Protein Expression and Functional Relevance of Efflux and Uptake Drug Transporters at the Blood-Brain Barrier of Human Brain and Glioblastoma

Abstract

The knowledge of transporter protein expression and function at the human blood-brain barrier (BBB) is critical to prediction of drug BBB penetration and design of strategies for improving drug brain/brain tumor delivery. This study determined absolute transporter protein abundances in isolated microvessels of human normal brain (N=30), glioblastoma (N=47), rat (N=10) and This article has been accepted for publication and undergone full peer review but has not been through the copyediting, typesetting, pagination and proofreading process, which may lead to differences between this version and the Version of Record. Please cite this article as doi:10.1002/CPT.1710 mouse brain (N=10), and cell membranes of MDCKII cell lines, using targeted proteomics. In glioblastoma microvessels, ABCB1, ABCG2, MCT1, GLUT1, Na/K ATPase, and Claudin-5 protein levels were significantly reduced, while LAT1 was increased and GLU1 remained the same, as compared to human normal brain microvessels. ABCC4, OATP1A2, OATP2B1, and OAT3 were undetectable in microvessels of both human brain and glioblastoma. Species difference in BBB transporter abundances was noted. Cellular permeability experiments and modeling simulations suggested that not a single apical uptake transporter, but a vectorial transport system consisting of an apical uptake transporter and basolateral efflux mechanism was required for efficient delivery of poor transmembrane permeability drugs from the blood to brain.

Introduction

Despite extensive efforts, progress in developing effective therapies for glioblastoma (WHO grade IV glioma), the most common and aggressive malignant primary brain cancer, has been disappointing. Lack of therapeutic efficacy may stem from the genetically heterogeneous nature of the tumor, the emergence of intrinsic resistance mechanisms, and particularly, poor penetration of potentially effective therapeutic agents across the human blood-brain barrier (BBB).¹ The integrity of the BBB in glioblastoma is disrupted to different degrees, ranging from completely compromised in bulky tumor areas and slightly “leaky” in invasive peripheral regions to complete intact in infiltrative tumor regions.² In spite of enhanced drug penetration into bulky tumor areas, insufficient penetration of potentially effective therapeutic agents into invasive, infiltrative tumor regions with an intact BBB represents a significant barrier to long-term, efficacious treatments.³

The BBB consists of both physical and biochemical barriers. The physical barrier is a continuous layer of brain microvascular endothelial cells connected by tight junctions, thus restricting paracellular exchange between the blood and brain. The biochemical barrier is composed of a wide range of transporters and receptors that facilitate transport of endogenous compounds and therapeutic drugs into or out of the brain. Small molecule drugs move across the BBB mainly via two modes: transcellular passive diffusion and transporter-mediated active transport. The efficiency of passive diffusion is dependent on drug physicochemical properties (e.g., molecular weight, lipophilicity, charge, and polar surface area).⁴ Nevertheless, the penetration of many drugs across the BBB, despite having physicochemical properties favorable to passive diffusion, is largely restricted by a group of ATP-binding cassette (ABC) efflux transporters expressed at the luminal endothelial cell membrane of the BBB, notably ABCB1 (P-glycoprotein) and ABCG2 (breast cancer resistance protein).^5,6 In addition, solute carrier transporters have been identified at the BBB to modulate the transport of nutrients and potentially therapeutic drugs into or out of the brain.^7,8

Mechanistic understanding and quantitative prediction of drug penetration across the human BBB is critical to rational development of novel therapeutics and better use of existing drugs for efficacious treatment of brain cancer especially glioblastoma. However, prospective, accurate prediction of drug brain/tumor penetration has been significantly hindered due to the lack of data on transporter protein expression levels at the human BBB and blood brain-tumor barrier. Accurate prediction of BBB penetration is further compromised by the incomplete understanding or misunderstanding of functional interplays of multiple transporters at the BBB, which also results in uncertainties and thus unlikely success on the development of BBB transporter-targeted strategies for improving drug CNS delivery. For example, extensive efforts have been focused on circumventing ABCB1-mediated efflux with the goal of enhancing drug CNS delivery. Despite significant increase of drug CNS exposure were observed in animal models and very potent ABCB1 inhibitors were clinically available, appreciable enhancement of CNS drug exposure as a result of ABCB1 inhibition has not been demonstrated in humans.⁹ On the other hand, targeting BBB uptake transporters such as OATP1A2 has been proposed as a strategy for more effective delivery of drugs to the brain.¹⁰ However, the functional expression of OATP1A2 at the human BBB and its ability to act as a facilitator of drug CNS delivery have been controversial.^11–14

As a step towards resolving these knowledge gaps, the goals of this study were: 1) to determine the absolute protein expression levels of major transporters in the BBB of glioblastomas, human and rodent brains, as well as in cell membranes of in vitro cell lines, by using liquid chromatography with tandem mass spectrometry (LC-MS/MS) based targeted proteomics; and 2) to better understand functional relevance of efflux and uptake transporters at the BBB, by leveraging in vitro cell permeability experiments and modeling simulation approach.

Materials and Methods

Human brain and glioblastoma specimens

Human glioblastoma specimens from 47 patients were obtained from the Biobank Core of St. Joseph’s Hospital and Medical Center (Protocol #05TS038) and the Department of Neurosurgery at Wayne State University (Protocol #1011009008). Tumor tissue collection was approved by the Institutional Review Boards of each institute, and in accordance with the Declaration of Helsinki. Written informed consents were obtained from individual patients. All tumor tissues were classified as glioblastoma (WHO grade 4 glioma) by a board-certified neuropathologist according to WHO criteria.

Human brain cortex tissue specimens from 30 donors who died naturally or accidentally were provided by the NeuroBioBank Brain and Tissue Repositories of National Institute of Health (Bethesda, MD) under an Agreement of Material Transfer. As no apparent neuropathological abnormality was identified, these specimens were considered as human normal brain tissues. Detailed information regarding demographics, health status, and cause of death for these donors is provided in Table S1.

Rat and mouse brain tissue specimens

Male Fischer 344 rats (5–8 weeks) and Balb/c mice (6–8 week) were purchased from Charles River Laboratories (Wilmington, MA). Animals were maintained on a 12 h light/dark cycle with free access to food and water. Animals were euthanized using CO₂, and whole brains were resected immediately. Cortex and non-cortex brain tissues were separated, snap-frozen, and stored at −80°C until isolation of brain microvessels. The animal protocols (# 16-09-139 for rats and # A10-01-15 for mice) were approved by the Institutional Animal Care and Use Committee at Wayne State University.

Cell lines

The parental Madin-Darby canine kidney II (MDCKII) cell line and stable cell lines with overexpression of ABCB1 (named MDCKII-ABCB1) or ABCG2 (named MDCKII-ABCG2) were provided by The Netherlands Cancer Institute (Amsterdam, Netherlands). MDCKII cell lines with stable expression of OATP1A2 or co-expression of ABCB1 and OATP1A were established and characterized in our lab, as described in Supplementary Materials. All cell lines were maintained in DMEM GlutaMAX (Gibco 10566) supplemented with 10% FBS and 1% penicillin/streptomycin.

Liquid chromatography with tandem mass spectrometry (LC-MS/MS) based quantitative targeted proteomics

We developed and optimized a LC-MS/MS based quantitative targeted proteomics method to quantitatively determine protein expression levels of major BBB transporters and markers in isolated brain microvessels and cellular membranes of in vitro cultured cells. Details on experimental procedures for the isolation of brain microvessels and cellular membranes, trypsin digestion, and LC-MS/MS analysis are provided in the Supplementary Materials. Table S2 summarizes the peptide probes and respective isotope-labeled internal standards that were used for determining efflux transporters (ABCB1, ABCG2, ABCC4) and solute carrier transporters (OATP1A2, OATP2B1, OAT3, GLUT3, SLC7A5/LAT1, and SLC16A1/MCT1), as well as key markers including GLUT1 (brain endothelial cell marker), Na⁺/K⁺-ATPase (plasma membrane maker), and claudin-5 (BBB tight junction marker). Transporter protein abundance was expressed as fmol per mg of microvessel protein or cellular membrane protein.

In vitro permeability experiments

Cellular permeability experiments for fexofenadine (a typical dual substrate for ABCB1 and OATP1A2) were performed using transwell systems with cell monolayers of parental MDCKII, MDCKII overexpressing ABCB1, and MDCKII co-expressing ABCB1 and OATP1A2, as described in Supplementary Materials.¹⁵

Data analysis

Transporter protein abundance data.

Because of non-symmetric distribution of the data, Mann-Whitney U test was used to compare the protein abundances of individual BBB transporters in isolated microvessels of human brain (n = 30) and glioblastoma (n = 47). Kruskal-Wallis test with post-hoc Dunn’s multiple comparison test was used to compare transporter abundances between different species and in vitro cell lines. P value was adjusted to account for multiple comparison, and adjusted P value < 0.05 was considered significantly different. Inter-species or in vitro-in vivo relative expression factor of a particular transporter was estimated as the ratio of median transporter protein abundance. Statistical analyses were performed using GraphPad Prism 8.0.1.

In vitro transcellular permeability data.

The rate of drug transcellular transport was assessed by the apparent permeability (P_app [cm/s]).¹⁵ The extent of drug penetration from the apical to basolateral side of cell monolayer was assessed by basolateral AUC, and the extent of drug cellular accumulation was assessed by cellular AUC, where AUC is the area under the drug concentration-time curve.

Three-compartment cellular model.

Simulations were performed to further illustrate functional relevance of uptake and efflux transporters to drug BBB penetration, using a three-compartment cellular model with NONMEM v7 and R v.3.5.1. The model incorporates bi-directional passive permeability at the apical and basolateral membranes, transporter-mediated active uptake and active efflux at the apical membrane, and transporter-mediated active efflux at the basolateral membrane. Simulations of drug intracellular and basolateral concentration time profiles were performed for drugs with low passive permeability (P_app = 0.5 × 10⁶ cm/s), moderate passive permeability (P_app = 5 × 10⁶ cm/s), and high passive permeability (P_app = 50 × 10⁶ cm/s). Modal equations and simulation parameters are provided in Supplementary Materials.

Results

Determination of BBB transporter protein abundances using quantitative targeted proteomics

LC-MS/MS based quantitative targeted proteomics enables quantitation of target proteins in complex biological matrix by monitoring specific peptide probes for individual target proteins with highly sensitive and specific multiple reaction monitoring (MRM) mode in a tandem mass spectrometer.^12,16,17 This approach has been used for determination of protein abundances of enzymes or transporters in cell or tissue samples.^12,16,18,19 Based on published methods,^12,16,20 we optimized LC-MS/MS conditions for separating and determining selected peptide probes. By using scheduled MRM mode and optimized mobile phase gradient, our LC-MS/MS method achieved equally good or better sensitivity and linear dynamic ranges while requiring half of running time, as compared to the published method.^12,20 Table S3 summarizes the lower limit of quantitation, calibration curve range, and linearity for the peptide probes of individual transporters and markers. As presented in Table S4, the overall, intra-day, and inter-day precision and accuracy of individual peptide probes were within the generally accepted criteria for bioanalytical method (< 15%). The reproducibility of the method was further evaluated by determining transporter abundances in the positive control samples (i.e., cellular membranes isolated from MDCKII cells with stable co-expression of ABCB1 and OATP1A2) in 8 different batches (days) of sample analyses (with triplicates in each batch). The overall precision, assessed by the coefficient variations (CVs) of all 24 measurements from 8 batches were less than 15% for individual transporters (including ABCB1, OATP1A2, GLUT1, LAT1, and Na/K ATPase), except for MCT1 with an overall CV of 20.8% (Figure S2), indicating a good reproducibility of the method. In addition, in each batch of sample analyses, we included both normal brain and glioblastoma samples, which effectively prevented potential statistics bias caused by inter-batch assay variation (if any).

We optimized sample preparation procedures to ensure the integrity and purity of isolated microvessels, solubilization of enriched microvessels, and sufficient trypsin digestion, which are critical to reproducible and reliable transporter quantitation.^21,22 The recovery (mean ± standard deviation) of microvessels isolated from glioblastoma, human brain, mouse brain, and rat brain cortex specimens were 309 ± 179 (n = 47), 190 ± 74 (n = 30), 89 ± 41 (n = 10), and 134 ± 87 (n = 10) μg protein per gram tissue, respectively, which were in line with published data.¹² Notably, the recovery of microvessels from glioblastoma was higher than that from normal human brain, supporting the notion that glioblastoma is one of the most angiogenic malignant tumors (which presents more, but dysfunctional blood vessels). The morphology and purity of isolated brain microvessels was demonstrated by microscopy image (Figure 1). Sufficient solubilization of enriched microvessels and trypsin digestion was confirmed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), as indicated by no bands over 20 KDa in the trypsin digested samples (Figure S1). It should be mentioned that different sample preparation procedures for microvessel isolation, cell membrane enrichment, protein solubilization, or trypsin digestion might attribute to the variability in targeted proteomic quantitation of transporters and enzymes.^23–25 Caution is needed when comparing and applying quantitative proteomics data from different laboratories. In our study, standardized sample preparation procedures were used for all samples (including human glioblastoma, human brain, mouse and rat brain, and cell samples). This allowed a better comparison of transporter abundances at the BBB of human brain and glioblastoma and also generation of reliable transporter relative expression factors for inter-species or in vitro-in vivo extrapolation of transporter-mediated clearance at the BBB.

Figure 1.

Microscopy image of microvessels isolated from (A) human glioblastoma, (B) human brain, (C) mouse brain cortex, and (D) rat brain cortex specimen by the nylon mesh method. 10x magnification, scale bar = 200 μm.

Table 1 summarizes the absolute protein abundances of major transporters and markers in microvessels isolated from human normal brain (n=30), glioblastoma (n=47), mouse brain (n=10), rat brain cortex and non-cortex (n=10), as well as in cell membranes of MDCKII cell lines. In isolated microvessels of 30 human brain specimens, the median protein abundances for ABCB1, ABCG2, GLUT1, GLUT3, LAT1, MCT1, Na/K ATPase, and Claudin-5 were 3.38, 6.21, 54.51, 7.17, 3.42, 5.71, 32.14, and 1.15 fmol/μg protein, respectively (Table 1). While a larger inter-individual variability was noted, these data were generally in agreement with published data in human brain microvessels of 7 donors and 5 patients with epilepsy or gliomas.^12,26 As compared to those in human normal brain microvessels, in isolated microvessels of glioblastoma, protein abundances of ABCB1, ABCG2, MCT1, GLUT1, Na/K ATPase, and Claudin-5 significantly reduced, while LAT1 significantly increased and GLUT3 remained the same (Figure 2). Specifically, in glioblastoma microvessels; the median ABCB1 and ABCG2 protein levels were reduced to 3.7% and 35% of the normal levels, respectively; the median GLUT1 and Na/K ATPase were reduced to ~ 50% of the normal levels; and claudin-5 was lost or significantly reduced. These data collectively suggested that the physical and biochemical barrier function of the BBB of human glioblastomas was largely disrupted. Notably, the protein expression levels of ABCC4, OATP1A2, OATP2B1, and OAT3 were below the lower limit of quantitation in isolated microvessels of both human brain and glioblastoma specimens.

Table 1.

Protein abundances (pmol/mg) of major transporters and markers in human glioblastoma microvessels, human brain microvessels, mouse brain microvessels, and rat brain microvessels, as well as in MDCKII cell membranes.

BBB Transporters	Human glioblastomaa (n=47)	Human braina (n=30)	Mouse braina (n=10)	Rat brain cortexa (n=10)	Rat brain non-cortexa (n=10)	MDCKII-ABCB1/OATP1A2b (n=4)	MDCKII-ABCB1b (n=4)	MDCKII-ABCG2b (n=4)
ABCB1	0.14 (BLQ-2.87)	3.38 (1.00–7.42)	8.57 (BLQ-12.82)	21.95 (19.63–25.93)	22.22 (17.16–29.73)	11.35 (11.18±1.62)	10.30 (10.30±0.20)	N/A
ABCG2	1.69 (BLQ-12.11)	6.21 (2.23–16.53)	3.84 (0.29–9.16)	4.59 (2.50–4.91)	5.57 (BLQ-7.43)	N/A	N/A	13.18 (12.85±0.94)
ABCC4	BLQ	BLQ	BLQ	BLQ	BLQ	N/A	N/A	N/A
OATP1A2	BLQ	BLQ	N/A	N/A	N/A	7.30 (7.36±0.80)	N/A	N/A
OATP2B1	BLQ	BLQ	N/A	N/A	N/A	0.72 (0.73±0.05)	1.03 (1.01±0.05)	0.96 (0.98±0.13)
OAT3	BLQ	BLQ	N/A	N/A	N/A	N/A	N/A	N/A
GLUT1	14.74 (BLQ-60.73)	54.51 (23.31–116.90)	83.96 (5.72–168.80)	201.59 (156.72–235.74)	157.78 (101.76–233.16)	19.56 (19.09±2.53)	19.34 (19.52±0.48)	22.48 (22.40±0.51)
GLUT3	8.08 (BLQ-36.56)	7.17 (3.59–15.71)	6.57 (1.23–11.98)	BLQ	BLQ	BLQ	BLQ	BLQ
LAT1	5.86 (BLQ-36.02)	3.42 (1.59–10.00)	3.44 (BLQ-4.75)	14.73 (13.82–15.12)	12.89 (8.57–18.65)	16.66 (17.22±2.36)	19.70 (20.34±2.50)	23.19 (23.80±2.30)
MCT1	1.25 (BLQ-16.13)	5.70 (0.62–11.83)	0.74 (0.28–2.01)	BLQ	BLQ	13.15 (13.06±2.62)	16.81 (16.88±0.35)	9.94 (10.10±2.10)
Na/K ATPase	11.64 (0.47–108.80)	32.14 (3.38–81.75)	45.51 (7.55 – 106.89)	22.91 (22.25–37.60)	28.24 (10.56–42.19)	34.19 (34.06±3.64)	31.90 (31.87±1.97)	31.59 (31.74±1.02)
Claudin-5	0.07 (BLQ-1.39)	1.15 (0.06–2.86)	NA	NA	NA	N/A	N/A	N/A

^aData are expressed as the median (range).

^bData are expressed a median (mean ± standard deviation). BLQ, below the lower limit of quantitation

NA, not determined because peptide probe is non-applicable.

Figure 2.

Comparative protein abundances (pmol/mg) of major transporters and markers in human normal brain microvessels (n = 30) and glioblastomas microvessels (n = 47). Symbols represent individual sample measurements. Lines and error bars represent the median and 95% confidence interval. Mann-Whitney U test was used for comparison of two groups.

Inter-species and in vitro-in vivo relative expression factors of individual transporters are summarized in Figure 3 and Table S5. Notably, ABCB1 protein abundance was significantly lower in human brain microvessels than that in microvessels of rat brain (P < 0.0001) and mouse brain (P = 0.0139), while ABCG2 levels were similar between human and rodent brain microvessels (P > 0.9999) (Figure 3). Solute carrier transporters including GLUT1, GLUT3, LAT1, and MCT1 were differentially expressed at the BBB of human, rat, and mouse (Table 1 and Table S5). Interestingly, the protein expression levels of Na/K ATPase (a plasma membrane marker) were similar in brain microvessels of different species (P > 0.9999) or between human brain microvessels and in vitro MDCKII cells (P > 0.9999) (Figure 3).

Figure 3.

Protein abundances (pmol/mg) of major transporters and markers in human brain microvessels (n = 30), rat brain cortex microvessels (n = 10), and mouse brain cortex microvessels (n = 10), as well as in cell membranes of MDCKII cell lines with stable expression of ABCB1 or ABCG2 (n = 4), and co-expression of ABCB1 and OATP1A2 (n = 4 cell samples, repeated measured in 8 batches). Symbols represent individual sample measurements. Lines and error bars represent the median and 95% confidence interval. Numbers are inter-species or in vitro-in vivo relative expression factors, as compared to transporter protein levels in human brain microvessels.

Functional relevance of efflux and uptake transporters at the human BBB: experiment and simulation data

The potential role of uptake transporters in facilitating drug penetration across the BBB is poorly understood due to the lack of appropriate in vitro models. The most commonly available cellular models with overexpression of uptake transporters are HEK293 or HeLa cell lines because of their high transfection efficiency. These cell lines, however, are not appropriate in vitro models for studying BBB permeability because they do not form polarized monolayers with tight junctions. In contrast, MDCKII cell line can form polarized monolayer with well-developed tight junctions after 3–5 day culture, and therefore it is widely used as an in vitro BBB model.²⁷ While MDCKII cell lines with stable expression of efflux transporter such as ABCB1 or ABCG2 or both have been well established and characterized (The Netherland Cancer Institute), MDCKII cell lines with stable expression of an uptake transporter especially with co-expression of efflux and uptake transporters are not commercially available.

OATP1A2 is the only uptake drug transporter that has been generally believed to present at the human BBB^7,11 although its protein expression level was undetectable using targeted proteomics by our group and others¹². We established MDCKII cell lines with stable expression of OATP1A2 or co-expression of ABCB1 and OATP1A2 as in vitro cellular models to study functional roles of uptake and efflux transporters at the human BBB. As shown in Figure 4, OATP1A2 protein expression in selected single clones was demonstrated by immunoblotting analysis; the functional activity was confirmed by cellular uptake assay with estrone 3-sulfate (a typical substrate of OATP1A2); and immunofluorescence confocal microscopy suggested OATP1A2 was largely co-localized with ABCB1 in MDCKII cells co-expressing ABCB1 and OATP1A2.

Figure 4.

Characterization of the established MDCKII cell line with stable co-expression of ABCB1 and OATP1A2. (A) Immunoblotting analysis of OATP1A2-V5/HIS protein expression in the control (MDCKII parental cell line) and selected OATP1A2-transfected clones. (B) Cellular uptake assay of estrone 3-sulfate (a typical OATP1A2 substrate) to assess OATP1A2 functional activity in the control and selected OATP1A2-transfected clones. Numbers show the fold-change of estrone 3-sufate cellular uptake in the absence of an OATP1A2 inhibitor (100 μM sulfobromophthalein) as compared to in the presence of the inhibitor. (C) Immunofluorescence confocal microscopy shows the protein expression and cellular location of OATP1A2 and ABCB1 in MDCKII parental cells and MDCKII cell lines with stable expression of ABCB1 and/or OATP1A2.

Using the established MDCKII cell lines, we assessed the functional role of ABCB1 and OATP1A2 in the intracellular accumulation and transcellular permeability of fexofenadine, a typical dual substrate for ABCB1 and OATP1A2. In the presence of ABCB1 alone (i.e., MDCKII-ABCB1 cells or MDCKII-ABCB1/OAPT1A2 cells with OATP1A2 inhibition), fexofenadine cellular accumulation and apical-to-basolateral transport were consistently reduced by ~ 40–50%, as compared to those in the control (i.e., parental MDCKII or MDCKII-ABCB1/OATP1A2 cells with inhibition of both ABCB1 and OATP1A2) (Figure 5). This observation was consistent with the known luminal location and efflux function of ABCB, and further supported the functional role of ABCB1 in restricting the blood-to-brain penetration of substrate drugs. When OATP1A2 was co-expressed with ABCB1 (i.e., MDCKII-ABCB1/OATP1A2 cells), fexofenadine apical-to-basolateral transport, assessed by basolateral AUC and apical-to-basolateral apparent permeability (P_app), increased by ~30% as compared to those in cells expressing ABCB1 only, suggesting OATP1A2 marginally counteracted ABCB1 efflux function (Figure 5). Nevertheless, the cellular accumulation and apical-to-basolateral transport of fexofenadine in the presence of both ABCB1 and OATP1A2 were still lower than that in the absence of both ABCB1 and OATP1A2, possibly due to a higher efflux efficiency of ABCB1. Interestingly, when ABCB1 was inhibited in MDCKII co-expressing ABCB1 and OATP1A2 cells, the presence of OATP1A2 doubled fexofenadine cellular accumulation (AUC) while having no apparent impact on the apical-to-basolateral transport (either basolateral AUC or apical-to-basolateral P_app) as compared to the control (Figure 5).

Figure 5.

Functional impact of ABCB1 and OATP1A2 on the intracellular accumulation and apical-to-basolateral transport of fexofenadine, a typical dual substrate of ABCB1 and OATP1A2. (A) Fexofenadine intracellular and basolateral concentration time profiles following incubation of MDCKII cell monolayers with 10 μM fexofenadine in the apical chamber for 15–240 min at 37°C, in the absence or presence of specific transporter inhibitor(s) (i.e., 0.5 μM elacridar for inhibition of ABCB1 and 100 μM sulfobromophthalein for inhibition of OATP1A2). Each symbol represents the mean of triplicate measurements in a representative experiment, with coefficient variation < 20%. (B) The extent of fexofenadine intracellular accumulation (cellular AUC) and apical-to-basolateral transport (basolateral AUC). Numbers represent AUC ratios relative to the control (MDCKII parental cell line). (C) Fexofenadine apical-to-basolateral apparent permeability, measured at 90-min incubation. Numbers represent permeability ratios relative to the control (MDCKII parental cell line).

To further illustrate functional interplays of efflux and uptake transporters in regulating drug transport across the BBB, we developed a 3-compartment cellular model (Figure 6A) to simulate time course of drug cellular accumulation and apical-to-basolateral transport at various scenarios. For a drug with low passive permeability (e.g., P_app = 0.5 × 10⁻⁶ cm/s), the presence of an apical uptake transporter considerably enhanced drug cellular accumulation, but had no apparent impact on basolateral drug concentration; a basolateral efflux mechanism is required for effective egress of intracellular drug into the basolateral compartment (Figure 6B and 6C). For a drug with moderate passive permeability (e.g., P_app = 5 × 10⁻⁶ cm/s), an apical uptake transporter facilitated drug cellular accumulation, and slightly increased basolateral drug concentration; the presence of a basolateral efflux transporter further promoted apical-to-basolateral drug transporter (Figure 6D and 6E). For a drug with high passive permeability (e.g., P_app = 50 × 10⁻⁶ cm/s), an apical uptake transporter enhanced drug cellular accumulation and basolateral drug concentration to a similar extent; a basolateral efflux transporter was not required for egress of intracellular drug to basolateral compartment since passive permeability was no longer the rate-limiting process (Figure 6F and 6G).

Figure 6.

Modeling simulations illustrate functional relevance of uptake and efflux transporters to drug BBB penetration. (A) Model structure of the 3-compartment cellular model, which incorporates bi-directional passive permeability (PSB) at the apical and basolateral membranes, transporter-mediated active uptake (C_Luptake) and active efflux (CL_efflux) at the apical membrane, and transporter-mediated active efflux (CL_BL,efflux) at the basolateral membrane. Simulations of drug intracellular and basolateral concentration time profiles were performed for drugs with low passive permeability (P_app = 0.5 × 10⁶ cm/s) (B and C), moderate passive permeability (P_app = 5 × 10⁶ cm/s) (D and E), and high passive permeability (P_app = 50 × 10⁶ cm/s) (F and G). In all simulations, K_m was fixed at 5 pmol/μL; V₁, V₂ and V₃ were fixed at 100, 13, and 220 μL respectively; fu₁, fu₂ and fu₃ were fixed at 1, 0.7 and 1 respectively; PSB is P_app × Surface area (i.e., 0.143 cm² for 96-well transwell membrane). In Simulation 1 to 7, V_max,efflux was set as 5, 5, 5, 5, 0.1, 0.1, and 0 pmol/min, respectively; V_max,uptake was set as 5, 5, 0.1, 0.1, 5, 5, and 0 pmol/min, respectively; V_max,BL was set as 0, 1, 0, 1, 0, 1, and 0 pmol/min, respectively.

Discussion

The absolute protein abundance data of transporters at the human BBB and blood-brain tumor barrier is essential to mechanistic and quantitative understanding of heterogeneous drug penetration into human normal brain and brain tumors. These data, however, are very limited mainly because of the difficulty to obtaining human brain/brain tumor specimens and the technical challenge of LC-MS/NS based targeted proteomics. Indeed, currently available data on the absolute transporter abundances in isolated human brain microvessels, which were determined by targeted proteomics, were obtained from only 7 donors¹² and 5 patients with epilepsy or gliomas²⁶, while there is no data on the absolute transporter abundances in isolated microvessels of human glioblastomas. Our study provided the first largest set of comparative quantitation data on the absolute protein abundances of major transporters and markers in human normal brain microvessels (N = 30) and glioblastoma microvessels (N = 47).

The physical and biochemical barriers of the BBB in glioblastomas were largely disrupted, as indicated by the loss or significant reduction in protein expression of claudin-5, GLUT1, Na/K ATPase, as well as the major efflux transporters (ing ABCB1 and ABCG2) in glioblastoma microvessels as compared to normal brain microvessels (Figure 2). Absence or significantly lower expression of claudin-5 (a tight junction marker) was consistent with the notion of leaky tight junctions in glioblastomas. Reduced protein expression levels of GLUT1 (a brain endothelial cell marker) and Na/K ATPase (a plasma membrane marker) may reflect the disorganized, abnormal, and dysfunctional neovascularity in glioblastomas.² The absence or markedly reduced ABCB1 and ABCG2 protein expression in glioblastoma microvessels indicated a disturbance of the integrity of the biochemical barrier with respect to efflux transporter expression. The disruption of the physical and biochemical barrier function of the BBB in glioblastomas can lead to increased tumor penetration of therapeutic drugs that are the substrates of ABCB1/ABCG2. For instance, AZD1775, a Wee1 inhibitor, exhibited good tumor penetration in glioblastoma patients (with median unbound drug tumor-to-plasma ratio of 3.2) although it was a good substrate for ABCB1 and ABCG2.¹⁵ Differential transporter protein abundances in the BBB of human brain and glioblastomas provided not only mechanistic insights but also quantitative basis for prediction of heterogeneous drug penetration into human brain and glioblastomas.

ABCB1 and ABCG2 are two predominant efflux transporters at the BBB of human and rodents. ABCB1 protein level was significantly lower at the human BBB than rodent BBB, while ABCG2 levels were similar between human and rodent BBB (Figure 3). Species difference in ABCB1 expression provided mechanistic explanations for observed species differences in the brain penetration of ABCB1 substrate drugs. For example, the brain-to-plasma ratio of ABCB1 substrates (including verapamil, GR205171 and altanserin) were several fold greater in humans as compared to rodents.²⁸ AZD1775, a good ABCB1 substrate, exhibited good brain and tumor penetration in glioblastoma patients, while it poorly penetrated into mouse brain.^15,29 Given the less functional importance of ABCB1 at human BBB, it is plausible that some ABCB1 substrates may retain sufficient BBB permeability in humans though they show poor brain penetration in preclinical models. Our data on transporter abundances in the BBB of human and rodents as well as in vitro cell models provided critical quantitative basis for inter-species and in vitro-in vivo extrapolation of BBB transporter-mediated active clearance.

Given the important role of ABCB1 in limiting drug brain penetration, efforts have been focused on circumventing BBB efflux by ABCB1 inhibitors. Nevertheless, co-administration of ABCB1 inhibitors with chemotherapeutic agents in clinical trials provided limited success in improving drug CNS penetration, while resulting in intolerable toxicities and undesirable systemic pharmacokinetic interactions.⁹ The limited success in improving CNS drug permeability by ABCB1 inhibitors has been primarily explained by the inability to achieve unbound systemic inhibitor concentrations sufficient to elicit appreciable inhibition in humans.⁹ In addition, it should be emphasized that ABCB1 and ABCG2 transporters act synergistically to restrict the brain penetration of dual ABCB1/ABCG2 substrate substrates.^30–32 Our data show that ABCG2 protein expression level at human BBB is ~ 2-fold greater than ABCB1 expression, suggesting that ABCG2 may play a more important role in limiting drug brain penetration in humans than previously thought. As such, inhibition of ABCB1 alone would unlikely be a successful strategy for improving CNS delivery of therapeutic drugs, many of which are dual substrates of ABCB1 and ABCG2.

The multidrug resistance-associated protein family (ABCC1–ABCC6, formally MRP1–MRP6) may play a role in limiting drug CNS delivery.³³ However, the protein expression, location, or functional role of ABCC transporters at the BBB of human or rodent are controversy due to the use of non-specific antibodies in immunoblotting assays, mRNA expression profiling in poorly purified brain microvessels, or the use of non-specific substrates or inhibitors in functional assays. Quantitative targeted proteomics indicated that only ABCC4 was expressed at quantifiable but low levels (< 5% of ABCG2 level) in isolated microvessels of human or rodent brain, while ABCC1,2, 3, 5, and 6 were undetectable.^12,26,34 Immunofluorescence microscopy analysis of protein expression of ABCC1–ABCC6 in human glioma specimens of different histologic subtypes suggested that ABCC1, ABCC3, or ABCC6 proteins were not detected in any human gliomas specimens; ABCC4 and ABCC5 proteins were detected in the endothelial cells and glioma cells of astrocytic tumors but not in the endothelial cells of glioblastomas or oligodendrogliomas.⁷ In our study, ABCC4 expression was under the lower limit of quantitation (< 0.1 fmol/μg) in isolated microvessels of all human brain and glioblastoma specimens. Collectively, the absent or very low protein expression levels of ABCC4 and other ABCC transporters in human brain microvessels and glioblastoma microvessels, as demonstrated by our and others’ work, indicate a limited role of ABCC transporters in controlling drug penetration across the BBB.

OATP1A2 is the only human OATP isoform whose expression at the human BBB is generally accepted. Using immunofluorescence microscopy, OATP1A2 protein was identified at the luminal membrane of human BBB,¹⁴ and OATP1A2 and OATP2B1 were detectable in endothelial cells of human gliomas.⁷ However, using quantitative targeted proteomics by others and our group, OATP1A2 or OATP2B1 protein levels were below the lower limit of quantitation (< 0.1 fmol/μg) in isolated microvessels of either human brain or glioblastoma specimens (Table 1).¹² In contrast, a high protein expression level of OATP1A2 (7.30 fmol/μg cell membrane) was detected in the positive control, MDCKII cells with co-expression of OATP1A2 and ABCB1 (Table 1). Thus, unquantifiable OATP1A2 protein levels in human brain microvessels by the sensitive, specific targeted proteomics approach likely indicated absent or very low protein expression of OATP1A2 at the human BBB. However, we could not exclude a small possibility that OATP1A2 protein in brain microvessels underwent some unknown posttranslational modifications, which could result in failure to detect the targeted probe peptides although two specific probe peptides for OATP1A2 (with the amino acid sequence of EGLETNADIIK and IYDSTTFR respectively) were used in our method.

Assuming the presence of OATP1A2 at the human BBB and given its ability to transport structurally and therapeutically diverse drugs, targeting OATP1A2-mediated drug influx has been advocated as a strategy for improving drug CNS delivery.¹⁰ However, the ability of human OATP1A2 (or rodent Oatp1a4) to act as a facilitator for drug delivery from the blood to brain remains controversial.^13,14 The common thoughts that OATP1A2 can mediate transport of substrate drugs across the BBB are largely derived from in vitro drug uptake experiments using oocytes, HEK293, or Hela cells. These cells, however, are not the appropriate in vitro cellular models for studying BBB permeability because they do not form polarized cell monolayers with well-developed tight junctions. Instead, MDCKII cell monolayer system is a widely accepted in vitro BBB model. Using MDCKII cells with stable expression of ABCB1 and/or OATP1A2, we demonstrated that OATP1A2 significantly enhanced cellular accumulation of fexofenadine (a model drug for ABCB1 and OATP1A2 substrates), but did not increase apical-to-basolateral drug transport (Figure 5). These in vitro observations could be translated into in vivo situations, suggesting while OATP1A2 can enhance the accumulation of a substrate drug within the BBB endothelial cells, but may not necessarily facilitate drug penetration from the blood to brain. This speculation is supported by clinical evidence that fexofenadine, as a third-generation non-sedative antihistamine drug, does not cause CNS side effects because of its poor penetration into human brain.³⁵ Furthermore, simulations with a 3-compartment model suggest that while an apical efflux transporter plays a key role in restricting drug BBB penetration, the functional relevance of an apical uptake transporter is dependent on drug transmembrane passive permeability (Figure 6). For drugs with poor transmembrane passive permeability (including a majority of OATP1A2 substrate drugs), not a single apical uptake transporter, but a vectorial transport system consisting of apical uptake transporter(s) and basolateral efflux mechanism at the BBB, is required for efficient drug delivery from the blood to brain. Our experiment and simulation data collectively suggest that targeting OATP1A2-medicated uptake at the BBB would unlikely be an effective strategy for improving drug CNS delivery.

Study Highlights.

• What is the current knowledge on the topic?

  Prediction of drug penetration across human BBB is critical to rational drug development and treatment for brain cancer especially glioblastoma. However, the prediction has been largely hindered due to the lack of quantitation data on transporter protein expression levels at human BBB. Accurate prediction is further compromised by the incomplete understanding of BBB transporter functional interplays, which also results in uncertainties and thus unlikely success on developing BBB transporter-targeted strategies for improving drug brain delivery.

• What question did this study address?

  This study was to determine BBB transporter protein expression levels and functional relevance.

• What does this study add to our knowledge?

  This study provides the first largest set of quantitation data on transporter protein expression levels at the BBB of human brain and glioblastomas, and sheds mechanistic insights into BBB transporter functional interplays.

• How might this change clinical pharmacology or translational science?

  Obtained knowledge is critical to better prediction of drug BBB penetration and rational design of therapeutic strategies for efficient delivery of drugs into human brain and brain tumors.