shown in Table 2. Seventeen of the 18 non-pgp substrate compounds (numbers 1 17) showed brain exposures comparable or several fold higher than unbound concentrations, with the brain-to-unbound concentration ratios ranging from 0.9 to 9.3 (Table 2). Compound number 18 appears to be an outlier, which in vitro showed good permeability and limited transport by Pgp but low in vivo brain exposure and a brain-to-unbound ratio of 0.3 (Table 2).

4 690 H. He et al. Table 1. Calculated physical properties and in vitro transport B to A/A to B ratios. Physical properties Compound ID Molecular weight Log P Log D (ph 7.4) P app 10 6 cm s 1 Control mdr 1a MDR1 The majority of the compounds identified in vitro as Pgp substrates showed brain exposures at less than onethird of the corresponding unbound concentrations (Table 2), a phenomenon that is consistent with active efflux of the compounds from mouse CNS back into systemic circulation. Three exceptions (compound numbers 22, 27 and 28) were observed in the Pgp substrate group with brain concentrations comparable with that in wild-type (Table 2). The higher brain concentrations for these Pgp substrates may be attributed to a variety of reasons including non-specific binding to brain tissue, variable receptor affinity, or Pgp saturation or inhibition. CSF-to- concentration ratios Following oral dosing to mice, the compound concentrations in mouse CSF and were measured, and the data are summarized in Table 2. CSF exposures correlate B to A/A to B transport ratios Non-Pgp substrates Pgp substrates Note: Pgp, P-glycoprotein. well with the unbound concentrations for those compounds categorized as non-pgp substrates, although the CSF concentrations were approximately three-fold lower than the corresponding unbound concentrations for 16 of 18 compounds (Table 2). The lower CSF exposures could be attributed to slow distribution of the compounds between mouse brain and CSF, as steadystate equilibrium may not have been achieved 1 h postdosing for some compounds, and may also be highly compound dependent. For the compounds identified as in vitro Pgp substrates (numbers 19 30), their CSF exposures were significantly lower (greater than three-fold) than the corresponding unbound concentrations; the former also showed no correlation with the latter (Table 2 and Figure 2). Compound numbers 20, 26 and 30 showed P app < cm s 1 ; thus their low CSF exposures could be attributed to poor membrane permeability (Tables 1 and 2).

5 Does unbound drug and Pgp transport predict CNS exposure? 691 Pgp substrate compounds were evaluated in Pgpdeficient (Pgp / ) mice, and the data are presented in Table 3. In this case, these compounds exhibited greatly increased CSF and brain exposures, as compared with that in wild-type mice (Table 2 and Figure 2). Discussion The central nervous system (CNS) is protected by the blood brain barrier (BBB) and blood cerebrospinal fluid barrier (BCSFB) from a large number of xenobiotic compounds, many of which are neurotoxins. Paradoxically, for a drug-discovery programme targeting CNS disorders, the goal is to overcome those barriers and deliver the compounds of interest to the site of therapeutic target. In this regard, an estimate of drug Table 2. In vivo exposures and in vitro protein binding data in wild-type mice. Compound ID Concentrations (µm) (n = 3) Total Brain Cerebrospinal fluid (CSF) f u () Unbound (µm) Non-Pgp substrates ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n.a ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n.a ± n.a ± n.a ± ± ± ± ± ± ± ± ± ± ± ± Pgp substrates ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± n.a ± ± ± ± ± ± ± ± ± ± ± ± ± Note: n.a., not available; Pgp, P-glycoprotein. exposures at the site of action is essential for establishing pharmacokinetic pharmacodynamic (PK/PD) relationships, and CSF drug concentrations have long been regarded as appropriate surrogates for CNS drug exposures. This is because: (1) CSF sampling is relatively straightforward as compared with sampling from other sites in the CNS, and (2) CSF drug concentrations approximate unbound drug exposures in the CNS, thus representing the true drug concentrations available for pharmacological activities. However, assessing CNS drug exposure, even with CSF sampling, is resource intense, representing relatively low throughput that usually cannot satisfy the need of fast-paced discovery programmes. On the other hand, rodents often serve as PD models in CNS programmes and exposure data are often available in the same species either from the PD experiment or from a separate PK study. In vitro P-glycoprotein (Pgp)- mediated transport can be evaluated on a routine basis for CNS programmes. It has been reported that in rats CSF drug concentrations correlated well with unbound drug concentrations normalized to in vitro Pgp transport ratios (Ohe et al. 2003). While this seems a reasonable approach, such a correlation remains empirical with only a limited number of compounds evaluated. The purpose of this study was therefore to investigate further the possibility of estimating CSF drug exposures based on a combined use of concentration, in vitro Pgp transport and protein binding data. Thus, 30 compounds from a drug-discovery programme were evaluated in vitro for Pgp-mediated transport. These compounds have molecular weights less than 500 Da and are relatively hydrophilic (log P < 3). The majority of the compounds exhibited greater than 20% unbound fraction in mouse and good LLC-PK1 permeability (P app > cm s 1 ) (Table 1). Twelve compounds were identified as mouse Pgp substrates using an arbitrary cut-off of the B to A/A to B ratio of 2 (Table 2). Based on a similar standard, these compounds also were determined to be human Pgp substrates, and the data demonstrated a qualitative agreement of substrate specificity between mouse and human Pgp (Table 2). These results suggest that mouse models may be of predictive value for Pgp-mediated transport in humans. In vivo, CSF and brain levels for non-pgp substrate compounds showed good correlation between unbound and brain exposures. Similarly, their CSF and unbound concentrations correlated well (Figure 1). Various regression calculations were attempted, with regression through zero yielding the greatest correlation for the data shown in Figures 1 3. These observations are consistent with the notion that, if permeability permits, xenobiotic molecules can

6 692 H. He et al. diffuse across the BBB to enter the CNS in the absence of active efflux. However, this hypothesis is nullified when in vitro Pgp substrates are included in the correlation (compound numbers 19 30). The majority of Pgp substrates showed substantially lower concentrations in mouse brain than in (Table 2), and their exposures in mouse CSF also were much lower than the unbound concentrations (Figure 2). In order to maintain clarity, three compounds (numbers 20, 26 and 30) in vitro showed P app < cm s 1 were excluded from data analysis because their low exposures in mouse brain can be attributed to low permeability (Table 1). The twelve Pgp substrates (numbers 19 30) then were evaluated for exposures in and CSF in Pgpdeficient (Pgp / ) mice. These compounds showed higher exposures in the CSF of Pgp / mice than Unbound (mm) y = 1.46x R = CSF Conc. (mm) non-pgp substates in wild-type mice Linear (non-pgp substates in wild-type mice) Figure 1. Correlation of mouse cerebrospinal fluid (CSF) versus unbound concentrations (, non-p-glycoprotein (Pgp) substrates in wild-type mice). Regressed through zero. Unbound ( mm ) y = 5.9x R = 0.45 y = 2.41x R = CSF Conc. (mm) Pgp substates in wildtype mouse Linear (Pgp substates in wild-type mouse) Pgp substrate in Pgp deficient mouse Linear (Pgp substrate in Pgp deficient mouse) Figure 2. Correlation of mouse cerebrospinal fluid (CSF) versus unbound concentrations (, P-glycoprotein (Pgp) substrates in wild-type mice;, Pgp substrates in Pgp-deficient (Pgp / ) mice). Regressed through zero. wild-type mice (Table 3). When plotted together, the CSF concentrations of the Pgp substrates determined in Pgp / mice were convoluted with those from non- Pgp substrates tested in wild-type mice, exhibiting a good correlation with the unbound fractions (Figure 3). Compound numbers 17, 18 and 19 appeared to be outliers and were not included in the correlation, although the reason behind the observation is not clear. This exercise reinforces that unbound and CSF drug concentrations correlate when the interference from active transporters is absent. This led to an effort to explore how would CSF exposures of Pgp substrates be better projected from their unbound concentrations. One possible approach is to plot CSF versus concentrations normalized to B to A/A to B transport ratios. This empirical method has been shown to produce an excellent correlation for 20 compounds between the CSF and systemic levels determined in rats (Ohe et al. 2003). Our results in mice Table 3. In vivo exposures and in vitro protein binding data in P-glycoprotein (Pgp)-deficient mice. Compound ID Unbound ( mm ) Concentrations (µm) (n = 3) Total Brain y = 1.04x R = 0.77 Cerebrospinal fluid (CSF) CSF Conc. (mm) Non-Pgp and PgP with Normalized Linear (Non-Pgp and PgP with Normalized) Figure 3. Correlation of mouse cerebrospinal fluid (CSF) versus unbound concentrations normalized to P-glycoprotein (Pgp) transport ratios for non-pgp and Pgp substrates. Regressed through zero. f u () Unbound (µm) Pgp substrates ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

7 Does unbound drug and Pgp transport predict CNS exposure? 693 also demonstrate CSF exposures of the 30 compounds of interest correlate well with the corresponding unbound concentrations when Pgp substrates were normalized using Pgp transport (B to A/A to B) ratios (Figure 3). Taken together, the present data collectively support the fact that CSF exposures correlate well with unbound drug concentrations normalized to in vitro Pgp transport ratios. Combined use of in vivo mouse models together with in vitro protein binding and Pgp transport data are of value in CNS-targeting drugdiscovery programmes. It should be noted, however, that selection of a Pgp substrate for development would likely lead to a low probability of success, since the compound would need high systemic exposures to achieve efficacious concentration in the CNS, potentially resulting in low safety margins. Acknowledgements The authors would like to acknowledge the chemists in MRL Rahway for providing the proprietary compounds for this research. Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper. References De Lange ECM. (2004). Potential role of ABC transporters as a detoxification system at blood CSF barriers. Adv Drug Deliv Rev 56: Kalvass JC, Maurer TS. (2002). Influence of nonspecific brain and binding on CNS exposure: implications for rational drug discovery. Biopharm. Drug Disposit 23: Kalvass JC, Maurer TS, Pollack GM. (2007). Use of and brain unbound to assess the extent of brain distribution of 34 drugs: comparison of unbound concentration ratios to in vivo P-glycoprotein efflux ratios. Drug Metab Disposit 35: Lin JH. (2004). How significant is the role of P-glycoprotein in drug absorption and brain uptake. Drug Today 40:5 22. Lin JH. (2008). CSF as surrogate for assessing CSF exposure: an industrial perspective. Curr Drug Metab 9: Maurer TS, DeBartolo DB, Tess DA, Obach SO. (2005). Relationship between exposure and nonspecific binding of thirty-three central nervous system drugs in mice. Drug Metab Disposit 33: Ohe T, Sato M, Tanaka S, Fujino N, Hata F, Shibata Y, Kanatani A, Fukami F, Yamazaki M, Chiba M, Ishii Y. (2003). Effect of P-glycoprotein-mediated efflux on cerebrospinal fluid/ concentration ratio. Drug Metab Disposit 31: Schinkel AH, Smit JJM, Van Tellingen O, Beijnen JH, Wagenaar E, Vand Deemter, Mol CAMM, der Valk V, Robanus-Maandag EC, Te Riele HPJ, Berns AJM, Borst P. (1994). Disruption of the mouse mdr1a P-glycoprotein gene leads to a deficiency in blood brain barrier and increased sensitivity to drugs. Cell 77: Street JA, Hemworth BA, Roach AG, Day MD. (1979). Tissue levels of several radiolabelled beta-adreno receptor antagonists after intravenous administration in rats. Arch Int Pharmacodyn Ther 237: Summerfield SG, Stevens AJ, Cutler L, Osuma MDC, Hammond B, Tang SP, Hersey TA, Spalding DJ, Jeffrey P. (2006). Improving the in vitro prediction of in vivo central nervous system penetration: integrating permeability, P-glycoprotein efflux and free fractions in blood and brain. J Pharmacol Exp Ther 316: