prestroke proteomic changes in cerebral microvessels in ... · metaprotein-pathway analyses of...

5 8 8 | B E R G E R A T E T A L . | M O L M E D 1 7 ( 7 - 8 ) 5 8 8 - 5 9 8 , J U L Y - A U G U S T 2 0 1 1

INTRODUCTIONStroke is the third leading cause of

death and the leading cause of disabilityin the United States with persistent riskdespite successful antihypertensive thera-pies (1), and minimal efficacy of therapies

targeting neuronal loss reduction (2).These unexpected shortfalls highlightthe need for the study of the brain microvasculature—composed of the mi-crovascular endothelial cell, basal laminamatrix (BLM), pericyte and astrocyte end

feet—in stroke pathogenesis (3,4). Clinicalevidence implicating cerebral cortical mi-crovessel (cMV) involvement in strokepathogenesis are deduced from cumula-tive MRI-based observations that cerebralmicrohemorrhages coexist with acutespontaneous intraparenchymal hemor-rhages (5,6) and predict new disabling orfatal strokes in patients with acute ische-mic stroke or transient ischemic attack(7). Altogether, these associations reiteratethe need to study cMVs as they comprisethe site for said microhemorrhages.

Given that microvessels are at thefrontline of oxidative stress, hypoxia/ ischemia and inflammation, and giventheir key role in the maintenance of the

Prestroke Proteomic Changes in Cerebral Microvessels inStroke-Prone, Transgenic[hCETP]-Hyperlipidemic, Dahl Salt-Sensitive Hypertensive Rats

Agnes Bergerat,1 Julius Decano,2,3 Chang-Jiun Wu,4 Hyungwon Choi,5 Alexey I Nesvizhskii,5

Ann Marie Moran,2,3 Nelson Ruiz-Opazo,2,3 Martin Steffen,1,6* and Victoria LM Herrera2,3*

1Department of Pathology and Laboratory Medicine; 2Whitaker Cardiovascular Institute; and 3Department of Medicine, BostonUniversity School of Medicine, Boston, Massachusetts, United States of America; 4Department of Medical Oncology, Dana-FarberCancer Institute, Boston, Massachusetts, United States of America; 5Department of Pathology, University of Michigan, Ann Arbor,Michigan, United States of America and 6Department of Biomedical Engineering, Boston University, Boston, Massachusetts, UnitedStates of America

Stroke is the third leading cause of death in the United States with high rates of morbidity among survivors. The search to fill theunequivocal need for new therapeutic approaches would benefit from unbiased proteomic analyses of animal models of spon-taneous stroke in the prestroke stage. Since brain microvessels play key roles in neurovascular coupling, we investigated prestrokemicrovascular proteome changes. Proteomic analysis of cerebral cortical microvessels (cMVs) was done by tandem mass spec-trometry comparing two prestroke time points. Metaprotein-pathway analyses of proteomic spectral count data were done toidentify risk factor–induced changes, followed by QSPEC-analyses of individual protein changes associated with increased strokesusceptibility. We report 26 cMV proteome profiles from male and female stroke-prone and non–stroke-prone rats at 2 months and4.5 months of age prior to overt stroke events. We identified 1,934 proteins by two or more peptides. Metaprotein pathway anal-ysis detected age-associated changes in energy metabolism and cell-to-microenvironment interactions, as well as sex-specificchanges in energy metabolism and endothelial leukocyte transmigration pathways. Stroke susceptibility was associated inde-pendently with multiple protein changes associated with ischemia, angiogenesis or involved in blood brain barrier (BBB) integrity.Immunohistochemical analysis confirmed aquaporin-4 and laminin-α1 induction in cMVs, representative of proteomic changeswith >65 Bayes factor (BF), associated with stroke susceptibility. Altogether, proteomic analysis demonstrates significant molecularchanges in ischemic cerebral microvasculature in the prestroke stage, which could contribute to the observed model phenotypeof microhemorrhages and postischemic hemorrhagic transformation. These pathways comprise putative targets for translationalresearch of much needed novel diagnostic and therapeutic approaches for stroke.© 2011 The Feinstein Institute for Medical Research, www.feinsteininstitute.orgOnline address: http://www.molmed.orgdoi: 10.2119/molmed.2010.00228

*VLMH and MS contributed equally as senior authors.

Address correspondence and reprint requests to Victoria LM Herrera, Whitaker Cardiovas-

cular Institute, Boston University School of Medicine, 700 Albany Street, Boston, MA 02118.

Phone: 617-638-4020; Fax: 617-638-4066; E-mail: [email protected]; or Martin Steffen, De-

partment of Pathology and Laboratory Medicine, Boston University School of Medicine,

670 Albany Street, Boston, MA 02118. Phone: 617-414-7935; Fax: 617-414-1646; E-mail:

[email protected].

Submitted November 16, 2010; Accepted for publication April 13, 2011; Epub

(www.molmed.org) ahead of print April 14, 2011.

R E S E A R C H A R T I C L E

M O L M E D 1 7 ( 7 - 8 ) 5 8 8 - 5 9 8 , J U L Y - A U G U S T 2 0 1 1 | B E R G E R A T E T A L . | 5 8 9

blood brain barrier (BBB), investigationof cerebral microvascular molecular andcellular changes could contribute criticalnew insight into mechanisms of strokepathogenesis. Analysis of cMVs couldaddress the translational roadblock incurrent stroke intervention strategies,which have focused on large arterypathology and neuronal death. However,owing to their inaccessibility for system-atic study in humans and the limitationsof in vitro systems, analysis of the puta-tive role(s) of cMVs in stroke pathogene-sis requires an animal model that recapit-ulates stroke paradigms which implicatemicrovascular pathology.

The rat stroke model of ischemic- hemorrhagic stroke induced by early lifesodium exposure in an inbred polygenic- hypertensive, transgenic-hyperlipidemicDahl salt-sensitive (Dahl-S) rat, Tg25spmodel with sex-specific differences (8)provides a strategic model for molecularanalysis of the cerebral microvasculature.This model is characterized by the pres-ence of both hypertension and hyperlipi-demia as risk factors, and by a disease-course continuum that recapitulatesmultiple paradigms of human stroke,such as nonocclusive carotid artery dis-ease with chronic low-flow ischemia, mi-crohemorrhages and hemorrhagic infarc-tions (8). This disease course implicatescMVs in the pathogenesis, albeit not asthe sole culprit. Because the stroke-pronemodel used is a genetically identical in-bred transgenic rat strain exhibiting a rel-atively confined temporal disease course(8), the prestroke stage can be definedand studied, thereby validating the studyof prestroke events.

Because the BLM contributes to cMVintegrity (9), neurovascular unit func-tion (4,10), and because BLM loss hasbeen implicated in cerebral postischemichemorrhagic transformation (9,11), pro-teomic analysis is necessary to assessprotein changes directly in both BLMand cMV cell components. The shotgunproteomic approach by tandem massspectrometry provides a strategy to in-terrogate the microvasculature along thedisease course of stroke. Only a few

proteomic studies of brain microvesselsto elucidate disease mechanisms havebeen done (12–14).

Studying cerebral microvascularchanges in stroke pathogenesis is impor-tant since the microvasculature is in-volved in effective paracrine cross-talk inthe neurovascular unit (15), and sincethey comprise accessible targets for newdiagnostic and therapeutic strategies.Using a high-throughput proteomicstrategy, we tested the hypothesis thatputative changes in cMVs precede theonset of stroke in the stroke-proneTg25sp model which exhibits carotid ar-tery disease, cerebral microhemorrhagesand hemorrhagic infarctions (8). Here,we report cMV prestroke proteomicchanges associated with age and strokesusceptibility, respectively, in both maleand female stroke-prone rats.

MATERIALS AND METHODS

Stroke-Prone Rat Model ofMicrohemorrhages and HemorrhagicInfarction

All animal model studies were donein accordance with the Institutional Ani-mal Care and Use Committee of BostonUniversity School of Medicine, BostonMA, USA. Endpoint euthanasia was ac-complished via vital tissue collectionunder general anesthesia. We used an in-bred, Dahl-S rat model of adult-onset is-chemic-hemorrhagic stroke with a com-bination of stroke risk factors:hypertension, hyperlipidemia and early-life sodium exposure (0.4% NaCl) (8).Stroke is induced in both nontransgenic(Tg–) and transgenic (Tg+) Dahl-S ratsexpressing the human cholesteryl estertransfer protein (CETP), Tg[hCETP]25

Dahl-S rats or Tg25 (8,16) maintained inour facility. Littermate male (M) and fe-male (F), Tg+ and Tg– Dahl-S rats wereused. Given identical genetic back-ground and developmental program-ming, the relative stroke susceptibilitymeasured as differences in stroke-onsettimepoints is: [Tg+F] > [Tg–F] = [Tg+M],in contrast to minimal to no strokes in[Tg–M] rats (8). We assessed two time

points prior to overt evidence of stroke,at 2 months of age and at 4.5 months ofage. With the exception of the 4.5-month-old Tg25+F (n = 4 rats), all otherstudy groups had n = 3 rats.

Isolation of Cerebral CortexMicrovessels

Brain microvessels from rat cerebrumwere isolated as described (8,17). IsolatedcMVs retained on the 40-micron cellstrainer were washed off, visualized andphotographed to confirm cMV isolation,then quantified by weight. Total proteinwas isolated from each rat cMV sampleand analyzed individually by tandemmass spectral analysis. Diagrams of ex-perimental design are shown in Supple-mentary Figures 1 and 2.

Separation of Proteins and MassSpectrometry Analysis

Processing of proteins and tandemmass spectral analysis was done as de-scribed (18). Tandem MS/MS spectrawere collected for the top 10 ions of eachMS scan. Spectra were queried against therat NCBI RefSeq protein database, down-loaded in October 2006. Tandem massspectra were analyzed using SEQUEST(Thermo Fisher Scientific, Waltham, MA,USA). Proteins were considered presentonly if two high-scoring unique- sequencepeptide matches were achieved in a singlegel slice, or if a single peptide was ob-served at two different charge states inthat gel slice. On the basis of previouswork evaluating false positive matches forpeptide identification (19), we adjustedXCorr scores until the false positive rateof protein identification was empiricallydetermined to be 2%, as calculated by theinclusion of two reversed peptide se-quences in a single gel slice. SEQUESTDelCn score was required to be ≥0.08.Raw data are available for anonymousdownload at proteome. commons.org with the following identifier:hvsCus+PP6NDxmBk+kWQBP3Vjl/8bnGCjUW3g9SYTkWMYFOFFPaK8puJsPx7TgOpA10t9VbBgT9WgaWvPQGGsMUWVvIAAAAAAANafQ== (http:// proteomecommons.org).


P R E S T R O K E C E R E B R A L M I C R O V A S C U L A R P R O T E O M I C C H A N G E S

Pathway-Specific Analysis of Age-and Sex-Specific Changes in cMVProteome

We employed a metaprotein analyticapproach recently developed for rela-tively sparse mass spectrophotometricproteomic data of phosphorylated pro-tein signatures in lung cancer (20). Anal-ysis of pathway-specific abundancechanges utilized the 200 pathways of theKyoto Encyclopedia of Genes andGenomes (KEGG) database included inthe “C2:curated pathways, CP:canonicalpathway-set” available in the MolecularSignatures Database (MSigDB, http://www.broadinstitute.org/gsea/msigdb/collections.jsp#C2). As described in Wu etal. (20), a metaprotein was defined as aspecific linear combination of proteinswithin a curated pathway protein set. Inthe first step, for each protein set P, weutilized principal component analysis on the spectral count profiles of all pro-teins belonging to P in all cMV samplesto define eigenvectors that reflect thevariation in observed spectral count pat-terns (21). Principal components wereconsidered only if they explained a sig-nificant amount of the variation inmetaprotein expression (>30%) and had20 or more proteins observed in the data.The metaprotein expression is the sum of the counts of individual proteinsweighted by the coefficients in the eigen-vectors corresponding to the principalcomponents.

Metaproteins with differential expres-sion between two comparative groupsfor age (4.5-month-olds versus 2-month-olds) and sex (F versus M) were deter-mined based on false discovery rates(FDR) q values less than 0.05. The Pear-son correlation coefficient between eachprotein and the metaprotein was com-puted to identify those proteins withinthe protein set that are most responsiblefor the observed pathway- specific varia-tion. This approach appears to be a novelmethodology for performing a variationof “gene set enrichment analysis” (22) onproteomic data, and should be broadlyapplicable to other types of sparse data.For differentially expressed pathways,

we also identified individual members ofthe pathway that are differentially ex-pressed between queried study groupsbased on a Kruskal-Wallis (KW) ranksum test P value <0.05.

Hierarchical Model-Based Analysis ofProteomic Changes Associated withStroke Susceptibility

Protein expression was compared be-tween stroke-susceptible and non–stroke-susceptible rats at 4.5 months of age todistinguish differential expressionunique to progressive stages of strokesusceptibility in the spTg25 rat model(8). The stroke-susceptible groupspanned three sets of rats that exhibitedstroke: Tg+ males, Tg+ females, Tg– females (n = 3 per set for each timepoint). The non–stroke-prone group iscomposed of littermate Tg– males (n = 3set for each time point).

The QSPEC method by Choi et al. (23)was used for the statistical inference fordifferential protein expression from spec-tral counts, with a modification that ac-counts for proteomic changes due to sex-and transgenic CETP-induced hyperlipi-demia. QSPEC is a hierarchical model-based method for statistical inference ofspectral count-based differential proteinexpression (23), parallel to gene set en-richment analysis of transcription profiles(22). Two statistical models are fit in eachprotein: (M1i) Yij = β0i + β1iCj + β2i Gj + β3iTj

and (M0i) Yij = β0i +β1iCj + β2iGj, where Yij

denotes the spectral count of protein i inrat j, Cj is the CETP mutation of rat j (1 iftransgenic, 0 otherwise), Gj is the genderof rat j (1 if female, 0 if male), and Tj is thestroke susceptibility (1 if susceptible, 0 otherwise). The differential expressionstatus is indicated by the presence or theabsence of parameterβ3i. Given the twocandidate models, differential expressionfor protein is tested based on the magni-tude of Bayes factor (BF), defined by alikelihood ratio of the two models, that is,BFi = P(M1i|Yi)|P(M0i|Yi) (24). Large BFsupports the hypothesis of differential ex-pression for the protein.

By following conventional suggestedguidelines (25), proteins with BF ≥30

were selected as differentially ex-pressed. Induced and de-induced pro-teins associated with stroke susceptibil-ity were separately listed. In addition,proteins with evidence of confoundingby sex and CETP transgene status wereremoved from the candidate list, whichwas also examined by their correspon-ding BF for the two parameters β1i andβ2i. Proteins were not included in thelist of differentially expressed proteinsassociated with stroke-susceptibility ifthey also exhibited a BF ≥30 for sex (For M) or transgenic hyperlipidemia sta-tus (Tg+ or Tg–).

Immunohistochemical Analysis of RatBrain Sections

Histological processing and immuno-histochemical analyses were done as de-scribed (8) (Supplementary Information-Methods for specific details).

All supplementary materials are availableonline at www.molmed.org.

RESULTS

Cerebral Microvascular Proteome—Insight into the Blood Brain Barrier

To determine proteins and pathologicprocesses associated with increasedstroke susceptibility, we used 1D-PAGE/LC/MS/MS spectrometry (18) to obtaintotal proteome profiles from cerebral cor-tex microvessels (cMVs). We reduced thecomplexity of stroke proteomes by focus-ing on changes in isolated cerebral mi-crovascular proteomes rather than thewhole brain. We studied rat subgroupswhich represent the spectrum of strokesusceptibility including the nonstrokephenotype. Stepwise analyses of cMVproteomes were done to distinguish pro-teomic changes due to specific risk fac-tors and those associated with strokesusceptibility independent of said riskfactors. We obtained cMV samples frominbred stroke-prone Tg[hCETP]25 Dahl Srat model, thus eliminating confoundersfrom genetic heterogeneity and environ-mental factor variation. We obtained andstudied cMV proteomes from eight rat



groups with varying combinations of themajor risk factors observed in this model:age, sex and combined hypertension- hyperlipidemia, all of which contributeto stroke susceptibility as measured byearlier stroke onset (8). The effects ofthese risk factors on the cMV proteomewere analyzed separately to gain insightinto risk factor–induced pathogenicmechanisms. Thereafter, exclusion ofthese proteins differentially expressedwith respect to age, sex or transgenic sta-tus allowed the identification of thoseproteins which change specifically withstroke susceptibility. (Experimental de-sign scheme shown in SupplementaryFigures 1 and 2.)

In total, we report 26 proteome profiles(Table 1). The average number of pro-teins identified from each rat within eachgroup is summarized (Table 1) with thecomprehensive listing of proteins persample in Supplementary Table 1. Weidentified 1,934 proteins by two or morepeptides in the 26 samples. Many pro-teins were observed in most or many ofthe samples: 637 proteins were identifiedin 20 or more samples, 1,033 were identi-fied in 10 or more. Only 122 proteins ofthe 1,934 total were identified only in asingle sample.

We also report the detection of rela-tively similar spectral counts among allcMV samples of some proteins charac-teristic of endothelial cells (ACE2, vWF),BLM (collagen [COL]-18A1, COL-6A2,COL-6A3, nidogen-1 and -2), pericytes(smooth muscle α actin) and astrocyteend feet (GFAP) (data not shown), thusindicating that cMVs isolated using thebrain microvessel procedure (17) repre-sent the BBB. The BBB is composed ofthe microvascular endothelial cell, BLM,pericytes and astrocyte end feet (26).This characterization of cMV composi-tion is concordant with the analysis of“brain microvessels” isolated by lasercapture microdissection which detectedidentical multicellular composition(12,27). Unsupervised cluster analysisconfirms the multiple cell types andECM representation of the BBB in allcMV samples (data not shown). Impor-

tantly, the detection of these proteinswith similar spectral counts in all cMVsamples indicate relatively similar BBBcell composition of cMV samples, thusvalidating the proteomic analysis ofcMVs and the detection of differentiallyexpressed proteins. We also note thatsome cMV isolates had red blood cell(rbc) contamination, and these rbc-spe-cific proteins, such as hemoglobin andcarbonic anhydrase-1 were excluded(data not shown).

Differential Effects of Age andHyperlipidemia on the cMV Proteome

We next analyzed cMV proteomicchanges associated with age or hyperlipi-demia. Integrating a functional context toour proteomic analysis, we used a path-way-based analytic approach developedfor proteomic spectral counts with thepathway protein set serving as themetaprotein and validated previously forrelatively sparse phosphoproteomic data(20). We explored pathway differencesbetween 2-month-old and 4.5-month-oldrats. Because most stroke-prone Tg+ fe-male rats succumb to stroke at 4 to 5

months, we were unable to study a latertime point. At 2 months, blood pressureand total plasma cholesterol and totalplasma triglyceride levels are equivalentand normal between male and femalerats, and stroke events are not detected(8). In contrast, at 4.5 months, hyperten-sion is moderate to severe in both Tg+and Tg– rats exposed to early life sodium(0.4% NaCl), while hyperlipidemia ispresent in Tg+ rats but not in Tg– rats,with Tg+ female rats exhibiting far lesshyperlipidemia than males (8,16; Supple-mentary Figure 3). Because all 4.5-month-old rats are compared with 2-month-old rats regardless of sex,transgenic-hyperlipidemic phenotypeand stroke predisposition in this 2-month versus 4.5-month analysis sub-set, differentially expressed proteins re-flect age-associated changes rather thanstroke susceptibility.

Metaprotein analysis using 200 KEGGpathways listed in the canonical pathwaydatabase (MSigDB) identified differentialexpression between 2-month-old and 4.5-month-old rats in six metaproteins orKEGG pathways within two unifying cat-

Table 1. Number of proteins detected per study group by proteomic analysis.

NumberNumber Stroke of

Group of rats Age CETP Lipid suscepti- proteins CVe

number studied Sex (months) transgenea levelsb BPc bilityd observed (%)

1 3 M 2 – nl Pre – 979 2.52 3 M 2 + + Pre – 1,063 2.93 3 F 2 – nl Pre – 1,021 9.64 3 F 2 + ± Pre – 1,024 2.65 3 M 4.5 – nl Hyp – 1,082 5.16 3 M 4.5 + +++ Hyp + 1,052 8.37 4 F 4.5 – nl Hyp ++ 1,177 6.78 4 F 4.5 + + Hyp +++ 1,063 14.6

Total 26 1,058 (avgf) 5.5 (avg)

a+, Human CETP transgene–positive expression; –, nontransgenic.b+, Mild increase in total plasma cholesterol (TC) and total plasma triglyceride (TG) asobserved previously (8); ++, moderate increases in TC and TG as observed previously (8);+++, moderate-severe increases in TC and TG as observed previously (8); nl, normal; ±, minimal increase in TC and TG levelscBP, blood pressure; Pre, prehypertensive; Hyp, moderate-severe hypertension.d+, ++, +++: Increasing stroke susceptibility marked by earlier onset.eCV, coefficient of variation.favg, Average.



egories: 1) energy metabolism, and 2)cell-microenvironment interactions. Allpathways listed exhibited significant dif-ferences with false discovery rates (FDR)≤0.05 (Table 2). Among differentially ex-pressed energy metabolism metaproteins,the glycolysis pathway is increased whilethe citrate cycle and pyruvate metabolismpathways are decreased (Table 2). Amongcell–microenvironment interactionmetaproteins, the cell communicationpathway is decreased and a net increasein ECM-receptor interaction pathway isobserved (Table 2). Furthermore, severalproteins within each pathway-specificprotein set exhibited significant correla-

tion to metaprotein changes (Pearson cor-relation P < 0.05) and statistically signifi-cant differences (KW P < 0.05) between4.5-month-old and 2-month-old rat cMVproteomes.

Surprisingly, transgenic hyperlipi-demia did not induce significant path-way changes in cMV proteomes usingfalse discovery rate (FDR) analysis (Sup-plementary Table S2). We note, however,that the top scoring pathway, ECM- receptor interactions, had individual pro-teins whose expression levels varied sig-nificantly (P < 0.05) and also correlatedwith the metaprotein (Pearson correla-tion P < 0.05) (Supplementary Table S2).

Pathways Differentially Expressedbetween Males and Females

To gain insight into putative mecha-nisms that underlie greater stroke sus-ceptibility in female rats compared withmales despite equivalent levels of hy-pertension and less hyperlipidemia, anda lower risk for coronary artery disease(8), we analyzed male-to-female differ-ences for young, old and both agegroups combined. We obtained qualita-tively similar results for both agegroups, but the highest statistical signif-icance was detected in the 4.5-month-old rat groups and are reported here(Table 3).

Table 2. Differentially expressed pathways with age: cMVs of 4.5-month-old versus 2-month-old rats.

Pearson 4.5- versuscorrelation 2-month- KW

Pathwaysa FDRb Protein IDc P valued old ratse P valuef

Energy metabolismHSA00010 glycolysis and gluconeogenesis 0.010 ALDH2: aldehyde dehydrogenase 2 0.019 ↓ 0.004

LDHB: lactate dehydrogenase 2 0.007 ↓ 0.005ALDOC: aldolase C 9.1 × 10–4 ↑ 0.050ENO2: enolase 2 1.4 × 10–5 ↑ 0.037HK1: hexokinase 1 4.2 × 10–7 ↑ 4.7 × 10–5

PFKP: phosphofructokinase, platelet 0.045 ↑ 1.9 × 10–3

PKM2: pyruvate kinase, muscle 7 × 10–7 ↑ 0.017HSA00020 citrate cycle 0.011 MDH2: malate dehydrogenase 2 6 × 10–32 ↓ 1.7 × 10–3

ACLY: ATP citrate lyase 4 × 10–4 ↑ 1.7 × 10–3

DLST: dihydrolipoamide S-succinyltransferase 1 × 10–3 ↑ 0.017HSA00620 pyruvate metabolism 0.011 ALDH2: aldehyde dehydrogenase 2 0.034 ↓ 0.004

LDHB: lactate dehydrogenase B 0.011 ↓ 0.005MDH2: malate dehydrogenase 2 0.000 ↓ 1.7 × 10-3

PDHB: pyruvate dehydrogenase β 3 × 10–5 ↓ 0.057Cell-microenvironment interactions

HSA01430 cell communication 0.011 ACTG1: actin, γ 1 0.002 ↓ 0.007HSA04512 ECM-receptor interaction 0.056 VIM: vimentin 2.8 × 10–9 ↓ 0.002

COL1A1: collagen, type I, α1 5.8 × 10–5 ↑ 0.016COL4A2: collagen, type IV, α2 1.3 × 10–4 ↑ 0.017LAMA1: laminin, α1 1.5 × 10–9 ↑ 0.019LAMA5: laminin, α5 1.1 × 10–8 ↑ 0.027LAMB2: laminin, β2 3.8 × 10–11 ↑ 0.035TNR: tenascin R 1.4 × 10–3 ↑ 0.004

aNumbered KEGG canonical pathways.bFDR, false discovery rate with significance defined as FDR <0.05.cEntrez Gene official symbol. Proteins may exist in multiple overlapping pathways; e.g., LAMA1 is a component of both cellcommunication and ECM-receptor interaction pathways.dPearson correlation P < 0.05, identifies proteins within the pathway protein set that are most responsible for the observed pathway-specific variation.eSee footnote f for information on the arrows.fKruskal-Wallis rank sum test of individual proteins in differentially expressed pathway identifying individual proteins that are increased (↑) ordecreased (↓) in 4.5-month-old rat cerebral microvessels.



Two KEGG pathways/metaproteinsexhibited significant differential expression—HSA00010 glycolysis andgluconeogenesis pathway andHSA04670 leukocyte trans endothelialmigration pathway (Table 3). In 4.5-month-old females, the glycolysis/ glu-coneogenesis metaprotein exhibited adecrease (FDR = 0.04) in contrast to 4.5-month-old males (Table 4). This de-crease is opposite of the increase ob-served with aging (Table 2). Individualprotein analyses detected five proteinswith significant correlation with themetaprotein (Pearson P < 0.002) and sig-nificant differences between sexes (KWP < 0.05) (Table 3). Interestingly, 3 of the5 proteins decreased in the female-spe-cific glycolysis pathway protein set areincreased in the protein set for the age-associated change in glycolysis/gluco-neogenesis pathway, PKM2, ENO2 andALDOC (Tables 2,3). The proteins de-tected to be changed significantly be-tween comparative groups and corre-lated with differential expression of theleukocyte transendothelial migrationpathway are vinculin and integrin β-1(Table 3).

Differentially Expressed Proteins inStroke-Prone Cerebral Microvessels

Having analyzed the proteomicchanges associated with age, sex-spe-cific differences and hyperlipidemia, we

next assessed cMV proteome alterationsthat are associated with stroke suscepti-bility independent of said riskfactor–induced changes. We limited theanalysis to 4.5-month-old rats to maxi-mize differences in stroke susceptibilityas observed (8). We utilized QSPEC toidentify differentially expressed pro-teins (23) as indicated by statisticallysignificant differences in spectral counts(28). We identified proteins that variedas a function of stroke susceptibilityusing analysis of variance (ANOVA),eliminating protein changes associatedwith sex-specific differences or hyper-lipidemia given that the latter two,while risk factors, also are confoundingvariables. Analysis was limited to 4.5-month-old rats that are closer tempo-rally to stroke onset, and because2-month-old rats do not display a hy-perlipidemic profile yet (8). We reportproteins that were significantly differ-ent between non–stroke-prone andstroke-prone 4.5-month-old rats (Table 1,group 5 versus groups 6,7,8), but notbetween nontransgenic-normolipidemicrats and transgenic-hyperlipidemic rats(Table 1, groups 5,7 versus groups 6,8),and not between male and female 4.5-month-old rats (Table 1, groups 5,6 ver-sus groups 7,8) (Supplementary Figure 2).With use of QSPEC analysis (23), pro-teins that exhibit BFs greater than 65(>65) for the first comparison, but less

than 30 for the transgenic-hyperlipidemiaand sex, comparisons were consideredas associated with stroke susceptibility.By using these criteria, stroke suscepti-bility is associated with 12 proteins that are increased (Table 4), and 14 thatare downregulated (SupplementaryTable S3).

Analysis of putative pathogenic im-plications of the protein set associatedwith stroke susceptibility was limited toinduced proteins (Table 4) since net de-induction could simply be due toloss from cell death. Interestingly, 6 of12 proteins in the induced-protein sethave been reported previously to be in-creased in ischemia or hypoxia, and/orassociated with proapoptosis effects(Table 4), and known to be present inendothelial cells or astrocytes. Addi-tionally, 9 of 12 proteins have been re-ported to be part of the BBB and theirincrease would be expected to alter BBBfunction, as well as result in loss of BBBintegrity promoting either brain edemaor leukocyte transmigration as ob-served in strokes. Five induced geneshave proangiogenic functions (Table 4)consistent with the notion of ischemia-induced angiogenesis. Altogether, thisprotein set indicates ongoing molecularchanges in cMVs prior to the onset ofstroke which could be expected to con-tribute to stroke susceptibility and post-stroke sequelae.

Table 3. Differentially expressed pathways in 4.5-month-old female versus male rat cMVs.

Pearson correlation KW

Pathwaysa FDRb Protein IDc P valued F versus Me P valuef

HSA00010 glycolysis and gluconeogenesis 0.040 PKM2: pyruvate kinase, muscle 0.003 ↓ 0.007ENO2: enolase 2 1.3 × 10–3 ↓ 0.014AKR1A1: aldoreductase family 1 1.2 × 10–3 ↓ 0.019ALDOC: aldolase C 1.1 × 10–3 ↓ 0.028

HAS04670 leukocyte transendothelial migration 0.040 VCL: vinculin 6.1 × 10–4 ↑ 0.014ITGB1: integrin, β1 0.006 ↑ 0.024

aNumbered KEGG canonical pathways.bFalse discovery rate with significance defined as FDR <0.05.cEntrez Gene official symbol.dP < 0.05 identifies proteins within the pathway protein set that are most responsible for the observed pathway-specific variation.e4.5-month-old female versus male rat cMVs. See footnote f for information on the arrows.fKruskal Wallis rank sum test of individual proteins in differentially expressed pathway identifying individual proteins that are increased (↑) ordecreased (↓) in 4.5-month-old female cMVs.



Tab

le 4

.Pro

tein

s in

du

ce

d w

ith in

cre

ase

d s

tro

ke s

usc

ep

tibili

ty in

4.5

-mo

nth

-old

rats

.

Baye

s Fa

cto

r

SP

Tg+

F

cM

V s

itePa

tho

ge

nic

Pa

thw

ay

Pro

tein

ve

rsu

s ve

rsu

sve

rsu

sIs

ch

em

ia/h

ypo

xia

; ID

an

on

SPb

Tg–c

Md

Pro

tein

na

me

EeM

fP

gA

hc

ell

de

ath

/de

ath

An

gio

ge

ne

sisBB

B st

ruc

tura

l an

d f

un

ctio

na

l in

teg

rity

AQ

P4

4,15

5,18

413

0A

qu

ap

orin

4E

AIn

cre

ase

d in

isc

he

mia

(36

)In

cre

ase

d in

bra

in e

de

ma

(36

)SL

C2A

116

,731

11

Glu

co

se

EA

Inc

rea

sed

in is

ch

em

ia,

tra

nsp

ort

er-1

hyp

oxi

a m

eta

ge

ne

(37

)G

NA

I26,

941

14

G-p

rote

in α

12E

Me

dia

tes

thro

mb

in in

du

ce

d E

C

pe

rme

ab

ility

(38

)C

OL1

2A1

2,78

01

12C

olla

ge

n X

IIM

BBB

inte

grit

y; in

cre

ase

d in

EC

s b

y sh

ea

r st

ress

(39

)C

A4

1,79

811

1C

arb

on

ic

EA

Inc

rea

sed

in h

ypo

xia

/Bu

ffers

bra

in e

xtra

ce

llula

r sp

ac

e (

41)

an

hyd

rase

-4a

no

xia

(40

)SL

C9A

3R2

1,39

17

11N

HE-

3 re

gu

lato

r-2E

Enh

an

ce

s P

DG

F-in

du

ce

d

Pro

ce

ll m

igra

tion

(43

)P

DZ

sca

ffold

, an

ch

ors

ion

ch

an

ne

ls a

nd

rec

ep

tors

to

cyt

osk

ele

ton

(44

)p

rolif

era

tion

(42

)A

NP

EP1,

258

11

Am

ino

pe

ptid

ase

-NE

PP

roa

ng

iog

en

esis

(45

)KR

AS

233

22

K-ra

sA

Inc

rea

sed

pK-

ras

in h

ypo

xia

, Pr

oa

ng

iog

en

esis

(46

)b

loc

ks a

po

pto

sis (

46)

SNTA

119

13

0Sy

ntr

op

hin

, α1

AA

nc

ho

rs A

QP4

in a

stro

cyt

e e

nd

fee

t (2

9)FL

NB

113

31

Fila

min

BE

Pro

an

gio

ge

ne

sis (

47)

Me

dia

tes

ICA

M-1

driv

en

wb

c

tra

nse

nd

oth

elia

l mig

ratio

n (

48)

TF73

124

Tra

nsf

err

inE

Iron

ac

cu

mu

latio

n c

au

ses

An

tioxi

da

nt

seq

ue

stra

tion

of

iron

(50

)o

xid

ativ

e s

tre

ss a

nd

n

eu

rod

eg

en

era

tion

(49

)LA

MA

169

92

Lam

inin

α1

EM

Pro

an

gio

ge

nic

Ke

y c

om

po

ne

nt

of

BLM

pe

ptid

es

fro

m

lam

a1

(51)

aEn

tre

z G

en

e o

ffic

ial s

ymb

ol.

bSt

roke

-pro

ne

ve

rsu

s n

on

–str

oke

-pro

ne

cM

Vs.

cTr

an

sge

nic

-po

sitiv

e v

ers

us

tra

nsg

en

ic-n

eg

ativ

e c

MV

s.dFe

ma

le v

ers

us

ma

le c

MV

s.ec

MV

en

do

the

lial c

ells

.f c

MV

BLM

.gc

MV

pe

ricyt

es.

hc

MV

ast

roc

yte

en

d fe

et.



Confirmation of Aquaporin-4 andLaminin-α1 Protein Changes

To confirm proteomic changes associ-ated with stroke susceptibility, we stud-ied AQP4 and laminin-α1 by immuno-histochemical analysis of serial rat brainsections from stroke-prone andnon–stroke-prone rats. AQP4 andlaminin-α1 were selected because theyrepresent the proteins with the highestand lowest BF with significance cutoff>65 BF for association with stroke sus-ceptibility, in contrast to <30 BF for eitherage or sex (Table 4).

In contrast to non–stroke-prone ratbrains (Figure 1A–D), AQP4 and lamininα1 immunostaining exhibited increasedsignal intensity and spatial distributionin stroke-prone rat arteriolar and post-capillary venular microvessels (Figure1E–H). Similarly, in contrast to capillarymicrovessels in non–stroke-prone ratbrains (Figure 2A, C), AQP4 also exhib-ited increased signal intensity and spatialdistribution in cMV capillaries (Figure2D, F). These immunohistochemical ob-servations confirm the proteomicchanges associated with stroke suscepti-bility (Table 4). Additionally, proteomicspectral count analysis of proteins thatanchor AQP4 in the dystrophin complexto the perivascular membrane of astro-cyte end feet revealed that key anchoringcomponents are present in the cMV pro-teome: dystroglycan, α1- syntrophin, dys-trophin- isoform Dp71, and dystrobrevin-α (Supplementary Table 1), thussupporting the detection of AQP4 incMV samples. Moreover, α1-syntrophinalso is increased in association withstroke susceptibility (Table 4). Interest-ingly, AQP4 is increased in arteriolar as-trocyte end feet, while increased AQP4appears to be increased in either or bothastrocyte end feet and endothelium inpostcapillary venules and capillaries,since light microscopy cannot distinguishthe two, and since abluminal appositionof AQP4 immunostaining to endothelialnuclei suggest abluminal endothelial-membrane localization (Figure 1C, G; 2D2F). Endothelial AQP4 expression hasbeen reported, with greater amounts in

Figure 1. Representative immunohistochemical analysis of AQP4 and laminin-α1 expressionin cerebral cortex 20- to 30-micron microvessels (cMVs). (A–D) control, non–stroke-prone ratbrains; (E–H) stroke-prone rat brain sections. (A) AQP4-positive immunostaining (green) is de-tected on one side of an arteriolar cMV BLM in an 18-month-old non–stroke-prone rat. RBCsexhibited red autofluorescence within the cMV lumen in the presence of double immunos-taining with AF594-fluorophore (red) labeled glial fibrillary acidid protein (GFAP). DAPI-stainednuclear DNA (blue) highlights the flat-ellipsoid nuclei of endothelial cells (ECs) lining thecMV lumen, with no AQP4 expression. DAPI-stained, round pericyte nuclei are detected be-tween ECs and AQP4-positive staining, consistent with astrocyte–end feet expression ofAQP4. (B) Diffusion contrast interference (DIC) image depicts structural outline of 25–30 mi-cron cMV in panel A, confirming cMV lumen with red blood cells. (C) Double immunostain-ing of AQP4 and AF594-labeled rat endothelial cell antigen1 (Reca1) detects partial AQP4immunostaining in abluminal side of endothelial cells lining an ~20 micron-cMV lumen in anon–stroke-prone 4.5-month-old female rat. Reca1- immunostaining (red) is not detected incMV endothelium; red autofluorescence is noted in RBCs. (D) Minimal to no laminin α1(Lama1) DAB-immunostaining (brown) of non–stroke-prone rat cMV. Hematoxylin-stained(blue), flat-ellipsoid EC-nuclei line cMV lumen filled with RBCs. (E,F) Increased intensity anddistribution of AQP4 immunostaining is detected bordering both sides of cMV consistentwith astrocyte–end feet localization. Reca1-immunostaining (red) is not detected in cMVendothelium; red autofluorescence noted in RBCs. (G) Increased intensity and distributionof AQP4 immunostaining borders both sides of cMV and abuts EC nuclei on the abluminalside. Reca1-immunostaining (red) is not detected in cMV endothelium; red autofluores-cence noted in RBCs. (H) Laminin α1-immunostaining (brown) of ECs and BLM is detectedin stroke-prone rat cMV. Hematoxylin-stained (blue), flat-ellipsoid EC-nuclei line cMV lumenwith RBCs. Bar: 10 microns in panels A–H.



the abluminal compared with the adlu-minal endothelial membrane (29).

DISCUSSION

Importance of Direct Study of cMVProteomes

Focus on rat stroke model-derivedcMV proteomes is important since thereis no in vitro system that can recapitulatethe critical cell-to-cell and cell-to-matrixinteractions in intact cMVs nor recapitu-late molecular-to-cellular events in strokepathogenesis. Additionally, studying thestatus of cMV proteomes per se gives in-sight into its critical roles in the mainte-

nance of BBB integrity and neurovascu-lar coupling (4). Detection of distinctpathways with differential expression iscontrasted to the many more pathwaysnot exhibiting statistically significant dif-ferential expression, thus giving com-pelling confidence to detection of saidobserved changes (Tables 2–4).

Insight into Risk-Factor InducedPrestroke Changes in cMVs

Metaprotein analysis identified in-creased glycolysis-pathway in 4.5-month-old rats compared with 2-month-old rats,similar to increased glycolysis associatedwith hypoxia (30). On the other hand,

sex-specific proteome analysis also de-tected increased expression of glycolysispathway proteins in 4.5-month-old malerats, which have been observed to haveless stroke susceptibility than female rats.These seemingly conflicting observationscan be reconciled with the hypothesisthat while increased glycolysis indicatesrelative low grade ischemia/hypoxia asrats age, males have the advantage overfemales because the increase in glycolysispathway proteins is thought to be anadaptive compensation seen in astrocytesbut not in neurons. The ability of astro-cytes to compensate and restore mito-chondrial membrane potential throughglycolytically generated ATP allows themto resist apoptosis in chronic low flow is-chemia in contrast to neurons, which areunable to increase glycolysis for energydemands during hypoxia, and hence pro-ceed to apoptosis given ischemia/hypoxia (30). Altogether, the greater in-crease in glycolysis pathway protein setin 4.5-month-old males could give them acompensatory advantage over 4.5-month-old females, thus accounting for thegreater susceptibility in females (8). Addi-tionally, 4.5-month-old females also exhib-ited a significant increase in the leukocytetransmigration pathway compared withmale cMVs, concordant with the observa-tion of greater stroke susceptibility in fe-males as observed in stroke-prone Tg25rat model (8).

Insight into Putative Stroke-PronePathogenic Mechanisms

Prestroke changes in cMV basal lam-ina matrix. In contrast to the analysis ofRNA changes, proteomic analysis ofcMVs gives a direct analysis of the cu-mulative net changes in the BLM of thecMV which is produced and modulatedby all three cellular components of theBBB—endothelial cell, pericyte and as-trocyte. Elucidation of changes in theBLM provides insight into intercellularcross-talk within the BBB and, evenmore importantly, within the neurovas-cular unit, which impacts not just the in-tegrity of microvascular permeability,but also neurovascular coupling, neutro-

Figure 2. Representative immunohistochemical analysis of AQP4 expression in cortical mi-crovascular capillaries (<10 microns). (A) Double immunofluorescence analysis with AQP4and GFAP detects low level AQP4-positive immunostaining (green) of cMV capillary en-dothelium (lining lumen), negative GFAP immunostaining, and RBC red autofluorescencein young 2-month-old stroke-prone rat capillaries. (B) DIC image of panel A showing capil-laries. (C) Double immunofluorescence analysis of AQP4 (green) and Reca1 (red) detectsminimal to no AQP4 expression in capillaries in non–stroke-prone rat brain and no Reca1immunostaining. Red autofluorescence of RBCs is noted. (D) Double immunostaining de-tects increased AQP4-positive immunostaining of cerebral micro vessel poststroke in a 4.5-month-old stroke-prone rat brain close to a microhemorrhage with red autofluorescenceof extravasated RBCs in the cortical parenchyma. (E) DIC image of panel 2D showingcapillaries and microhemorrhage. (F) Increased AQP4 immunostaining in cortical capillar-ies (≤10 microns) with red autofluorescence of luminal RBCs. DAPI-nuclear DNA stain(blue) Bar: 10 microns for panels A–F.



transmitter inactivation, neurotrophiccoupling and angiogenic and neurogeniccoupling (31). More specifically, the de-tected net increase in collagen-12A1 andlaminin-1 associated with stroke suscep-tibility suggests the hypothesis that al-tered basement membrane compositionand amount could lead to a loss of theability of the vessels to autoregulate inresponse to changing hemodynamic con-ditions, thus leading to chronic low flowischemia and vulnerability to microhem-orrhages and/or postischemic infarctionhemorrhagic transformation (8). Thesestroke susceptibility-specific changes arereinforced with ECM-receptor- interaction pathway changes indepen-dently induced by age, thus altogetherleading to changes in the collagen net-work affecting 4 types of collagen,COL12A1, COL4A1, COL4A2 andCOL1A1, and the laminin network af-fecting several laminin isoforms:LAMA1, LAMA5, LAMB2 (Table 2).

Prestroke pathway changes associatedwith stroke susceptibility. Concordantly,proteins that are increased in associationwith stroke susceptibility are known tobe expressed in the different BBB celltypes: ECs, pericytes or astrocytes(Table 4). Induced cMV proteins that areassociated previously with ischemia(Table 4) confirms the presence ofchronic low flow ischemia first detectedin this stroke-prone rat model by mag-netic resonance microimaging (8). Eluci-dation of significant proteomic changesin the prestroke stage confirms the path-ogenic role of low flow ischemia in thisstroke-prone Tg25 rat model. Addition-ally, proteomic changes which can leadto changes in BBB integrity be it edema,leukocyte transmigration or BBB dys-function (Table 4), provide molecular in-sight into putative mechanisms thatcould contribute to the observed strokefeatures of neutrophil transmigration,microhemorrhages and/or predisposi-tion to hemorrhagic transformation asobserved in this model (8).

The association of stroke susceptibilitywith increased AQP4 expression in bothspatial distribution and intensity, in all

cMV components (arterioles, postcapil-lary venules and capillaries, as well as inboth astrocyte end feet and abluminalendothelium) altogether suggests the fol-lowing hypotheses: first, that AQP4 in-crease in astrocyte end feet provides amolecular mechanism for a preset vul-nerability to poststroke brain edema thatis initiated at the prestroke stage and notjust at the onset of stroke, since AQP4 isthe major water channel underlyingbrain water balance and implicated in is-chemic cytogenic edema (32). Secondly,although the role of endothelial AQP4has not been elucidated, we hypothesizethat parallel to known AQP4 roles in as-trocytes (32–34), AQP4 increases instroke-prone cMV venular and capillaryendothelia also result in altered water-ion homeostasis and signaling in saidcMV endothelium. This perturbationcould then contribute to endothelial dys-function with cascading effects on theneurovascular unit, thereby resulting inaltered neurovascular coupling. The re-sultant putative dysfunction of neurovas-cular coupling then sustains, if not wors-ens, the chronic low grade ischemiaobserved in the stroke-prone, hyperten-sive rat strain, which coupled withAQP4-mediated cytogenic edema, to-gether contribute to stroke-susceptibility.These hypotheses are supported by ob-servations of increased AQP4 in stroke-prone SHRSP rats (35) and in poststrokehuman brains with cerebral infarction(36), and by the detection of a hypoxiainducible factor-1α response element inthe 5′ flanking region of the humanAQP4 gene.

Limitations of the study. While keyinsights were obtained, delineation oflimitations of the study that could im-pact biological context of observations isimportant. Shotgun proteomic analysisby tandem mass spectrometry is limitedby the smaller total number of proteinsanalyzed compared with transcriptionprofiling. Hence while positive findingsare informative, nondetection of a pro-tein does not equate with noninvolve-ment in stroke susceptibility. As with all“-omics” analysis of complex tissues val-

ued for critical biological context, detec-tion of protein level alterations notcaused by cell composition changes doesnot pinpoint cell-specific origin(s) ofsaid changes. Additionally, the temporalvariation in disease course of adult-onsetstroke models, despite identical geneticbackgrounds, adds further variabilityonto the inherent variation in shotgunproteomic analysis by tandem massspectrometry, thus reiterating that nega-tive results are noninformative. How-ever, these limitations emphasize the ro-bustness of positive observations asreported.

CONCLUSIONSAltogether, stepwise proteomic analy-

ses of cMVs detected prestroke pathwaychanges (ischemia, BBB-integrity, angio-genesis pathways) associated withstroke susceptibility and pathwaychanges associated with age- and sex-specific changes (glycolytic pathways,cell-microenvironment interactions, andtransendothelial migration pathways).These observations suggest that a para-digm shift toward the prestroke stagecould be central to overcoming the“translational roadblock to effectivestroke therapies” (2). Observed pre-stroke pathway changes provide a prioriputative targets for new diagnostic andtherapeutic approaches, and can alsogive critical pathophysiological contextto help address the current conundrumof “Janus-faced stroke treatment targetswhich have limited efficacy combinedwith serious side effects” (2).

ACKNOWLEDGMENTSGrant funding from the National Insti-

tutes of Health, RO1 AG32649-02 to VLMHerrera; Whitaker Cardiovascular Insti-tute T32 HL07224 to A Bergerat.

DISCLOSURESThe authors declare that they have no

competing interests as defined by Molec-ular Medicine, or other interests thatmight be perceived to influence the re-sults and discussion reported in thispaper.



REFERENCES1. Dahlof B. (2007) Prevention of stroke in patients

with hypertension. Am. J. Cardiol. 100(Suppl):17J–24J.

2. Endres M, et al. (2008) Improving outcome afterstroke: overcoming the translational roadblock.Cerebrovasc. Dis. 25:268–78.

3. Hawkins BT, Davis TP. (2005) The blood brainbarrier neurovascular unit in health and disease.Pharmacol. Rev. 37:173–85.

4. del Zoppo GJ, Milner R. (2006) Integrin-matrixinteractions in the cerebral microvasculature. Arterioscler. Thromb. Vasc. Biol. 26:1966–75.

5. Lee SH, et al. (2004) Cerebral microbleeds are re-gionally associated with intracerebral hemor-rhage. Neurology. 62:72–6.

6. Alemany M, Stenborg A, Terent A, Sonninen P,Raininko R. (2006) Coexistence of microhemor-rhages and acute spontaneous brain hemorrhage:correlation with signs of microangiopathy andclinical data. Radiology. 238:240–7.

7. Boulanger JM, et al. (2006) Cerebral microhemor-rhages predict new disabling or fatal strokes inpatients with acute ischemic stroke or transientischemic attack. Stroke. 37:911–4.

8. Decano JL, et al. (2009) Early-life sodium expo-sure unmasks susceptibility to stroke in hyper-lipidemic, hypertensive heterozygous Tg25 ratstransgenic for human cholesteryl ester transferprotein. Circulation. 119:1501–9.

9. Hamann GF, Okada Y, del Zoppo GJ. (1996) He-morrhagic transformation and microvascular in-tegrity during focal cerebral ischemia/reperfu-sion. J. Cereb. Blood Flow Metab. 16:1373–8.

10. Rhodes JM, Simons M. (2007) The extracellularmatrix and blood vessel formation: not just ascaffold. J. Cell Mol. Med. 11:176–205.

11. Wang CX, Shuaib A. (2007) Critical role of mi-crovasculature basal lamina in ischemic brain in-jury. Prog. Neurology. 83:140–8.

12. Haqqani AS, et al. (2005) Characterization of vas-cular protein expression patterns in cerebral is-chemia/reperfusion using laser capture mi-crodissection and ICAT-nanoLC-MS/MS. FASEBJ. 19:1809–21.

13. Soreghan BA, et al. (2005) Using proteomics andnetwork analysis to elucidate the consequences ofsynaptic protein oxidation in a PS1+AβPP mousemodel of Alzheimer’s disease. J. Alzheimers Dis.8:227-41.

14. Calabria AR, Shusta EV. (2006) Blood-brain bar-rier genomics and proteomics: elucidating phe-notype, identifying disease targets and enablingbrain drug delivery. Drug Discov. Today. 11:792–9.

15. del Zoppo GJ. (2010) The neurovascular unit inthe setting of stroke. J. Intern. Med. 267:156–71.

16. Herrera VL, et al. (2004) Analysis of gender-spe-cific atherosclerosis susceptibility intransgenic[hCETP]25DS rat model. Atherosclero-sis. 177:9–18.

17. Song L, Pachter JS. (2003) Culture of murinebrain microvascular endothelial cells that main-tain expression and cytoskeletal association of

tight junction-associated proteins. In Vitro Cell.Dev. Biol. Anim. 39:313–20.

18. Ahmad QR, Nguyen DH, Wingerd MA, ChurchGM, Steffen MA. (2005) Molecular weight assess-ment of proteins in total proteome profiles using1D-PAGE and LC/MS/MS. Proteome Sci. 3:6.

19. Peng J, Elias JE, Thoreen CC, Licklider LJ, Gygi SP.(2003) Evaluation of multidimensional chromatog-raphy coupled with tandem mass spectrometry(LC/LC-MS/MS) for large-scale protein analysis:the yeast proteome. J. Proteome Res. 2:43–50.

20. Wu CJ, et al. (2009) A predictive phosphorylationsignature of lung cancer. PlosOne. 4:e7994.

21. Kumar N, Wolf-Yadlin A, White FM, Lauffen-burger DA. (2007) Modeling HER2 effects on cellbehavior from mass spectrometry phosphotyro-sine data. PLoS Comput. Biol. 3:e4.

22. Subramanian A, et al. (2005) Gene set enrichmentanalysis: a knowledge-based approach for inter-preting genome-wide expression profiles. Proc.Natl. Acad. Sci. U. S. A. 102:15545–50.

23. Choi H, Fermin D, Nesvizhskii AI. (2008) Signifi-cance analysis of spectral count data in label-freeshotgun proteomics. Mol. Cell. Proteomics.7:2373–85.

24. Kass RE, Raftery AE. (1995) Bayes Factors. J. Am.Stat. Assoc. 90:773–95.

25. Jeffreys H. (1961) The Theory of Probability. OxfordUniversity Press, 1961, p 432.

26. Ballabh P, Braun A, Nedergaard M. (2004) Theblood brain barrier: an overview: structure, regu-lation, and clinical implications. Neurobiol. Dis.16:1–13.

27. Agarwal N, Lippmann ES, Shusta EV. (2010)Identification and expression profiling of bloodbrain barrier membrane proteins. J. Neurochem.112:625–35.

28. Liu H, Sadygov RG, Yates JR. (2004) A model forrandom sampling and estimation of relative pro-tein abundance in shotgun proteomics. Anal.Chem. 76:4193–201.

29. Amiry-Moghaddam M, et al. (2004) Alpha syn-trophin deletion removes the perivascular butnot endothelial pool of aquaporin-4 at the bloodbrain barrier. FASEB J. 18:542–4.

30. Bolanos JP, Almeida A, Moncada S. (2010) Gly-colysis: a bioenergetic or a survival pathway?Trends Biochem. Sci. 35:145–9.

31. Zlokovic BV. (2008) The blood-brain barrier inhealth and chronic neurodegenerative disorders.Neuron. 57:178–201.

32. Verkman AS. (2009) Knock-out models revealnew aquaporin 4 functions. In E. Beitz (ed),Aquaporins, Handb. Exp Pharmacol. 190:359–81.

33. Nagelhus EA, Mathiisen TM, Ottersen OP. (2004)Aquaporin-4 in the central nervous system: cellu-lar and subcellular distribution and coexpressionwith Kir4.1. Neuroscience. 129:905–13.

34. Nicchia GP, et al. (2004) The role of aquaporin-4in the blood-brain barrier development and in-tegrity: studies in animal and cell culture models.Neuroscience. 129:935–45.

35. Tomassoni D, Bramanti V, Amenta F. (2010) Expres-

sion of aquaporins 1 and 4 in the brain of sponta-neously hypertensive rats. Brain Res. 1325:155–63.

36. Aoki K, et al. (2003) Enhanced expression ofaquaporin 4 in human brain with infarction.Acta. Neuropathol. 106:121–4.

37. Bufa FM, Harris AL, West CM, Miller CJ. (2010)Large meta-analysis of multiple cancers reveals acommon, compact and highly prognostic hy-poxia metagene. Br. J. Cancer. 102:428–35.

38. Gavard J, Gutkind JS. (2008) Protein kinase C- related kinase and ROCK are required for throm-bin-induced endothelial cell permeability down-stream from Galpha12/13 and Galpha11/q. . Biol. Chem. 283:29888–96.

39. Jin X, et al. (2003) Shear stress-induced collagenXII expression is associated with atherogenesis.Biochem. Biophys. Res. Commun. 308:152–8.

40. Nogradi A, et al. (2003) Up-regulation of cere-bral carbonic anhydrase by anoxic stress inpiglets. J. Neurochem. 85:843–50.

41. Shah GN, et al. (2005) Carbonic anhydrase IV andXIV knockout mice: roles of the respective carbonicanhydrases in buffering the extracellular space inbrain. Proc. Natl. Acad. Sci. U. S. A. 102:16771–6.

42. Jung Kang Y, et al. (2004) NHERF2 increasesplatelet-derived growth factor-induced prolifera-tion through PI-3-kinase/Akt-, and Src family ki-nase-dependent pathway. Cell. Signal. 16:791–800.

43. Thiesen CS, Wahl JK 3rd, Johnson KR, WheelockMJ. (2007) NHERF links the N-cadherin/catenincomplex to the platelet-derived growth factor re-ceptor to modulate the actin cytoskeleton andregulate cell motility. Mol. Biol. Cell. 18:1220–32.

44. Terwaki S, Maesaki R, Hakoshima T. (2006)Structural basis for NHERF recognition by ERMproteins. Structure. 14:777–89.

45. Rangel R, et al. (2007) Impaired angiogenesis inaminopeptidase N-null mice. Proc. Natl. Acad.Sci. U. S. A. 104:4588–93.

46. Zeng M, Kikuchi H, Pino MS, Chung DC. (2010)Hypoxia activates the K-ras proto-oncogene tostimulate angiogenesis and inhibit apoptosis incolon cancer cells. PLoSOne 5(6):e10966.

47. Del Valle-Perez B, et al. (2010) Filamin B plays akey role in VEGF-induced endothelial cell motil-ity through its interaction with Rac-1 and Vav-2.J. Biol. Chem. 285:10748–60.

48. Kanters E, et al. (2008) Filamin B mediatesICAM-1-driven leukocyte transendothelial mi-gration. J. Biol. Chem. 283:31830–9.

49. Moos T, Morgan EH. (2004) The metabolism ofneuronal iron and its pathogenic role in neurologi-cal disease: review. Ann. N.Y. Acad. Sci. 1012:14–26.

50. van Campenhout A, van Campenhout CM, Lagrou AR, Manuel-Y-Keenov B. (2007) Transferrinmodifications and lipid peroxidation: implicationsin diabetes mellitus. Free Radic. Res. 37:1069–77.

51. Ponce ML, Kleinman HK. (2003) Identification ofredundant angiogenic sites in laminin alpha1and gamma1 chains. Exp. Cell Res. 285:189–95.

prestroke proteomic changes in cerebral microvessels in ... · metaprotein-pathway analyses of...

Documents