XBP1 mRNA Splicing Triggers an Autophagic Response in Endothelial Cells through BECLIN-1 Transcriptional Activation*

Background: Apoptosis and autophagy are two closely related systems that induce cell death. Results: X-box-binding protein 1 (XBP1) mRNA splicing regulates BECLIN-1 transcriptional activation, a fundamental player in the initiation of autophagy. Conclusion: XBP1 splicing induces an autophagic response in endothelial cells. Significance: XBP1 could be used as an important pharmacological target that can regulate the autophagic machinery and endothelial cell death. Sustained activation of X-box-binding protein 1 (XBP1) results in endothelial cell (EC) apoptosis and atherosclerosis development. The present study provides evidence that XBP1 mRNA splicing triggered an autophagic response in ECs by inducing autophagic vesicle formation and markers of autophagy BECLIN-1 and microtubule-associated protein 1 light chain 3β (LC3-βII). Endostatin activated autophagic gene expression through XBP1 mRNA splicing in an inositol-requiring enzyme 1α (IRE1α)-dependent manner. Knockdown of XBP1 or IRE1α by shRNA in ECs ablated endostatin-induced autophagosome formation. Importantly, data from arterial vessels from XBP1 EC conditional knock-out (XBP1eko) mice demonstrated that XBP1 deficiency in ECs reduced the basal level of LC3β expression and ablated response to endostatin. Chromatin immunoprecipitation assays further revealed that the spliced XBP1 isoform bound directly to the BECLIN-1 promoter at the region from nt −537 to −755. BECLIN-1 deficiency in ECs abolished the XBP1-induced autophagy response, whereas spliced XBP1 did not induce transcriptional activation of a truncated BECLIN-1 promoter. These results suggest that XBP1 mRNA splicing triggers an autophagic signal pathway through transcriptional regulation of BECLIN-1.

Sustained activation of X-box-binding protein 1 (XBP1) results in endothelial cell (EC) apoptosis and atherosclerosis development. The present study provides evidence that XBP1 mRNA splicing triggered an autophagic response in ECs by inducing autophagic vesicle formation and markers of autophagy BECLIN-1 and microtubule-associated protein 1 light chain 3␤ (LC3-␤II). Endostatin activated autophagic gene expression through XBP1 mRNA splicing in an inositol-requiring enzyme 1␣ (IRE1␣)-dependent manner. Knockdown of XBP1 or IRE1␣ by shRNA in ECs ablated endostatin-induced autophagosome formation. Importantly, data from arterial vessels from XBP1 EC conditional knock-out (XBP1eko) mice demonstrated that XBP1 deficiency in ECs reduced the basal level of LC3␤ expression and ablated response to endostatin. Chromatin immunoprecipitation assays further revealed that the spliced XBP1 isoform bound directly to the BECLIN-1 promoter at the region from nt ؊537 to ؊755. BECLIN-1 deficiency in ECs abolished the XBP1-induced autophagy response, whereas spliced XBP1 did not induce transcriptional activation of a truncated BECLIN-1 promoter. These results suggest that XBP1 mRNA splicing triggers an autophagic signal pathway through transcriptional regulation of BECLIN-1.
Although tremendous recent advances have been made in the field, the functional significance of autophagy in human disease remains incompletely understood. Diseases such as atherosclerosis show evidence of cellular senescence characterized by reduced cell proliferation, irreversible growth arrest, and apoptosis. Nonetheless, a growing body of evidence suggests that elucidation of the signaling pathways regulating autophagy may provide novel therapeutic approaches in the prevention or treatment of atherosclerosis (1). Recent studies have reported that autophagy becomes dysfunctional in atherosclerosis, and its deficiency promotes atherosclerosis in part through inflammasome hyperactivation (2). Moreover, it has also been shown that macrophage autophagy plays a protective role in advanced atherosclerosis (3), whereas transmission electron microscopy of smooth muscle cells in the fibrous cap of advanced plaques revealed ultrastructural features of autophagy, such as vacuolization and formation of myelin figures (4). Endothelial cell (EC) 3 dysfunction or death are also key events in the pathophysiology of the vessel wall and the development of atherosclerosis (5,6). However, the regulation of autophagy in ECs in atherosclerotic plaques has not been fully elucidated yet. Autophagy is a multistep catabolic process, which has been shown to be involved in a variety of pathophysiological conditions (1). During the autophagic process, long-lived proteins and organelles are sequestered in a double membrane-bound autophagosome (1) and degraded via lysosome (7)(8)(9)(10)(11). Under physiological con-* This work was supported by Grants from the British Heart Foundation and the Oak Foundation. □ S This article contains supplemental Figs. S1 and S2. 1 To whom correspondence may be addressed. Tel.: 44-20-7848-5322; Fax: 44-20-7848-5296; E-mail: andriani.margariti@kcl.ac.uk. 2 To whom correspondence may be addressed. E-mail: lingfang.zeng@ kcl.ac.uk.
ditions, autophagy is a reparative and life-maintaining process.
In cases of excessive induction of autophagy by environmental or intracellular stress, it becomes a cell death pathway. However, the mechanisms that are involved in the regulation of autophagy are largely unknown. We have previously shown that sustained activation of X-box-binding protein 1 (XBP1) mRNA splicing, a crucial signal transducer in endoplasmic reticulum stress response, leads to EC apoptosis and atherosclerosis development (12). XBP1 is a transcription factor that belongs to the basic region/leucine zipper family (13,14), and it has an unspliced and a spliced form. The unspliced form has a molecular mass of 29 kDa and acts as a dominant negative of spliced XBP1 (15,16). In response to endoplasmic reticulum (ER) stress, XBP1 mRNA undergoes unconventional splicing through the inositol-requiring enzyme 1␣ (IRE1␣), giving rise to a 56-kDa spliced form with intact transcriptional activity (17,18). XBP1 mRNA splicing is implicated in affecting plasma cell differentiation (19), fetal survival, and liver development (20,21). In the present study, we provide strong evidence that XBP1 mRNA splicing is involved in the regulation of autophagy in ECs through BECLIN-1 transcriptional activation. These findings provide insights into the cell death machinery of ECs, which is a key event in atherosclerotic diseases.

EXPERIMENTAL PROCEDURES
Cell culture media and serum were purchased from Invitrogen, and cell culture supplements were purchased from Sigma. Antibodies against XBP1 (M-186), CD31, and GAPDH were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA), whereas antibody against FLAG was from Abcam, and anti-LC3 (M186-3) was from MBL. Antibodies against IRE1␣, pIRE1␣, and BECLIN-1 (Autophagy Kit-4445) were purchased from Cell Signaling. Endostatin was from Calbiochem. Chloroquine was from Sigma. All secondary antibodies were from DAKOCytomation, and all other chemicals were from Sigma.
Cell Culture-Human umbilical vein endothelial cells were isolated from the human umbilical cord and cultured on collagen I-coated flasks in EGM-2 medium (Clonetics) or in M199 medium supplemented with 1 ng/ml ␤-EC growth factor (Sigma), 3 g/ml EC growth supplement from bovine neural tissue (Sigma), 10 units/ml heparin, 1.25 g/ml thymidine, 10% fetal bovine serum (FBS, PAA, A15-108) (12), and 100 g/ml penicillin and streptomycin, as described previously (22). 293T cells were maintained in DMEM supplemented with 10% FBS and penicillin/streptomycin. Living cell images were assessed by a Nikon Eclipse TS100 microscope with Ph1 ADL ϫ10/0.25 objective lenses and Nikon DS-Fil camera at room temperature and processed by Adobe Photoshop software.
Adenoviral Gene Transfer-Ad-FLAG-XBP1s and Ad-FLAG-XBP1u were created and amplified as described previously (12). The FLAG tag is located in the N terminus. Empty vector-derived Ad-null virus was used as control virus. For adenoviral gene transfer, confluent human umbilical vein endothelial cells were infected with Ad-XBP1s or Ad-XBP1u at a multiplicity of infection (MOI) of 5 for 6 h followed by replacement with complete growth medium and further culture for the time indicated.
Transmission Electron Microscopy-Human umbilical vein endothelial cells were infected with Ad-null, Ad-XBP1u, and Ad-XBP1s viruses at 5 MOI for 48 h, or ECs were infected with non-target (shNT) or XBP1-targeting (shXBP1) shRNA lentiviruses for 48 h following by 24 h of endostatin. The cells were grown to confluence on 13-mm diameter glass coverslips. For transmission electron microscopy, cells on coverslips were fixed by removing the medium and replacing it immediately with 2.5% (v/v) glutaraldehyde in 0.1 M phosphate buffer. The cells were fixed for 2 h at 4°C. After fixation, the coverslips were rinsed several times with 0.1 M phosphate buffer, followed by postfixation with 1% (w/v) osmium tetroxide in 0.1 M phosphate buffer (pH 7.3) for 20 min at 4°C. The coverslips were rinsed (10 min) in phosphate buffer and dehydrated in a graded series of ethanol to 100%. The coverslips were then flooded with TAAB epoxy resin and left to infiltrate for 4 h at room temperature. To section in the plane of the monolayers, the coverslips were embedded by inverting them (monolayer side down) onto a capsule (TAAB) overfilled with resin and then polymerized for 24 h at 70°C. Ultrathin sections (70 -90 nm) were prepared using a Reichert-Jung Ultracut E ultramicrotome, mounted on 150-mesh copper grids, contrasted using uranyl acetate and lead citrate, and examined on an FEI Tecnai 12 transmission microscope operated at 120 kV. Images were acquired with an AMT 16000M digital camera.
Indirect Immunofluorescence Assay-ECs were seeded on collagen I-coated slides 24 h before infection with Ad-XBP1s or Ad-XBP1u viruses at 5 MOI. Three days later, the slides were fixed with 4% paraformaldehyde, PBS for 15 min and permeabilized with 0.1% Triton X-100, PBS for 10 min, followed by incubation with primary and secondary antibodies, as described previously (22). Briefly, incubation with the primary antibodies, FLAG (goat), BECLIN-1 (rabbit), LC3␤ (rabbit), and XBP1 (rabbit), was performed for 1 h at 37°C. The bound primary antibodies were revealed by incubation with an ALEXA 546-conjugated donkey anti-goat IgG and an FITCconjugated swine anti-rabbit IgG at 37°C for 30 min. Cells were counterstained with 4Ј,6-diamidino-2-phenylindole (DAPI) (Sigma), mounted with Floromount-G (DAKOCytomation, Glostrup, Denmark), and examined under an SP5 confocal microscope (Leica, Germany). Magnification in the figures is indicated by the scale bars.
Immunoblotting-Cells were harvested and washed with cold PBS, resuspended in lysis buffer (25 mM Tris-Cl, pH 7.5, 120 mM NaCl, 1 mM EDTA, pH 8.0, 0.5% Triton X-100) supplemented with protease inhibitors (Roche Applied Science), and lysed by ultrasonication (twice, 6 s each) (Branson Sonifier 150), as described previously (22). The protein concentration was determined using the Bio-Rad protein assay reagent. 50 g of whole lysate was applied to SDS-PAGE and transferred to Hybond PVDF membrane (GE Healthcare), followed by a standard Western blot procedure. The bound primary antibodies were detected by the use of horseradish peroxidase (HRP)-conjugated secondary antibody and the ECL detection system (GE Healthcare).
Lentiviral Particle Transduction-Lentiviral particles were produced using MISSION shXBP1, shIRE1␣ and shBECLIN-1 plasmids DNA (Sigma) according to the protocol provided. The shRNA non-targeting vector was used as a negative control. Briefly, 293T cells were transfected with the lentiviral vector and the packaging plasmids, pCMV-dR8.2 and pCMV-VSV-G (both obtained from Addgene), with Fugene 6. The supernatant containing the lentivirus was harvested 48 h later and filtered, and the transducing unit was calculated, as described previously. For lentiviral infection, ECs were seeded overnight on collagen I-coated plates or flasks. The next day, cells were incubated with shXBP1 or shIRE1␣ or non-targeting control (1 ϫ 10 7 transducing units/ml) in complete medium supplemented with 10 g/ml Polybrene for 24 h. Subsequently, fresh medium in the absence or presence of endostatin (20 ng/ml) was added, and the cells were harvested 48 h later.
Chromatin Immunoprecipitation Assay-ECs (1 ϫ 10 6 cells) were infected with Ad-XBP1s or Ad-XBP1u at 5 MOI for 6 h and further cultured for 48 h, or normal ECs were used for endogenous assays. The chromatin immunoprecipitation assay was performed with a commercial kit (EZ ChIP, Millipore) according to the protocol provided and as described previously (12,24). In brief, cells were treated with 1% (v/v) formaldehyde at room temperature for 10 min and then quenched with glycine at room temperature. The medium was removed, and cells were harvested in lysis buffer. Following a short incubation on ice, chromatin was sheared by sonication. The sheared samples were diluted into 0.9 ml of ChIP dilution buffer and precleared with Protein G-agarose/salmon sperm DNA beads for 1 h. Subsequently, immunoprecipitation was conducted with anti-FLAG antibody (to precipitate XBP1s and XBP1u and their bound chromatin) or XBP1 antibody (Santa Cruz Biotechnology, Inc.) to precipitate the endogenous XBP1. Normal IgG was used as a control. Immunocomplexes were collected the following morning using Protein G-agarose/salmon sperm DNA beads pelleted by centrifugation and washed with low salt buffer, high salt buffer, and Tris-EDTA buffer (25 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, pH 7.2) to remove any nonspecific binding. The immunocomplexes were eluted from the beads using 200 l of elution buffer (100 mM NaHCO 3 , 1% SDS), and the cross-links of the protein-DNA complexes were reversed by an overnight incubation of the elute products at 65°C. A total of 2 l of proteinase K (10 g/l) was subsequently added to the solution, and samples were incubated at 45°C for 1 h. DNA was then purified using the spin columns provided. Aliquots of chromatin were also analyzed before immunoprecipitation and served as an input control. BECLIN-1 gene promoter sequences were amplified by PCR with primer sets 1-4: set 1, 5Ј-AACC-CTCACGGCTCTTATTGGAGT-3Ј and 5Ј-AGAGGCAACA-TTAGGGAGAAGCGA-3Ј; set 2, 5Ј-TTGTTCATCCGCTGA-AGCCCGCTG-3Ј and 5Ј-ACTCCCTTCTAAGGTCCCACC-TC-3Ј; set 3, 5Ј-ACCACGCCCGGCTAGTTT-3Ј and 5Ј-TTA-CCACGGGAAGTGTAGGAGCCAA-3Ј; set 4, 5Ј-TGGTCT-CGAACTCCTGACCTTGTGAT-3Ј and 5Ј-ATCCCAGCTT-AATCCCAGCTACTC-3Ј. The PCR products were analyzed in 2% agarose gel, and images were assessed with the BioSpectrum AC imaging system and Vision-WorksLS software.
Luciferase Activity Assay-The BECLIN-1 promoter (bp Ϫ206 to Ϫ910) region was amplified by PCR with a primer set of 5Ј-CAGCATCTCGAGTGTTGTTCTGAGATGGAG-3Ј and 5Ј-GCCGGTAAGCTTACTCCAATAAGAGCCGTG-3Ј and cloned into the XhoI/HindIII sites of pGL3-Luc reporter vector and verified by DNA sequencing. Additionally, pGL3-Luc-BECLIN-1 plasmid was digested with the restriction enzymes PmlI and StuI, and a truncated construct lacking the XBP1sspecific binding site was generated. In brief, ECs were transfected with full-length or truncated pGL3-Luc-BECLIN-1 (0.33 g/well) together with expression plasmid (0.16 g/well) encoded from XBP1s and XBP1u. pGL3-Luc Renilla (0.1 g/well) was included in all transfection assays as an internal control, and pShuttle2 vector was also used as a mock control. Luciferase and Renilla (Promega) activity assays were run 24 and 48 h after transfection using a standard protocol. A relative luciferase unit was defined as the ratio of luciferase activity to Renilla activity, with that of control set as 1.0.
Ex Vivo Experiments-For the generation of XBP1 EC conditional knock-out (XBP1eko) mice, an 11.5-kb DNA fragment encompassing XBP1 gene exons 1-5 was isolated from the mouse genome. Two LoxP sites were inserted into the promoter and intron 2 regions, respectively. An FRT-flanked Neo cassette was inserted upstream of the second LoxP site in the intron 2 for positive clone selection. Recombinant 129Sv/Pas ES cell clones were injected into recipient blastocysts isolated from pregnant C57BL/6J females. The injected blastocysts were then reimplanted into OF1 pseudopregnant females and allowed to develop to term. The F1 generation of chimeras was cross-bred with CAG-Cre deleter to create global knock-out XBP1 mice, whereas they were cross-bred with FLPe transgenic mice to create XBP1 loxP/ϩ mice (25,26). All of the gene cloning, ES clone selection and injection, and the XBP ϩ/Ϫ and XBP1 loxP/ϩ mouse creation were performed by genOway. The following breeding procedures were performed in the King's College London animal facility. All animal experiments in this study were performed according to protocols approved by the Institutional Committee for Use and Care of Laboratory Animals. Arterial vessels from aortas were isolated from wild type and XBP1eko mice and cut into 2 ϫ 2-mm sections. The sections were incubated with M199 supplemented with 10% FBS in the presence or absence of endostatin (20 ng/ml) for 24 h. The segments were fixed with 4% paraformaldehyde, followed by immunofluorescence staining with anti-XBP1, anti-CD31, and anti-LC3␤ antibodies, as described in the supplemental material. The segments were mounted on a slide with the vessel lumen facing up. Images were taken under an SP5 confocal microscope (Leica). Magnification is indicated in the figures by scale bars.
Statistical Analysis-Data are expressed as the means Ϯ S.E. and were analyzed with a two-tailed Student's t test for two groups or pairwise comparisons. A value of p Ͻ 0.05 was considered to be significant.

RESULTS
Sustained Activation of XBP1 mRNA Splicing Induces Autophagy in ECs-In the present study, we tested the hypothesis that XBP1 splicing induced autophagy in ECs. The study was initiated by infected ECs with XBP1 spliced form adenovirus (Ad-XBP1s), where a number of structures that could be either droplets or autophagic vacuoles were observed (Fig. 1, A-C).
No such structures or very few of them were observed in ECs infected with unspliced XBP1 form (Ad-XBP1u) or null control (Ad-null) (Fig. 1, A-C). Oil Red O staining demonstrated that these structures were not oil droplets. To explore whether these structures had an autophagic phenotype, ECs were infected with Ad-XBP1s, Ad-XBP1u, or Ad-null viruses for 48 h, and samples were prepared and observed using transmission electron microscopy. The results confirmed the presence of autophagosomes in ECs infected with Ad-XBP1s (Fig. 1, D-I). The overexpression of the unspliced and spliced XBP1 is confirmed by conventional and real-time PCR in Fig. 1, J and K, respectively. These results indicate that XBP1s may trigger an autophagic response in ECs.
Ad-XBP1s-infected ECs Show a Typical "Punctate Staining" for the LC3␤ Marker-Immunofluorescence staining of LC3␤ showed a typical punctate staining for the LC3␤ marker in Ad-XBP1s-infected cells ( Fig. 2A). In order to answer the question of whether XBP1 and LC3␤ proteins are found in the same cell, more fields have been examined, as shown in Fig. 2B, and the percentage of the XBP1 and LC3␤ double-positive cells has been calculated and found to be around 95% (Fig. 2C). Addi-tional experiments have revealed that low levels of XBP1s overexpression are sufficient to trigger an autophagic response in ECs by increasing the expression of BECLIN-1

XBP1 Triggers Autophagy in Endothelial Cells
and LC3␤ (Fig. 2D). Importantly, when ECs were infected with Ad-XBP1s and treated with chloroquine, an inhibitor of the fusion of autophagosomes with lysosomes, a further increase in LC3␤II expression level was observed (Fig. 2E). These results further confirm that XBP1s triggers an autophagic response in ECs.
Overexpression of XBP1s Increases Autophagy Gene Expression-To verify that XBP1 mRNA splicing induces autophagy, XBP1s and XBP1u were overexpressed in ECs by adenoviral gene transfer, and the expression of autophagy genes was assessed at the mRNA and protein levels. Real-time quantitative RT-PCR showed that XBP1s increases BECLIN-1 and LC3␤ autophagy genes when compared with control virus-infected cells or Ad-XBP1u-infected cells in a time course exper-iment (Fig. 3A). Routine RT-PCR confirmed the overexpression of the two isoforms of XBP1 in ECs (Fig. 3B). XBP1 mRNA splicing was detected by routine RT-PCR, followed by PstI digestion. There is a PstI site in the 26-nucleotide intron of XBP1u but not in XBP1s mRNA. Digestion of the RT-PCR product with a PstI restriction enzyme allows XBP1s (not digested) and XBP1u (digested into two smaller bands) to be distinguished. Western blot analysis showed induced expression of BECLIN-1 and autophagosome-specific isoform LC3␤II, indicating activation of autophagy (Fig. 3C). Furthermore, immunofluorescence staining revealed that Ad-XBP1s increased BECLIN-1 expression in ECs (Fig. 3D). These results suggest that XBP1 splicing is involved in the induction of the autophagy gene transcription in ECs. FIGURE 2. Ad-XBP1s-infected ECs showed a typical "punctate staining" for the LC3␤ marker. ECs were infected with Ad-XBP1u or Ad-XBP1s at 5 MOI and stained with LC3␤ antibody and anti-FLAG antibody to detect exogenous XBP1s. A, immunofluorescence staining of LC3␤ showed a typical punctate staining for the LC3␤ marker in Ad-XBP1s-infected cells. B, more fields were examined and confirmed that XBP1 and LC3␤ proteins are found in the same cell. C, the percentage of the XBP1-and LC3␤-positive cells has been calculated and found to be around 95%. *, p Ͻ 0.05. D, ECs were infected with Ad-null or Ad-XBP1s at 0.5-3 MOI, and real time quantitative RT-PCR was performed for XBP1, BECLIN-1, and LC3␤II. *, p Ͻ 0.05. E, ECs were infected with Ad-null or Ad-XBP1s and treated with chloroquine for 6 h, and Western blot for LC3␤II, an inhibitor of the fusion of autophagosomes with lysosomes, is shown. Data presented are representative of or an average from three independent experiments. Error bars, S.E.

BECLIN-1 Deficiency in ECs Abolishes the XBP1-induced
Autophagy Response-To shed light on the mechanisms involved in XBP1s-induced autophagy, further experiments were performed to explore the effect of XBP1s on BECLIN-1 expression. BECLIN-1 has been shown to play an important role in the vesicle nucleation during the initial steps of autophagy (27). Importantly, when BECLIN-1 was knocked down by shRNA in ECs, the induction of autophagy mediated by XBP1 was ablated because autophagic vacuole structures were not observed (Fig. 4A). Further experiments revealed that knockdown of BECLIN-1 abolished XBP1-induced LC3␤II expression at the protein level (Fig. 4, B and C, quantification). Recent reports have shown that 14-3-3T and E2F1 regulate the BECLIN-1 gene and initiate autophagy (28). Therefore, we F, ChIP assays demonstrated that XBP1s bound directly on the BECLIN-1 gene promoter. Anti-FLAG antibody was used to pull down exogenous XBP1s and its bound DNA fragments. To define the XBP1 binding site to the BECLIN-1 gene promoter, four sets of primers covering the Ϫ17 to Ϫ1009 nt promoter region were used to amplify the precipitated DNA fragments. G, ChIP assays were performed with an anti-XBP1 antibody to pull down the endogenous XBP1s and its bound DNA fragments in normal ECs. H, when ECs were co-transfected with plasmids encoded by XBPIu, XBP1s, or control empty vector together with a truncated construct of pGL3-Luc-Beclin promoter lacking the Ϫ537 to Ϫ755 region, luciferase activity assays demonstrated an ablated transcriptional activation of BECLIN-1 induced by XBP1s, p Ͻ 0.05. Data presented are representative of or an average from three independent experiments. Error bars, S.E. explored whether XBP1s could regulate BECLIN-1 through 14-3-3T and E2F1. However, neither XBP1s nor XBP1u could regulate 14-3-3T and E2F1 gene expression (supplemental Fig.  S1), suggesting that XBP1s-regulated BECLIN-1 expression is not through these two genes. Because XBP1s is a transcription factor, we hypothesized that it may directly regulate BECLIN-1 gene expression at the transcriptional level. Luciferase assays demonstrated that overexpression of XBP1s induced the expression of the BECLIN-1 promoter in ECs (Fig. 4E). This transcriptional regulation of BECLIN-1 was specific for the XBP1s and not for the XBP1u (Fig. 4E). These results demonstrate that knockdown of BECLIN-1 abolished XBP1-induced autophagy, highlighting the importance of BECLIN-1 in induction of autophagy mediated by XBP1s.
XBP1s Directly Binds to the BECLIN-1 Gene Promoter-To investigate whether XBP1s directly binds to the BECLIN-1 promoter region, chromatin immunoprecipitation assays were performed with anti-FLAG antibody. A map of the BECLIN-1 promoter showing the relative regions covered by the primer sets 1-4 is illustrated in Fig. 4D. As expected, a binding of XBP1s to the BECLIN-1 gene promoter was detected, which locates at region Ϫ537 to Ϫ755 nt upstream of the transcription initiation site (Fig. 4F). Moreover, the binding of XBP1 to the BECLIN-1 gene promoter was verified in experiments with antibody against endogenous XBP1 (Fig. 4G). Further experiments were performed to explore whether the direct binding of XBP1s to the BECLIN-1 promoter is essential for BECLIN-1 transcription. A truncated construct of the BECLIN-1 promoter lacking the Ϫ537 to Ϫ755 region was created. Transient transfections with the intact and the truncated pGL3-Luc-BECLIN-1 promoter constructs were performed. As shown in Fig. 4H, deletion of the Ϫ537 to Ϫ755 region abolished XBP1sinduced BECLIN-1 reporter gene expression. These results suggest that XBP1 mRNA splicing triggers an autophagic response in ECs through BECLIN-1 transcriptional activation by direct binding to the Ϫ537 to Ϫ755 nt region in BECLIN-1 gene promoter.
Endostatin Induces Autophagy in ECs through IRE1␣-mediated XBP1 mRNA Splicing-Angiogenesis inhibitors, such as endostatin, are reported to induce autophagy in ECs (29). In this study, ECs treated with endostatin induced autophagy, as demonstrated by the appearance of autophagic vacuole structures (Fig. 5A). Further experiments revealed that endostatin induced the expression levels of genes related to autophagy at both the RNA level in a time course experiment (Fig. 5B) and at the protein level (Fig. 5C) after 24 h upon endostatin stimulation. Importantly, when ECs were treated with endostatin in the presence of chloroquine, an accumulation of LC3␤II was observed (Fig. 5D). To determine whether XBP1 mRNA splicing is involved in the endostatin-induced autophagy process, the effect of endostatin on XBP1 mRNA splicing was assessed. Endostatin treatment robustly increased XBP1 splicing but had no effect on the total XBP1 mRNA level (Fig. 5E). XBP1 mRNA undergoes unconventional splicing through the phosphorylation of IRE1␣ (17,18). Therefore, we further tested whether endostatin could alter the phosphorylation status of IRE1␣. Indeed, endostatin induced IRE1␣ phosphorylation, as revealed by Western blot analysis (Fig. 5F). These results suggest that endostatin can activate XBP1 mRNA splicing through IRE1␣ phosphorylation. To investigate whether XBP1 mRNA splicing is necessary for endostatin-induced autophagy gene expression, ECs were infected with non-target (shNT) or XBP1targeting (shXBP1) shRNA lentiviruses for 72 h and then treated with endostatin for 4 h. Real-time PCR analysis demonstrated that knockdown of XBP1 abolished endostatininduced autophagy gene expression (Fig. 5G). Similarly, knockdown of IRE1␣ also suppressed endostatin-induced autophagy-related gene expression (Fig. 5, H and I). These results suggest that IRE1␣-mediated XBP1 mRNA splicing is necessary for endostatin-induced autophagy in ECs. In order to address the important question of whether knockdown of XBP1 could inhibit the induction of autophagy, ECs were infected with Non-target (shNT) or XBP1-targeting (shXBP1) shRNA lentiviruses for 48 h, followed by 24-h endostatin treatment. The cells were fixed, and samples were prepared to be observed under a transmission electron microscope. It is shown that knockdown of XBP1 by shRNA abolished the presence of autophagosomes in ECs mediated by endostatin treatment (Fig.  6, A-D). A real-time PCR assay confirmed the knockdown of XBP1 by shRNA in ECs (Fig. 6E). These results reveal that knockdown of XBP1 abolished the induction of autophagy mediated by endostatin, highlighting the important role of XBP1s in the induction of autophagy.
XBP1 Deficiency in ECs Abolishes the Endostatin-induced Autophagy Response-The above data have demonstrated that XBP1 mRNA splicing is involved in autophagy response in cultured ECs. To further explore the role of XBP1 in autophagy in intact endothelium, we generated a conditional null allele of XBP1 in endothelial cells. Targeting of XBP1 was performed by introducing LoxP sites into the promoter and intron 2 regions through homologous recombination (supplemental Fig. S2A). The introduction of a LoxP site into the promoter region does not affect XBP1 transcription. The mutation mediated by Cre recombinase deletes a 2.2-kb DNA fragment comprising part of the promoter and the exons 1 and 2, totally ablating XBP1 transcription. Germ line transmission was detected by Southern blot (supplemental Fig. S2B), and deletion of XBP1 was confirmed by PCR at the genomic level (supplemental Fig. S2, C and D). XBP1 neo-loxP mice were bred to CAG-Cre (25) transgenic mice, which express Cre recombinase ubiquitously, allowing for the generation of XBP1 ϩ/Ϫ mice. XBP1 loP/loxP mice were bred to Tie2-Cre (26) transgenic mice, which express Cre recombinase in ECs and bone marrow progenitor cells, allowing for the generation of EC conditional knock-out mice (Tie2Cre/XBP1 loxP/loxP ). Supplemental Fig. S2D shows the genotypes of the littermates from the cross-breeding of Tie2Cre/ XBP1 loxP/ϩ and Tie2Cre/XBP1 loxP/loxP (XBP1eko) mice. The XBP1eko mice can survive, showing defects in angiogenesis. 4 To this end, ex vivo experiments were performed on arterial vessel sections isolated from wild type and XBP1eko mice in which XBP1 is specifically disrupted in ECs. Arterial vessels were isolated from wild type and XBP1eko mice, dissected into 2 ϫ 2-mm sections, and incubated with M199 supplemented with 10% FBS in the absence or presence of endostatin for 24 h, followed by immunofluorescence staining with anti-XBP1, anti-CD31 (Fig. 7A), and anti-LC3␤ (Fig. 7B) antibodies. As expected, no XBP1 can be detected in ECs on vessels from XBP1eko mice (Fig. 7A, bottom). Importantly, XBP1 deficiency not only reduced the basal level of LC3␤ (Fig. 7B, left panels) but also abolished the endostatin-induced LC3␤ expression (Fig.  7B, right panels). In XBP1eko mice, a low level of LC3␤ could be detected after endostatin treatment, indicating that endostatininduced LC3␤ expression and autophagy response can also be activated in XBP1-independent pathways. These results confirm that XBP1 is involved in endostatin-induced autophagy.

DISCUSSION
The findings of this study reveal that XBP1 mRNA splicing is implicated in the induction of autophagy in ECs through transcriptional regulation of BECLIN-1. In line with this finding, endostatin activated autophagic gene expression through XBP1 mRNA splicing in a IRE1␣-dependent manner. Targeting of XBP1 or IRE1␣ by shRNA in ECs ablated endostatin-induced autophagy, whereas XBP1 deficiency in ECs reduced the basal level of LC3␤ expression and ablated response to endostatin stimulation in XBP1 EC conditional knock-out (XBP1eko) mice.
Among the causes that stimulate autophagy are intracellular stress conditions and hypoxia (11,30). In response to ER stress, XBP1 mRNA undergoes unconventional splicing through IRE1␣, giving rise to a spliced form (17,18). A number of previous reports have already highlighted the link of ER stress response with the induction of autophagy (31)(32)(33)(34). For instance, Yorimitsu et al. (35) have shown that ER stress induces autophagy, which can protect against cell death. Furthermore, hypoxia has been reported to activate an autophagylysosomal degradation pathway, and unfolded protein response enhances the capacity of hypoxic tumor cells to carry out autophagy and induces cell survival (36). Recent evidence is also providing mechanistic insights into this concept by demonstrating that activating transcription factor 4 (ATF4) is required for ER stress and hypoxia-induced expansion of autophagy through direct binding to a cyclic AMP-response element binding site in the LC3␤ promoter. This results in LC3␤ up-regulation and implication in cell survival (37). In contrast, inhibition of ER stress by knockdown of IRE1 inhibited autophagy and adipogenesis (38). In the present study, we provide novel findings that suggest that XBP1s, a crucial signal transducer of ER stress response, is involved in the induction of autophagy in ECs, providing insight into EC survival.
Autophagy has also been activated in response to angiogenesis inhibitors (39). Endostatin, which is a well characterized inhibitor of angiogenesis that induces apoptosis (29,40), has been shown to induce autophagy in ECs (40) and increase BECLIN-1 expression through ␤-cateninand Wnt-mediated signaling pathways (29,41). In the present study, we provide evidence that endostatin stimulation is implicated in autophagic gene expression through XBP1 mRNA splicing in an IRE1␣-dependent manner. When XBP1 or IRE1␣ was targeted by shRNA in ECs or in XBP1eko mice, the endostatininduced autophagic gene expression and autophagosome for-

XBP1 Triggers Autophagy in Endothelial Cells
mation were ablated. These results further support the notion that XBP1s is involved in the regulation of autophagy in ECs.
BECLIN-1 is an essential autophagic protein, which has a crucial role in the initial stages of autophagy (27). Previous studies have shown that BECLIN-1 levels and autophagic vesicle formation were regulated by Bcl-2 and Bcl-x L (29), which are important players in a cell's decision to progress to an apoptotic or necrotic death (27,(42)(43)(44). In order to shed light on the mechanistic insights into the regulation of autophagy by XBP1 splicing in ECs, further experiments revealed that activation of XBP1 splicing triggers BECLIN-1 transcriptional induction through directly binding to the BECLIN-1 promoter at the region of nt Ϫ537 to Ϫ755. A number of potential transcription factor binding sites in the region of nt Ϫ537 to Ϫ755 have been identified using TFMATRIX software, including CREBP. Therefore, XBP1s may function as homodimer or heterodimer with another transcription factor, such as CREBP. Importantly, further experiments have demonstrated that XBP1s did not induce transcriptional activation of a truncated construct of the pGL3-Luc-BECLIN-1 promoter lacking the Ϫ537 to Ϫ755 region, providing further insights into the BECLIN-1 transcriptional activation mechanism through XBP1s. XBP1 has been reported to regulate histone H4 acetylation (45), whereas XBP1s is also a target of acetylation and deacetylation mediated by p300 and SIRT1 (sirtuin 1), respectively (46). In addition, it has been recently shown that the acetylation status of the ER is regulated by IRE1/XBP1, by controlling the influx of acetyl-CoA through the membrane transporter AT-1, which results in the regulation of autophagy (47). Thus, XBP1s may control the transcriptional activation of BECLIN-1 through recruiting other co-factors that induce acetylation and protein stability and/or inhibit deacetylation in a cell-dependent manner. Further experiments will elucidate these pathways and enhance our knowledge on this aspect.
Vidal et al. (50) have recently reported that targeting XBP1 protects against Huntington disease through the regulation of FoxO1 and autophagy. The authors also showed that XBP1 deficiency enhanced the severity of spinal cord injury in a mouse model (48), whereas it did not alter prion pathogenesis (49), which indicates that the XBP1-regulated autophagy pathway contributes only to certain diseases with distinct outputs (50). Therefore, based on those findings and taking into account our data, we can now describe important aspects in the regulation of autophagy, for example, the activation of XBP1 splicing has a precise role in the autophagy, and autophagymediated by XBP1 splicing in ECs is cell type-dependent. For instance, activation of XBP1 splicing in vascular smooth muscle cells does not induce autophagic response. 4 Evidence is accumulating that autophagy occurs in advanced atherosclerotic plaques (51). For instance, in advanced atherosclerosis, the role of autophagy in macrophages (3) and vascular smooth muscle cells (52) has been shown. In addition, Ox-LDL can also activate autophagy (53) in ECs through the LC3␤/ BECLIN-1 pathway, leading to the degradation of ox-LDL through lysosomes (54), whereas induction of autophagy promotes angiogenesis by activating the vascular endothelial growth factor (55). Therefore, autophagy may be defined as "a final attempt for survival" based on an adaptive response to fight against cellular stress. However, it remains elusive whether autophagy is a harmful or a protective mechanism in atherosclerotic plaques. The key questions are "how long" and "at what levels" autophagy can be activated without causing detrimental consequences (56,57). Recent studies have shown that caspase activation regulates the extracellular export of autophagic vacuoles, which reveals that apoptosis and autophagy are two closely related and tightly regulated processes (58). We have previously shown that sustained activation of XBP1 results in EC apoptosis and atherosclerosis development (12). In the present study, we provide evidence that XBP1s triggers an autophagy response in ECs. Thus, it seems that the tight regulation of the expression levels and duration of splicing activation of key molecules, such as XBP1s, could determine the threshold of the autophagic responses through regulation of precise mechanisms in a cell-specific manner. Cross-breeding of the mouse model for atherosclerosis apolipoprotein E with XBP1eko mice is currently in progress, and it will undoubtedly help to answer these questions and demonstrate the role of XBP1s in the regulation of autophagy in ECs and the development of atherosclerosis. Further studies will elucidate whether XBP1 mRNA splicing could be used as an important pharmacological target that can regulate the autophagic machinery.
In summary, endostatin and other stimuli activate IRE1␣ phosphorylation, which in turn removes the 26-nt intron from XBP1 mRNA, causing the open reading frame shift and giving rise to the spliced isoform. The spliced XBP1 translocates into the nucleus and binds as homodimer or heterodimer to the BECLIN-1 gene promoter, leading to BECLIN-1 transcription. BECLIN-1 activates autophagic response, leading to cell survival or apoptosis (Fig. 8). This study supports the notion that XBP1 induced an autophagic response in ECs through transcriptional regulation of BECLIN-1 gene expression, which mediates activation of LC3␤II and in turn induces autophagy, providing key insights into the threshold of the autophagic responses. Under normal conditions, ER-resident IRE1␣ associates with Grp78. Upon treatment with endostatin or other stimuli, IRE1␣ dissociates from Grp78 and oligomerizes, leading to its autophosphorylation. The activated IRE1␣ catalyzes the excision of a 26-nucleotide unconventional intron from XBP1 mRNA, which causes an open reading frameshift in the XBP1 mRNA, giving rise to a 371-amino acid, 54-kDa protein with an activation domain, the XBP1s isoform. The XBP1s translocates into the nucleus and binds as homodimers or heterodimers to the BECLIN-1 gene promoter region at Ϫ537 to Ϫ755 nt upstream of the transcription initiation site, leading to the activation of BECLIN-1 transcription, which mediates activation of LC3␤II and in turn induces autophagic response, leading to EC survival or apoptosis.