Tuning the predictive capacity of the PAMPA-BBB model
Graphical abstract
Introduction
The blood–brain barrier (BBB) characterized by a complex membrane structure is composed of brain endothelial cells with tight junctions. Its physiological role is considered as dual: separating the brain from the systemic circulation, and controlling the molecular traffic of endogenous and exogenous molecules into the central nervous system (CNS). The strict and complex nature of this control is based on the unique combination of physical (lipid membrane with tight junctions), transport (efflux and influx transporters) and metabolic (metabolic enzymes) barrier properties (Abbott, 2004).
Screening for BBB permeation in the early phase of drug discovery is of outstanding importance. In the last decades, great effort was focused on the methodological improvement of in silico and in vitro BBB transport models as well, however, no superior (i.e., widely accepted industrial standard) assay has emerged for the prediction of BBB permeation (Wilhelm and Krizbai, 2014). Cell-based assays are capable of simultaneously modeling both active and passive transport mechanisms; however they are considered as labor-intensive, time consuming and low-throughput. Moreover, the transporters in different cell lines are often overexpressed, and can be easily saturated. Widely used and patented cell-based assays for the prediction of BBB permeation are reviewed by Toth et al (Toth et al., 2011).
It has been reported that efforts were made to generate IAM-HPLC (Immobilized Artificial Membrane — IAM) systems that mimicked the properties of biological membranes, especially BBB, to use as a permeability screening technique. However, IAM methods are not widely used (Grumetto et al., 2012).
The parallel artificial membrane permeability assay (PAMPA), developed by Kansy et al., 1998, is a non-cell-based, high-throughput permeation model (Kansy et al., 1998). PAMPA is widely used in the early phase of drug discovery for the prediction of passive diffusion of drug molecules across phospholipid membranes as it is a cost effective and robust method with good reproducibility (Avdeef, 2005). The original PAMPA model utilized phosphatidylcholine as membrane; since then new organ-specific models appeared which can predict the gastrointestinal absorption (Sugano et al., 2001), the BBB (Di et al., 2003) and the skin penetration (Sinko et al., 2012). During the last decade the PAMPA system has undergone several modifications and improvements as reviewed recently by Bicker et al. (Bicker et al., 2014). First, Di et al. (Di et al., 2003) developed a porcine brain lipid extract (PBLE) based artificial membrane dissolved in n-dodecane. 30 compounds were investigated and classified into two classes based on effective permeability data: CNS +/− molecules (which can and cannot cross the BBB, respectively). Composition and volume of the membrane were varied and optimized for the discrete separation of CNS + and CNS − compounds, however an in silico logBB scale was used only in the assay validation. Tsinman et al. (Tsinman et al., 2011) published a model called in combo PAMPA. This method was developed by combining in vitro measured PAMPA permeability data with a H-bond descriptor. Applicability of this in combo approach was demonstrated primarily for basic molecules. As far as we know, however, in combo PAMPA has not became a widespread method due to its computational cost. Jhala et al. (Jhala et al., 2012) evaluated the impact of the incubation time on the in vitro permeability. Campbell et al. (Campbell et al., 2014) reported differences between the properties of PBLE and human brain microvessel lipid (MVL) based membranes, and microvessel lipids with cholesterol. Carrara et al. (Carrara et al., 2007) demonstrated the importance of the concentration of lipid components and the thickness of the lipid membrane. It was claimed that n-dodecane overdosed the effect of phospholipids in permeability measurements. The mixture of n-dodecane:n-hexane 1:1 (v/v) was recommended to dissolve phospholipids; hence as the n-hexane evaporated, the membrane layer became thinner and more phospholipid-specific. Due to this modification an overall increase in permeability was observed. However, the relation between the permeability results and corresponding logBB values was not investigated.
Lipophilicity is one of the most important physicochemical parameters in drug discovery and development, and reported as a key determinant of CNS exposure (Rankovic, 2015). However, n-octanol–water partition cannot accurately describe the hydrogen-bonding and ionic interactions between drugs and lipid components of biological membranes (e.g. BBB). In anisotropic systems like IAM and PAMPA methods, the partition into a water–phospholipid system is not only governed by lipophilic/hydrophobic intermolecular effects but also influenced by electrostatic interactions between charged species and polar headgroups of phospholipids (Grumetto et al., 2012). Thus, transport (and accumulation) depends on the affinity of drug molecules for headgroups (cephalophilicity), core (lipophilicity) or both phases (amphiphilicity) of the bilayer membrane (Balaz, 2012). Moreover, in the PAMPA system the lipid solvent n-dodecane modifies the various interactions between drug molecules and phospholipids.
Like the n-octanol–water partition and liposome–water partition different effects prevail. In the case of n-octanol–water partition there is a lack of ionic interaction and the presence of lipophilic interactions. In contrast to this, in the case of the liposome–water system ionic interactions occur and there are H-bond donor and acceptor atoms. Similarly, the exclusive use of n-dodecane results in solvent-driven permeability (Avdeef, 2001). Reducing the amount of n-dodecane increased the interaction between phospholipids and drug molecules hence the permeability was rather driven by the lipid–drug interactions. Using BBB specific phospholipids the model mimicked specifically BBB–drug interaction. These considerations could be summarized as follows: n-dodecane poses the diffusion-barrier in the case of solvent-driven mechanism, while, if the vigorous effect of the solvent could be avoided and the diffusion is governed by phospholipids, the dominant mechanism of permeability can be considered as lipid-driven.
Summarized the pros and cons of previous studies related to PAMPA variables and based on our earlier investigations on BBB specific PAMPA (Konczol et al., 2013), we aimed to investigate the diffusion mechanisms and further optimize the PAMPA-BBB model by (i) finding the optimal lipid solvent (type, composition and volume) in order to avoid the non-specific overdosing effect of n-dodecane; (ii) studying the effect of cholesterol and lipid concentration in the artificial membrane; and (iii) evaluating physiology-mimicking assay conditions.
Section snippets
Chemicals
Analytical grade DMSO, acetonitrile, and formic acid were from Merck (Darmstadt, Germany). All other solvents and reference compounds were of analytical grade and purchased from Sigma-Aldrich (St Louis, MO, USA), except for porcine polar brain lipid extract, which was obtained from Avanti Polar Lipids Inc. (Alabaster, AL, USA). Water was obtained from a Milli-Q water-purification system (Millipore, Bedford, MA, USA) and used for all aqueous solutions.
PAMPA method
The PAMPA method was previously published (
The original method
27 structurally diverse and commercially available drugs were chosen with a wide range of logBB scale (Table 1). In terms of molecular weight and lipophilicity a broad range was covered (138 < Mw < 385; − 2.25 < logD7.4 < 3.48). Acidic (n = 5), basic (n = 17) and neutral (n = 5) compounds were also included. Due to the critical selection none of them are known as influenced by active transporters to the CNS, the molecules which cross the BBB permeate by passive diffusion. This property is important in a model
Conclusions
In the early phase of CNS drug discovery screening for blood–brain barrier penetration is a key issue. The pharmaceutical industry needs robust, fast, well-designed and reliable HTS models. Predictive value of these models can be improved by strict compound selection and optimizing the critical parameters. The PAMPA model is a robust and versatile tool for predicting membrane penetration which gives a platform for fine tuning of conditions. Several PAMPA methods have been reported in the
Acknowledgment
The authors thank M. Meszlényi Sipos for the GC-FID measurements. J. Müller thanks the Gedeon Richter Talentum Foundation for the financial support.
References (39)
Prediction of blood–brain barrier permeation in drug discovery from in vivo, in vitro and in silico models
Drug Discov. Today Technol.
(2004)Does transbilayer diffusion have a role in membrane transport of drugs?
Drug Discov. Today
(2012)- et al.
pH-gradient PAMPA-based in vitro model assay for drug-induced phospholipidosis in early stage of drug discovery
Eur. J. Pharm. Sci.
(2013) - et al.
Blood–brain barrier models and their relevance for a successful development of CNS drug delivery systems: a review
Eur. J. Pharm. Biopharm.
(2014) - et al.
Significance of lipid composition in a blood–brain barrier-mimetic PAMPA assay
J. Biomol. Screen.
(2014) - et al.
Evaluation of in vitro brain penetration: optimized PAMPA and MDCKII-MDR1 assay comparison
Int. J. Pharm.
(2007) - et al.
High throughput artificial membrane permeability assay for blood–brain barrier
Eur. J. Med. Chem.
(2003) - et al.
Cholesterol metabolism in the central nervous system during early development and in the mature animal
J. Lipid Res.
(2004) - et al.
Potential of biopartitioning micellar chromatography as an in vitro technique for predicting drug penetration across the blood–brain barrier
J. Chromatogr. B Anal. Technol. Biomed. Life Sci.
(2004) - et al.
Lipophilic and electrostatic forces encoded in IAM-HPLC indexes of basic drugs: their role in membrane partition and their relationships with BBB passage data
Eur. J. Pharm. Sci.
(2012)
Correlation of blood–brain penetration using structural descriptors
Bioorg. Med. Chem.
Drug-like properties and the causes of poor solubility and poor permeability
J. Pharmacol. Toxicol. Methods
Application of PAMPA-models to predict BBB permeability including efflux ratio, plasma protein binding and physicochemical parameters
Int. J. Pharm.
Correlation and prediction of a large blood–brain distribution data set—an LFER study
Eur. J. Med. Chem.
Skin-PAMPA: a new method for fast prediction of skin penetration
Eur. J. Pharm. Sci.
Optimized conditions of bio-mimetic artificial membrane permeation assay
Int. J. Pharm.
Prediction of passive blood–brain partitioning: straightforward and effective classification models based on in silico derived physicochemical descriptors
J. Mol. Graph. Model.
PAMPA study of the temperature effect on permeability
Eur. J. Pharm. Sci.
Prediction of blood–brain barrier permeation of α-adrenergic and imidazoline receptor ligands using PAMPA technique and quantitative-structure permeability relationship analysis
Eur. J. Pharm. Sci.
Cited by (33)
Extending the limitations in the prediction of PAMPA permeability with machine learning algorithms
2023, European Journal of Pharmaceutical SciencesA review of pregnancy-induced changes in opioid pharmacokinetics, placental transfer, and fetal exposure: Towards fetomaternal physiologically-based pharmacokinetic modeling to improve the treatment of neonatal opioid withdrawal syndrome
2022, Pharmacology and TherapeuticsCitation Excerpt :The maternal and fetal circulation are separated from each other by syncytiotrophoblasts, cytotrophoblast, and connective tissue (Fig. 3). Buprenorphine, fentanyl, and methadone are lipophilic compounds, with logD7.4 = +3.93, +2.81, and +1.82, respectively (Avdeef, Barrett, Shaw, Knaggs, & Davis, 1996; Kung et al., 2019; Müller, Esső, Dargó, Könczöl, & Balogh, 2015), which promotes passive diffusion across the placenta. Oxycodone and morphine are, in contrast, somewhat hydrophilic, with logD7.4 = −0.07 and −0.15, respectively (Avdeef et al., 1996; Pöyhiä & Seppälä, 1994).
Development of Ussing model coupled with artificial membrane for predicting intestinal absorption mechanisms of drugs
2020, International Journal of PharmaceuticsCitation Excerpt :All initial drug solution was prepared in DMSO and then diluted with 10 mM phosphate buffer solution (PBS) to obtain the final concentrations of 0.5 mM for acyclovir, famotidine, atenolol, furosemide, ketoprofen, propranolol, verapamil, ofloxacin, naproxen, penicillin, pravastatin, metformin, ranitidine, cimetidine, doxorubicin, sulfasalazine, ampicillin, ciprofloxacin, digoxin, chlorpheniramine and 0.2 mM for carbamazepine, dexamethasone and ibuprofen because of their low solubility. Muller’s found shows the most effective DMSO concentration on PAMPA was 5% v/v, so the final concentration of DMSO was kept at 5% v/v (Müller et al., 2015). The structure of Ussing chamber was shown in Fig. 1A (Sjoberg et al., 2013).
Neuropharmacokinetics: a bridging tool between CNS drug development and therapeutic outcome
2019, Drug Discovery TodayCorneal-PAMPA: A novel, non-cell-based assay for prediction of corneal drug permeability
2019, European Journal of Pharmaceutical SciencesA new approach to examining the extraction process of Zhishi and Zhiqiao considering the synergistic effect of complex mixtures by PAMPA
2018, Journal of Chromatography B: Analytical Technologies in the Biomedical and Life SciencesCitation Excerpt :It has been shown that DMSO induces water pores in the dipalmitoyl-phosphatidylcholine (DPPC) bilayers and thus can be a possible pathway for the enhancement of drug molecule penetration through lipid membranes. Recently, Muller [29] investigated the effect of DMSO concentration and the incubation temperature on PAMPA. It was found that the most effective DMSO concentration is 5%(v/v) and incubation at room temperature produces weaker results than incubating at the more physiologically relevant temperature of 37 °C.