Sage Journals: Discover world-class research

Abstract

The blood–brain barrier (BBB) is a multicellular vascular structure separating blood from the brain parenchyma that is composed of endothelial cells with tight intercellular junctions, surrounded by a basal lamina, astrocytes, and pericytes. Previous studies have generated detailed databases of the microvessel transcriptome; however, less information is available on the BBB at the protein level. In this study, we specifically focused on characterization of the membrane fraction of cells within the BBB to generate a more complete understanding of membrane transporters, tight junction proteins, and associated extracellular matrix proteins that are functional hallmarks of the BBB. We used Multidimensional Protein Identification Technology to identify a total of 1,143 proteins in mouse brain microvessels, of which 53% were determined to be membrane associated. Analyses of specific classes of BBB-associated proteins in the context of recent transcriptome reports provide a unique database to assess the relative contribution of genes at the level of both RNA and protein in the maintenance of normal BBB integrity.

Keywords

astrocyte basal lamina blood–brain barrier endothelial cell proteomics vascular permeability

Introduction

The blood–brain barrier (BBB) is a multicellular structure that primarily consists of endothelial cells and astrocyte endfeet separated by a basal lamina (Abbott et al, 2006). The hallmark of this multicellular structure, based on anatomic and physiologic studies, is the neurovascular unit in which multiple cell types (i.e., astrocytes, pericytes, and neurons) have been identified in association with the vascular endothelium depending on the location within the brain (Abbott et al, 2006; Enerson and Drewes, 2006; Wolburg et al, 2009). The function of these specialized endothelial-associated cells, their secreted proteins, the tight interendothelial barrier itself, the unique transcellular transport, and active efflux pumps combine to yield a unique vascular barrier (Agarwal et al, 2009; Daneman et al, 2010; Enerson and Drewes, 2006). Breakdown of the BBB is associated with many different pathologies of the brain, including stroke and brain tumor invasion, leading to changes in the expression of intracellular and extracellular proteins that mediate integrity (Pottiez et al, 2009a). This study focuses on the identification of BBB protein components, especially membrane and extracellular components, at the protein level from freshly isolated microvessels. This procedure eliminates potential fixation, cell culture, and cell suspension artifacts, thus obtaining the most complete resource of BBB-associated proteins to date.

Previous studies have examined the gene and protein expression profiles of cultured brain endothelial cells, which have yielded important expression profile data cataloging gene families using in vitro models (Agarwal and Shusta, 2009; Haseloff et al, 2006; Shusta et al, 2002). However, cell culture itself influences the relative abundance of various BBB cell subtypes, gene expression profiles, and/or the extracellular matrix (ECM) microenvironment. Microdissection strategies to recover microvessels from brain tissues have yielded useful, comparative information on protein expression profiles from an intact BBB. However, this approach requires fixation, laser dissection, immunostaining, and subsequent gel electrophoresis analysis. Therefore, we focused on fresh microvessel isolation to characterize the BBB proteome from mouse brains and then validated the isolation technique using immunoblotting to establish quality control.

On the basis of our interest in developing a resource identifying ECM components and BBB membrane proteins that contribute to ECM interactions, we focused on a membrane-enrichment strategy for mass spectrometry-based protein profiling. Historically, BBB protein analyses have relied on various gel electrophoretic-based techniques for protein separation, followed by antibody or mass spectrometry for protein identification. However, protein separation is limited because of the insolubility of membrane/ECM proteins and gel electrophoretic resolution. The use of Multidimensional Protein Identification Technology (MudPIT) enables the discrete identification of a large number of proteins from complex protein mixtures obtained from freshly isolated tissues and does not require fixed/frozen samples or single cell suspension-based isolation techniques, i.e., flow cytometry or magnetic beads (Aebersold and Mann, 2003; Pottiez et al, 2009a; Washburn et al, 2001; Wu et al, 2003). Moreover, a recent study has used mass spectrometry with the addition of isotope-labeled peptide standards to provide quantitation for known proteins (Uchida et al, 2011).

In our study, freshly isolated normal mouse brain microvessels were disrupted, membranes fractionated, which then underwent proteolysis, and was followed by protein profiling using MudPIT. We have performed multiple MudPIT runs to maximize coverage of membrane proteins of low and high abundance and to establish standards for technical replicates within the same samples and between different tissue collections. Identified proteins were assigned to various BBB classes, based on classical markers for specific cell types, as well as recent proteomic and transcriptome studies (Daneman et al, 2010; Enerson and Drewes, 2006). Our strategy enabled the identification and classification of BBB membrane transporters, tight junction proteins, integrins, ECM proteins, and proteoglycans. We discuss the classes of proteins identified and their relative abundance for the purpose of providing new information on membrane protein constituents of the normal BBB.

Materials and methods

Mouse Brain Microvessel Isolation

Mouse brain microvessels were isolated using a modified protocol based on previously published studies (Hartz et al, 2006; Yousif et al, 2007). All animal studies were IACUC (Institutional Animal Care and Use Committee) approved by the University of California San Diego and The Scripps Research Institute. For each preparation, mice (n=6) were overdosed with 5% isoflurane and their brains harvested. All subsequent steps were carried out at 4°C or on ice. After a wash in phosphate-buffered saline, the brains were homogenized using a dounce (0.25 mm clearance) (whole brain is defined as the S0 fraction; the nomenclature of S0, S1, and S5 describes the microvessel fractionation steps consistent with the microvessel isolation procedure as described below and previously by Yousif et al (2007). Overall, 15 mL of 30% Ficoll was added to 10 mL of the homogenate and mixed thoroughly. The resulting density gradient was spun at 5,800 × g for 20 minutes (the pellet was defined as S1 fraction). The pellet was resuspended in 1 mL phosphate-buffered saline with 1% bovine serum albumin and passed through a glass bead column, with 100-μm nylon mesh filter on the top and a 20-μm nylon mesh filter at the bottom. The glass beads were gently agitated in phosphate-buffered saline with 1% bovine serum albumin to obtain microvessels. The resulting sample was washed with bovine serum albumin-free phosphate-buffered saline before use in experiments, and defined as the S5 fraction.

Endothelial Cell Lysis

At 4°C, ∼40 μL of cells were suspended in 0.5 mL of relaxation buffer (pH 7.25, 50 mmol/L KCl, 1.5 mmol/L NaCl, 5 mmol/L PIPES, 1.75 mmol/L MgCl₂, 1 mmol/L ATP, 0.62 mmol/L EGTA). Cell suspension was placed in a 15 mL conical tube, cavitation bomb tubing attached, and nitrogen pressure (500 to 700 psi) applied for 10 minutes. After placing a clean 15-mL receiving tube at the end of the pressure-relief tubing, the pressure-relief valve was gently opened to collect lysed cells. The tubing was then purged to collect an additional 0.5 mL of relaxation buffer and pooled with the lysate.

Endothelial Cell Lysate Fractionation and Protein Digest

At 4°C, the lysate was placed in a microfuge tube and spun at 750 × g for 5 minutes. The supernatant was placed in a 5 mL ultracentrifuge tube with relaxation buffer to a total volume of 4.8 mL, and centrifuged at 100,000 × g for 90 minutes. The supernatant, containing cytosolic proteins, was discarded, whereas the pellet, containing membrane proteins, was resuspended in 15 μL 8 mol/L urea followed by addition of 20 μL 0.2% ProteaseMAX acid-labile surfactant (Promega Corporation, Madison, WI, USA) in 50 mmol/L ammonium bicarbonate. Aliquots of resuspended membrane proteins were analyzed for total protein content by Bradford protein assay (Bio-Rad Corporation, Hercules, CA, USA). Proteins were then reduced with dithiothreitol and carbamidomethylated with iodoacetamide before tryptic digestion according to the ProteaseMAX technical bulletin (ProteaseMAX surfactant, Trypsin Enhancer; no. TB373; Promega Corporation). Digestion was stopped by addition of 90% formic acid to obtain a final in-digest concentration of 5% formic acid.

Multidimensional Liquid Chromatography: Mass Spectrometry (Multidimensional Protein Identification Technology)

Samples in 1.5 mL microfuge tubes were diluted 1:2 with 5% acetonitrile 0.1% formic acid (buffer A), centrifuged at 13,000 × g for 10 minutes, and supernatant proportionate to 15 μg of protein was transferred to a new microfuge tube for loading to a MudPIT precolumn. The MudPIT precolumn (250 μm internal diameter) with 4 cm of strong cation exchange 5 μm resin (Partisphere SCX, GE Healthcare, Piscataway, NJ, USA) was preceded by 3 cm of C18 stationary-phase column (5 μm particle size, Jupiter C18, Phenomenex Corporation, Torrance, CA, USA). The resolving column dimensions were 100 μm (internal diameter) with 10 cm of C18 stationary phase. Operating parameters of the MudPIT experiment are provided in Supplementary Table 1 and Table 2, based on the standards provided by the MIAPE (Minimum Information About a Proteomics Experiment (http://psidev.sourceforge.net/miape/).

Bioinformatic Analyses

Spectra were searched using SEQUEST, and filtered using DTASelect2 (http://fields.scripps.edu). Spectral count data were then processed using PatternLab for Proteomics (Carvalho et al, 2008). The spectral count information stored in the DTAselect-filter.txt files was extracted using the Perl script dtarray2.pl (Kaschani et al, 2009). The resulting dtarray2.txt file was loaded into Excel (Microsoft Corp., Redmond, WA, USA) and further ordered. This arrangement is the source for the master list of protein identifications in Supplementary Table 1. Information for IPI is located at http://www.ebi.ac.uk/IPI/IPIhelp.html. Information on Uniprot Knowledgebase is located at http://www.uniprot.org/help/uniprotkb. For SEQUEST searches followed by DTASelect2 processing, the following criteria were used: (1) a 1% false positive discovery rate; (2) a minimum of two unique tryptic peptides per identification were required; (3) oxidation of methionine (+16Da) was allowed as a differential modification; and (4) cysteines were required to be carbamidomethylated (+57 Da).

Statistical validation was performed for every protein hit using statistical functions and parameters set within the SEQUEST search as we have previously described in detail (Tabb et al, 2002). Additional information on the use of DTASelect2 is available at http://www.scripps.edu/chemphys/cravatt/protomap/dtaselect_instructions.html, along with additional information on data analysis of shotgun proteomics (Cociorva et al, 2006).

Results

Microvessel Isolation and Validation

The cortices of normal mouse brains were dissected, tissues pooled, and subjected to a multistep isolation protocol (see overview in Figure 1A) that has been previously described and used for transcriptome analysis, BBB transporter, and signaling studies (Enerson and Drewes, 2006; Hartz et al, 2006; Yousif et al, 2007). To verify the enrichment of brain microvessel material, we first examined the integrity of the microvessel lumens by staining the vasculature before harvest. Mice were subjected to an intravenous injection of the fluorescently labeled endothelial cell-binding lectin, Griffonia simplicifolia (Figure 1B). After harvest of labeled brains, and purification of the microvessels on a Ficoll gradient (termed ‘S1’ as described by Yousif et al; see the ‘Materials and methods’ section for further details), samples were subjected to imaging by laser scanning confocal microscopy to reveal stained endothelia based on fluorescent lectin binding to the microvessels. Further enrichment of microvessel preparations over the glass bead column led to a reduction of cellular debris and was verified by imaging microvessels by phase microscopy (i.e., S5 fraction as described previously) (Yousif et al, 2007). Nuclear integrity was verified by staining nuclei with DAPI, followed by fluorescent imaging (Figure 1C). Consistent with previous studies (Yousif et al, 2007), microvessels were enriched in the S5 preparation (i.e., microvessels) versus the S1 preparation (i.e., microvessels with cellular debris), as evidenced by the loss of cellular debris in the enriched S5 microvessels. To further confirm the enrichment of microvessels, lysates of whole-brain samples (S0) were compared with enriched microvessels (i.e., S5), in which both were subjected to immunoblotting with antibodies to the endothelial (i.e., P-glycoprotein, Pgp/Abcb1), neuronal (i.e., Synaptophysin, Syn), and astrocyte endfoot (i.e., Aquaporin-4, Aqp4) compartments of the BBB. Although both Pgp and Aqp4 were increased as predicted in microvessels (Figure 1D), indicating a relative enrichment of endothelial cells and vessel-associated astrocyte endfeet, levels of the neuronal marker (Syn) were decreased in microvessels (i.e., S5 fraction) versus the whole brain. These quality control techniques (i.e., lectin staining, microscopy, and immunoblotting) were performed on parallel samples or aliquots of preparations used for proteomic analysis, to ensure the most rapid processing of unstained microvessels for proteomic analysis.

Figure 1

Scheme of brain microvessel enrichment. (A) Flow chart of microvessel enrichment procedure in which S0 is the whole-brain lysate, S1 is the brain pellet after Ficoll centrifugation, and S5 is obtained from purified microvessels. (B) Confocal imaging of brain microvessels after mouse intravenous injection with FITC-lectin followed by microvessel extraction and isolation. Bar=50 μm. (C) Fluorescent and phase imaging of DAPI-stained nuclei from S1 versus S5 microvessel fractions. (D) Immunoblotting of S0 versus S5 fractions with anti-P-glycoprotein (Pgp), anti-synaptophysin (Syn), or anti-aquaporin-4 (Aqp4) antibodies. BSA, bovine serum albumin; FITC, fluorescein isothiocyanate.

Proteomic Profile of Brain Microvessel Membranes

Enriched S5 fraction microvessels were subjected to nitrogen cavitation lysis, membrane isolation, tryptic digestion, and prepared for mass spectrometry analysis as described in the ‘Materials and methods’ section. Digested proteins were separated by liquid chromatography using microcapillary columns followed in-line by tandem mass spectrometry as described previously (i.e., MudPIT (Chen et al (2006)). Using the three distinct, S5 microvessel membrane fractionation samples, independent mass spectrometry runs enabled the identification of 1,645 microvessel proteins, which were operationally termed ‘BBB proteins’. The standards and settings used for data acquisition (Taylor et al, 2007) (i.e., MIAPE data) for these studies are found in Supplementary Table 1 along with a complete list of proteins and their spectral counts from all mass spectrometry runs.

Mass spectrometry is a sampling technique in which probability of detection is a function of protein abundance. By counting the numbers of spectra that map to identified proteins, spectral counting can be used as a surrogate for protein abundance (Liu et al, 2004). The criteria used for protein inclusion for the study was that proteins must be identified by MudPIT in three technical replicates (i.e., the same sample analyzed at three different times: 153-1, 153-2, and 153-3), which was considered as a single biologic sample, and then in an additional two biologic samples (118 and 153). Moreover, the average of the spectral counts from the three technical replicates and the two biologic samples must be 5 (Figure 2 and Supplementary Table 1). In some cases, proteins were found in technical replicates, but not in biologic replicates, as well as the converse (Supplementary Table 1). Each protein sample, 118, 153, and 155, was extracted from isolated, pooled microvessels obtained from different mice.

Figure 2

Information on proteins identified by MudPIT from technical or biologic replicates from the S5 microvessel fraction samples. (A) Shown in a Venn diagram are the numbers of proteins identified by MudPIT analyses common to or distinct from each other in three technical replicates of the S5 microvessel fraction, 153-1, 153-2, and 153-3. Three single samples were prepared and then stored individually, and then analyzed independently by MudPIT. Inclusion criteria for the 1,355 proteins common to all 3 replicates were 5 spectral counts and common identity to all 3 replicates, with 1,833 proteins (common+unique proteins) identified from all samples. Individual protein outliers correspond to 247 (153-1), 160 (153-2), and 121 (153-3) proteins from each replicate. Supplementary Table 1 contains the complete listing of all proteins identified and the number of spectra from each replicate and all S5 microvessel fraction samples. See the ‘Materials and methods’ section for information on mass spectrometry and bioinformatic analyses. (B) A Venn diagram shows the numbers of proteins common to or distinct from each other in three biologic replicates of the S5 microvessel fraction, 118, 154, and 153 (proteins common to 153-1, 153-2, and 153-3, respectively). Each S5 microvessel fraction sample was prepared and then stored as an individual sample before analyses. The criteria used for inclusion of proteins common to 1,149, or distinct from each sample, 78 (118), 150 (154), and 274 (153) were as discussed in panel A, 1,645 proteins (common+unique proteins) were identified from all samples. (C) A comparison of the average of all spectral counts obtained from all mass spectrometry runs was compared with individual spectral counts obtained from each mass spectrometry run. The log-scale regression plot presented the average of all spectral counts for all microvessel membrane fraction 5 analyses as compared with the spectral counts for individual analyses. The legend is on the plot for each technological replicate, 1A, 1B, and 1C, and the two biologic replicates, 2 and 3. R²= 0.98624. (D) The cellular location of proteins identified from microvessel fractions, with numbers of proteins given after the category in the bar chart, for a given subcellular location relative to the total number of location-annotated Uniprot IDs for all five MudPIT analyses of the S5 membrane factions. For each identified protein, the International Protein Index (IPI) (Kersey et al, 2004) identifiers were first mapped to Uniprot identifiers (IDs). The Uniprot Knowledgebase was then used to correlate IDs with subcellular location. Membrane refers to all proteins identified that can be designated as a membrane protein. The percentages do not add to 100%, as many proteins have more than one subcellular location. For additional information, please see the ‘Materials and methods’ section. MudPIT, Multidimensional Protein Identification Technology.

To first determine the intrasample variation of proteins identified by MudPIT, three samples from the same S5 microvessel fraction preparation were analyzed (i.e., technical replicates), and then the distribution of the proteins identified was compared (Figure 2A). For technical replicates (i.e., 153-1, 153-2, and 153-3), we found 1,355 proteins (74% of all proteins identified) common to all three samples, with 247 proteins unique to replicate 153-1, 160 proteins unique to replicate 153-2, and 121 proteins unique to replicate 153-3 (Figure 2A). The distribution of proteins common to any two samples was found to be 102 proteins within 153-1 and 153-2, 93 proteins within 153-2 and 153-3, and 164 proteins within 153-1 and 153-3. Thus, a significant degree of protein overlap exists, 74%, between all replicates from the same starting microvessel preparation.

Next, we assessed the distribution of proteins identified among the three different S5 fraction microvessel samples, 118, 153 (proteins common to 153-1, 153-2, and 153-3), and 155, in which each sample was from pooled tissues of different mice, which we termed ‘biologic replicates’ (Figure 2B). A total of 1,143 proteins (69% of all proteins identified) were common to all three samples, whereas 78 proteins were unique to 118, 274 proteins were unique to 153, and 150 proteins were unique to 155. The distribution of proteins common to any two samples was found to be 67 proteins within 118 and 153-3, 249 proteins within 153-3 and 155, and 58 proteins within 155 and 118. As found with technical replicates, a significant portion of the proteins identified for biologic replicates, 69%, was common to all three samples and was only slightly less than that identified from technical replicates. Moreover, a scatter plot comparison of the proteins identified following the parallel analyses of replicate samples established a significant degree of overlap (Figure 2C).

Finally, Figure 2D, a summary of the bioinformatic analysis of BBB proteins identified from the three samples obtained from different groups of mice (Figure 2B) provides insights as to the number of membrane proteins recovered (53%), as well as mitochondrial (15%) and cell junction proteins (5%) recovered. A complete listing of all proteins identified is found in Supplementary Table 1 and the proteins are grouped in Tables 1, 2, 3 and 4 based on previously published reports and a classical understanding of their function.

Table 1

Transporter and channel proteins

Protein	Gene	Mean	s.d.	Reference
IPI00752412	Atp1a3	1,954.7	211.5	Lu et al, 2008a, 2008b
IPI00420569	Atp1a2	1,249.7	99.1	Daneman et al, 2010^a; Cahoy et al, 2008^b
IPI00311682	Atp1a1	1,135	79.9
IPI00468481	Atp5b	1,033	318
IPI00130280	Atp5a1	849	158.4
IPI00378485	Atp1a4	639.7	48
IPI00115564	Slc25a4	568	50.5	Enerson and Drewes, 2006
IPI00875419	Slc1a2	500	110	Cahoy et al, 2008 ^b
IPI00127841	Slc25a5	397.3	11.9
IPI00857092	Slc4a4	211	13.1	Cahoy et al, 2008 ^b
IPI00308162	Slc25a12	209	53.7
IPI00114279	Slc1a3	208.3	35.3	Uchida et al, 2011
IPI00121550	Atp1b1	184.3	46.2
IPI00119113	Atp6v1b2	178.7	35
IPI00831180	Atp2b2	175	16.1
IPI00313475	Atp5c1	166.3	53.3
IPI00553834	Abcb1a	150	46.7	Enerson and Drewes, 2006; Daneman et al, 2010; Yousif et al, 2007; Kamiie et al, 2008; Uchida et al, 2011^c
IPI00556827	Atp2b1	140.7	19.8
IPI00407692	Atp6v1a	130.7	29.7
IPI00341282	Atp5f1	121.3	35
IPI00407499	Abat	118	41.6
IPI00338964	Atp2a2	118	18.7	Lu et al, 2008a, 2008b
IPI00828950	Atp6v0a1	113	19.3
IPI00751474	Atp2b4	107.7	28.9
IPI00265299	Atp2b3	107	18
IPI00877254	Slc12a5	91	6.1	Enerson and Drewes, 2006
IPI00124771	Slc25a3	88.3	8.7	Lu et al, 2008a, 2008b
IPI00109275	Slc25a22	79.7	16.3
IPI00114641	Slc3a2	75.3	15	Enerson and Drewes, 2006; Daneman et al, 2010; Kamiie et al, 2008; Uchida et al, 2011^c
IPI00313841	Atp6v0d1	69.7	23.7
IPI00230507	Atp5h	58.7	5
IPI00230680	Atp4a	56.3	8.1	Kamiie et al, 2008
IPI00118986	Atp5o	56	9.2
IPI00315999	Atp6v1f	52.7	7.4
IPI00123704	Atp1b2	51	14.9	Daneman et al, 2010 ^a
IPI00230754	Slc25a11	49.3	10.1
IPI00119115	Atp6v1e1	49.3	5.5
IPI00752501	Slc4a10	42	5
IPI00120761	Slc4a1	39	9.2	Lu et al, 2008a, 2008b
IPI00311654	Atp2a1	32	27.9
IPI00308691	Slc2a1	29.7	4.5	Enerson and Drewes, 2006; Daneman et al, 2010; ^cYousif et al, 2007; Kamiie et al, 2008; Uchida et al, 2011
IPI00626793	Cacna2d1	26.3	5.9
IPI00131666	Slc1a6	25.7	3.1	Enerson and Drewes, 2006
IPI00111770	Atp5k	25	7.8
IPI00135324	Slc12a2	21.7	4	Enerson and Drewes, 2006; Daneman et al, 2010^a
IPI00875016	Slc25a18	18.7	2.3	Cahoy et al, 2008 ^b
IPI00227928	Slc6a1	17.7	2.3	Enerson and Drewes, 2006
IPI00322156	Slc38a3	17	7.5	Enerson and Drewes, 2006; Daneman et al, 2010^c
IPI00111686	Abcc4	17	3.6	Enerson and Drewes, 2006, Daneman et al, 2010^c; Kamiie et al, 2008; Uchida et al, 2011
IPI00308063	Slc9a3r2	16.7	4.7	Enerson and Drewes, 2006; Daneman et al, 2010^c
IPI00828762	Abca1	15.7	5.1	Uchida et al, 2011
IPI00311200	Atp2a3	12.7	2.5	Daneman et al, 2010 ^a
IPI00137194	Slc16a1	11.7	4.7	Enerson and Drewes, 2006; Daneman et al, 2010^c; Kamiie et al, 2008; Uchida et al, 2011
IPI00134191	Slc2a3	11.3	2.1	Enerson and Drewes, 2006; Uchida et al, 2011
IPI00109311	Slc9a3r1	11.3	1.5	Enerson and Drewes, 2006
IPI00553576	Abcd3	11	3	Enerson and Drewes, 2006
IPI00109153	Slc17a7	10	4.6	Enerson and Drewes, 2006
IPI00229554	Slc6a17	10	2.6	Daneman et al, 2010 ^c
IPI00468924	Slc25a24	9.3	5.5	Daneman et al, 2010 ^c
IPI00221921	CD9	8	6.6	Uchida et al, 2011
IPI00265568	Slc25a25	7.3	1.5	Daneman et al, 2010 ^c
IPI00120166	Slc30a1	7.3	5.9	Enerson and Drewes, 2006; Daneman et al, 2010^c
IPI00420244	Slc1a1	7	3.6	Daneman et al, 2010 ^c
IPI00468691	Abcg2	6.7	3.2	Enerson and Drewes, 2006; Daneman et al, 2010^c; Yousif et al, 2007; Kamiie et al, 2008; Uchida et al, 2011
IPI00120097	Slc27a1	6.3	1.5	Enerson and Drewes, 2006
IPI00624501	Asna1	6	5.6	Enerson and Drewes, 2006
IPI00121634	Slc7a1	6	5.3	Enerson and Drewes, 2006; Daneman et al, 2010^c; Uchida et al, 2011
IPI00122522	Ggt1	5.3	3.1	Kamiie et al, 2008 ^c
IPI00225534	Atp13a5	4.7	1.5	Daneman et al, 2010 ^a
IPI00622815	Slc9a1	4.3	1.5	Enerson and Drewes, 2006
IPI00125970	Abca7	4	4	Enerson and Drewes, 2006; Uchida et al, 2011
IPI00459383	Abcb8	4	2	Enerson and Drewes, 2006
IPI00136334	Abcc9	3.7	3.2	Enerson and Drewes, 2006; Daneman et al, 2010^a; Armulik et al, 2010
IPI00317074	Slc25a10	3.7	2.1	Enerson and Drewes, 2006
IPI00828826	Slc44a1	3.7	1.5	Daneman et al, 2010 ^c
IPI00553414	Slc5a6	3.3	2.9	Enerson and Drewes, 2006
IPI00131584	Slc25a20	3	1.7	Enerson and Drewes, 2006; Daneman et al, 2010^c
IPI00128358	Insr	3	3.6	Uchida et al, 2011
IPI00858037	Abca9	2.7	3.1	Daneman et al, 2010 ^a
IPI00273801	Slc39a10	2.7	0.6	Daneman et al, 2010 ^c
IPI00750694	Slc39a12	2.7	3.1	Cahoy et al, 2008 ^b
IPI00169896	Slc44a2	2.7	2.3	Daneman et al, 2010 ^c
IPI00470963	Slc4a3	2.3	2.1	Enerson and Drewes, 2006
IPI00120094	Abcb6	2	1.7	Enerson and Drewes, 2006
IPI00112616	Abca2	1.7	1.5	Enerson and Drewes, 2006; Uchida et al, 2011
IPI00331175	Slc12a7	1.7	1.5	Daneman et al, 2010 ^c
IPI00387413	Slc22a8	1.7	1.5	Daneman et al, 2010^c; Kamiie et al, 2008
IPI00742415	Abca3	1.3	2.3	Uchida et al, 2011
IPI00353962	Slc16a2	1.3	1.2	Daneman et al, 2010 ^c
IPI00331577	Slc7a5	1.3	2.3	Daneman et al, 2010^c; Kamiie et al, 2008
IPI00314205	Slc6a20	1	1.7	Enerson and Drewes, 2006
IPI00624726	Abcc8	0.7	1.2	Enerson and Drewes, 2006; Daneman et al, 2010; Uchida et al, 2011
IPI00830169	Atp7a	0.7	1.2	Daneman et al, 2010 ^c
IPI00453817	Slc38a2	0.7	1.2	Enerson and Drewes, 2006; Daneman et al, 2010^c; Kamiie et al, 2008
IPI00109242	Slc7a3	0.7	1.2	Daneman et al, 2010 ^c

Pericyte.

Astrocyte.

Endothelial.

Italics: mean spectral count <5.

Table 2

(A) Tight junction transmembrane and cytoplasmic adapter proteins and integrins, (B) tight junction cytoplasmic adapters, and (C) integrins

Protein	Gene	Mean	s.d.	Reference
(A)
IPI00122971	Ncam1	205	51.2	Lu et al, 2008a, 2008b
IPI00850457	Cadm2	42.7	9.1
IPI00380273	Gja1	35.3	7	Enerson and Drewes, 2006
IPI00121378	Alcam	32.7	7.4	Enerson and Drewes, 2006
IPI00115762	L1cam	31.7	2.5	Lu et al, 2008a, 2008b
IPI00395042	Nrcam	21.7	4.2	Enerson and Drewes, 2006
IPI00127556	Ncam2	21.7	2.5
IPI00471176	Hepacam	21.3	8.5
IPI00856723	Cadm1	19.3	7.1
IPI00857780	Cadm3	19	6.1
IPI00877299	Icam5	18	3.6	Enerson and Drewes, 2006
IPI00229475	Jup	17	1	Enerson and Drewes, 2006; Lu et al, 2008a, 2008b
IPI00473693	Pkp4	16.7	5	Enerson and Drewes, 2006
IPI00749860	Cdh2	14.3	4	Daneman et al, 2010
IPI00875877	CD47	10.7	1.2
IPI00153840	Cadm4	9.7	4.9
IPI00406901	Pecam1	8	3.5
IPI00120295	Cldn11	7.7	1.2	Enerson and Drewes, 2006; Lu et al, 2008a, 2008b
IPI00154057	Pcdh1	6.7	1.5	Cayrol et al, 2008
IPI00117828	Ocln	4.3	2	Daneman et al, 2010; Krishna and Redies, 2009
IPI00775975	Cdh13	4.3	1	Enerson and Drewes, 2006; Daneman et al, 2010
IPI00117424	Icam2	3.3	2	Enerson and Drewes, 2006; Daneman et al, 2010; Krishna and Redies, 2009
IPI00410802	CD44	3	1	Yousif et al, 2007; Cayrol et al, 2008
IPI00138180	Cdh5	2.3	2	Enerson and Drewes, 2006; Daneman et al, 2010; Cayrol et al, 2008; Krishna and Redies, 2009
IPI00111902	Gjb6	2.3	2	Yousif et al, 2007
IPI00138190	Cdh11	2	2	Enerson and Drewes, 2006; Daneman et al, 2010; Cayrol et al, 2008
IPI00755700	Pcdh9	2	2	Cayrol et al, 2008
IPI00469123	Mcam	1.7	2	Enerson and Drewes, 2006; Yousif et al, 2007
IPI00848632	Pcdh10	1.3	2	Cayrol et al, 2008
IPI00122973	Icam1	1.3	1	Krishna and Redies, 2009
IPI00356667	Pcdh17	1.3	1	Cayrol et al, 2008
IPI00346482	Cdh10	1	2	Daneman et al, 2010
IPI00775835	Cdh6	1	2	Daneman et al, 2010; Cayrol et al, 2008
IPI00119893	Cldn5	1	2	Enerson and Drewes, 2006; Daneman et al, 2010
IPI00109330	Cdh15	0.7	1	Daneman et al, 2010
IPI00622727	Jam2	0.7	1	Daneman et al, 2010
IPI00653571	Vcam1	0.7	1	Yousif et al, 2007
(B)
IPI00125899	Ctnnb1	55	4.4	Daneman et al, 2010; Yousif et al, 2007
IPI00119870	Ctnna2	49	3
IPI00112963	Ctnna1	43.3	7.6
IPI00874522	Tjp1	24.3	10.1	Enerson and Drewes, 2006; Daneman et al, 2010; Lu et al, 2008a, 2008b
IPI00136135	Ctnnd2	22.3	3.8
IPI00125861	Dlg1	20	3	Daneman et al, 2010
IPI00752108	Ctnnd1	16.7	4.2
IPI00126827	Esam1	9.7	5.5	Enerson and Drewes, 2006; Daneman et al, 2010
IPI00136498	Lin7c	9	1.7	Daneman et al, 2010
IPI00137706	Mpp1	8.7	3.1	Daneman et al, 2010
IPI00323349	Tjp2	6	2.6	Enerson and Drewes, 2006; Daneman et al, 2010
IPI00124051	Mpp5	6	2	Daneman et al, 2010
IPI00121362	F11r	5	3	Enerson and Drewes, 2006; Daneman et al, 2010
IPI00330186	Dlgap4	5	1	Daneman et al, 2010
IPI00877180	Cgnl1	3.3	1.5	Daneman et al, 2010
IPI00761693	Mpp7	2	2	Daneman et al, 2010
IPI00129388	Scrib	2	2	Daneman et al, 2010
IPI00399953	Wnk1	1	1.7	Daneman et al, 2010
IPI00761515	Magi3	0.7	1.2	Daneman et al, 2010
IPI00762889	Tjp3	0.7	1.2	Daneman et al, 2010
(C)
IPI00132474	Itgb1	44.7	9.6	Yousif et al, 2007
IPI00620873	Itga6	25.3	4.2
IPI00466371	Itga1	22.3	6.4	Yousif et al, 2007
IPI00877232	Itga2b	11.7	5.8
IPI00466276	Itga7	11.3	1.5
IPI00857195	Itgav	8.7	1.5	Krishna and Redies, 2009
IPI00468674	Itga3	5.3	0.7

Italics: Mean spectral count <5.

Table 3

Extracellular matrix proteins

Protein	Gene	Mean	s.d.	Reference
IPI00123058	Cntn1	125	16.5	Li et al, 2009 ^a
IPI00515360	Hspg2	60.3	44
IPI00227126	Tnr	53.7	4.2
IPI00854028	Cntnap1	46	14.8
IPI00400016	Lamc1	43	36.4
IPI00109612	Lamb2	33	24.6
IPI00756745	Lama2	30.3	34.5	Daneman et al, 2010 ^b
IPI00648938	Agrn	26.7	20.4
IPI00111793	Nid1	24.7	17.5
IPI00676957	Lama5	24	23.6
IPI00420554	Cntnap2	18.7	3.8
IPI00620601	Astn1	16.7	0.6
IPI00122273	Dag1	16.7	9
IPI00113726	Lama1	13.3	7.4
IPI00279079	Fgb	13	6.1
IPI00119970	Cntn2	12.7	3.8
IPI00338785	Lamb1	12.3	6
IPI00223446	Lama4	12	5.3
IPI00109588	Col4a1	11.7	15.3
IPI00121038	Vcan	11	3.5
IPI00338452	Col4a2	10.7	11
IPI00719919	Nid2	10	9.2
IPI00270915	Cntn3	8.3	3.5
IPI00869394	Bcan	6.3	2.1
IPI00115522	Fga	6.3	1.2
IPI00113539	Fn1	6.3	5.5
IPI00830749	Col6a3	5	5.6
IPI00469123	Mcam	1.7	1.5	Cayrol et al, 2008 ^a
IPI00480306	Cntn4	0.7	1.2	Cayrol et al, 2008 ^a
IPI00129240	Vtn	0.7	1.2	Daneman et al, 2010 ^b

Endothelial

Pericyte

Italics: Mean spectral count <5.

Table 4

(A) Endothelial proteins, (B) astrocyte proteins, and (C) pericyte proteins

Protein	Gene	Mean	s.d.	Reference
(A)
IPI00227299	Vim	191.3	63	Lu et al, 2008a, 2008b; Cayrol et al, 2008; Pottiez et al, 2009a, 2009b
IPI00115240	Mbp	75.3	28.4	Enerson and Drewes, 2006; Lu et al, 2008a, 2008b
IPI00119063	Lrp1	57	12.2	Uchida et al, 2011
IPI00228618	Gnaq	37.3	0.6	Li et al, 2009
IPI00553798	Ahnak	37.3	22.9	Cayrol et al, 2008
IPI00121378	Alcam	32.7	7.4	Cayrol et al, 2008
IPI00224776	Hspa12b	31.7	14.6	Daneman et al, 2010
IPI00620256	Lmna	24.7	6.7	Agarwal et al, 2009
IPI00131138	Flna	24	6.9	Pottiez et al, 2009a, 2009b
IPI00313884	Podxl	23	1	Agarwal et al, 2009
IPI00124700	Tfrc	22.3	12.9	Enerson and Drewes, 2006; Uchida et al, 2011
IPI00408495	Bsg	20	3.5	Daneman et al, 2010; Cayrol et al, 2008
IPI00653398	Flot2	18	4	Li et al, 2009
IPI00118569	Gna13	16	3.5	Cayrol et al, 2008
IPI00648927	Clta	14.3	2.5	Cayrol et al, 2008
IPI00311671	Nos3	13	2.6	Cunnea et al, 2010
IPI00130589	Sod1	12.7	4.7	Cunnea et al, 2010
IPI00134131	Scp2	12.3	1.5	Agarwal et al, 2009
IPI00114594	Rras	11.7	3.8	Cayrol et al, 2008
IPI00127942	Dstn	10.7	1.5	Agarwal et al, 2009
IPI00331609	Vwa1	10.7	6	Daneman et al, 2010
IPI00762435	Fyn	10.3	2.3	Li et al, 2009
IPI00117829	Cav1	9.7	5	Cayrol et al, 2008; Li et al, 2009
IPI00129253	Ly75	8	3	Daneman et al, 2010
IPI00131881	Adam10	7.7	4.9	Li et al, 2009
IPI00278804	Scamp1	7.7	1.5	Agarwal et al, 2009
IPI00315280	Sema7a	6.7	0.6	Daneman et al, 2010
IPI00120066	Prom1	6.3	1.5	Daneman et al, 2010
IPI00133608	Ociad1	5.7	0.6	Agarwal et al, 2009
IPI00849927	Rpl10	5.7	4	Agarwal et al, 2009
IPI00113377	Rplp1	5	4.4	Agarwal et al, 2009
IPI00338094	Bmpr2	4	2	Lu et al, 2008a, 2008b
IPI00762331	Cobll1	4	3.6	Daneman et al, 2010
IPI00828826	Slc44a1	3.7	1.5	Daneman et al, 2010
IPI00322245	Fmo2	3.3	2.9	Daneman et al, 2010
IPI00222369	Itih5	3.3	3.5	Daneman et al, 2010
IPI00111600	Spock2	3.3	0.6	Daneman et al, 2010
IPI00321477	Eng	3	2.6	Daneman et al, 2010; Lu et al, 2008a, 2008b
IPI00128358	Insr	3	3.6	Uchida et al, 2011
IPI00129423	Thbd	3	3.6	Cunnea et al, 2010
IPI00798576	Vwf	2.7	0.6	Enerson and Drewes, 2006
IPI00876541	Gpr116	2.3	2.5	Daneman et al, 2010; Li et al, 2009
IPI00221547	Tek	2.3	2.1	Enerson and Drewes, 2006
IPI00458230	Mfsd2	2	2	Daneman et al, 2010
IPI00131498	Cd93	1.7	1.5	Daneman et al, 2010
IPI00467127	Chid1	1.7	2.9	Agarwal et al, 2009
IPI00625441	Tpd52	1.7	1.5	Daneman et al, 2010
IPI00133257	CD34	1.3	2.3	Enerson and Drewes, 2006
IPI00124497	Flt1	1.3	1.2	Enerson and Drewes, 2006; Daneman et al, 2010; Cunnea et al, 2010
IPI00230191	Gna12	1.3	2.3	Li et al, 2009
IPI00123996	Nrp1	1.3	2.3	Li et al, 2009
IPI00137172	Ppapdc2	1.3	2.3	Daneman et al, 2010
IPI00226730	Clic5	1	1.7	Daneman et al, 2010
(B)
IPI00117042	Gfap	282	5.2	Enerson and Drewes, 2006; Lu et al, 2008a, 2008b; Cahoy et al, 2008; Yousif et al, 2007
IPI00119458	Aldoc	66.3	5.7	Cahoy et al, 2008
IPI00230738	Aqp4	51.3	12.1	Cahoy et al, 2008
IPI00153317	Aldh1l1	17	6.9	Cahoy et al, 2008
IPI00453834	Acsbg1	16	4	Cahoy et al, 2008
IPI00131995	Hapln1	11	7.2	Cahoy et al, 2008
IPI00322760	Prodh	7.7	2.1	Cahoy et al, 2008
IPI00119130	1500005I02Rik	6.7	2.5	Cahoy et al, 2008
IPI00830222	Mlc1	4.7	2.5	Cahoy et al, 2008
IPI00118069	F3	1.3	1.2	Cahoy et al, 2008
IPI00329896	Pla2g7	1.3	2.3	Cahoy et al, 2008
(C)
IPI00117043	Acta2	133.3	116.4	Enerson and Drewes, 2006
IPI00416906	Gnas	33.3	2.5	Armulik et al, 2010
IPI00134585	Enpep	31	12.8	Daneman et al 2010; Armulik et al, 2010
IPI00756745	Lama2	30.3	34.5	Daneman et al, 2010
IPI00113517	Ctsb	26	4.6	Armulik et al, 2010
(C)
IPI00153202	Ace2	21.7	6.4	Daneman et al, 2010
IPI00136716	Grm3	14.3	0.6	Daneman et al, 2010
IPI00126186	Mrc1	12	9.2	Armulik et al, 2010
IPI00133208	Hspa1a	6	10.4	Armulik et al, 2010
IPI00344142	Lin7a	5.3	5	Daneman et al, 2010
IPI00118880	Gucy1b3	5	1	Daneman et al, 2010
IPI00128915	Cspg4	4.3	1.5	Daneman et al, 2010
IPI00229516	Itgb5	4.3	2.3	Daneman et al, 2010
IPI00121827	Pdgfrb	2.7	1.2	Daneman et al, 2010
IPI00329896	Pla2g7	1.3	2.3	Daneman et al, 2010
IPI00123125	Xtrp3s1	1	1.7	Daneman et al, 2010
IPI00221980	Pde8b	0.7	1.2	Daneman et al, 2010; Armulik et al, 2010

Italics: Mean spectral count <5.

Membrane Transporter and Channel Proteins

The BBB is characterized by the expression of a complex range of solute carrier proteins (SLCs), ATP-binding cassette (ABC) proteins, voltage-gated channel proteins, and other transmembrane proteins that are essential for normal function of the brain, i.e., import of nutrients and export of metabolites across the endothelium. SLC/ABC proteins were ranked based on the number of spectral counts. We identified 84 SLC/ABC transmembrane proteins by MudPIT, with proteins having a spectral count of <5 peptides indicated in italics below the dashed line. All of the detailed supporting spectral count data are provided in Supplementary Table 1. Of these microvessel-derived membrane proteins identified, we noted that 21 of these proteins (i.e., Slc25a4, Abcb1a, Slc12a5, Slc3a2, Slc2a1, Slc1a6, Slc12a2, Slc6a1, Slc38a3, Abcc4, Slc9a3r2, Slc16a1, Slc2a3, Slc9a3r1, Abcd3, Slc17a7, Slc30a1, Slc27a1, Asna1, Abcg2, and Slc7a1) were identified in an earlier study based on mRNA expression profiling in rat brain microvessels (Enerson and Drewes, 2006). An additional 13 SLC/ABC proteins (i.e., Slc9a1, Abca7, Abcb8, Abcc9, Slc25a10, Slc5a6, Slc25a20, Slc4a3, Abcb6, Abca2, Slc6a20, Abcc8, and Slc38a2) were detected in mouse microvessels with a spectral count of <5 that were also present at the mRNA level in rat microvessels (Enerson and Drewes, 2006). Moreover, five out of six ABC proteins recently quantified in a study of human BBB transporters (Uchida et al, 2011) were also identified in our study, some with spectral counts 5 (i.e., Abcb1a, Abcc4, and Abcg2) and others with <5 spectral counts (i.e., Abca2 and Abcc8). In a similar comparison, 7 out of 10 SLC proteins had spectral counts 5 (i.e., Slc1a3, Slc3a2, Slc2a1, Slc2a3, Slc16a1, and Slc7a1), whereas Slc7a5 had a spectral count <5 (see Table 1 and Supplementary Table 1). An additional 37 transmembrane proteins identified by MudPIT in our study were also detected at the mRNA level from isolated mouse endothelial cells (i.e., Abcb1a, Slc3a2, Slc2a1, Slc38a3, Abcc4, Slc9a3r2, Slc16a1, Slc6a17, Slc25a24, Slc25a25, Slc30a1, Slc1a1, Abcg2, Slc7a1, Slc44a1, Slc25a20, Slc39a10, Slc44a2, Slc12a7, Slc22a8, Slc16a2, Slc7a5, Atp7a, Slc38a2, and Slc7a3; see Table 1), pericytes (i.e., Atp1a2, Atp1b2, Slc12a2, Atp2a3, Atp13a5, Abcc9, and Abca9; see Table 1), and astrocytes (i.e., Atp1a2, Slc1a2, Slc4a4, Slc25a18, Slc39a12; see Table 1) (Cahoy et al, 2008; Daneman et al, 2010). Prototypic examples of BBB proteins in the endothelium, such as Abcb1a (i.e., PgP), insulin receptor, the transferrin receptor, and α2 macroglobulin receptor, were detected among the top 50 proteins in our MudPIT studies and from mRNA profiling of microvessels (Enerson and Drewes, 2006), isolated endothelium (Daneman et al, 2010), and human microvessel studies (Uchida et al, 2011). Table 1 provides a listing of the 101 transmembrane SLC/ABC and voltage-gated channel proteins identified by MudPIT with relevant references to other complementary profiling studies. Additional SLC/ABC and voltage-gated channel proteins are listed in Supplementary Data Table 1.

Tight Junction Proteins and Integrins

The physical integrity of the endothelial component of the BBB is generally defined by the expression of tight junction proteins; therefore, we examined their relative abundance following MudPIT of mouse brain microvessels (Table 2). In our analyses, 19 tight junction proteins were identified with a spectral count 5 (Table 2A), including Pecam1 (i.e., CD31), occludin, cadherins (types 2 and 13), and claudin (type 11). An additional 18 tight junction proteins were identified including Vcam1, Jam 2, and claudin 5, although having limited representation among recovered sequences (i.e., <5 spectral counts). Cytoplasmic tight junction adapter proteins were also found (14 protein hits with 5 spectral counts; Table 2B) and included 5 catenins (types α1, α2, β1, δ1, and δ2) Esam1, F11r, and Tjp1 and Tjp2, as well as 6 proteins with low representation (Table 2B). In addition to tight junction proteins, seven integrins were identified (α1, α2b, α3, α6, α7, αv, and β1), with an additional seven integrins detected with low recovery (α2, α5, αm, β2, β3, β5, and β8; Supplementary Table 1). These findings provide a resource for further analysis of cell–cell and cell–ECM mediators expressed in the membrane fraction of brain microvessels.

ECM and Basal Lamina Proteins

To better characterize the ECM microenvironment and basal lamina proteins, we examined the MudPIT proteomic profile from microvessels. We identified 27 ECM proteins including tenascin-R, fibrinogen β-chain, 6 laminins (types α1, α2, α4, β1, β2, and γ1), and 5 proteoglycans (agrin, brevican, dystroglycan, perlecan, and versican). In addition, the neuronal-associated proteins contactin (types 1 and 2), and neural cell-adhesion molecule L1 were detected. Although only collagens VIα3, IVα1, and IVα2, were detected with spectral counts 5, several additional collagens (types Iα2, IVα3, VIα1, XIIα1, XVα1, and XVIIIα1) were also detected with low spectral abundance (see Supplementary Table 1). In combination, our findings provide a more complete resource of the relevant ECM proteins present in the brain microvessel microenvironment than has been previously reported.

Blood–Brain Barrier-Specific Cell Types

The analysis of microvessel proteome included the identification of proteins associated with endothelial cells and their associated astrocytes and pericytes. A total of 24 proteins were associated with endothelial cells based on previously published studies of the RNA expression in isolated cells (Daneman et al, 2010). Of these, the ABC transporter Abcb1a was the most abundant, along with several SLC proteins (Slc2a1, Slc1a1; see Table 1), von Willebrand factor-related proteins (Vwf and Vwa1, see Table 4A), as well as CD31 and Occludin (Pecam1 and Ocln, see Table 2A). A total of 31 endothelial proteins were identified (Table 4A), in addition to the 36 transmembrane endothelial transporter/channel proteins annotated in Table 1. Vascular endothelial growth factor receptors (Flt1 and Kdr), were also detected, although below the 5 spectral count threshold. Also present were the prototypic astroycte markers such as glial fibrillary acidic protein and Aqp4, as well as the more recently characterized astrocyte marker, Aldh1L1. Similarly, recovery of pericyte markers such as Slc1a2, Slc4a4, and laminin α2 (Table 3) indicates that membrane fractionation of microvessels recovers both transmembrane proteins and associated ECM proteins.

Discussion

In this study, we validated a combination of brain microvessel isolation, membrane fractionation, proteolytic digestion, liquid chromatography, mass spectrometry, and bioinformatic techniques to identify and catalog the protein expression profile in the BBB. The specific strategy enabled the profiling of transmembrane and membrane-associated markers of the BBB at the protein level (Kamiie et al, 2008; Ohtsuki and Terasaki, 2007). We focused on the identification of the canonical classes of membrane-associated BBB proteins that are associated with solute transport functions and normal brain physiology (e.g., SLC/ABC transporters and voltage-gated channel proteins), tight junction proteins maintaining integrity (i.e., transmembrane proteins and associated cytoplasmic adapters), and basal lamina extracellular proteins that constitute the ECM microenvironment and indirect intercellular contacts (i.e., integrin-mediated interactions with the basal lamina) (Hawkins and Davis, 2005). For ECM proteins in particular, we were interested in an approach to identify basal lamina proteins from intact microvessels to provide the most physiologically relevant in vivo profile, in contrast to in vitro assays in which cell culture has the potential to introduce artifacts with regard to ECM protein expression.

The intent of our studies was to generate a BBB protein database using MudPIT, which provides a reference for identification of BBB membrane proteins, ECM, and basal lamina. Membrane fractionation of the microvessels followed by a urea-assisted, acid-labile, surfactant-mediated protease digestion was used to identify: (1) the most abundant proteins (i.e., a standard of comparison for proteomic analysis and a measure of contaminants from other cellular compartments); (2) the profile of classical transmembrane transport proteins; and (3) the profile of tight junction and ECM proteins associated with microvessels (i.e., a measure of proteins that define the physical barrier properties of the neurovascular unit (Wolburg et al, 2009)).

Previous studies by Enerson and Drewes (2006) examined the mRNA expression of isolated rat microvessels and validated the techniques to isolate intact fresh microvessels . We used a similar isolation strategy as described by several laboratories to isolate BBB components, and then interrogated which proteins were associated with the BBB membrane fraction using MudPIT. We annotated our proteomic findings for comparison with mRNA expression profiling of SLC/ABC transporters (Table 1), tight junction proteins (Table 2), and endothelial genes (Table 4A) leading to the observation that as expected, many of the genes expressed at the mRNA level in the rat brain were also expressed in membrane protein-focused proteomic profiling. For example, 62% of the SLC/ABC transporters expressed at the RNA level by Enerson and Drewes were also detected in our proteomic study. Although a total of 217 SLC/ABC transporters and voltage-gated channel proteins were detected by our MudPIT approach, only 58% of these proteins satisfied our selection threshold for identity in three different biologic samples and with the average spectral count 5. Although additional membrane transporters were identified, further investigation will be necessary to validate the expression and biologic relevance of the less abundant proteins that did not meet the stringent spectral count cutoff criteria. We have compiled a complete listing of all peptides identified with accompanying spectral data (Supplementary Table 1). Similarly, 53% of the tight junction proteins identified by Enerson and Drewes were also detected by MudPIT, whereas an additional 21 tight junction proteins/adapters (Daneman et al, 2010) were identified along with 7 integrins that have not been previously identified through microvessel proteomics. Tables 1, 2, 3 and 4 are further annotated with gene expression data based on protein (Agarwal et al, 2009; Cayrol et al, 2008; Cunnea et al, 2010; Krishna and Redies, 2009; Li et al, 2009; Lu et al, 2008b; Ohtsuki and Terasaki, 2007; Pottiez et al, 2009b; Yousif et al, 2007) and mRNA (Cahoy et al, 2008; Daneman et al, 2010; Enerson and Drewes, 2006) detection strategies. A recent quantitative targeting proteomics strategy, reported as sensitive to the fmol range, for human blood–brain transporters and receptors yielded a similar ABC/SLC transporter profile, with 5 of 6 ABC proteins and 7 of 10 SLC proteins identified by MudPIT present in the reported human study (Uchida et al, 2011). Interestingly, Uchida et al compared expression levels of ABC transporters between human and mouse microvessel and determined that Abca8, which was not detected in our study, and Abca2, which had a spectral count <5, were present in their analyses at relatively low concentrations in mouse microvessels. Thus, it seems that our findings compare favorably with those of Uchida et al (2011) consistent with a high degree of protein detection sensitivity in our studies.

The recent mRNA gene profiling report (Daneman et al, 2010) on isolated endothelial cells, pericytes, and astrocytes provided an opportunity to compare and annotate our protein-based findings. We observed that by focusing on the membrane fraction for proteomic analyses, we were able to obtain an unexpectedly large number of SLC/ABC transporters present in microvessels, of which the mRNA expression of specific family members of transporters was similar, but not identical to that observed in isolated cells (Daneman et al, 2010). For example, 5 of the 6 transporters Apcdd1, Abcb1a, Slco1a4, Slc22a8, Slc7a5, but not Slc01c1 were highly expressed at the mRNA level by Daneman et al (2010) and were also identified by proteomics, although with low spectral counts. Similarly, although we obtained a large number of transporters as identified by RNA-based expression studies, the exact expression profile of transporters varied. We recognized the inherent difficulty of protein isolation from complex tissue sources and identification of low abundance proteins, as compared with RNA transcript analysis of low-abundance mRNAs. However, the identification of proteins with low spectral counts provides a starting point for further studies. Most importantly, further understanding of the interdependent roles of the ∼120 transporters recovered with the most abundant spectral counts will provide insights into BBB integrity and function.

Interestingly, among the most abundant transmembrane transporter/channel proteins was Atp1a2, the gene expression of which was detected in both isolated pericytes (Daneman et al, 2010) and astrocytes (Cahoy et al, 2008). The expression of other pericyte markers such as platelet-derived growth factor receptor-β was low, or in the case of CD13 and NG2, not represented in the proteomic database generated herein. Expression profiling is another complementary technique, which has been used by Shusta and colleagues to characterize membrane BBB proteins using expression cloning (Agarwal et al, 2009; Agarwal and Shusta, 2009). Other in vitro proteomic screening studies have focused on triton-soluble proteins (Pottiez et al, 2009b) and proteins lipid membrane rafts in cultured human endothelial cells where Alcam, AHNAK (desmoyokin), and intracellular trafficking proteins (clathrin light chain and Cav-1) were identified (Cayrol et al, 2008). In vivo proteomic analysis of integral membrane proteins with colloidal silica recovery technique also identified Cav-1, GPR116, G-αq, G-α12, Flot-2, Fyn, ADAM10, and contactin-1 (Li et al, 2009). Several of the proteins identified with these complementary approaches have also been identified in our study and are annotated in Table 4A. In combination, these findings underscore that our MudPIT-based approach for microvessel protein identification is useful for high coverage of membrane proteins and identification of physiologically relevant proteins on specific BBB cell types (Pottiez et al, 2009a).

Although transcriptome profiling provides a standard from which to assess mRNA expression levels in microvessels or isolated primary cells, the generation and characterization of a BBB gene expression profile at the protein level is essential for comparative purposes. Therefore, developing techniques to better analyze unknown constituents of the BBB proteome is essential as shown by the laser capture microdissection of microvessels from mouse brain slices using cryopreservation, fixation, and immunostaining followed by gel electrophoresis and then mass spectrometry for protein identification (Lu et al, 2008b). In a comparison with our MudPIT strategy of using fresh intact microvessels, 58% of BBB-enriched/endothelial proteins identified by laser capture microdissection/gel electrophoresis were also evident in our studies as indicated in Tables 1, 2, 3 and 4. Quantification of the relative abundance of known membrane transporters from human and mouse microvessels (Uchida et al, 2011), along with our results identifying proteins shows that a MudPIT approach is highly appropriate for proteomic identification from isolated BBB membranes and provides increased protein identification coverage. Each spectral count is a selected peptide precursor ion selected for fragmentation that is subsequently assigned for protein identification. For spectral counts between 5 and 500, abundance is linearly proportionate to spectral count (Collier et al, 2010; Cooper et al, 2010; Liu et al, 2004). Therefore, although there are limitations with relying exclusively on spectral counts for relative quantitation, once protein identifications have been made using a procedure such as MudPIT, other techniques can be deployed to quantify known proteins (Uchida et al, 2011). The identification of ECM proteins and transmembrane integrins in our study using freshly isolated tissues provides a profile of the membranes in contact with the extracellular microenvironment is not influenced by potential tissue culture-induced artifacts (McCarty, 2009). Although many ECM proteins were identified by MudPIT (i.e., six different specific laminin subtypes and five different proteoglycans), the mRNA encoding these and other ECM proteins seemed to be underrepresented in transcriptome studies, suggesting the possibility that ECM proteins with a long half-life may help explain the low abundance of their coding mRNAs. Therefore, to better understand the ECM microenvironment in the BBB, a MudPIT strategy may provide the most complete protein expression profile to date of ECM proteins.

In contrast to classical two-dimensional isoelectric focusing/SDS-PAGE, in which solubilization and resolution of hydrophobic proteins are limited, our solubilization and enzymatic fragmentation yielded experimental material sufficient for a relatively high-content analyses of a large number of peptide fragments. This was due, in part, to the use of: (1) a cell membrane lysis approach that minimized nuclear lysis; (2) a relaxation buffer to better unfold membranes (Borregaard et al, 1983) before lysis and centrifugal fractionation; and (3) use of urea as a solubilizing agent and ProteaseMAX acid-labile surfactant to increase the efficiency of tryptic digestion. Similar approaches applied to the analysis of membrane proteins in yeast, heart, and human parotid exosomes have led to the identification of high-content membrane protein data sets (Gonzalez-Begne et al, 2009; Lu et al, 2008a; Washburn et al, 2001). As with most sensitive analytical tools such as MudPIT, an important consideration is the purity of the starting material (i.e., microvessels versus other cell types) and the quality of membrane fractionation (i.e., the yield of membrane proteins). Therefore, we have relied on validation techniques consistent with DeCleves and colleagues (Yousif et al, 2007) and our own immunoblotting of starting microvessel material to validate the expression of membrane proteins, enrichment of endothelium/astrocyte endfeet, and a decrease in neuronal proteins.

In a comparison of proteins identified in this study of mouse brain blood vessels by MudPIT versus RNA gene profiling and immuohiostochemistry, it remains a challenge to determine how source, isolation, and analytical techniques may influence the conclusions (Uchida et al, 2007). For example, in the study of the BBB, expressions of claudin 5 and cadherin 11 are widely described (Hawkins and Davis, 2005; Nitta et al, 2003), and yet neither was a major component of the mass spectrometry spectral counts analyzed. In contrast, cadherins 1 and 2 and cadherin 11 predominated in our analysis suggesting that if extraction techniques affect different members of the same family, then such differences in apparent abundance may have other unknown explanations. These differences may reflect the limitations recovery techniques, the source of the lysate (i.e., heterogeneity of brain microvessels), the posttranslational status, ex vivo proteolytic activity, and the possibility that other tight junction proteins may deserve closer study in terms of relative abundance and physiologic relevance. For example, a recent in situ mRNA study of blood vessel cadherins in developing ferret brains indicated that although many cadherins were expressed in development, surprisingly no mRNA was detected in adult brain capillaries solubilization. This finding further supports the need to characterize BBB components at the protein level (Krishna and Redies, 2009).

It has become increasingly clear that a better understanding of the BBB requires a more comprehensive knowledge of the protein constituents of this important biologic barrier (Czeisler and Janigro, 2006; Neuwelt et al, 2008; Pardridge, 2010; Soni et al, 2010). In our study, we have used MudPIT on fresh unfixed microvessels to provide a comprehensive database resource for membrane transporter, tight junction and ECM proteins of the BBB. Our microvessel-based analysis of proteome provides a novel bioinformatic resource to help place into perspective transcriptome and proteomic findings from intact microvessels and isolated cells of the BBB.

Footnotes

Acknowledgements

This is publication MEM no. 20592 from The Scripps Research Institute. The authors thank Dr Sergio Catz (The Scripps Research Institute, La Jolla, CA, USA) for his suggestions and support regarding cell lysis and the use of relaxation buffer with cell membranes. They are also grateful to Dr Gabriel Simon (Washington University, Saint Louis, MO, USA) for providing dtarray2.pl.

The authors declare no conflict of interest.

References

Abbott

Ronnback

Hansson

(2006) Astrocyte-endothelial interactions at the blood-brain barrier. Nat Rev Neurosci 7:41–53

Aebersold

Mann

(2003) Mass spectrometry-based proteomics. Nature 422:198–207

Agarwal

Lippmann

Shusta

(2009) Identification and expression profiling of blood-brain barrier membrane proteins. J Neurochem 112:625–35

Agarwal

Shusta

(2009) Multiplex expression cloning of blood-brain barrier membrane proteins. Proteomics 9:1099–108

Armulik

Genové

Mäe

Nisancioglu

Wallgard

Niaudet

Norlin

Lindblom

Strittmatter

Johansson

Betsholtz

(2010) Pericytes regulate the blood-brain barrier. Nature 468:557–61

Borregaard

Heiple

Simons

Clark

(1983) Subcellular localization of the β-cytochrome component of the human neutrophil microbicidal oxidase: translocation during activation. J Cell Biol 97:52–61

Cahoy

Emery

Kaushal

Foo

Zamanian

Christopherson

Xing

Lubischer

Krieg

Krupenko

Thompson

Barres

(2008) A transcriptome database for astrocytes, neurons, and oligodendrocytes: a new resource for understanding brain development and function. J Neurosci 28: 264–78

Carvalho

Fischer

Chen

Yates

3rd Barbosa

(2008) PatternLab for proteomics: a tool for differential shotgun proteomics. BMC Bioinformatics 9:316

Cayrol

Wosik

Berard

Dodelet-Devillers

Ifergan

Kebir

Haqqani

Kreymborg

Krug

Moumdjian

Bouthillier

Becher

Arbour

David

Stanimirovic

Prat

(2008) Activated leukocyte cell adhesion molecule promotes leukocyte trafficking into the central nervous system. Nat Immunol 9:137–45

10.

Chen

Hewel

Felding-Habermann

Yates

3rd (2006) Large scale protein profiling by combination of protein fractionation and Multidimensional Protein Identification Technology (MudPIT). Mol Cell Proteomics 5:53–6

11.

Cociorva

Yatess

III (2006) DTASelect 2.0: Improving the confidence of peptide and protein identification. 54th ASMS Conference; Session: Bioinformatics II-#640 (conference).

12.

Collier

Sarkar

Franck

Rao

Dean

Muddiman

(2010) Direct comparison of stable isotope labeling by amino acids in cell culture and spectral counting for quantitative proteomics. Anal Chem 82:8696–702

13.

Cooper

Feng

Garrett

(2010) Relative, label-free protein quantitation: spectral counting error statistics from nine replicate MudPIT samples. J Am Soc Mass Spectrom 21:1534–46

14.

Cunnea

McMahon

O'Connell

Mashayekhi

Fitzgerald

McQuaid

(2010) Gene expression analysis of the microvascular compartment in multiple sclerosis using laser microdissected blood vessels. Acta Neuropathol 119:601–15

15.

Czeisler

Janigro

(2006) Reading and writing the blood-brain barrier: relevance to therapeutics. Recent Pat CNS Drug Discov 1:157–73

16.

Daneman

Zhou

Agalliu

Cahoy

Kaushal

Barres

(2010) The mouse blood-brain barrier transcriptome: a new resource for understanding the development and function of brain endothelial cells. PLoS One 5:e13741

17.

Enerson

Drewes

(2006) The rat blood-brain barrier transcriptome. J Cereb Blood Flow Metab 26:959–73

18.

Gonzalez-Begne

Han

Hagen

Hand

Melvin

Yates

(2009) Proteomic analysis of human parotid gland exosomes by Multidimensional Protein Identification Technology (MudPIT). J Proteome Res 8:1304–14

19.

Hartz

Bauer

Fricker

Miller

(2006) Rapid modulation of P-glycoprotein-mediated transport at the blood-brain barrier by tumor necrosis factor-alpha and lipopolysaccharide. Mol Pharmacol 69:462–70

20.

Haseloff

Krause

Bigl

Mikoteit

Stanimirovic

Blasig

(2006) Differential protein expression in brain capillary endothelial cells induced by hypoxia and posthypoxic reoxygenation. Proteomics 6:1803–9

21.

Hawkins

Davis

(2005) The blood-brain barrier/neurovascular unit in health and disease. Pharmacol Rev 57:173–85

22.

Kamiie

Ohtsuki

Iwase

Ohmine

Katsukura

Yanai

Sekine

Uchida

Ito

Terasaki

(2008) Quantitative atlas of membrane transporter proteins: development and application of a highly sensitive simultaneous LC/MS/MS method combined with novel in-silico peptide selection criteria. Pharm Res 25: 1469–83

23.

Kaschani

Niessen

Hoover

Cravatt

van der Hoorn

(2009) Diversity of serine hydrolase activities of unchallenged and botrytis-infected Arabidopsis thaliana. Mol Cell Proteomics 8:1082–93

24.

Kersey

Duarte

Williams

Karavidopoulou

Birney

Apweiler

(2004) The International Protein Index: an integrated database for proteomics experiments. Proteomics 4:1985–8

25.

Krishna

Redies

(2009) Expression of cadherin super-family genes in brain vascular development. J Cereb Blood Flow Metab 29:224–9

26.

Wang

Griffin

Long

Shore

Schnitzer

(2009) Enhancing identifications of lipid-embedded proteins by mass spectrometry for improved mapping of endothelial plasma membranes in vivo. Mol Cell Proteomics 8:1219–35

27.

Liu

Sadygov

Yates

3rd (2004) A model for random sampling and estimation of relative protein abundance in shotgun proteomics. Anal Chem 76: 4193–201

28.

McClatchy

Kim

Yates

3rd (2008a) Strategies for shotgun identification of integral membrane proteins by tandem mass spectrometry. Proteomics 8:3947–55

29.

Park

McClatchy

Liao

Yates

3rd (2009) Shotgun protein identification and quantification by mass spectrometry in neuroproteomics. Methods Mol Biol 566:229–59

30.

Murugesan

Macdonald

Pachter

Hancock

(2008b) Analysis of mouse brain microvascular endothelium using immunolaser capture microdissection coupled to a hybrid linear ion trap with Fourier transform-mass spectrometry proteomics platform. Electrophoresis 29:2689–95

31.

McCarty

(2009) Cell adhesion and signaling networks in brain neurovascular units. Curr Opin Hematol 16: 209–14

32.

Neuwelt

Abbott

Abrey

Banks

Blakley

Davis

Engelhardt

Grammas

Nedergaard

Nutt

Pardridge

Rosenberg

Smith

Drewes

(2008) Strategies to advance translational research into brain barriers. Lancet Neurol 7:84–96

33.

Nitta

Hata

Gotoh

Seo

Sasaki

Hashimoto

Furuse

Tsukita

(2003) Size-selective loosening of the blood-brain barrier in claudin-5-deficient mice. J Cell Biol 161:653–60

34.

Ohtsuki

Terasaki

(2007) Contribution of carrier-mediated transport systems to the blood-brain barrier as a supporting and protecting interface for the brain; importance for CNS drug discovery and development. Pharm Res 24:1745–58

35.

Pardridge

(2010) Biopharmaceutical drug targeting to the brain. J Drug Target 18:157–67

36.

Pottiez

Flahaut

Cecchelli

Karamanos

(2009a) Understanding the blood-brain barrier using gene and protein expression profiling technologies. Brain Res Rev 62:83–98

37.

Pottiez

Sevin

Cecchelli

Karamanos

Flahaut

(2009b) Actin, gelsolin and filamin-A are dynamic actors in the cytoskeleton remodelling contributing to the blood brain barrier phenotype. Proteomics 9: 1207–19

38.

Shusta

Boado

Pardridge

(2002) Vascular proteomics and subtractive antibody expression cloning. Mol Cell Proteomics 1:75–82

39.

Soni

Jain

Khare

Gulbake

Jain

(2010) Potential approaches for drug delivery to the brain: past, present, and future. Crit Rev Ther Drug Carrier Syst 27:187–236

40.

Tabb

McDonald

Yates

III (2002) DTASelect and Contrast: tools for assembling and comparing protein identifications from shotgun proteomics. J Proteome Res 1:21–6

41.

Taylor

Paton

Lilley

Binz

Julian

Jr Jones

Zhu

Apweiler

Aebersold

Deutsch

Dunn

Heck

Leitner

Macht

Mann

Martens

Neubert

Patterson

Ping

Seymour

Souda

Tsugita

Vandekerckhove

Vondriska

Whitelegge

Wilkins

Xenarios

Yates

3rd Hermjakob

(2007) The minimum information about a proteomics experiment (MIAPE). Nat Biotechnol 25:887–93

42.

Uchida

Kamiie

Ohtsuki

Terasaki

(2007) Multichannel liquid chromatography-tandem mass spectrometry cocktail method for comprehensive substrate characterization of multidrug resistance-associated protein 4 transporter. Pharm Res 24: 2281–96

43.

Uchida

Ohtsuki

Katsukura

Ikeda

Suzuki

Kamiie

Terasaki

(2011) Quantitative targeted absolute proteomics of human blood-brain barrier transporters and receptors. J Neurochem 117:333–45

44.

Washburn

Wolters

Yates

3rd (2001) Large-scale analysis of the yeast proteome by Multidimensional Protein Identification Technology. Nat Biotechnol 19:242–7

45.

Wolburg

Noell

Mack

Wolburg-Buchholz

Fallier-Becker

(2009) Brain endothelial cells and the glio-vascular complex. Cell Tissue Res 335:75–96

46.

MacCoss

Howell

Yates

3rd (2003) A method for the comprehensive proteomic analysis of membrane proteins. Nat Biotechnol 21:532–8

47.

Yousif

Marie-Claire

Roux

Scherrmann

Decleves

(2007) Expression of drug transporters at the blood-brain barrier using an optimized isolated rat brain microvessel strategy. Brain Res 1134:1–11

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

2.95 MB

0.02 MB

The Proteome of Mouse Brain Microvessel Membranes and Basal Lamina

Abstract

Keywords

Introduction

Materials and methods

Mouse Brain Microvessel Isolation

Endothelial Cell Lysis

Endothelial Cell Lysate Fractionation and Protein Digest

Multidimensional Liquid Chromatography: Mass Spectrometry (Multidimensional Protein Identification Technology)

Bioinformatic Analyses

Results

Microvessel Isolation and Validation

Proteomic Profile of Brain Microvessel Membranes

Membrane Transporter and Channel Proteins

Tight Junction Proteins and Integrins

ECM and Basal Lamina Proteins

Blood–Brain Barrier-Specific Cell Types

Discussion

Footnotes

Acknowledgements

References

Supplementary Material