Comparison of brain capillary endothelial cell-based and epithelial (MDCK-MDR1, Caco-2, and VB-Caco-2) cell-based surrogate blood–brain barrier penetration models

Abstract

An accurate means of predicting blood–brain barrier (BBB) penetration and blood–brain partitioning of NCEs (new chemical entities) would fulfill a major need in pharmaceutical research. Currently, an industry-standard BBB drug penetration model is not available. Primary brain capillary endothelial cells, optionally co-cultured with astrocytes and/or pericytes, are the most valued models of BBB. For routine use, establishing and maintaining a co-culture system is too costly and labor intensive. Alternatively, non-cerebral cell lines such as MDCK-MDR1 are used, and most recently, the suitability of native and modified Caco-2 for predicting brain penetration has also come under investigation. This study provides comparative data on the morphology and functionality of the high integrity brain capillary endothelial BBB model (EPA: triple culture of brain capillary endothelial cells with pericytes and astrocytes) and the epithelial cell-based (native Caco-2, high P-glycoprotein expressing vinblastine-treated VB-Caco-2 and MDCK-MDR1) surrogate BBB models. Using a panel of 10 compounds VB-Caco-2 and MDCK-MDR1 cell lines show restrictive paracellular pathway and BBB-like selective passive permeability that makes them comparable to the rat brain BBB model, which gave correlation with the highest r² value with in vivo permeability data. In bidirectional assay, the VB-Caco-2 and the MDCK-MDR1 models identified more P-glycoprotein drug substrates than the rat brain BBB model. While the complexity and predictive value of the BBB model is the highest, for the screening of NCEs to determine whether they are efflux substrates or not, the VB-Caco-2 and the MDCK-MDR1 models may provide a simple and inexpensive tool.

Introduction

The insufficient presence of drugs at their brain targets due to the barrier function of brain capillary endothelial cells is a common cause of failure of drugs that target the central nervous system (CNS). An important feature of the brain capillary endothelial cells that form the blood–brain barrier (BBB) is that they exert a strict control over molecular movements between the brain and periphery through the expression of a wide range of brain-specific, high activity uptake, and efflux transporters [1], [2], [3], [4].

It is clear that BBB transporters such as P-glycoprotein (P-gp, MDR1), and drug binding to plasma proteins or brain tissue, may drastically modify the distinct processes of (1) the rate, (2) the extent of drug penetration, and (3) the intra brain drug distribution, which all affect the success of drug therapy [5], [6].

In drug discovery for the prediction of brain penetration, several types of models are used. Such models are in silico prediction, PAMPA (parallel artificial membrane permeability assay), cell culture-based approaches, and also animal models (BUI, in situ perfusion, etc.) as reviewed recently [7], [8]. PAMPA is a high throughput and low cost method for prediction of passive brain penetration in early phase of drug research for screening compounds [9]. Due to the higher complexity of information derived on both passive penetration and active transport processes, cell cultures are the next favored tools for BBB drug penetration modeling [7]. Basically, there are two types of cell culture models. Firstly, the “real BBB models” that are based on primary cultures of brain capillary endothelial cells or cell lines alone or supplemented and/or co-cultured with astrocytes/pericytes [7]. Primary culture-based models include porcine brain microvessel endothelial-cell (PBMEC) model by Zhang et al. [10], 4D/24w bovine brain capillary endothelial-cell model of Culot et al. [11], and the rat brain capillary model established by triple co-culture with pericytes and astrocytes by Nakagawa et al. [12]. Immortalized brain endothelial cell lines are easier to handle and less costly models. For many of the cell lines, sufficient integrity of monolayers to use them in permeability assays could not be reached [13]. While brain endothelial cell lines t-BBEC (bovine brain capillary endothelial), bEnd5 (mouse), RBE4 (rat) and the hCMEC/D3 (human) [14] are well characterized and very valuable tools for BBB research, their application for drug transport studies is limited as reviewed recently [7].

Second type of cell culture models are the surrogate BBB models that use epithelial-like cells like Madin–Darby Canine Kidney (MDCK) cells transfected with human MDR1 gene and the recently challenged human colon carcinoma cell line (Caco-2) [8], [15], [16].

As a rule of thumb, in vitro BBB models have to be free of leakiness as proven by low penetrability for paracellular low molecular size tracer molecules. Alternatively, they have to display a transepithelial electric resistance (TEER) that is preferably greater than or equal to 150–200 Ω cm² [13], [17]. The activity of BBB transporters, most importantly that of P-gp, is a prerequisite in the models, due to the clinical importance of this efflux transporter [18], [19], [20] and the large number of its substrates.

Understandably, brain capillary endothelial cell-based models are believed to be the best in vitro BBB models. Unlike the epithelial cell-based surrogate models, these cells stem from brain microvessels and their genetic programs define most BBB features. Under in vivo conditions, the development, maintenance, and function of the capillary endothelial cells are under the complex influence of the surrounding astrocytes, pericytes, and even neuronal contacts. This complexity results in the BBB functions. Many BBB functions are downregulated in vitro, and the supplementation of mono cultures with BBB-sourced factors and/or co-culturing with astrocytes/pericytes can greatly improve a number of diminished properties [11], [21], [22], [23], [24], [25], [26]. Unfortunately, these models are not convenient for routine industrial use because they are labor intensive, monolayer integrity is sensitive for experimental conditions, and the model itself is expensive. The characterization of these models is still uneven; while monolayer tightness is in high focus, the transporter functionality is not fully characterized even in the best models of BBB penetration in use. Data on the efflux functionality in brain capillary endothelial models are also scarce. Bachmeier et al. [27] reported four times higher P-gp and MRP functionality characterized by the efflux kinetics of 2′,7′,-bis(2-carboxyethyl)-5(6)-carboxyfluorescein acetoxymethyl ester in freshly isolated bovine microvessels than in cultured endothelial cells or MDCK, and little data have been published on the efflux ratio of P-gp and MRP substrate rhodamine 123 [10], [12].

The under-representation of transporters in the models is reflected in the few available studies where high in vitro–in vivo permeability correlations appear more frequently if transporter substrates, especially uptake substrates, are excluded [10], [28], [29]. Correcting in vivo data with brain and plasma protein binding has been reported to improve the strength of correlations [30], [31], [32].

Highly comparable in vitro–in vivo BBB permeability correlations were also achievable with epithelial cell-based Caco-2 and MDCK-MDR1 models [28], [29]. This may be surprising as these cells originate from the periphery with the appropriate organ-specific sets of membrane proteins [33], [34], [35], and they have a cell membrane lipid composition that differs from that of brain capillary endothelial cells [36].

MDCK-MDR1 has been identified as a surrogate BBB penetration model [31], [32], [37], [38]. Native Caco-2, which is the preferred choice of the industry for the prediction of intestinal absorption [39], [40], [41], [42], is also increasingly being investigated in comparative studies for BBB permeability prediction [28], [29], [43]. A serious disadvantage of Caco-2 is that the activity of P-gp in native culture is low and highly variable [44], [45], [46]. In a previous study, we reported the development of a vinblastine-treated Caco-2 cell line expressing a high level of P-gp [47], enabling this new model to detect both passive mechanism and P-gp substrate drugs. The VB-Caco-2 model shows identical passive permeability to native Caco-2 model, but it sensitively identifies P-gp substrates. The P-gp expression and activity remains consistently high in this cell line over 150 passages.

This study provides comparative data on the morphology and functionality of the high integrity primary brain capillary endothelial BBB model (EPA: triple culture of brain capillary endothelial cells with pericytes and astrocytes) and the epithelial cell-based (native Caco-2, high P-glycoprotein expressing vinblastine-treated VB-Caco-2 and MDCK-MDR1) surrogate BBB models. We also present here a large set of in vitro permeability data obtained in parallel in the widely recognized surrogate MDCK-MDR1 and the recently challenged high P-gp expressing VB-Caco-2 models of drug penetration.