Sustained activation of XBP1 splicing leadsto endothelial apoptosis and atherosclerosisdevelopment in response to disturbed flowLingfang Zenga,1, Anna Zampetakia,1, Andriana Margaritia, Anna Elena Pepea, Saydul Alama, Daniel Martina,Qingzhong Xiaoa, Wen Wangb, Zheng-Gen Jinc, Gillian Cockerilld, Kazutoshi Morie, Yi-shuan Julie Lif, Yanhua Hua,Shu Chienf,2, and Qingbo Xua,2

aCardiovascular Division, King’s College London BHF Centre, London SE5 9NU, United Kingdom; bMedical Engineering Division, School of Engineering andMaterials Science, Queen Mary, University of London, London E1 4NS, United Kingdom; cAab Cardiovascular Research Institute, University of RochesterSchool of Medicine and Dentistry, Rochester, NY 14642; dDepartment of Cardiovascular Medicine, St George’s University of London, London SW17 0RE,United Kingdom; eDepartment of Biophysics, Graduate School of Science, Kyoto University, Kyoto 606-8502, Japan; and fDepartments of Bioengineeringand Medicine, University of California at San Diego, La Jolla, CA 92093

Contributed by Shu Chien, March 26, 2009 (sent for review January 4, 2009)

X-box binding protein 1 (XBP1) is a key signal transducer inendoplasmic reticulum stress response, and its potential role in theatherosclerosis development is unknown. This study aims to ex-plore the impact of XBP1 on maintaining endothelial integrityrelated to atherosclerosis and to delineate the underlying mech-anism. We found that XBP1 was highly expressed at branch pointsand areas of atherosclerotic lesions in the arteries of ApoE�/� mice,which was related to the severity of lesion development. In vitrostudy using human umbilical vein endothelial cells (HUVECs) indi-cated that disturbed flow increased the activation of XBP1 expres-sion and splicing. Overexpression of spliced XBP1 induced apopto-sis of HUVECs and endothelial loss from blood vessels during exvivo cultures because of caspase activation and down-regulation ofVE-cadherin resulting from transcriptional suppression and matrixmetalloproteinase-mediated degradation. Reconstitution of VE-cadherin by Ad-VEcad significantly increased Ad-XBP1s-infectedHUVEC survival. Importantly, Ad-XBP1s gene transfer to the vesselwall of ApoE�/� mice resulted in development of atheroscleroticlesions after aorta isografting. These results indicate that XBP1plays an important role in maintaining endothelial integrity andatherosclerosis development, which provides a potential thera-peutic target to intervene in atherosclerosis.

caspase � endothelial integrity � Ve-cadherin � vessel graft � mouse model

A therosclerosis is a leading cause of death worldwide (1, 2).Accumulating evidence suggests that atherosclerosis is a

multifactorial disease that can be initiated by risk factors (3–6).An important feature of atherosclerosis is its geographic distri-bution along the artery wall, i.e., occurring more frequently atcurved or branching points in the vasculature, indicating that theflow pattern exerts an important role in the development ofatherosclerotic lesions (7, 8).

Endothelial cells (ECs) are key cellular components of bloodvessels, functioning as selectively permeable barriers betweenblood and tissues. It is believed that risk factors induce ECapoptosis, leading to the denudation or dysfunction of the intactendothelial monolayer, which causes lipid accumulation, mono-cyte adhesion, and inflammatory reactions that initiate athero-sclerotic lesion (5, 9–12). Although information on risk factor-induced atherosclerosis has been accumulating, the underlyingmechanism remains unclear.

The X-box binding protein 1 (XBP1) was originally identifiedas a bZIP protein capable of binding to the cis-acting X boxpresent in the promoter regions of human major histocompat-ibility complex class II genes (13) and is known to be essentialfor liver growth and B lymphocyte differentiation (14, 15). Inmammalian cells, XBP1 is a key signal transducer in the endo-plasmic reticulum (ER) stress response. It has also been reported

that there is a link between XBP1 and human disease (16, 17).Although ER stress is reported to be involved in atherosclerosis(18–22), the role of XBP1 in vascular disease has not beenexamined in detail. In the present study, we demonstrated thatdisturbed flow induces XBP1 splicing and sustained activationthat led to EC apoptosis and the formation of atheroscleroticlesion in ApoE�/� mice.

ResultsExpression of XBP1 Is Related to Atherosclerotic Lesions. To explorethe potential role of XBP1 in the development of atherosclerosis,XBP1 expression on the aorta was stained in 18-months-oldwild-type (ApoE�/�, C57BL/6J) and ApoE�/�/Tie2-LacZ(C57BL/6J) mice by en face preparation. X-gal staining showeddifferent morphology of endothelial cells in the linear (Fig. 1A)and branching (Fig. 1B) regions. Immunostaining indicates thatvery little XBP1 protein was detected in normal aorta (data notshown) and the linear regions of 18-month-old ApoE�/� mice(Fig. 1C), but abundant amount of XBP1 was detected in thebranch curve and lesion areas (Fig. 1 D and E). There are 2isoforms of XBP1, a 29KDa unspliced and a 56KDa splicedisoform. As the XBP1 antibody (M186), which recognizes theinternal part (aa76–263) shared by both isoforms, could not tellwhich isoform was expressed in en face staining, Western blotwas then performed to detect the isoform levels in whole aortictissues from wild-type and ApoE�/� mice at different ages. Asshown in Fig. 1F, the 56KDa spliced isoform was detected at asmall amount in wild-type mice (18 months old), but at highlevels in older ApoE�/� mice (18 or 24 months old). The 29KDaunspliced isoform was only detected in old ApoE�/� mice atrelatively low level as compared to spliced one. These results maysuggest that both isoforms exist in the branch curve and lesionareas as shown in Fig. 1 D and E with the spliced isoform as themain one. The lack of significant differences in PECAM1 levelssuggests a similar ratio of ECs exist in all tissue samples. Theseresults suggest that XBP1 expression is related to atheroscleroticlesion location.

Author contributions: L.Z., A.Z., S.C., and Q. Xu designed research; L.Z., A.Z., A.M., A.E.P.,S.A., D.M., Q. Xiao, Y.J.L., and Y.H. performed research; Z.-G.J., G.C., and K.M. contributednew reagents/analytic tools; L.Z., A.Z., and W.W. analyzed data; and L.Z and Q. Xu wrotethe paper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.

1L.Z. and A.Z. contributed equally to this work.

2To whom correspondence may be addressed. E-mail: qingbo.xu@kcl.ac.uk orshuchien@ucsd.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0903197106/DCSupplemental.

8326–8331 � PNAS � May 19, 2009 � vol. 106 � no. 20 www.pnas.org�cgi�doi�10.1073�pnas.0903197106

Dow

nloa

ded

by g

uest

on

July

31,

202

1

XBP1 Splicing Is Related to EC Proliferation. As elevated XBP1proteins were only detected in the branch curve and lesion areasof aortas in ApoE�/� mice, it seemed that the expression of XBP1responded to flow pattern. To test this hypothesis, laminar anddisturbed flow were applied to HUVECs, followed by XBP1protein assessments. When laminar flow was applied, both thespliced and unspliced XBP1 proteins were decreased (Fig. 2A).In contrast, disturbed flow caused an increase in both isoformsof XBP1 protein (Fig. 2B).

It is well-known that laminar flow is related to EC quiescentand survival, while disturbed flow links to EC proliferation andapoptosis. The flow responding pattern of XBP1 expression andsplicing suggests that XBP1 may be involved in EC proliferation.Indeed, Western blot analysis showed higher level of XBP1 (bothspliced and unspliced) proteins in proliferating HUVECs ascompared to quiescent cells (supporting information (SI) Fig.S1 A). To further investigate the involvement of XBP1 in ECproliferation, knockdown experiments were performed withXBP1 shRNA lentivirus and IRE1� siRNA, respectively. Uponinfection, the different XBP1 shRNA lentiviruses decreasedXBP1 mRNA level after 24 h at different efficiency. Theproliferation rate has a parallel relationship with the XBP1 level.In Fig. S1B, the lower panel showed decreased spliced andunspliced XBP1 proteins by one of the XBP1 shRNA lentiviruses72 h after infection; the upper panel showed the average of therelative 5-Bromo-2�-deoxy-Uridine (Br-dU) incorporation by 3different XBP1 shRNA lentiviruses. Further experiments

showed that knockdown of IRE1� by siRNA transfection de-creased XBP1 splicing (Fig. S1C, Lower) and Br-dU incorpora-tion (Fig. S1C, Upper) in HUVECs. Under this condition,unspliced XBP1 remained constant (data not shown). Theseresults suggest that transient activation of XBP1 splicing mayincrease EC proliferation.

Overexpression of Spliced XBP1 Induces EC Apoptosis Through Down-Regulation of VE-cadherin. To explore the effect of high level ofXBP1 on EC, we overexpressed XBP1s in HUVECs by adeno-viral gene transfer to mimic the endogenous high levels of XBP1.Morphology observation revealed that overexpression of un-spliced XBP1 (Ad-XBP1u) exerted no significant effect onHUVECs compared to empty virus (Ad-tTA) (Fig. S2 A). How-ever, overexpression of the spliced XBP1 (Ad-XBP1s) causedHUVECs to become round in shape and to detach 72 h afterinfections (Fig. S2 A). A proliferation assay using the MTTmethod revealed that unspliced XBP1 slightly increased cellproliferation, while spliced XBP1 dramatically decreased cellsurvival (Fig. S2B). As only spliced XBP1 showed a significanteffect on HUVEC, further experiments were mainly focused onthis isoform.

We then studied the effect of overexpression of XBP1 on ECsurvival in intact vessel walls. Arterial vessels were isolated fromTie2-LacZ transgenic mice and cut into segments, which werethen infected with different amount of viruses and cultured invitro for 4 days, followed by X-gal staining to determine ECsurvival. As shown in Fig. 3A, Ad-XBP1s induced EC loss fromthe vessel wall in a dose-dependent manner compared to thesame titer of empty virus.

VE-cadherin is one of the most important molecules in themaintenance of endothelium integrity via its role in adherensjunctions. Western blot analysis showed that overexpression ofspliced XBP1 by Ad-XBP1s gene transfer decreased VE-cadherin protein levels in a dose-dependent manner (Fig. 3B).Immunofluorescence staining revealed that in Ad-XBP1s-infected cells VE-cadherin was decreased and translocated frompericellular junctions to cytosol (Fig. 3C). To explore whetherXBP1s-induced EC apoptosis was related to the decrease inVE-cadherin, experiments were conducted using Ad-VEcad (23)gene transfer to overexpress VE-cadherin. Although overexpres-sion of exogenous VE-cadherin slightly decreased cell prolifer-ation as compared to control virus-infected cells, Ad-VEcadincreased Ad-XBP1s-treated HUVEC survival, as demonstrated

29

56

XBP1u

XBP1s

3m18m

ApoE-/-

GAPDH

130

37

ApoE+/+

OB

F

PECAM-1

18m 18m 3m 3m 18m 18m 24m

OB

B C

D E

KD

A

Fig. 1. XBP1 expression level was related to atherosclerotic lesion develop-ment. (A–E) Aortas from Tie2-LacZ/ApoE�/� mice were harvested, preparedfor en face staining; A and B were developed with X-gal showing the differentmorphology of endothelial cells in the linear (A) and branching (B) regions.(C–E) Immunostaining for XBP1. Note that XBP1 was highly expressed inbranch point (D) and lesion areas (E) but not in straight part (C) in ApoE�/�

mice arteries. (Scale bar: 100 �m.) OB indicates opening of a branch. (F)Western blot analysis of protein extracts from mouse aortas indicates that theprotein levels of both spliced and unspliced XBP1 were related to the severityof atherosclerosis. High levels of spliced XBP1 exist in aged ApoE�/� aortas,little in same age wild-type mice. The data presented is the representative of3 independent experiments, respectively.

A

B

XBP1s

XBP1u

GAPDH

29

37

KD56 1.0

0.5

HD

PA

G/u1

PB

X

1.0

0.5

HD

PA

G/s1P

BX

SS - + SS - +

0 1 2 4 8 24 h

XBP1u

GAPDH

Disturbed Flow

29

37

KDXBP1s56

0 1 2 4 8 24 h

1.0

2.0

3.0HD

PA

G/1P

BX

XBP1uXBP1s

- + SS*

**

* * *

Fig. 2. Disturbed flow activated XBP1 splicing. (A) Laminar shear stressdecreased XBP1 protein level. HUVECs were subjected to 12dynes/cm2 steadyflow for 2 h. (Right) Average of band density from 3 independent experiments(*P � 0.05). (B) Disturbed flow increased XBP1 protein level. (Right) averageof band density from 3 independent experiments (*P � 0.05).

Zeng et al. PNAS � May 19, 2009 � vol. 106 � no. 20 � 8327

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

July

31,

202

1

by increasing attached-cell numbers (Fig. S2C). Proliferationassay also showed that Ad-VEcad increased Ad-XBP1s-treatedHUVEC survival (Fig. S2D). These results indicate that splicedXBP1-mediated decrease in VE-cadherin at least partially con-tributes to EC apoptosis and cell loss from the vessel wall.

Ad-XBP1s Down-Regulates VE-Cadherin Through Transcriptional Inhi-bition and MMP-Mediated Degradation. VE-cadherin can be de-graded through several signal pathways, such as proteasome,caspase, matrix metalloproteinase (MMP), and lysosome pro-teases (24–27). To determine which pathway might be involved,the effect of different inhibitors was compared. Proteasomeinhibitors (MG132 and ALLN, Fig. S3A), lysosome proteaseinhibitor [chloroquine (ChQ), Fig. S3B] and caspase inhibitor(Pan-FMK) (Fig. S3C) could not block XBP1s-induced VE-cadherin degradation, although MG132 and ALLN blocked thedegradation of XBP1s itself as expected. Only the MMP inhib-itor (GM6001) partially blocked XBP1s-induced VE-cadherindegradation (Fig. 4A). Moreover, GM6001 could also partiallyblock XBP1s-induced EC loss from the vessel wall in ex vivoexperiments (Fig. 4B).

RT-PCR analysis indicates that overexpression of spliced XBP1decreased VE-cadherin mRNA level in a dose- and time-dependentmanner (data not shown). Luciferase activity assay with VE-cadherin gene promoter (pGL3-VEcad-Luc reporter) showed thatoverexpression of XBP1s significantly decreased the reporter geneexpression (Fig. 4C). Unspliced XBP1 (XBP1u) exerted a slightlyinhibitory effect, while mature ATF6 (ATF6N), another ER stresstransducer (28), had no effect on VE-cadherin gene expression(Fig. 4C). To explore whether XBP1 was directly involved inVE-cadherin gene transcription, ChIP assay was performed. Asspliced XBP1 was unstable, and no appropriate antibody forimmunoprecipitation was available to pull down endogenousXBP1, we infected HUVECs with Ad-XBP1 and used antiflagantibody to pull down exogenous XBP1 and its associated DNA

fragments instead. Six primer sets covering the �121�2027ntpromoter region (Table S1) were used to amplify the pull-downDNA fragments. Only primer set 3 demonstrated that both splicedand unspliced XBP1 bound to the VE-cadherin gene promoter inliving cells (Fig. 4D), while the other primer sets did not detect anybinding (data not shown). These results suggest XBP1 binds to the�374��672nt region in VE-cadherin promoter. Considering his-tone acetylation and methylation played a switch role in controllingchromatin structure and gene transcription (29–32), we performedChIP assay with anti-acH3 (lysine 9 acetylated) and H3K4DM(lysine 4 double methylated) antibodies. As shown in Fig. 4E,the acetylation and methylation of histone H3 in VE-cadheringene promoter region were significantly decreased in Ad-XBP1s-infected HUVECs (Left) but not in Ad-XBP1u-infectedcells (Right), indicating that XBP1s may recruit histone deacety-lases/demethylases to the VE-cadherin gene promoter. Theseresults suggest that spliced XBP1 regulates VE-cadherin genetranscription.

Overexpression of Spliced XBP1 Induces EC Apoptosis ThroughCaspase Activation. To further explore the mechanisms of XBP1-induced EC apoptosis, the Pan-FMK was used in ex vivoexperiments. As shown in Fig. 5A, Pan-FMK inhibitor signifi-cantly reduced XBP1s-induced EC loss from blood vessels.Caspase-2, -3, -9 and pan-caspase inhibitors could partially blockthe XBP1s-induced decrease in HUVEC viability (Fig. 5B),indicating that these caspases were activated. Indeed, Westernblot analysis revealed the activation of these caspases as cleavedbands were detected. Although caspase-8 and -12 inhibitorscould not block Ad-XBP1s’ effect, the activation of both caspaseswas also identified (data not shown). Fig. 5C showed theactivation of caspase-2 and -3 as demonstrated by the presenceof p12 and p18 bands, respectively. These results indicate thatoverexpression of spliced XBP1 activates multiple caspases that

Fig. 3. Overexpression of spliced XBP1 induced EC apoptosis through down-regulation of VE-cadherin (A) Overexpression of spliced XBP1 induced EC loss fromthe vessel wall in a dose dependent manner. Artery segments from Tie2-LacZ/ApoE�/� mice were infected with Ad-tTA or Ad-XBP1s virus at indicated multiplicityof infection (MOI) and cultured for 4 days. The surviving ECs were revealed by X-gal staining. (Left) Representative images of X-gal staining of vessel segments.(Scale bar: 100 �m.) (Right) The statistical data of cell loss from 6 samples of each group (*P � 0.05; **P � 0.01). (B) Ad-XBP1s decreased VE-cadherin proteinlevel in a dose dependent manner. HUVECs were infected with Ad-XBP1s at MOI indicated and followed by Western blot analysis 72 h after infection. Ad-tTAvirus was included as control and to compensate the MOI. Exogenous XBP1s was revealed by anti-Flag antibody. (C) Immunofluorescence staining showed thatAd-XBP1s induced VE-cadherin translocation from cell surface to cytosol. Images were taken 72 h after Ad-XBP1s (5 MOI) infection. (Scale bar: 50 �m.) The dataare representative or means � SEM of 3 independent experiments, *P � 0.05.

8328 � www.pnas.org�cgi�doi�10.1073�pnas.0903197106 Zeng et al.

Dow

nloa

ded

by g

uest

on

July

31,

202

1

may serve as mediators between VE-cadherin decrease andendothelial cell apoptosis.

Overexpression of Spliced XBP1 Induces Atherosclerosis in an AorticIsograft Model. To further investigate the potential role of XBP1splicing in atherosclerosis development, spliced XBP1 was over-expressed by adenoviral gene transfer in ECs in the straight partof blood vessels to mimic high levels of spliced XBP1 in branchareas. Artery isograft is an appropriate model to study ECfunction in atherosclerosis, as the isograft itself does not inducelesion development (33). In this model, a monolayer of endo-thelial cells was found in grafted vessels 4 weeks after grafting(Fig. S4). The thoracic aortas were harvested from donorApoE�/� mice and infected with Ad-XBP1s virus in vitro,followed by isografting into recipient ApoE�/� mice. Four weekslater, the grafted vessels were harvested, sectioned, and stainedwith haematoxylin eosin. No (4/6) or little (2/6) neointimaformation was detected in empty virus-infected artery grafts, butall (6/6) Ad-XBP1s-infected grafts formed significant neointimallesions. Fig. 6 shows typical images of uninfected (Fig. 6A),empty virus (Ad-tTA, Fig. 6 B and C) and Ad-XBP1s virus (Fig.6 D, E, and F) infected grafts. The lumen was significantlyreduced by overexpression of spliced XBP1 with concomitantincrease of lesion area (Fig. 6 G and H). The lesion displayedmononuclear cell infiltration and cell proliferation (Fig. 6F).

These results suggest that sustained activation of XBP1 splicingin the vessel wall induces atherosclerotic lesions.

DiscussionAtherosclerosis is a multistep process involving multiple genesand signal pathways. The initiation of the pathology is theperturbation of the endothelium triggered by multiple riskfactors. In this study, we have found that XBP1 expression andsplicing was highly increased in the atherosclerosis prone area invessel walls and activated by disturbed flow in endothelial cellsin vitro. We demonstrated that transient activation of XBP1splicing is related to EC proliferation, while sustained activationinduced EC apoptosis, cell loss from vessel walls, and athero-sclerotic lesion development in aorta isograft model. Thus,XBP1 splicing has a pro-atherogenic effect and may serve as apotential therapeutic target for treatment of atherosclerosis.

Under normal conditions, XBP1 exists as a 29KDa unsplicedisoform. In response to ER stress, XBP1 mRNA undergoes un-conventional splicing, giving rise to a 56KDa spliced isoform withtranscriptional activity (28, 34). XBP1 splicing is essential for cellsurvival under stress condition. However, long term ER stress willinduce apoptosis. Besides functioning as an ER stress transducer,XBP1 is also involved in other physiological or pathological pro-cesses (14, 15, 35, 36). In this study, we demonstrate a novel functionof XBP1, i.e., XBP1 splicing is involved in EC proliferation. First,indirect evidence came from the observation that both unsplicedand spliced XBP1 were highly expressed in atherosclerosis prone

VE-cad

Flag

GAPDH

- - + + Ad-XBP1s- + - + GM6001

DMSO GM6001A B

Ad-XBP1uAd-XBP1s

Flag IgG Flag IgG

ChIP

input

VE-cad promoter

C D

dA

-A

Ttd

A-

s1P

BX

1.0

0.5

UL

R

kcoMs1PBX

u1PBXN6FTA

LucVE-cad

**

VE-cad promoter

ChIP

input

E

37

56

130KD

Ac K4DM IgG Ac K4DM IgG Ac K4DM IgG Ac K4DM IgG

Ad-tTA Ad-tTAAd-XBP1s Ad-XBP1u

Fig. 4. Ad-XBP1s down-regulated VE-cadherin through MMP-mediateddegradation and transcriptional suppression. (A) MMP inhibitor partiallyattenuated Ad-XBP1s-induced VE-cadherin decrease. HUVECs were infectedwith Ad-XBP1s at 5 MOI for 72 h, and GM6001 (5 �M) was included in themedium 24 h before harvesting the cells. Ad-tTA and DMSO were included asvirus and vehicle controls, respectively. (B) GM6001 partially rescued Ad-XBP1s-induced endothelial cell loss from the vessel wall. Artery segments fromTie2-LacZ/ApoE�/� mice were infected with 5 � 107pfu/ml Ad-XBP1s orAd-tTA viruses and cultured in the absence (DMSO) or presence of 5 �MGM6001 for 4 days. (Scale bar: 100 �m.) (C) Spliced XBP1 decreased pGL3-VEcad-Luc reporter gene expression in HUVECs (**P � 0.01). (D) ChIP assayrevealed that both spliced and unspliced XBP1 bound to VE-cadherin genepromoter region. (E) ChIP assay indicated that overexpression of spliced XBP1(Left) decreased the acetylation and methylation of histone H3 in VE-cadheringene promoter region but unspliced XBP1 (Right) did not. The data arerepresentative or means � SEM of 3 independent experiments.

A

dA

-A

Ttd

A-

s1P

BX

DMSO Pan-FMK

1.0

0.5

noitc

ud

nI fo

dlo

F

DMSOPan-FMK

**

Ad-tTA Ad-XBP1s

B

1.0

0.5

noitc

ud

nI fo

dlo

F

* **

DMSOCas-2Cas-3

Pan

Cas-8

Cas-12Cas-9

0 2 5 10 20 MOI

Ad-XBP1s

Cas-2

Cas-3

Flag

GAPDH

37

2015

56

37

37

2015

KD

C

Ad-tTA Ad-XBP1s

Fig. 5. Sustained activation of XBP1 splicing induced EC apoptosis throughcaspase activation. (A) Pan-FMK rescued Ad-XBP1s-induced EC loss from thevessel wall. Five microMolar Pan-FMK was included in ex vivo experiments inwhich artery segments from Tie2-LacZ/ApoE�/� mice were infected withviruses at 5 � 107pfu/ml. The same volume of DMSO was included as vehiclecontrols. (Left) The representative image of segments. (Scale bar: 100 �m.)(Right) The statistical data of cell loss from 6 samples for each group (**P �0.01). (B) Caspase inhibitors partially reduced Ad-XBP1s-induced cell prolifer-ation decrease in HUVECs. The data are means � SEM from 3 independentexperiments, *P � 0.05. (C) Ad-XBP1s induced caspase activation in HUVECs.HUVECs were infected with Ad-XBP1s viruses at MOI as indicated and culturedfor 72 h. Ad-tTA was included as controls and to compensate MOI. Arrowindicates cleaved bands. The data are the representative of 3 independentexperiments.

Zeng et al. PNAS � May 19, 2009 � vol. 106 � no. 20 � 8329

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

July

31,

202

1

areas in older ApoE�/� mice but not in linear regions of the vesselwall, and that both isoforms were up-regulated by disturbed flowbut decreased by laminar flow. ECs in atherosclerotic lesion proneareas or under disturbed flow are believed to be undergoingproliferation and apoptosis, while cells in the linear regions of vesselwalls or under laminar flow are in a quiescent state (37, 38). Indeed,such relationship was observed in in vitro-cultured HUEVCs. Arelatively high level of spliced XBP1 was detected in proliferatingcells as compared to confluent quiescent cells. On the other hand,the suppression of Br-dU incorporation by IRE1� siRNA andXBP1 shRNA in HUVECs gives direct evidence for this notion.The slight increase of HUVEC proliferation by overexpression ofunspliced XBP1 also supports this concept, under which splicedXBP1 is increased accordingly (data not shown). As spliced XBP1can increase HUVEC size and cell size increase is an essential stepfor cell division, it is postulated that XBP1 regulates EC prolifer-ation through modulation of cell growth. However, the underlyingmechanism deserves further detailed investigation.

Cascade activation of caspases plays an important role in theregulation of cell apoptosis in response to different stimuli. Severalsignal pathways have already been established. All these pathwaysactivate the effector caspase, caspase-3 (39, 40). In this study,

caspase-2, -3, -8, -9, and -12 were activated by overexpression ofspliced XBP1 in ECs, suggesting that multiple signal pathways havebeen triggered. The overall activation of these caspases contributedto EC dysfunction, as pan-caspase inhibitor could block Ad-XBP1s-induced cell loss from blood vessels in ex vivo experiments.

As a transcription factor, the spliced XBP1 is not only involvedin the transcriptional regulation of genes essential for cellsurvival or apoptosis in response to stress stimuli, but is alsoinvolved in other physiological processes (14, 15, 35, 41–45). Inthis study, we identify another candidate target gene for XBP1,VE-cadherin. However, in this case, XBP1 functions as a tran-scriptional co-repressor. Both spliced and unspliced XBP1 canbind to the promoter of VE-cadherin gene, but only splicedXBP1 exerts a significant inhibitory effect. Both isoforms ofXBP1 have common N-terminal and internal DNA bindingdomain but differ in the C-terminals; the spliced isoform has amuch longer C-terminal domain. Analyzing the DNA sequenceof the promoter region (�374��672nt) to which XBP1 binds,it seems there is no consensus binding site for XBP1 (28, 46).Thus, the binding of XBP1 to the promoter of VE-cadherin genemay be through indirect binding via N-terminal-mediated inter-action with other DNA binding proteins and may function as aco-repressor. However, the direct binding of XBP1 cannot beexcluded. The C-terminal domain of spliced XBP1 may recruitdeacetylases and demethylases, as the acetylation and methyl-ation status of histone H3 in the VE-cadherin gene promoterarea is significantly decreased by overexpression of splicedXBP1. Therefore, XBP1 inhibits VE-cadherin gene transcrip-tion. The transcriptional inhibitory effect may be specific toXBP1, and not relating to the secondary effect of the ER stressresponse, as active ATF6 (ATF6N), another ER stress trans-ducer (34), has no effect on VE-cadherin gene expression.Although unspliced XBP1 could also bind to VE-cadherin genepromoter and luciferase reporter analysis also showed slightlyinhibitory effect, overexpression of unspliced XBP1 by Ad-XBP1u gene transfer did not decrease VE-cadherin proteinlevel. In fact, XBP1u could partially rescue XBP1s-inducedVE-caherin decrease in coinfected cells (Fig. S5). This studyprovides novel insights into the VE-cadherin gene transcrip-tional regulation and offers additional evidence for its role in themaintenance of endothelial integrity.

Endothelial cell dysfunction is the initial step of atheroscle-rosis development. In this process, XBP1 splicing may play a veryimportant role in endothelial cell dysfunction. High level ofspliced XBP1 was detected in atherosclerosis prone areas, andoverexpression of spliced XBP1 could induce EC apoptosis invitro and EC loss from vessel wall ex vivo. Importantly, whenspliced XBP1 was overexpressed in EC in the straight part ofartery vessel in a mouse isograft model mimicking the high levelof spliced XBP1 in prone areas, neointima formation wastriggered, featuring smooth muscle cell proliferation and mono-cytes infiltration, a similar characteristic of atherosclerosis.Normal vessels consist of ECs, smooth muscle cells, and peri-cytes, while in atherosclerotic lesion, monocytes, macrophages,and foam cells are also included. The high levels of spliced XBP1in aged ApoE�/� aorta tissues are not only derived from ECs butalso from other cell types. Thus, the role of spliced XBP1 in othercell types and its contribution to atherosclerosis developmentneeds further investigation.

In summary, this study demonstrates for the first time thatatherosclerotic risk factors, such as disturbed flow, can activateXBP1 splicing. Transient activation of XBP1 splicing may in-crease EC proliferation, while sustained activation leads to ECapoptosis, endothelium denudation, and atherosclerotic lesiondevelopment via multiple caspases activation and down-regulation of VE-cadherin at gene transcriptional level andMMP-mediated degradation. This study provides novel insightsinto understanding how the atherosclerosis process is initiated,

control Ad-tTA Ad-tTA

Ad-XBP1s Ad-XBP1s Ad-XBP1s

50

100

50

100

)%( ezis

nem

uL

a ide

m+amit

ni/ami t

nioe

N)

%(

Ctl Ad-tTA Ad-XBP1s Ctl Ad-tTA Ad-XBP1s

**

**

G H

A B C

D E F

Fig. 6. Overexpression of spliced XBP1 induced atherosclerosis develop-ment. Thoracic aortas were isolated from donor ApoE�/� (C57BL/C) mice andun-infected (A) or infected with Ad-tTA virus (B and C) or Ad-XBP1s virus (D–F)at 1 � 106pfu/ml in vitro and isograft into same background recipient ApoE�/�

(C57BL/C) mice. Neointima formation was checked on the grafts 4 weeks later.(A) A typical image of uninfected vessels. (B) Ad-tTA virus induced slight lesiondevelopment. (C) Higher magnification of (B). (D) Ad-XBP1s significantlyinduced lesion development; (E and F) higher magnification images. (G) Themeans � SEM of lumen size from 6 grafts for each group presented aspercentage with that of uninfected vessels set as 100%. (H) The means � SEMof neointima area from 6 grafts for each group presented as percentage of thearea of neointima to the whole area of intima plus lumen. **Significantdifference between Ad-tTA virus and Ad-XBp1s groups, P � 0.01.

8330 � www.pnas.org�cgi�doi�10.1073�pnas.0903197106 Zeng et al.

Dow

nloa

ded

by g

uest

on

July

31,

202

1

and targeting XBP1 splicing may provide a new therapeuticstrategy for vascular disease.

Materials and MethodsCell Culture. ECs were isolated from postnatal human umbilical vein (HUVECs)and cultured on collagen I-coated flasks in M199 medium supplemented with1 ng/ml �-endothelial cell growth factor, 3 �g/ml EC growth supplement frombovine neural tissue, 10�/ml heparin, 1.25 �g/ml thymidine, 10% fetal bovineserum (FBS), 100�/ml penicillin, and streptomycin in humidified incubatorsupplemented with 5% CO2. The cells were split every 3 days at a ratio of 1:4.Cells up to passage 10 were used in this study. All other cell types weremaintained in DMEM supplemented with 10% FBS and penicillin/streptomy-cin. Living cell images were assessed by Nikon Eclipse TS100 microscope withPh1 ADL 10�/0.25 objective lenses and Nikon DS-Fil camera at room temper-ature and processed by Adobe Photoshop software.

Animal Model. Tie2-LacZ/ApoE�/� (C57BL/C) or Tie2-LacZ/ApoE�/� (C57BL/C) orApoE�/� (C57BL/C) mice were used for en face staining and ex vivo or artery

isografting experiments as described in detail in SI Text. All animal experi-ments in this study were performed according to protocols approved by theInstitutional Committee for Use and Care of Laboratory Animals.

Generation of Adenoviral and Lentiviral Vectors. The following adenoviral andlentiviral vectors were used in this study: Ad-XBP1s and Ad-XBP1u werecreated from cDNA cloning. XBP1 shRNA lentiviruses and non-target shRNAlentivirus were purchased from Sigma. Ad-tTA virus is commercially available.The viral vector construction and transduction of viruses are described in detailin SI Text.

Statistical Analysis. Data expressed as the mean � SEM were analyzed with atwo-tailed student’s t test for two-groups or pair-wise comparisons. A value ofP � 0.05 was considered to be significant.

Other materials and methods are described in detail in SI Text.

ACKNOWLEDGMENTS. This work was supported by grants from the BritishHeart Foundation and the Oak Foundation.

1. Murray CJ, Lopez AD (1997) Alternative projections of mortality and disability by cause1990–2020: Global Burden of Disease Study. Lancet 349:1498–1504.

2. Jaffer FA, Libby P, Weissleder R (2006) Molecular and cellular imaging of atheroscle-rosis: Emerging applications. J Am Coll Cardiol 47:1328–1338.

3. Libby P (2002) Inflammation in atherosclerosis. Nature 420:868–874.4. Li C, Xu Q (2007) Mechanical stress-initiated signal transduction in vascular smooth

muscle cells in vitro and in vivo. Cell Signalling 19:881–891.5. Ross R (1999) Atherosclerosis–an inflammatory disease. N Engl J Med 340:115–126.6. Steinberg D (1997) Low density lipoprotein oxidation and its pathobiological signifi-

cance. J Biol Chem 272:20963–20966.7. Febbraio M, et al. (2000) Targeted disruption of the class B scavenger receptor CD36

protects against atherosclerotic lesion development in mice. J Clin Invest 105:1049–1056.

8. Traub O, Berk BC (1998) Laminar shear stress: Mechanisms by which endothelial cellstransduce an atheroprotective force. Arterioscler Thromb Vasc Biol 18:677–685.

9. Xu Q (2006) The impact of progenitor cells in atherosclerosis. Nat Clin Pract CardiovascMed 3:94–101.

10. Dardik A, et al. (2005) Differential effects of orbital and laminar shear stress onendothelial cells. J Vasc Surg 41:869–880.

11. World CJ, Garin G, Berk B (2006) Vascular shear stress and activation of inflammatorygenes. Curr Atheroscler Rep 8:240–244.

12. Zeng L, Zhang Y, Chien S, Liu X, Shyy JY (2003) The role of p53 deacetylation in p21Waf1regulation by laminar flow. J Biol Chem 278:24594–24599.

13. Liou HC, et al. (1990) A new member of the leucine zipper class of proteins that bindsto the HLA DR alpha promoter. Science 247:1581–1584.

14. Reimold AM, et al. (2000) An essential role in liver development for transcription factorXBP-1. Genes Dev 14:152–157.

15. Reimold AM, et al. (2001) Plasma cell differentiation requires the transcription factorXBP-1. Nature 412:300–307.

16. Marciniak SJ, Ron D (2006) Endoplasmic reticulum stress signaling in disease. PhysiolRev 86:1133–1149.

17. Ozcan U, et al. (2004) Endoplasmic reticulum stress links obesity, insulin action, andtype 2 diabetes. Science 306:457–461.

18. Austin RC, Lentz SR, Werstuck GH (2004) Role of hyperhomocysteinemia in endothelialdysfunction and atherothrombotic disease. Cell Death Differ 11 Suppl 1:S56–S64.

19. Lawrence de Koning AB, Werstuck GH, Zhou J, Austin RC (2003) Hyperhomocysteine-mia and its role in the development of atherosclerosis. Clin Biochem 36:431–441.

20. Gargalovic PS, et al. (2006) The unfolded protein response is an important regulator ofinflammatory genes in endothelial cells. Arterioscler Thromb Vasc Biol 26:2490–2496.

21. Sharma P, et al. (2006) Mining literature for a comprehensive pathway analysis: A casestudy for retrieval of homocysteine related genes for genetic and epigenetic studies.Lipids Health Dis 5:1.

22. Zhou J, et al. (2004) Association of multiple cellular stress pathways with acceleratedatherosclerosis in hyperhomocysteinemic apolipoprotein E-deficient mice. Circulation110:207–213.

23. Ha CH, Bennett AM, Jin ZG (2008) A novel role of vascular endothelial cadherin inmodulating c-Src activation and downstream signaling of vascular endothelial growthfactor. J Biol Chem 283:7261–7270.

24. Herren B, Levkau B, Raines EW, Ross R (1998) Cleavage of beta-catenin and plakoglobinand shedding of VE-cadherin during endothelial apoptosis: Evidence for a role forcaspases and metalloproteinases. Mol Biol Cell 9:1589–1601.

25. Xiao K, et al. (2003) Mechanisms of VE-cadherin processing and degradation inmicrovascular endothelial cells. J Biol Chem 278:19199–19208.

26. Everson WV, Smart EJ (2001) Influence of caveolin, cholesterol, and lipoproteins onnitric oxide synthase: Implications for vascular disease. Trends Cardiovasc Med 11:246–250.

27. Tsou TC, et al. (2005) Arsenite induces endothelial cytotoxicity by down-regulation ofvascular endothelial nitric oxide synthase. Toxicol Appl Pharmacol 208:277–284.

28. Lee K, et al. (2002) IRE1-mediated unconventional mRNA splicing and S2P-mediatedATF6 cleavage merge to regulate XBP1 in signaling the unfolded protein response.Genes Dev 16:452–466.

29. Eberharter A, Becker PB (2002) Histone acetylation: A switch between repressive andpermissive chromatin. Second in review series on chromatin dynamics. EMBO Rep3:224–229.

30. Hayashi K, Yoshida K, Matsui Y (2005) A histone H3 methyltransferase controls epige-netic events required for meiotic prophase. Nature 438:374–378.

31. Jenuwein T, Allis CD (2001) Translating the histone code. Science 293:1074–1080.32. Lachner M, Jenuwein T (2002) The many faces of histone lysine methylation. Curr Opin

Cell Biol 14:286–298.33. Bentzon JF, et al. (2006) Smooth muscle cells in atherosclerosis originate from the local

vessel wall and not circulating progenitor cells in ApoE knockout mice. Arterioscler,Thromb, Vasc Biol 26:2696–2702.

34. Yoshida H, Matsui T, Yamamoto A, Okada T, Mori K (2001) XBP1 mRNA is induced byATF6 and spliced by IRE1 in response to ER stress to produce a highly active transcriptionfactor. Cell 107:881–891.

35. Lee AH, Scapa EF, Cohen DE, Glimcher LH (2008) Regulation of hepatic lipogenesis bythe transcription factor XBP1. Science 320:1492–1496.

36. Kaser A, et al. (2008) XBP1 links ER stress to intestinal inflammation and confers geneticrisk for human inflammatory bowel disease. Cell 134:743–756.

37. Akimoto S, Mitsumata M, Sasaguri T, Yoshida Y (2000) Laminar shear stress inhibitsvascular endothelial cell proliferation by inducing cyclin-dependent kinase inhibitorp21(Sdi1/Cip1/Waf1). Circ Res 86:185–190.

38. Cunningham KS, Gotlieb AI (2005) The role of shear stress in the pathogenesis ofatherosclerosis. Lab Invest 85:9–23.

39. Riedl SJ, Salvesen GS (2007) The apoptosome: Signaling platform of cell death. Nat RevMol Cell Biol 8:405–413.

40. Troy CM, Shelanski ML (2003) Caspase-2 redux. Cell Death Differ 10:101–107.41. Acosta-Alvear D, et al. (2007) XBP1 controls diverse cell type- and condition-specific

transcriptional regulatory networks. Mol Cell 27:53–66.42. Kim R, Emi M, Tanabe K, Murakami S (2006) Role of the unfolded protein response in

cell death. Apoptosis 11(1):5–13.43. Koumenis C (2006) ER stress, hypoxia tolerance and tumor progression. Current Mol

Med 6:55–69.44. Yoshida H (2007) Unconventional splicing of XBP-1 mRNA in the unfolded protein

response. Antioxid Redox Signaling 9:2323–2333.45. Sriburi R, Jackowski S, Mori K, Brewer JW (2004) XBP1: A link between the unfolded

protein response, lipid biosynthesis, and biogenesis of the endoplasmic reticulum.J Cell Biol 167:35–41.

46. Mai B, Breeden L (1997) Xbp1, a stress-induced transcriptional repressor of the Sac-charomyces cerevisiae Swi4/Mbp1 family. Mol Cell Biol 17:6491–6501.

Zeng et al. PNAS � May 19, 2009 � vol. 106 � no. 20 � 8331

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

July

31,

202

1