Prediction of brain:blood unbound concentration ratios in CNS drug discovery employing in silico and in vitro model systems

1. Download high-res image (134KB)
2. Download full-size image

     

Determination of myocardium to plasma concentration ratios of five antipsychotic drugs: comparison with their ability to induce arrhythmia and sudden death in clinical practice

Reviewing available data shows that most of antipsychotic drugs are associated with arrhythmia and sudden death. Experimental studies have shown a HERG channel blockade, a dose-dependent increase in duration of action potential or of QT interval, with various degrees of indicators of serious arrhythmogenicity. However, it seems difficult to relate these in vitro and in vivo preclinical models to clinical findings, in part, because the relationship between concentrations used and in vivo tissue concentrations during treatment in man is not known. Consequently, we established the myocardium to plasma concentration ratios for a series of antipsychotic drugs by intraperitoneal administration of different level doses to the guinea pig. Then, we compared these values to their ability to induce arrhythmia or torsade de pointes in clinical practice. The myocardium to plasma concentration ratios were 2.2 for clozapine, 2.7 for olanzapine, 3.1 for sertindole, 4.5 for risperidone, and 6.4 for haloperidol. These data suggest that when the ratio is higher than 4, arrhythmia and sudden death may be expected. On the contrary, when the ratio is less than 3, little effect may be predicted. These results underscore the importance of interpreting HERG channel data and electrophysiological data in the context of other pharmacokinetic parameters such as myocardium to plasma distribution.     

Methods to assess tissue-specific distribution and metabolism of drugs

Most drugs exert their effects not within the plasma compartment, but in defined target tissues into which drugs have to distribute from the central compartment. Unfortunately, a complete and lasting equilibration between blood and tissue cannot always be taken for granted. Drug distribution processes may be characterized by a high intertissue- and intersubject variability and target site drug levels may substantially differ from corresponding plasma levels. Suboptimal target site concentrations may have important clinical implications, as it is a potential explanation for therapeutic failures. Therefore, determination of drug tissue penetration plays an important role in clinical drug development. In recent years, the assessment of tissue concentrations after administration of very low, sub-pharmacological drug amounts (microdosing) has attracted major interest in early clinical drug development, which calls for the availability of highly sensitive analytical methods. The present article will review the most important techniques that are currently available for studying drug disposition in humans. These can be classified as semi-invasive (microdialysis, MD) and non-invasive (positron emission tomography, PET, and magnetic resonance spectroscopy, MRS). We will discuss individual strengths and shortcomings of each method and provide some recent examples with particular focus on antiinfective and anticancer drugs. Whereas MD and MRS also lend themselves to the assessment of tissue-specific drug metabolism, PET usually does not provide metabolic information. For some drugs, such as the anticancer agents 5-fluorouracil and capecitabine, measurement of drug metabolites is particularly important as these represent the therapeutically active species.     

Use of hepatocytes to assess the contribution of hepatic uptake to clearance in vivo.

The wealth of information that has emerged in recent years detailing the substrate specificity of hepatic transporters necessitates an investigation into their potential role in drug elimination. Therefore, an assay in which the loss of parent compound from the incubation medium into hepatocytes ("media loss" assay) was developed to assess the impact of hepatic uptake on unbound drug intrinsic clearance in vivo (CL(int ub in vivo)). Studies using conventional hepatocyte incubations for a subset of 36 AstraZeneca new chemical entities (NCEs) resulted in a poor projection of CL(int ub in vivo) (r2 = 0.25, p = 0.002, average fold error = 57). This significant underestimation of CL(int ub in vivo) suggested that metabolism was not the dominant clearance mechanism for the majority of compounds examined. However, CL(int ub in vivo) was described well for this dataset using an initial compound "disappearance" CL(int) obtained from media loss assays (r2 = 0.72, p = 6.3 x 10(-11), average fold error = 3). Subsequent studies, using this method for the same 36 NCEs, suggested that the active uptake into human hepatocytes was generally slower (3-fold on average) than that observed with rat hepatocytes. The accurate prediction of human CL(int ub in vivo) (within 4-fold) for the marketed drug transporter substrates montelukast, bosentan, atorvastatin, and pravastatin confirmed further the utility of this assay. This work has described a simple method, amenable for use within a drug discovery setting, for predicting the in vivo clearance of drugs with significant hepatic uptake.     

Prediction of unbound serum valproic acid concentration by using in vivo binding parameters.

In a previous study, we determined the in vivo binding parameters of valproic acid (VPA) to serum proteins in seven healthy young adults at steady state by using the Scatchard equation. To evaluate the ability of the Scatchard binding equation to predict steady-state unbound serum VPA concentrations (Cf), 39 adult patients receiving VPA monotherapy and ranging in age from 16 to 68 years were studied. The correlation between predicted and observed Cf was high (r = 0.865). Mean prediction error, mean absolute error (MAE), and root mean squared error (RMSE) were calculated, and served as a measure of prediction bias and precision. The MAE and RMSE were low (MAE = 12.9 mumol/L, RMSE = 17.7 mumol/L). It is feasible to use the Scatchard binding equation to predict Cf in patients receiving VPA monotherapy.     

Species comparison of in vivo P-glycoprotein-mediated brain efflux using mdr1a-deficient rats and mice

The experiments described herein compared the extent of in vivo P-glycoprotein (P-gp)-mediated brain efflux between rats and mice for a set of known central nervous system compounds. With use of newly introduced genetically modified mdr1a-deficient rats and their gene-competent counterparts, the brain to plasma distribution was assessed and compared with the distribution pattern in mdr1a-deficient and wild-type mice. Four compounds (aripiprazole, citalopram, risperidone, and venlafaxine) were administered using a continuous subcutaneous osmotic minipump infusion paradigm. Steady-state brain and plasma concentrations of the compounds, including selected metabolites (9-hydroxyrisperidone, O-desmethyl-venlafaxine and N-desmethyl-venlafaxine) were measured in mdr1a-deficient rats and mice and their wild-type counterparts along with their free fractions to determine total and unbound brain to plasma distribution between genotypes within and between species. The results revealed qualitative as well as quantitative similarities between P-gp functionality in vivo at the blood-brain barrier level in rats and mice. All compounds tested were shown to have a significantly higher brain to plasma distribution in both mdr1a-deficient rats and mice compared with that in their wild-type counterparts. Moreover, the relative enhancement in extent of brain penetration between mdr1a-deficient and wild-type rats could be directly correlated to the enhancement ratios obtained in mice. From the unbound brain to unbound plasma distributions, the impact of P-gp on the overall brain penetration capabilities showed minor differences between rats and mice for the compounds tested. In conclusion, a comparable functional role of P-gp between rats and mice with respect to brain efflux mediated by this transporter is suggested.     

Consideration of the linear concentration increase of the unbound drug fraction in plasma

The concentration of drug in plasma may not necessarily be much less than that of the drug binding proteins, as often considered. Thus the unbound drug fraction could be concentration dependent, which should be taken into account in the interpretation of drug pharmacokinetics and modeling. It is shown that the increase of the unbound drug fraction, f(u), can be very accurately considered proportional to the drug plasma concentration for a relatively wide range of concentrations. Equations for the calculation of f(u) in this linear range are obtained, as well the limiting drug concentration when the linear approximation of f(u) is applicable. The suggested approach greatly simplifies the calculation of f(u) and can be ready used in pharmacokinetic calculations and PK-PD models, as well as for the prediction of the change of f(u) due to the variation of protein concentrations in plasma. Naproxen protein binding in human plasma is considered as an illustration of the method.     

In vitro and in vivo investigations on fluoroquinolones; effects of the P-glycoprotein efflux transporter on brain distribution of sparfloxacin.

The role of mdr1a-encoded P-glycoprotein on transport of several fluoroquinolones across the blood-brain barrier was investigated. In vitro, P-glycoprotein substrates were selected by using a confluent monolayer of MDR1-LLC-PK1 cells. The inhibition of fluoroquinolones (100 microM) on transport of rhodamine-123 (1 microM) was compared with P-glycoprotein inhibitors verapamil (20 microM) and SDZ PSC 833 (2 microM). Subsequently, transport polarity of fluoroquinolones was studied. Sparfloxacin showed the strongest inhibition (26%) and a large polarity in transport, by P-glycoprotein activity. In vivo, using mdr1a (-/-) and wild-type mice, brain distribution of pefloxacin, norfloxacin, ciprofloxacin, fleroxacin and sparfloxacin was determined at 2, 4, and 6 h following intra-arterial infusion (50 nmol/min). Brain distribution of sparfloxacin was clearly higher in mdr1a (-/-) mice compared with wild-type mice. Sparfloxacin was infused (50 nmol/min) for 1, 2, 3 and 4 h in which intracerebral microdialysis was performed. At 4 h, in vivo recovery (dynamic-no-net-flux method) was 6.5+/-2.2 and 1.5+/-0.5%; brain(ECF) concentrations were 5.1+/-0.2 and 26+/-21 microM; and total brain concentrations were 7.2+/-0.3 and 23+/-0.3 microM in wild-type and mdr1a (-/-) mice, respectively. Plasma concentrations were similar (18.4+/-0.7 and 17.9+/-0.5 microM, respectively). In conclusion, sparfloxacin enters the brain poorly mainly because of P-glycoprotein activity at the blood-brain barrier.     

Use of plasma proteins as solubilizing agents in in vitro permeability experiments: correction for unbound drug concentration using the reciprocal permeability approach

The purpose of the present study was to explore the applicability of the reciprocal permeability approach to correct for changes in thermodynamic activity when in vitro permeability data are generated in the presence of plasma proteins. Diazepam (DIA), digoxin (DIG), and propranolol (PRO) permeability was assessed in the presence of bovine serum albumin (BSA) and bovine alpha-1-acid glycoprotein (AAG). The reciprocal permeability approach was subsequently employed to calculate the true permeability coefficient (Papp(corr)) and the operational protein association constant (nK(a)). For BSA binding, good agreement was observed between the Papp(corr) values and Papp values obtained in the absence of protein. For PRO and AAG, where binding affinity was high, deviation in the reciprocal permeability plots was evident suggesting ligand depletion at low drug/high protein concentrations. Bidirectional DIG permeability data in the presence of either BSA or AAG indicated that neither protein had an effect on the efflux transporters involved in DIG permeability. The data suggest that plasma proteins can be utilized in permeability experiments with no adverse effects on transporter function and that the reciprocal permeability approach can be used to correct permeability data for changes in unbound drug concentration.     

On the distribution of drugs in saliva and blood plasma.

The equilibration of drug concentrations between blood plasma (unbound part) and saliva was studied for selected drugs with different physicochemical properties: Quinidine, sulfamerazine, paracetamol, diazepam and ethanol. From these and other experimental results it is suggested: 1. The concentration ratio between saliva and blood plasma (unbound) can only equal one for basic drugs with a pK a lower than 5.5 and acid drugs with a pK a higher than 8.5 and for pH indifferent drugs. 2. The drug must have a sufficient permeation ability through lipid membranes, valuable by the lipid water partition coefficient.     

Use of solution calorimetry to determine the extent of crystallinity of drugs and excipients

A solution calorimetry method was developed to quantitatively examine mixtures of amorphous and crystalline sucrose or warfarin sodium based on the energy differences between their solid forms. The heats of solution of crystalline sucrose, amorphous sucrose, clathrate warfarin sodium and amorphous warfarin sodium were 1474.08±37.78 cal/mol, −3550.47±51.04 cal/mol, −1.701±0.041 cal/g and −7.386±0.226 cal/g, respectively. The observed linear relationship between the heat of solution and the percent of the crystalline form present in the sample provided a rapid and convenient way to quantitatively determine the crystallinity of a common drug excipient (sucrose) and/or a complex system, such as the clathrate warfarin sodium (a complex of warfarin sodium, isopropyl alcohol and water). The solid state conversion process could also be monitored by measuring the energy changes associated with changes in crystallinity.     

Naproxen free plasma concentrations and unbound fractions in patients with osteoarthritis: relation to age, sex, efficacy, and adverse events.

Plasma samples from 237 patients with osteoarthritis treated with 750 mg naproxen daily were obtained after four weeks of active therapy. The samples were subjected to equilibrium dialysis, analyzed by high-performance liquid chromatography, and free concentrations and unbound fractions were determined. Within the free plasma concentration range in this large patient group, we could detect no association between free naproxen concentration and efficacy score or between free concentrations and adverse events. Free concentrations were 0.295 +/- 0.260 microgram/ml (mean +/- SD) and unbound fractions were 0.33 +/- 0.24%. Females had 65% higher free concentration compared to males (p less than 0.001). For females, but not for males, there was a statistically significant correlation (p less than 0.005) between age and free concentration. The free concentration was estimated to be 88% higher in an 80-year-old female compared to a 50-year-old. Females had a 41% higher unbound fraction than males (p less than 0.005). For females, but not for males, a statistically significant relationship between age and unbound fraction was found. The unbound fraction was estimated to be 62% higher in an 80-year-old female than in a 50-year-old.     

Use of the cassette-dosing approach to assess brain penetration in drug discovery

The objective of the present study was to examine the cassette dosing method in determination of brain-to-plasma concentration ratio (area under the concentration-time profiles for plasma/area under the concentration-time profiles for brain, K(p)). Eleven model compounds, amprenavir, citalopram, digoxin, elacridar, imatinib, (3S,6S,12aS)-1,2,3,4,6,7,12,12a-octahydro-9-methoxy-6-(2-methylpropyl)-1,4-dioxopyrazino[1',2':1,6]pyrido[3,4-b]indole-3-propanoic acid 1,1-dimethylethyl ester (Ko143), loperamide, prazosin, quinidine, sulfasalazine, and verapamil, were selected to compare their K(p) determined from discrete dosing in wild-type mice and their K(p) from cassette dosing in wild-type, Mdr1a/1b(-/-), Bcrp1(-/-), and Mdr1a/1b(-/-)/Bcrp1(-/-) mice at 1 to 3 mg/kg. The mice brain and plasma were collected at 0.25, 1, and 3 h and were analyzed using high-performance liquid chromatography-tandem mass spectrometry methods. The K(p) determined from discrete dosing versus cassette dosing in the wild-type mice were within 2-fold for all the compounds except sulfasalazine and Ko143. The brain concentrations of sulfasalazine and Ko143 and the plasma concentrations of Ko143 were below the lower limit of quantitation. In addition, the K(p) values estimated by mass spectrometry responses, namely the ratio of compound peak area to internal standard peak area, were within 2-fold of the K(p) observed from the actual concentrations. Furthermore, the ratios of K(p) in Mdr1a/1b(-/-), Bcrp1(-/-), and Mdr1a/1b(-/-)/Bcrp1(-/-) mice versus the K(p) in the wild-type mice from cassette dosing were consistent with the ones reported in the literature where the compounds were dosed discretely. These results demonstrate that drug-drug interactions at the blood-brain barrier are unlikely at a subcutaneous dose of 1 to 3 mg/kg and support the use of the cassette dosing approach to assess brain penetration in drug discovery.     

Influence of maternal plasma protein binding on fetal unbound plasma concentration of propranolol in the pregnant ewe

According to theory, for a drug of nonrestrictive, flow-limited clearance, a change during pregnancy of the unbound fraction (fu) of drug in maternal plasma should cause a change in steady-state unbound plasma drug concentration (Cu) in maternal plasma, which should also cause a change in fetal Cu. This theory was examined in 14 chronically cannulated, unanesthetized pregnant ewes in which 28 separate experiments were performed during the latter part of gestation. An initial bolus dose and 3-h constant rate infusion of propranolol were administered via the maternal jugular vein and steady-state maternal and fetal carotid arterial plasma total and unbound propranolol concentrations were measured. Fetal Cu (32 +/- 21 ng/mL) was significantly less than maternal Cu (78 +/- 52 ng/mL), due to previously demonstrated fetal hepatic extraction of propranolol. Notwithstanding fetal elimination, there was a significant correlation between fetal Cu and maternal Cu (r = 0.41, p less than 0.025). There was also a strong correlation between fetal Cu and the maternal unbound fraction of drug (fu; r = 0.75, p less than 0.001). We conclude that for propranolol, a drug of nonrestrictive, flow-limited clearance, changes in maternal fu can have a significant influence on fetal Cu, and therefore would be expected to influence the pharmacological effect of the drug in the fetus.     

Autoradiographic analysis of the in vivo distribution of 3H-imipramine and 3H-desipramine in brain: comparison to in vitro binding patterns

Using high resolution autoradiographic techniques, the distribution of radioactivity in forebrain and brainstem was assessed after IV injection of 3H-impramine or 3H-desipramine. Results were compared with regional binding of the drugs to brain sections in vitro. Similar topographic binding of 3H-imipramine and 3H-desipramine was observed in vitro among brain regions, except in the paraventricular nucleus of the hypothalamus and locus coeruleus, where binding was greater for 3H-desipramine. For both 3H-desipramine and 3H-imipramine, some brain regions that exhibited high binding in vitro also showed high accumulation after in vivo injection. However, certain regions that contained high densities of binding sites for the antidepressant drugs as measured by in vitro binding showed very low accumulation of radioactivity after in vivo treatment. Such regions included the dentate gyrus of the hippocampus, layer 1 of piriform cortex, caudate-putamen, pontine and midbrain central gray, and cerebellar granular layer. Compared to in vitro binding of the drugs, the distribution of imipramine and desipramine in vivo appears more anatomically selective. For imipramine, primary sites of action in vivo, as indicated by the topographic distribution in brain, appear to be the locus coeruleus, hippocampus, lateral septal nucleus, and amygdala. For desipramine, the greatest accumulation in vivo was found in the locus coeruleus, paraventricular nucleus of the hypothalamus, and anterior thalamic nuclei.     

Substance P: in vitro inactivation by rat brain fractions and human plasma.

The in vitro degradation of substance P (SP) by rat brain fractions and human plasma was studied by means of polyacrylamide gel electrophoresis and ³H-[Nle]¹¹SP. Substance P was degraded by neutral metalloendopeptidase systems of the crude synaptosomal fraction and plasma with K_m values of 3.6 × 10^?5 and 8 × 10^?6 M, respectively. The peptide was also degraded by other subcellular brain fractions. The degradation rates corresponded with the biological inactivation as assayed on guinea pig ileum. The basic N-terminal sequence of the peptide was found to be important for this inactivation. No specific uptake of the peptide in the synaptosomal brain fraction could be observed.It is concluded that in vivo substance P is degraded with first order dependence and that this proteolytic breakdown is responsible for the termination of the biological activity of substance P as a putative neurotransmitter or neuromodulator.