Proteomic differences in brain vessels of Alzheimer’s disease mice: Normalization by PPARγ agonist pioglitazone

Abstract

Cerebrovascular insufficiency appears years prior to clinical symptoms in Alzheimer’s disease. The soluble, highly toxic amyloid-β species, generated from the amyloidogenic processing of amyloid precursor protein, are known instigators of the chronic cerebrovascular insufficiency observed in both Alzheimer’s disease patients and transgenic mouse models. We have previously demonstrated that pioglitazone potently reverses cerebrovascular impairments in a mouse model of Alzheimer’s disease overexpressing amyloid-β. In this study, we sought to characterize the effects of amyloid-β overproduction on the cerebrovascular proteome; determine how pioglitazone treatment affected the altered proteome; and analyze the relationship between normalized protein levels and recovery of cerebrovascular function. Three-month-old wildtype and amyloid precursor protein mice were treated with pioglitazone- (20 mg/kg/day, 14 weeks) or control-diet. Cerebral arteries were surgically isolated, and extracted proteins analyzed by gel-free and gel-based mass spectrometry. 193 cerebrovascular proteins were abnormally expressed in amyloid precursor protein mice. Pioglitazone treatment rescued a third of these proteins, mainly those associated with oxidative stress, promotion of cerebrovascular vasocontractile tone, and vascular compliance. Our results demonstrate that amyloid-β overproduction perturbs the cerebrovascular proteome. Recovery of cerebrovascular function with pioglitazone is associated with normalized levels of key proteins in brain vessel function, suggesting that pioglitazone-responsive cerebrovascular proteins could be early biomarkers of Alzheimer’s disease.

Introduction

Cognitive decline in Alzheimer’s disease (AD) is the end result of chronic brain synaptic and neuronal alterations, as well as early changes in the cerebrovasculature.^1–3 Cerebral blood vessels in AD exhibit alterations in both structure and function.^4,5 Chronic cerebral hypoperfusion typically appears prior to measurable mnemonic impairment and persists well into the latest stages of the disease.^1,3 Recommended as a robust preclinical biomarker of AD,³ cerebrovascular hypoperfusion has also been correlated with negative treatment outcome in AD.² Similarly, reduced cerebral blood flow (CBF) is an early feature of AD transgenic mouse models, including transgenic mice expressing a mutated form of the human amyloid precursor protein (APP mice), and precedes Aβ plaque deposition and cognitive deficits.^6–8

Increased levels of Aβ peptide, especially soluble Aβ species, are the primary instigator of cerebrovascular dysfunction in APP mice. Young APP mice lacking Aβ plaque deposition, but harboring increased levels of soluble Aβ, exhibit oxidative stress in their brain vasculature,^9,10 a major contributor of vascular abnormalities, including reductions in endothelial-mediated dilatory function^9,10 and CBF.¹¹ Oxidative stress and reductions in CBF are also observed in the brain vasculature of wildtype (WT) mice exposed to Aβ.⁹ Consistent with its detrimental effects on mouse cerebrovasculature, excess Aβ has also been shown to disrupt cerebral hemodynamics in isolated human cerebral arteries.¹² We previously demonstrated that the peroxisome proliferator-activated receptor gamma (PPARγ) agonist pioglitazone, which belongs to the thiazolidinedione (TZD) class of anti-diabetic drugs, fully reversed Aβ-induced cerebrovascular dysfunction in APP mice.^13,8 Pioglitazone exerts these benefits likely through vascular transcription factor PPARγ, which is capable of activating or trans-repressing a multitude of genes.¹⁴

While a previous study using proteomics on microvessel-enriched fractions of whole-brain homogenate in an AD mouse model found age-related changes,¹⁵ to date, no study has characterized the differential protein expression patterns in surgically extracted and individually isolated cerebral arteries between WT and APP mice, nor tested the impact of effective pharmacotherapy on cerebrovascular function. Addressing this gap in knowledge, our aims were to (a) characterize the effect of Aβ overproduction on the cerebrovascular proteome of APP mice; (b) determine the extent to which pioglitazone rescues the Aβ-altered cerebrovascular proteome; and (c) determine the link between normalized protein levels by pioglitazone and cerebrovascular function recovery. Our study is timely considering pioglitazone has recently entered a Phase III trial for delay of onset of AD in high risk, cognitively normal participants.¹⁶

Material and methods

Mice

Six-month-old male APP mice (line J20, n = 18) and their wildtype (WT, n = 18) littermates (C57BL/6J) were used. The APP/J20 line carries one copy of the familial early onset AD Swedish (670/671KM→NL) and Indiana (717V→F) APP mutations on the C57BL/6J background,¹⁷ which leads to increased production of APP-derived Aβ peptides. Mice were housed under a 12-h light-dark cycle, in a room with controlled temperature (23℃) and humidity (50%). All experiments were approved by the Animal Ethics Committee of the Montreal Neurological Institute and complied with the Canadian Council on Animal Care and the Animal Research: Reporting In Vivo Experiments (ARRIVE) guidelines.

Drug treatment

A 14-week treatment regime was initiated in three-month-old mice, an age when soluble Aβ is present in APP mice, but not deposited forms of Aβ peptide.¹⁷ It should be noted that vascular Aβ deposits or cerebral amyloid angiopathy (CAA) is clearly detectable in J20 APP mice at 12-months of age in the distal segments and small branches of the arteries of the circle of Willis.¹⁸ The treated WT and APP groups (n = 9 each) received a pioglitazone-containing-diet [20 mg/kg/day (Takeda Pharmaceuticals)], while the untreated groups (n = 9 each) received non-medicated control-diet. This dose of pioglitazone was selected since it previously showed efficacy in restoring cerebrovascular function in APP mice.¹³ Both diets were mixed in Teklad Rodent chow (Research Diets Inc., New Brunswick, NJ, USA). Mice had access to tap water and food ad libitum.

Surgical extraction of cerebral arteries

The circle of Willis and major cerebral arteries along with their main branches were surgically removed from mice and individually stripped from the attached pia matter to obtain a clean preparation of vascular tissue, as previously described.¹⁹ Arteries extracted from three mice were combined to constitute one biological replicate and stored at −80℃. Three biological replicates (B1, B2, B3) were prepared for each of the four groups. Such a preparation of readily accessible cerebral blood vessels that are bathed in cerebrospinal fluid offers a reliable window to the brain microvasculature since both vascular beds are exposed to increased levels of soluble Aβ proteins, and previous proteomic analyses have revealed a very high level of similarities between these two segments of the brain circulation.¹⁹

Cerebrovascular proteomics workflow

Protein extraction from mouse cerebral arteries was performed using our recently published and validated protocol.¹⁹ A brief description of the procedure is provided in Supplemental information.

Generation of statistically significant protein lists

For generation of protein lists, statistical analyses were performed on proteins with ≥2 peptides, peptide score ≥35, and fold-change ≥1.5. In addition, proteins detected using both gel and gel-free approaches were retained only if the change in protein expression level was in the same direction. Visual validation of average peptide intensity distribution per biological and technical replicate was performed on all statistically identified proteins.

Proteins differentially expressed between WT and APP cerebral arteries were identified through parametric Student t and non-parametric Mann–Whitney U tests, with p < 0.05 considered statistically significant. Similarly, to identify differentially expressed proteins in the cerebral vasculature of APP mice that were affected by pioglitazone, we performed (a) parametric Student t-test, (b) non-parametric Mann–Whitney U test between APP and pioglitazone-treated APP mice, and (c) a two-way ANOVA taking transgene and treatment as the two factors (MATLAB, http://www.mathworks.com/products/matlab/). Proteins were considered fully rescued by pioglitazone treatment if they did not significantly differ from WT protein levels. Significance was considered to be p < 0.05.

Characterization of proteins

APP peptide alignment was performed using ClustalW2, a sequence alignment program for DNA or proteins (http://www.ebi.ac.uk/Tools/msa/clustalw2/). Proteins known to directly or indirectly (secondary and tertiary interactions only) interact with APP and/or with each other were identified by querying PubMed, the STRING database that contains known and predicted protein interactions (http://string-db.org, version 9.1, confidence: medium to high), the IntAct database (http://www.ebi.ac.uk/intact/), and an in-house database that is a compendium of several databases, including BIND, BioGRID, HPRD, HIMAP, and EcoCyc. Species selected for the analyses were Mus musculus and/or Homo sapiens. APP-associated proteins, i.e. proteins co-mentioned with APP or amyloid-β peptides in the literature that may or may not be a direct interacting partner, were identified by querying the PubMed reference library. Proteins were indicated as associated with increased AD risk if present in AlzGene (http://alzgene.org), a database of AD genetic association studies, and/or by querying PubMed. Subcellular localizations of proteins were determined using data from Gene Ontology (http://geneontology.org/) and the Universal Protein Resource (www.uniprot.org) databases combined. The PANTHER Classification System (http://www.pantherdb.org/pathway/) was used to classify proteins by ‘molecular function’. Genes (of proteins) harboring peroxisome proliferator responsive element (PPRE) and PACM (PPAR-associated conserved motif) were identified using a published database.²⁰ Proteins associated with extracellular microvesicles (or exosomes) released from brain endothelial cells (BECs) into the blood were identified using the BEC Exosome database.²¹ These proteins are a rich source of brain vasculature-specific biomarkers, as well as receptors known for delivering molecules across the blood–brain barrier.

Results

Proteins differentially expressed in brain vessels of APP and WT mice

We identified APP in both gel-free and gel-based proteomic approaches from one peptide (LVFFAEDVGSNK, Figure 1(a)) with protein scores of 93 and 71, respectively. Compared to WT mice, APP protein levels were significantly (p < 0.01) increased in the cerebral arteries of APP mice (Figure 1(b) and (c)). We identified an additional 192 proteins, corresponding to 3% of total proteins detected, differentially expressed between WT and APP mice cerebral arteries (Supplementary Table 1). The majority (131 proteins, 68.2%) of these proteins were identified using both parametric Student’s t and non-parametric Mann–Whitney U distributions (p < 0.05), while 50 (26.1%) and 11 (5.7%) proteins were detected using only parametric and non-parametric analyses, respectively (Figure 2(a)). Moreover, 78 (40.6%) of the 192 proteins were also identified using two-way ANOVA (Supplementary Table 1). The majority (62.5%) of proteins in the APP cerebral vasculature were upregulated with fold-changes ranging from ±1.5 to ±5.0 (Figure 2(b), Supplementary Table 1). As much as 122 (63.2%) of the differentially expressed proteins were found to interact with each other, and 32 (16.7%) were also present in the BEC Exosome database (Supplementary Table 1).

Figure 1.

Quantification of APP protein levels in cerebral arteries. (a) Alignment of the detected APP peptide (LVFFAEDVGSNK) in relation to the full-length APP protein. Also, indicated are the cleavage sites for the α, β, and γ secretases, as well as the AD-associated Swedish (KM) and Indiana (V) mutations. (b) Peptide intensity distribution of WT and APP mice as visualized using MSight. For both genotypes, biological replicate 1, each consisting of vessels extracted from three different mice is shown. The shaded grey area and/or the purple cross hair represent APP protein level. (c) As expected, APP mice (blue) display significantly elevated cerebrovascular APP protein levels relative to WT controls (green). Pioglitazone therapy moderately rescued APP protein levels in treated APP (red), but had no effect on WT (grey) mice. Error bars denote standard error or SEM and ⋆⋆ p < 0.01.

PIO: pioglitazone.

Figure 2.

Proteins differentially expressed in brain vessels of APP mice. (a) Distribution (in percentage) of proteins identified with either parametric or non-parametric Student t-test or both statistical analyses. (b) Fold-change distribution of differentially expressed proteins along with the percentage of upregulated or downregulated proteins in the APP vasculature. (c) Protein interactors of APP or proteins known to directly (black line) or indirectly (secondary: purple line and tertiary: grey line interactions only) interact with APP were identified using the STRING, IntAct, and in-house, and/or PubMed reference library. Note that only a few indirectly interacting proteins have been shown. Proteins associated with increased AD risk in the AlzGene database and/or by querying the PubMed reference library are identified in red and underlined (e.g. APOE). Average (d) APOE, and (e) HSP90B1, AQP1, and ACTN4 peptide/s intensities for WT (green), WT(pio) (grey), APP(blue), and APP(pio) (red). Error bars denote standard error or SEM, ⋆ p < 0.05 and ⋆⋆⋆ p < 0.001 denote significant genotype effect using two-way ANOVA. PIO: pioglitazone.

Proteins interacting or associated with APP

We found that over a third (68 proteins) of the differentially expressed proteins were either (a) protein interactors of APP (60 proteins) and/or (b) associated with APP (eight unique proteins, namely AIFM1, ATP5F1, BSN, CLOCK, DCN, PLA2G4A, PPP2CB, and SRF; protein names and additional information in Table 1). Of the established interactors, the 40 identified to interact directly with APP are as follows: ACTN4, ALDH7A1, APBB3, APOE, AQP1, ASPH, CCNG1, CYFIP, DEPDC6, GABRA4, GAK, GAS2, HAVCR2, HSP90AA1, HSP90BI, HSPA5, HTR2C, KIF2C, MAPRE1, MYH9, NAPA, NFASC, PCBP1, PGM1, PRDX5, RAB5C, RNF151, SENP1, SENP2, SERBP1, SHOC2, SOD1, STX12, TCP11L2, TMCC2, TPI1, TRIM41, USP5, VIM, and VTN (Table 1, Figure 2(c)). Of these, all except AQP1, CYFIP1, DEPDC6, HAVCR2, HTR2C, MAPRE1, NAPA, PGM1, SENP2, STX12, and TRIM41 were upregulated in APP mice (Figure 2(c)). Representative increased peptide levels of APOE, ACTN4, and HSP90B1 and decreased levels of AQP1 are illustrated in Figure 2(d) and (e). Nineteen established interactors of APP, 14 directly interacting (ACTN4, ASPH, CYFIP1, HSP90AA1, HSP90B1, HSPA5, MAPRE1, MYH9, PCBP1, RAB5C, SERBP1, SHOC2, TPI1, and VIM) and 5 indirectly interacting (CCT2, DES, HSPB1, RDX, RPLP0) were also present in the BEC Exosome database (Table 1).

Table 1.

Cerebral arterial proteins differentially expressed between WT and APP mice.

Protein ID	Protein Symbol	Protein Name	Protein Interactors of APP	Associated with APP or AD (*) or both (bold) in Pubmed (PMID provided)	Present in AlzGene Db	Linked to APP and AD	Direction of change (APP)
Q8QZT1	ACAT1	Acetyl-CoA acetyltransferase, mitochondrial	inD	20398792 24587158	y	b	↑
P57780	ACTN4	Alpha-actinin-4	D	23055000*	–	b	↑
Q9Z0X1	AIFM1	Apoptosis-inducing factor 1, mitochondrial	–	14767566	–	–	↓
Q9DBF1	ALDH7A1	Alpha-aminoadipic semialdehyde dehydrogenase	D	21832049	–	–	↑
O70423	AOC3	Membrane primary amine oxidase	–	11461168*	–	–	↑
Q8R1C9	APBB3	Amyloid-beta A4 precursor protein-binding family B member 3	D	14527950 16973241	–	b	↑
P08226	APOE	Apolipoprotein E	D	multiple	y	b	↑
Q02013	AQP1	Aquaporin-1	D	16871401* 17123487* 18509662 21107133* 21955788* 22805778* 19687153	–	b	↓
Q8BSY0	ASPH	Aspartyl/asparaginyl beta-hydroxylase	D	21832049	–	–	↑
Q9CQQ7	ATP5F1	ATP synthase subunit b, mitochondrial	–	24587158	–	–	↑
P50516	ATP6V1A	V-type proton ATPase catalytic subunit A	inD	24587158	–	–	↓
O88737	BSN	Protein bassoon	–	24587158	–	–	↓
P51945	CCNG1	Cyclin-G1	D	21832049 12901840* 17006763*	–	b	↑
P80314	CCT2	T-complex protein 1 subunit beta	inD	11441917*	–	b	↑
P53566	CEBPA	CCAAT/enhancer-binding protein alpha	–	18405661*	–	–	↑
O08785	CLOCK	Circadian locomoter output cycles protein kaput	–	22634208 23781009*	–	b	↑
Q9CPQ1	COX6C	Cytochrome c oxidase subunit 6C	inD	–	–	–	↑
Q7TMB8	CYFIP1	Cytoplasmic FMR1- interacting protein 1	D	–	–	–	↓
P28654	DCN	Decorin	–	1370306* 7793988* 17321057* 17660861* 21310214	–	b	↓
Q570Y9	DEPDC6	DEP domain-containing mTOR-interacting protein	D	21832049	–	–	↓
P31001	DES	Desmin	inD	–	–	–	↑
O08749	DLD	Dihydrolipoyl dehydrogenase, mitochondrial	inD	–	y	b	↑
Q9D832	DNAJB4	DnaJ homolog subfamily B member 4	inD	–	–	–	↑
P70372	ELAVL1	ELAV-like protein 1	–	19221430*	–	–	↓
Q8VHX6	FLNC	Filamin-C	–	9437013* 20847418* 22815492*	–	–	↑
Q9D6F4	GABRA4	Gamma-aminobutyric acid receptor subunit alpha-4	D	21832049	–	–	↑
Q99KY4	GAK	Cyclin G-associated kinase	D	21832049	–	–	↑
P11862	GAS2	Growth arrest-specific protein 2	D	21832049	–	–	↑
Q01097	GRIN2B	Glutamate [NMDA] receptor subunit epsilon-2	inD	–	y	b	↓
Q8BH60	GOPC	Golgi-associated PDZ and coiled-coil motif-containing protein	–	–	–	–	↑
Q8BMS1	HADHA	Trifunctional enzyme subunit alpha, mitochondrial	–	20538040*	–	–	↑
Q8VIM0	HAVCR2	Hepatitis A virus cellular receptor 2 homolog	D	21832049	–	–	↓
P07901	HSP90AA1	Heat shock protein HSP 90-alpha	D	19449461 16049941	–	b	↑
P08113	HSP90B1	Endoplasmin	D	17934213	–	–	↑
P20029	HSPA5	78 kDa glucose-regulated protein	D	9748217 17132139 17934213	y	b	↑
P14602	HSPB1	Heat shock protein beta-1	inD	16635482* 21670152*	–	b	↑
P34968	HTR2C	5-hydroxytryptamine receptor 2C	D	–	y	b	↓
Q9Z0Y7	IRS4	Insulin receptor substrate 4	inD	24274089*	–	b	↓
A2ARA8	ITGA8	Integrin alpha-8	inD	–	–	–	↑
Q9Z0R4	ITSN1	Intersectin-1	–	24573290 16442855* 21876463*	–	b	↓
Q922S8	KIF2C	Kinesin-like protein KIF2C	D	21832049	–	–	↑
Q6PFD6	KIF18B	Kinesin-like protein KIF18B	–	–	y	–	↑
Q80U28	MADD	MAP kinase-activating death domain protein	–	15007167* 16253995* 22678883*	–	–	↓
Q61166	MAPRE1	Microtubule-associated protein RP/EB family member 1	D	21832049	–	–	↓
O70583	MID1	Midline-1	–	21098287* 22613722*	–	–	↓
Q3V1L6	MTMR11	Myotubularin-related protein 11	–	–	–	–	↑
Q8VDD5	MYH9	Myosin-9	D	16049941	–	b	↑
Q9DB05	NAPA	Alpha-soluble NSF attachment protein	D	21832049 24573290	–	–	↓
Q810U3	NFASC	Neurofascin	D	17934213	–	–	↑
P60335	PCBP1	Poly(rC)-binding protein 1	D	21832049	–	–	↑
P08003	PDIA4	Protein disulfide-isomerase A4	inD	–	–	–	↑
Q9D0F9	PGM1	Phosphoglucomutase-1	D	21832049 17522440*	–	b	↓
P47713	PLA2G4A	Cytosolic phospholipase A2	–	17713604 22188721 multiple*	–	b	↑
Q3UYC0	PPM1H	Protein phosphatase 1H	–	–	y	–	↓
Q6R891	PPP1R9B	Neurabin-2	–	–	–	–	↑
P62715	PPP2CB	Serine/threonine-protein phosphatase 2A catalytic subunit beta isoform	–	23020770 multiple*	–	b	↑
P99029	PRDX5	Peroxiredoxin-5, mitochondrial	D	21832049 24573290	–	–	↑
P15331	PRPH	Peripherin	inD	–	–	–	↑
P35278	RAB5C	Ras-related protein Rab-5C	D	21832049	–	–	↑
P60764	RAC3	Ras-related C3 botulinum toxin substrate 3	inD	–	–	–	↓
P26043	RDX	Radixin	inD	–	–	–	↓
Q9CQ29	RNF151	RING finger protein 151	D	21832049	–	–	↑
P14869	RPLP0	60S acidic ribosomal protein P0	inD	–	–	–	↑
P59110	SENP1	Sentrin-specific protease 1	D	21832049 24778618*	–	b	↑
Q91ZX6	SENP2	Sentrin-specific protease 2	D	21832049	–	–	↓
Q9CY58	SERBP1	Plasminogen activator inhibitor 1 RNA-binding protein	D	21832049	–	–	↑
Q8R0X7	SGPL1	Sphingosine-1-phosphate lyase 1	–	–	y	–	↑
O88520	SHOC2	Leucine-rich repeat protein SHOC-2	D	21832049	–	–	↑
P08228	SOD1	Superoxide dismutase [Cu-Zn]	D	10195200 22790929 multiple*	–	b	↑
Q9JM73	SRF	Serum response factor	–	22912719 17215356* 19098903* 22850315*	–	b	↑
Q9ER00	STX12	Syntaxin-12 (alias Syntaxin-13)	D	20925061	–	–	↓
Q8JZP2	SYN3	Synapsin-3	–	–	y	–	↑
Q8K1H7	TCP11L2	T-complex protein 11-like protein 2	D	21832049	–	–	↑
P39447	TJP1	Tight junction protein ZO-1	inD	22745485*	–	b	↑
Q80W04	TMCC2	Transmembrane and coiled-coil domains protein 2	D	21832049 23409049	–	b	↑
Q3UHK8	TNRC6A	Trinucleotide repeat- containing gene 6A protein	–	–	–	–	↓
P17751	TPI1	Triosephosphate isomerase	D	21832049	–	–	↑
Q5NCC3	TRIM41	E3 ubiquitin-protein ligase TRIM41	D	21832049	–	–	↓
O08747	UNC5C	Netrin receptor UNC5C	–	24866402*	–	–	↑
P56399	USP5	Ubiquitin carboxyl-terminal hydrolase 5	D	21832049 24573290	–	–	↑
P20152	VIM	Vimentin	D	21832049 23038755	–	–	↑
P29788	VTN	Vitronectin	D	21832049	–	–	↑
Q60989	XIAP	Baculoviral IAP repeat- containing protein 4	inD	16599295* 17292615*	–	b	↑
O88967	YME1L1	ATP-dependent metalloprotease YME1L1	inD	–	–	–	↓

Note: Only proteins mentioned in the manuscript are listed, while the complete list of proteins differentially expressed between APP and WT mice is provided in Supplementary Table 1. All proteins were identified with ≥2 peptides, exhibiting peptide scores ≥35. Protein IDs used follow the UniProt assignment (www.uniprot.org). In COLUMN 4 (Protein Interactors of APP), proteins known to directly interact with APP are indicated as ‘D’ and those known to indirectly (secondary or tertiary only) interact with APP are indicated as ‘inD’. These protein interactors of APP were identified using STRING, IntAct, an in-house database, and by querying PubMed. D indicates that the protein was listed as a direct interactor (medium to high confidence; accessed March 14, 2016) in the STRING database. COLUMN 5 provides PubMed identifiers or PMIDs of proteins/genes associated with APP only (PMID or multiple), AD only (PMID* or multiple*), or both (PMID or multiple*). For each of the three categories, PMIDs are listed in numerical order. COLUMN 7 (Linked to APP and AD) identifies proteins (‘b’) that interact/or are associated with APP and are associated with AD. AD: Alzheimer’s disease. Symbols: ↑, up; ↓, down; y, protein present in the AlzGene Db.

Proteins associated with AD

We identified 39 (20.3%) AD-associated proteins in our differentially expressed list (Table 1). Ten, namely, ACAT1, APOE, DLD, GRIN2B, HSPA5, HTR2C, KIF18B, PPM1H, SGPL1, and SYN3, are products of genes present in the AlzGene database and linked with increased AD risk (Table 1). Of these, APOE, HSPA5, and HTR2C are known to directly interact with APP (Figure 2(c)), while ACAT1, DLD, and GRIN2B demonstrate indirect interaction. Overall, we found that more than half (27/39 proteins, 69.2%) of the proteins identified as “associated with AD” were also known to “interact or associate with APP” (refer to subsection “Proteins interacting or associated with APP” and Table 1).

Molecular functions disrupted by mutant APP overexpression

The top five locations of differentially expressed proteins were cytoplasm (40.0%), membrane (36.5%), nucleus (29.2%), mitochondrion (8.9%), and cell junction (7.3%) (Supplemental Figure 1(a)). 143 (75%) of these proteins also had known molecular functions in the PANTHER database, which corresponded to (i) proteins binding to nucleic acids (27 proteins), (ii) cytoskeletal filaments (14 proteins), (iii) receptors (14 proteins), (iv) regulatory molecules (13 proteins), and (v) oxidoreductases (12 proteins) (Supplementary Table 2).

Differentially expressed proteins rescued by pioglitazone in APP mice

We found that pioglitazone induced a near-significant reduction of APP levels (p = 0.07) in APP mice, but had no detectable effect on WT mice (Figure 1(c)). Pioglitazone further rescued 60 of the 192 (31%) differentially expressed proteins, with 45 (75%) proteins demonstrating full normalization, and 15 (25%) exhibiting partial albeit significant recovery to WT levels (Table 2).

Table 2.

Cerebral arterial proteins normalized by pioglitazone.

Protein ID	Protein symbol	Protein name	Direction of change (APP)	Normalization by Pioglitazone	Two-way ANOVA Interaction Term
Q8VCH0	ACAA1B	3-ketoacyl-CoA thiolase B, peroxisomal	↓	full	–
Q8JZV7	AMDHD2	Putative N-acetylglucosamine-6- phosphate deacetylase	↓	full	–
Q9CZK6	ANKS3	Ankyrin repeat and SAM domain-containing protein 3	↑	full	–
Q99PT1	ARHGDIA	Rho GDP-dissociation inhibitor 1	↑	full	–
Q8C9S8	ATG4A	Cysteine protease ATG4A	↓	full	–
Q9D4K7	CCDC105	Coiled-coil domain-containing protein 105	↑	full	–
Q99PF4	CDH23	Cadherin-23	↑	partial	–
Q69ZA1	CDK13	Cell division protein kinase 13	↓	full	–
P53566	CEBPA	CCAAT/enhancer-binding protein alpha	↑	full	–
Q9R0M0	CELSR2	Cadherin EGF LAG seven-pass G-type receptor 2	↓	full	–
Q6PG95	CRAMP1L	Protein cramped-like	↓	partial	–
P56395	CYB5A	Cytochrome b5	↑	full	–
O08749	DLD	Dihydrolipoyl dehydrogenase, mitochondrial	↑	full	–
Q6XUX1	DSTYK	Dual serine/threonine and tyrosine protein kinase	↑	full	y
P70372	ELAVL1	ELAV-like protein 1	↓	full	–
Q99JW1	FAM108A	Abhydrolase domain-containing protein FAM108A	↑	partial	–
Q8VHX6	FLNC	Filamin-C	↑	full	–
Q9CYL5	GLIPR2	Golgi-associated plant pathogenesis-related protein 1	↑	full	–
Q8BH60	GOPC	Golgi-associated PDZ and coiled-coil motif-containing protein	↑	full	–
Q8BMS1	HADHA	Trifunctional enzyme subunit alpha, mitochondrial	↑	full	–
P62806	HIST1H4A	Histone H4	↑	full	–
Q64523	HIST2H2AC	Histone H2A type 2-C	↓	full	–
P20029	HSPA5	78 kDa glucose-regulated protein	↑	full	–
P14602	HSPB1	Heat shock protein beta-1	↑	full	–
Q9Z0R4	ITSN1	Intersectin-1	↓	full	–
Q80V70	MEGF6	Multiple epidermal growth factor-like domains protein 6	↑	full	–
Q791V5	MTCH2	Mitochondrial carrier homolog 2	↓	full	–
Q3V1L6	MTMR11	Myotubularin-related protein 11	↑	partial	–
Q8VDD5	MYH9	Myosin-9	↑	full	–
Q7TSZ8	NACC1	Nucleus accumbens-associated protein 1	↑	full	–
Q9DB05	NAPA	Alpha-soluble NSF attachment protein	↓	partial	–
Q9Z1P6	NDUFA7	NADH dehydrogenase [ubiquinone] 1 alpha subcomplex subunit 7	↓	partial	–
P29341	PABPC1	Polyadenylate-binding protein 1	↑	partial	y
Q8CEE6	PASK	PAS domain-containing serine/ threonine-protein kinase	↓	partial	–
P08003	PDIA4	Protein disulfide-isomerase A4	↑	full	–
Q9D0F9	PGM1	Phosphoglucomutase-1	↓	full	–
Q6R891	PPP1R9B	Neurabin-2	↑	partial	y
P99029	PRDX5	Peroxiredoxin-5, mitochondrial	↑	full	y
P15331	PRPH	Peripherin	↑	full	–
P35278	RAB5C	Ras-related protein Rab-5C	↑	partial	y
Q8BIV3	RANBP6	Ran-binding protein 6	↑	full	–
P26043	RDX	Radixin	↓	partial	–
P50543	S100A11	Protein S100-A11	↑	full	–
Q8R2U0	SEH1L	Nucleoporin SEH1	↓	full	–
Q91ZX6	SENP2	Sentrin-specific protease 2	↓	full	–
O88520	SHOC2	Leucine-rich repeat protein SHOC-2	↑	partial	–
Q924W5	SMC6	Structural maintenance of chromosomes protein 6	↑	full	–
Q91ZR2	SNX18	Sorting nexin-18	↓	partial	–
P08228	SOD1	Superoxide dismutase [Cu-Zn]	↑	full	–
Q9D489	SOHLH2	Spermatogenesis- and oogenesis- specific basic helix-loop- helix-containing protein 2	↑	partial	–
Q9DC40	TELO2	Telomere length regulation protein TEL2 homolog	↑	full	–
Q80W04	TMCC2	Transmembrane and coiled-coil domains protein 2	↑	full	–
Q3UHK8	TNRC6A	Trinucleotide repeat-containing gene 6A protein	↓	full	–
P17751	TPI1	Triosephosphate isomerase	↑	full	y
O08747	UNC5C	Netrin receptor UNC5C	↑	full	–
P20152	VIM	Vimentin	↑	full	–
Q8BX70	VPS13C	Vacuolar protein sorting-associated protein 13C	↑	full	–
Q60989	XIAP	Baculoviral IAP repeat-containing protein 4	↑	full	y
P23607	ZFA	Zinc finger autosomal protein	↑	full	y
Q5DU37	ZFYVE26	Zinc finger FYVE domain-containing protein 26	↑	partial	–

Note: Listed are cerebral arterial proteins significantly rescued (p < 0.05, t-test) by pioglitazone in APP mice compared to untreated APP mice. Only proteins identified with ≥ 2 peptides and exhibiting peptide scores ≥35 are included. Protein IDs used are those assigned by UniProt (www.uniprot.org). In COLUMN 2, Protein Symbol, Protein Symbol, and Protein Symbol, denotes associated with Alzheimer’s disease, interacting with and/or associated with APP, or both respectively. COLUMN 4 (two-way ANOVA) identifies (‘y’) proteins that are rescued by pioglitazone when analyzed using a two-way ANOVA taking genotype and treatment as factors. Shaded rows identify proteins present in the brain endothelial cell Exosome database and normalized by pioglitazone. Symbols: ↑, up; ↓, down.

Of the pioglitazone-rescued proteins, 8 (13%) had known PPARγ-PPRE/PACM binding sites (Figure 3(a), Table 1), 16 (27%) were found in circulating BEC microvesicles (Table 2, Figure 4(a)), and 4, namely FLNC, RAB5C, TNRC6A, and TPI1 exhibited both features. Of the aforementioned APP binding proteins, levels of primary interactors HSPA5, MYH9, PGM1, PRDX5, SENP2, SOD1, TMCC2, TPI1, and VIM, and secondary interactors DLD, HSPB1, PDIA4, PRPH, and XIAP were normalized (Figure 3(a), Table 1). Levels of AD-associated AlzGene proteins DLD and HSPA5 were also rescued by pioglitazone treatment (Figure 3(a), Table 1). Figure 3(b) demonstrates peptide levels of five representative proteins, namely SOD1, XIAP, TPI1, RAB5C, and PPP1R9B, normalized by pioglitazone.

Figure 3.

Pioglitazone-recovered proteins in the APP vasculature. (a) Pioglitazone-rescued protein that (i) directly or indirectly interact with APP, and/or (ii) harbor known PPARγ-PPRE/PACM binding sites in their genomic sequences. (b) Average SOD1, XIAP, TPI1, RAB5C, and PPP1R9B peptide intensities for WT (green), WT(pio) (grey), APP (blue), and APP(pio) (red) mice.

Note: The function listed for RAB5C is probable, but not confirmed. Error bars denote standard error or SEM and ⋆ p < 0.05 denote significant interaction effect using two-way ANOVA. NO: nitric oxide; PIO: pioglitazone.

Figure 4.

Cerebrovascular functional recovery by pioglitazone in APP mice. (a) 16 cerebrovascular shed proteins (or proteins present in circulating BEC microvesicles) constituting 27% of pioglitazone-rescued proteins were identified. Fourteen of these shed proteins were upregulated (↑) in APP mice and normalized by pioglitazone: ARHGDIA, FLNC, GLIPR2, HIST1H4A, HSPA5, HSPB1, MYH9, PABPC1, RAB5C, S100A11, SHOC2, TPI1, VIM, and VPS13C, whereas two were downregulated (↓) in APP mice and normalized by treatment: RDX and TNRC6A. These proteins could serve as surrogates to indicate efficacy and/or select responders in pioglitazone clinical trials. (b) APP overexpression and consequent Aβ increase in APP mice alter levels of proteins involved in mediating (i) oxidative stress (both ROS generators and scavengers), (ii) vascular elasticity and compliance, and (iii) increased basal tone, which then promotes impaired vasodilation. Pioglitazone treatment normalizes levels of these proteins and recovers cerebrovascular function. Of the proteins shown, FLNC, HSPB1, MYH9, TPI1, and VIM are shed proteins. Symbols: ↑, up and ↓, down.

The top five locations of pioglitazone-rescued proteins were cytoplasm (43.0%), membrane (33.3%), nucleus (33.3%), mitochondrion (11.7%), and endoplasmic reticulum (6.7%) (Supplemental Figure 1(a)). The majority (>75%) of disturbed molecular functions were normalized to various degrees by pioglitazone, the top five categories being: isomerase activity, calcium binding, signaling, oxidation-reduction, and regulatory processes (Supplemental Figure 1(b)).

Discussion

Using two complementary and previously validated proteomic approaches,¹⁹ we found perturbations in the cerebrovascular proteome of transgenic mice overexpressing the AD-associated mutant-APP gene. We detected increase in the protein-level of APP, and alterations in the levels of an additional 192 proteins. Breakdown of the differentially expressed proteins by subcellular localization was in good agreement with the relative sizes of cellular compartments (e.g. membrane > mitochondria),¹⁹ demonstrating that our proteomic approach was unbiased towards any one cellular compartment. Our finding that 82 (43%) of these proteins are linked to the APP and/or AD literature strengthens our confidence that the protein list contains numerous novel cerebrovascular candidates affected by APP overexpression. We identified proteins involved in/or responding to APP overexpression associated with molecular pathogenic changes, such as RNA/DNA damage, vascular cytoskeleton alterations, and deregulation of the oxidoreductase system. We further demonstrated that approximately a third of the APP-altered cerebrovascular proteins were partially or fully normalized to WT levels following pioglitazone treatment. Since more than a quarter (27%) of pioglitazone-rescued proteins corresponded to BEC microvesicles,²¹ they could serve as markers of efficacy in pioglitazone clinical trials (Figure 4(a)). While prior study suffered from limited sensitivity due to suboptimal cerebral vascular tissue extraction technique from brain homogenates,¹⁵ our use of surgically extracted cerebral arteries, individually isolated and cleaned of the attached pial membrane,¹⁹ has allowed us to identify a large number of relevant vascular proteins that could be tested in functional assays to validate the identified pathways as key therapeutic targets.

We previously showed that pioglitazone is highly effective at reversing soluble Aβ-induced cerebrovascular dysfunction in APP mice, in particular, impaired vasodilation and reductions in neurovascular coupling.^13,8 Since oxidative stress, alterations in vascular elasticity and increased basal tone can impair vascular dilation, we propose that rescue of these three categories of proteins by pioglitazone underlies the documented pioglitazone-mediated normalization of cerebrovascular function (Figure 4(b)). In the current study, we found that pioglitazone showed a near-significant ability to reduce APP protein levels, a result that was not observed in our previous study in older APP mice.¹³ The observed benefits of pioglitazone may thus be mediated both by reduction of cerebrovascular APP protein levels, and direct effects on the cerebrovasculature, as discussed below.

Mutant APP overexpression, oxidative stress, and pioglitazone

Pioglitazone reduces cerebrovascular APP/Aβ toxicity

The increase in cerebrovascular APP levels in APP mice was moderately reduced following early, long-term (three months) treatment with pioglitazone. Our finding is in line with a previous report that PPARγ activation by another TZD, rosiglitazone, increases APP protein ubiquination and degradation in cell cultures.²² Moreover, pioglitazone normalized the upregulated AD-susceptibility protein, HSPA5,²³ a physiological binding partner of APP suggested to facilitate correct folding of APP and limit access of APP to β-/γ-secretases.²⁴ Thus, the increase in cerebrovascular HSPA5 protein levels in APP mice might be a physiological attempt at reducing the amyloidogenic processing of APP and the ensuing Aβ toxicity. Normalized HSPA5 level in pioglitazone-treated APP mice likely results from reduced APP protein levels in cerebral vessels, and decrease in oxidative stress.

Pioglitazone normalizes altered levels of reactive oxygen species scavenging enzymes

We found upregulation of SOD1 and PRDX5 in the APP cerebrovasculature. Increase in these two reactive oxygen species (ROS)-scavenging enzymes signpost the presence of excess ROS. SOD1 reduces cytoplasmic ROS by catalyzing the dismutation of superoxide, and prevention of Aβ-induced cerebrovascular dysfunction has been demonstrated in APP transgenic mice co-overexpressing SOD1.²⁵ Conversely, PRDX5 promotes decrease in mitochondrial ROS via reduction of hydrogen peroxide (H₂O₂) and other peroxide substrates.²⁶ Our finding of increased mitochondrial ROS in the young APP vasculature is coherent with previous studies showing increased ROS production and concurrently elevated SOD2 protein levels^10,13 in brain vessels from adult and aged APP mice. Furthermore, normalized SOD1 and PRDX5 protein levels in young APP mice by pioglitazone parallels the restored levels of SOD2 protein in pioglitazone-treated older mice,¹³ supporting pioglitazone as a potent restorer of antioxidant protection in the brain vasculature (Figure 4(b)).

Pioglitazone rescues altered levels of ROS promoting proteins

We identified two upregulated ROS promoting proteins, namely DLD and HADHA, in the APP brain vasculature that were normalized by pioglitazone (Figure 4(b)). DLD, an indirect protein interactor of APP, is a component of the pyruvate dehydrogenase complex (PDC) and oxidizes pyruvate to acetyl–coenzyme A.²⁷ The enzymatic activity of DLD is coupled to the production of NADH,²⁷ a major substrate for ROS generation by the mitochondrial electron transport chain and cytosolic NOX oxidases in arterial SMC.²⁸ Increase in total NADH has also been associated with the initiation of SMC contraction in peripheral arteries.²⁹ Since the increase in total NADH parallels the rise in PDC activity,²⁹ it is possible that surplus DLD in APP arteries contributes to the destabilization of NADH homeostasis and, consequently, increased ROS. Previous findings implicating DLD as a source of ROS support our hypothesis.³⁰ Single-nucleotide polymorphism variations in the DLD gene have been linked to increased AD risk in certain populations.³¹ In addition, upregulation of the inner mitochondrial membrane protein HADHA also contributes to increased NADH/NAD⁺ ratio in the APP vasculature.³² In contrast, reduced HADHA levels were reported in the tunica media of amyloid-laden pial arteries from AD patients,³³ suggesting that HADHA upregulation in the cerebral arteries of our six-month-old APP mice may be a response to increased soluble Aβ peptide in the absence of detectable amyloid angiopathy at this age.

Pioglitazone rescues altered levels of proteins involved in regulating vascular tone and compliance

We identified three differentially expressed proteins in the APP cerebral vasculature, namely, HSPB1 (or HSP27), TPI1, and ELAVL1 (or HuR), known to promote increased vascular basal tone and, consequently, reduce vascular compliance (Figure 4(b)). The upregulated protein HSP27, a secondary interactor of APP, is known to mediate contraction in brain arteries upon phosphorylation,³⁴ while increased levels of ROS induce HSP27 phosphorylation.³⁵ Taken together, our findings suggest that in the milieu of excess, and potentially phosphorylated HSP27 in the APP vasculature, the response of vascular SMC to vasoconstrictors acting via the p38 mitogen-activated protein kinase (MAPK)/HSP27 pathway, such as angiotensin II,³⁶ thrombin,³⁷ and endothelin-1,³⁴ is augmented. Since normal vascular tone requires balanced vasoconstrictor and vasodilator influences, selective augmentation of vasoconstrictor effects by HSP27 is likely to increase basal tone. The glycolytic enzyme TPI1, known to indirectly interact with APP, was also upregulated in APP mice. TPI1 is a component of the ATP-sensitive potassium (K_ATP) channel complex and plays a role in regulating channel function by altering ATP levels in the intracellular microenvironment.³⁸ TPI1 activity and the subsequent increase in intracellular ATP reportedly inhibit activity at K_ATP channels,³⁸ which are key mediators of vasodilatory responses in brain arteries.³⁹ Notably, inhibition of K_ATP channel activity in SMC promotes increases in vascular basal tone and contraction.⁴⁰ Protein level of HuR, a RNA-binding protein that regulates gene expression by modulating mRNA stability, was decreased in APP mice. In arterial SMC, binding of HuR to the β1-subunit mRNA of soluble guanylate cyclase (sGC) increases its stability, expression, and activity.⁴¹ Since sGC facilitates NO-dependent relaxation in cerebral arteries, decreased HuR levels in the APP vasculature and the accompanying reduction in sGC activity may also contribute to increased basal tone. In line with this hypothesis is the observation that NO-responsive sGC activity is reduced in AD brain.⁴² All three proteins normalized to WT levels in pioglitazone-treated APP mice.

Pioglitazone rescues altered levels of proteins reducing vascular elasticity

Increased cerebral arterial rigidity resulting in decreased dilatory capacity and compromised brain perfusion is present in AD.⁴³ Impaired arterial elasticity most likely stems from soluble Aβ-triggered early onset vascular inflammation,⁴⁴ as seen in AD transgenic mouse models.⁴⁵ In our six-month old APP mice lacking vascular Aβ deposition, we identified four upregulated proteins potentially contributing to impaired arterial elasticity (Figure 4(b)), all of which were normalized by pioglitazone. Two were actin-cross-linking proteins with known PPARγ-PPRE/PACM binding sites.²⁰ The muscle-specific FLNC that regulates cell membrane elasticity⁴⁶ and PPP1R9B that inhibits protein phosphatase 1,⁴⁷ a promoter of SMC relaxation, were both upregulated in APP brain vessels. Pioglitazone also normalized the intermediate filament protein VIM that promotes SMC rigidity and contractile state,⁴⁸ and MYH9, an inherently contractile protein⁴⁹ that acts through increased formation of the potent vasoconstrictor angiotensin II and degradation of the vasodilator bradykinin.⁵⁰ MYH9 and VIM are known interactors of APP, and VIM has also been shown to co-localize with Aβ42 in brain tissues of AD patients and APP mice.⁵¹ Normalization of these upregulated proteins in pioglitazone-treated APP mice strongly supports the beneficial effects of TZD drugs in the reversal of decreased cell membrane elasticity via actin remodeling in the vasculature,⁵² and proper functioning of the perivascular drainage pathway within arterial and arteriolar basement membranes.^53,54

Conclusions

In conclusion, our findings demonstrate that APP overexpression and increased soluble Aβ adversely impacts the cerebrovascular proteome. It should be noted that this detrimental effect occurs in the absence of detectable vascular Aβ deposits or CAA. We also show that pioglitazone reduces oxidative stress and normalizes the levels of several proteins involved in the reduced dilatory capacity of the APP brain vasculature. Moreover, pioglitazone-facilitated improvement in cerebral elasticity has the potential to diminish lymphatic congestion and improve brain Aβ clearance. Together, our findings indicate that pioglitazone significantly alleviates Aβ-induced cerebrovascular impairment and provides a link between protein rescue and cerebrovascular functional recovery by pioglitazone.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work is supported by grants (E.H.) from the Canadian Institutes of Health Research (CIHR, MOP-84275 and MOP-126001) and Takeda Pharmaceuticals North America, Inc., and a CIHR Banting and Best Canada Graduate Scholarship (A.B.).

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Authors’ contributions

APB designed the study, performed experiments, analyzed and interpreted the data, and wrote the manuscript; RB assisted in experiments; DS helped in the design of the study, data interpretation and preparation of the manuscript; ASH supervised data analysis and interpretation, and helped in the preparation of the manuscript; and EH designed the study, contributed to experiments, interpreted the data and contributed in the writing and revision of the manuscript. All co-authors meet conditions required for authorship credit as indicated on the JCBFM website, and all have approved the final manuscript.

Supplementary material

Supplementary material for this paper can be found at http://jcbfm.sagepub.com/content/by/supplemental-data