Promyelocytic Leukemia Protein (PML) Regulates Endothelial Cell Network Formation and Migration in Response to Tumor Necrosis Factor α (TNFα) and Interferon α (IFNα)*

Background: PML is highly expressed in endothelial cells (ECs), but its role in ECs remains largely unexplored. Results: PML is critical for TNFα- and IFNα-mediated inhibition of EC network formation and migration. Conclusion: PML is an angiogenesis inhibitor following inflammation-associated cytokine signaling. Significance: PML is a novel regulator of endothelial cell physiology. Promyelocytic leukemia protein (PML) is a tumor suppressor that is highly expressed in vascular endothelium and inflamed tissues, yet its role in inflammation-associated cytokine-regulated angiogenesis and underlying mechanism remains largely unclear. We show that tumor necrosis factor α (TNFα) and interferon α (IFNα) stimulate PML expression while suppressing EC network formation and migration, two key events during angiogenesis. By a knockdown approach, we demonstrate that PML is indispensable for TNFα- and IFNα-mediated inhibition of EC network formation. We further demonstrate that signal transducer and activator of transcription 1 (STAT1) binds PML promoter and that is an important regulator of PML expression. Knockdown of STAT1 reduces endogenous PML and blocks TNFα- and IFNα-induced PML accumulation and relieves TNFα- and IFNα-mediated inhibition of EC network formation. Our data also indicate that PML regulates EC migration, in part, by modulating expression of downstream genes, such as negatively regulating integrin β1 (ITGB1). In addition, knockdown of STAT1 or PML alleviates TNFα- and IFNα-mediated inhibition of ITGB1 expression. Antibody blockade demonstrates that ITGB1 is functionally important for PML- and STAT1-regulated EC migration. Taken together, our data provide novel mechanistic insights that PML functions as a negative regulator in EC network formation and migration.


Promyelocytic leukemia protein (PML) is a tumor suppressor that is highly expressed in vascular endothelium and inflamed
tissues, yet its role in inflammation-associated cytokine-regulated angiogenesis and underlying mechanism remains largely unclear. We show that tumor necrosis factor ␣ (TNF␣) and interferon ␣ (IFN␣) stimulate PML expression while suppressing EC network formation and migration, two key events during angiogenesis. By a knockdown approach, we demonstrate that PML is indispensable for TNF␣-and IFN␣-mediated inhibition of EC network formation. We further demonstrate that signal transducer and activator of transcription 1 (STAT1) binds PML promoter and that is an important regulator of PML expression. Knockdown of STAT1 reduces endogenous PML and blocks TNF␣-and IFN␣-induced PML accumulation and relieves TNF␣-and IFN␣-mediated inhibition of EC network formation. Our data also indicate that PML regulates EC migration, in part, by modulating expression of downstream genes, such as negatively regulating integrin ␤1 (ITGB1). In addition, knockdown of STAT1 or PML alleviates TNF␣-and IFN␣-mediated inhibition of ITGB1 expression. Antibody blockade demonstrates that ITGB1 is functionally important for PML-and STAT1-regulated EC migration. Taken together, our data provide novel mechanistic insights that PML functions as a negative regulator in EC network formation and migration.
The promyelocytic leukemia protein (PML), 2 initially identified as a fusion partner of retinoic acid receptor ␣ (RAR␣) involved in acute promyelocytic leukemia (APL), regulates several cellular processes including proliferation, apoptosis, tran-scription, virus infection, and DNA damage signaling in response to extracellular stimuli (1)(2)(3)(4)(5)(6)(7). PML is the core scaffold component of a proteinaceous nuclear structure called PML nuclear bodies (NBs) (8). The precise function of PML NBs is not clear, but they are believed nuclear depots that sequester other factors to control cellular processes such as transcription (9,10). While it is known that PML is highly expressed in vascular endothelium and inflamed tissues (8), its role in endothelial cells (ECs) is largely unknown.
The inflammation-associated cytokines tumor necrosis factor ␣ (TNF␣) and interferons (IFNs) exert their cellular effects through signaling cascades that involve extracellular receptors, and intracellular mediators (11)(12)(13). TNF␣ is released by activated macrophage at sites of inflammation and binds to TNF␣ receptors on the local vascular endothelium triggering multiple responses in these cells, including up-regulation of leukocyte adhesion molecules and increased endothelial permeability. TNF␣ possesses pro-apoptotic and anti-viral activities and is capable of inhibiting tumorigenesis; however it is also able to stimulate cell survival. As such, intracellular responses to TNF␣ are controversial and long-debated (13). Type I (IFN␣ and IFN␤) and type II (IFN␥) IFNs are induced in the innate and adaptive immune responses, respectively and are generally considered to be anti-viral and pro-apoptotic (12,14). IFN␣, released by leukocyte from inflamed tissues, activates macrophage, which in turn promotes the release of TNF␣. Both TNF␣ and IFN␣ affect gene transcription in local ECs; however, the molecular networks downstream of TNF␣ and IFN␣ in ECs are still not fully elucidated.
Angiogenesis is a hallmark of tumor development and chronic inflammatory diseases (15)(16)(17)(18). It is believed that cytokines, including TNF␣, and IFN␣, are important regulators of angiogenesis in local endothelium (13,14,17). Angiogenesis is a complex process and represents a systemic morphological alteration that involves enzymatic degradation and remodeling of the extracellular matrix, migration, apoptosis, and proliferation of ECs (16). IFNs elicit anti-angiogenic activity (14), whereas TNF␣ has pro-or anti-angiogenic activity probably depending on the dose, duration of treatment, and cell type (19 -21). Therefore, identification of downstream effectors that mediate TNF␣ and IFN␣ activity in ECs will help understand the mechanisms by which these two cytokines regulate physiological and pathological angiogenesis.
We have recently demonstrated that TNF␣ potently induces PML and regulates the expression of the matrix-associated metalloproteinase MMP-10 in ECs (9). Our current study supports a model in which STAT1 and PML play a role in TNF␣and IFN␣-mediated inhibition of EC network formation and migration.
Inhibition of NF-B by IKK Inhibitor VII-HUVECs were concurrently treated with TNF␣ (20 ng/ml) in the presence of vehicle, 100 nM, or 200 nM IKK inhibitor VII. Cells were collected, and aliquots of the cells were subjected to whole cell extract preparation, immunofluorescence microscopy, and total RNA preparation.
Total RNA Extraction, RT-PCR, and Real-time PCR-Cells were harvested, and total RNA was extracted with a PrepEase kit (USB/Affymetrix), quantified by A 260 /A 280 spectrometry. 2 g of total RNAs were used for RT-PCR. The cDNA pool was generated from each RNA sample with Superscript3 Reverse Transcriptase (Invitrogen) according to the manufacturer's instructions. The specific cDNAs of interest and internal controls were quantified by real-time PCR using an iCycler (Bio-Rad) platform with 2ϫiQ SYBR Green Supermix (Bio-Rad) and appropriate primer sets. The PCR program ran for 40 cycles with three steps of 95°C for 10 s, 55°C for 20 s, and 72°C for 30 s. Melting curves were acquired after PCR to ensure the homogeneity of the PCR products. The relative quantities of genes of interest were normalized to an internal control (GAPDH or 18s rRNA) and depicted as mean Ϯ S.D. from three independent experiments. The primer sequences are shown in supplemental Table S1.
Western Blotting and Quantification of Band Intensity-To prepare whole cell extracts, cells were harvested and lysed in RIPA lysis buffer (1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS in 1ϫ PBS) supplemented with a protease inhibitor mixture (Roche). The extracts were fractionated by SDS-PAGE and transferred to PVDF membranes (Millipore). Membranes were blocked in 10% nonfat milk in 1ϫ PBS supplemented with 0.1% Tween-20 (PBST) for 1 h and followed by incubation with primary antibodies in PBST for 1 h. The membranes were rinsed and washed in PBST and incubated with HRP-conjugated secondary antibodies in PBST for 1 h. The membranes were incubated with chemiluminescent substrate for HRP (Thermo Sci. 34080) and exposed on autoradiography film (Denville Sci.). For quantification of band density, films were scanned, and the integrated density of each band was quantified with the "Analyze: Gels" tools in ImageJ software (v1.42a, NIH). The densities of proteins of interest were normalized to that of an internal control, and the first lane was set as 1 to reflect the fold change in the remaining lanes.
Immunofluorescence Microscopy-HUVECs, plated on glass cover slips, were treated with or without TNF␣ and IFN␣ for 16 h, and the same protocol was followed for HUVECs transfected with siRNAs. The cells were fixed in 1% paraformaldehyde in 1ϫ PBS for 30 min at room temperature, permeabilized in 1ϫ PBS supplemented with 0.1% Triton X-100 and 10% goat serum for 10 min, washed three times with 1ϫ PBS, and blocked in 1ϫ PBS containing 10% goat serum and 0.1% Tween-20 for 1 h followed by incubation with primary antibodies for 1 h. After washing, Alexa Fluor secondary antibodies were added for 1 h in the dark. Cover slips were mounted on slides using Vectashield mounting medium with DAPI (Vector Laboratories), visualized and images captured on a Leica immunofluorescence microscopy. Unless specified, all images were taken under same microscope setting.
In Vitro EC Network Formation Assay-The assays were performed following the manufacturer's protocol (Lonza ECM625). Under our experimental conditions, we did not observe significant differences in apoptosis or viability of HUVEC transfected with or without siRNAs against PML or a control siRNA (data not shown). Briefly, HUVECs or HMVECs were transfected with control siRNA or siRNAs against PML for 72 h, and followed by a 16 -20 h treatment with TNF␣ (10 ng/ml), IFN␣ (10 3 units/ml), or IFN␥ (10 3 units/ml). Subsequently, the cells were trypsinized and counted. Equal numbers of HUVECs were plated on matrix gel (Chemicon ECM625) pre-coated 96-well plate (1 ϫ 10 5 /well) or chamber system (2.5 ϫ 10 5 /chamber, Lab-Tek 4808). A fraction of the cells was plated for Western blotting to examine PML knockdown efficiency. After seeding the cells on the ECM, the images of network formation from randomly chosen fields (plate, n ϭ 12; chamber, n ϭ 8) were taken at 3, 8, and 20 h. The trends of change in network formation are similar for these time points.
The images taken at 20 h are presented. The numbers of branch points were quantified and depicted as mean Ϯ S.D. For statistics used in Figs. 1, 2, 3, and 4, unpaired two-tail t-tests were applied, and the p values were presented as *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; #, p Ͻ 0.00001; and ns (not significant, p Ͼ 0.05).
Chromatin Immunoprecipitation (ChIP) Assays-This protocol was adapted with modification as previously described (22). Briefly, 1 ϫ 10 6 HUVECs were used for each immunoprecipitation. Crosslinked nuclei were sonicated on a sonic dismembrator (Fisher 550␣) with power setting 4 of 120 s total processing time in a 10s-pulse-50s-idle manner to shear the chromatin to around 600 bp. The retrieved DNA was quantified by qPCR. Each ChIP experiment had 2-3 biological repeats. The antibodies used were listed under "Reagents and Antibodies." The input signal was set to 1, and others were normalized as percentage of input. The primer sequences are listed in supplemental Table S1.
In Vitro Wound-healing Assay-The assay was adapted from a previous protocol with modification (23). Confluent HUVECs plated in a 6-well plate, 16 -20 h after treatment of TNF␣, IFN␣, or IFN␥ and 72 h post-transfection of siRNA if applied, were scratched with a pipette tip to create wounds. Normal IgG or blocking antibody against integrin ␤1 (␣-ITGB1, 10 g/ml) was supplemented in media during the assay if specified. Randomly chosen fields (n ϭ 16) were used for imaging, and the images were taken at identical locations at time 0 and overnight. The wound widths were measured by Photoshop (Adobe) software, normalized and represented as the percentage of wound measured at time 0 (mean Ϯ S.D.). All statistics were performed with unpaired two-tailed t test.
Transwell Migration Assay-Boyden chambers with 8 m pore size were purchased from Corning (3422). HUVECs were transfected with control or PML siRNAs for 72 h. The cells were trypsinized, counted and 1 ϫ 10 5 cells/well were plated on 6.5 mm diameter inserts in the upper chamber of the apparatus. Migration was allowed to proceed overnight. For migration toward serum (10% FBS), the cells were starved in medium containing 1% FBS for 24 h and migrated toward medium supplemented 10% FBS placed in the lower chamber. For migration toward extracellular matrix (ECM), the bottom of the inserts was coated with ECM (Chemicon, ECM625), and both upper and lower chambers contained normal culture medium. After migration, the cells that had migrated through the inserts and attached to the underneath of the inserts were fixed in 4% paraformaldehyde for 10 min and stained in crystal violet for 15 min. The inserts were then air dried and subjected to cell counting under light microscopy with 40ϫ magnification. The experiments were done in duplicate, and 8 random fields were counted from each insert. The unpaired two-tail t-tests were used to determine the significance of differences in migration between the different conditions.
Microarray Analyses-Human umbilical vein endothelial cells (HUVECs) were transfected with control siRNA or two different PML siRNA for 72 h, and total RNA was extracted by a PrepEase Kit (USB/Affymetrix). The total RNAs were sent to the Genomics Core at Cleveland Clinic Foundation for microarray with HumanRef-8_V2_0_R0_11223162_A (Illu-mina) chip. Each sample had technical duplicates. All statistical analysis was done in R/Bioconductor (24,25).
Statistical Analysis of Microarray Data-Data were analyzed in the R/Bioconductor environment (24,25). Briefly, the raw data were exported by BeadStudio (Illumina, v3.2.3) with background subtracted and imported and preprocessed by lumi package (26,27). Data were transformed by the Variance-Stabilizing Transformation (VST) package (28) and then normalized by the Robust Spline Normalization (RSN) package (28). The significantly changed gene list was retrieved by a general linear model and empirical Bayes method through the Linear Models for Microarray Data (limma) package (29). The false discovery rate (FDR) was adjusted by Benjamini and Hochberg's method. Changes in expression level of Ͼ2 fold and p Ͻ 0.01 were considered significant. Gene ontology analysis was done by a hypergeometric test through the GeneAnswers package (30).
Accession Number of Microarray Data-The microarray data are deposited in the Gene Expression Omnibus (GEO) database. The accession number of the data set is GSE29238.

RESULTS
To examine the role of PML in EC network formation, we carried out PML knockdown followed by in vitro network formation assay to assess the ability of ECs to undergo angiogenesis. We found that that knocking down PML potentiated the in vitro angiogenesis of primary human umbilical vein endothelial cells (HUVECs) (Fig. 1, A and B). TNF␣ has been shown to have an ambiguous role in endothelial angiogenesis (20,21), cell migration, and adhesion (21,(31)(32)(33). Interestingly, we have previously shown that TNF␣ induces PML protein accumulation in HUVECs (9). To better understand the role of PML in TNF␣-regulated angiogenesis, we determined the effect of TNF␣ on EC network formation and found that TNF␣ inhibited network formation in ECs in a dose-dependent manner ( Fig. 1, C and D). Similarly, knockdown of PML increased network formation in microvascular endothelial cells (HMVECs) (Fig. 1E). PML is highly expressed in inflamed tissues and normal vascular endothelium (8). Part of the inflammatory response includes changes in the vascular endothelium induced by inflammation-associated cytokines. We were therefore interested in whether PML is up-regulated in ECs treated with other cytokines, such as IFNs and what effects these cytokines had on ECs. To test this, HUVECs were treated with TNF␣, IFN␣, or IFN␥ for 16 h followed by in vitro EC network formation assays. We found that these cytokines inhibited in vitro network formation and induced PML protein levels in HUVECs (Fig. 2, A-C). A similar observation was obtained in HMVECs (Fig. 2, D and E).
The observations that PML knockdown promotes EC network formation and that TNF␣, IFN␣, and IFN␥ induce PML expression and inhibited EC network formation raised the possibility that induction of PML is mechanistically involved in the ability of these cytokines to inhibit network formation. To test this, we performed siRNA knockdown of PML followed by TNF␣, IFN␣, or IFN␥ treatment and in vitro network formation assays in HUVECs and HMVECs. We found that knockdown of PML significantly alleviated TNF␣-, IFN␣-, and IFN␥-mediated inhibition of network formation in HUVECs (Fig. 3, A  and B). A fraction of cells were plated and harvested for Western blotting to determine PML knockdown efficiency (Fig. 3C). Furthermore, TNF␣ and IFN␣ were more potent inducers of PML accumulation than IFN␥, which correlates with the efficiency of inhibition in the network formation assays. In addition, these cytokines significantly inhibited network formation in HMVECs (Fig. 3D). Similarly, a fraction of HMVECs from experiments in Fig. 3D was harvested to verify PML knockdown efficiency (data not shown). No significant differences in apoptosis of HUVEC treated with vehicle, TNF␣, or IFN␣ were observed under our experimental settings (data not shown).
The signal transducers and activators of transcription (STAT) are the primary transcription factors that mediate type I interferon activity (12,14). By microarray analysis, we found that STAT1 is the most abundant of the STAT family member transcripts in HUVECs. In addition, PML is known as an IFNsstimulated gene (34). We suspected that STAT1 regulates PML transcription in HUVECs following IFN␣ treatment. Indeed, IFN␣ was capable of inducing STAT1 and PML mRNA (Fig.  4A) and protein levels (Fig. 4B) as well as PML NBs formation (supplemental Fig. S1). Knockdown of STAT1 by siRNA significantly abolished the ability of IFN␣ to induce PML mRNA, protein levels, and formation of PML NBs (Fig. 4, A, C, D). Surprisingly, we also observed that TNF␣, when treated solely at non-apoptotic dose (20 ng/ml), modestly induced STAT1 mRNA accumulation and protein levels (Fig. 4, A-D). Although the mechanism by which TNF␣ induces STAT1 in HUVECs is currently unclear, knockdown of STAT1 abolished TNF␣-mediated induction of PML (Fig. 4, A, C, D). These data indicate that STAT1 positively regulates PML expression following TNF␣ and IFN␣ treatment. We analyzed PML promoter and identified two putative STAT binding sites upstream of the transcription start site predicted by Genomatix (Fig. 4E). These two putative STAT binding sites are different from a previously mapped STAT responsive element in the exon 2 of the PML  JULY 6, 2012 • VOLUME 287 • NUMBER 28 gene (34). By chromatin immunoprecipitation assays, we show that STAT1 binds to these two putative sites following TNF␣ and IFN␣ treatment and that TNF␣ and IFN␣ treatment also increased the acetylation of histone H3 in these two sites compared with vehicle (Fig. 4E). To test whether STAT1 plays a role in TNF␣and IFN␣-mediated inhibition of EC network formation, we transiently knocked down STAT1 and assessed the ability of TNF␣ and IFN␣ to inhibit EC network formation. We found that knockdown of STAT1 significantly relieved TNF␣and IFN␣-mediated inhibition of network formation (Fig. 4F). These data indicate that STAT1 is critical for basal and TNF␣and IFN␣-induced expression of PML mRNA and protein and that STAT1 is important for TNF␣-and IFN␣-mediated inhibition of EC network formation.

The Role of PML in Endothelial Cells
EC migration is an essential step during network formation. To further assess the role of PML in EC angiogenesis, we analyzed how PML regulates EC migration. The observation that knockdown of PML promoted HUVEC migration (supplemental Fig. S2) indicated that PML may regulate expression of genes involved in cell migration. We carried out a microarray analysis using total RNA extracted from HUVECs either treated with a control siRNA or with two different PML siRNAs. We identified a panel of altered genes known to be involved in cell migration and angiogenesis following PML knockdown. A heatmap presentation (Fig. 5A) shows that some of the known angiostatic genes were down-regulated while other known angiogenic genes were up-regulated after knockdown of PML. Of the significantly up-regulated genes, we focused our study on integrin ␤1 (ITGB1). ␤ integrins play an important role in endothelial cell migration during angiogenesis (31). Among the 8 human ␤ integrins (namely ITGB1-8), ITGB1, ITGB3, and ITGB5 are expressed in endothelial cells and ITGB1 is the pre-dominant form (31, 33). Indeed, our microarray analysis showed that ITGB1 is the most highly expressed ␤ integrin in HUVECs, although expression of ITGB3 and ITGB5 was also detectable. We also found that knockown of PML specifically up-regulates ITGB1 but not ITGB3 or ITGB5 (data not shown). We therefore focused on ITGB1. ITGB1 has been implicated in ischemic neovascularization (35), cell-to-cell and cell-to-extracellular matrix adhesion and cell migration (36,37). We hypothesized that PML regulates EC migration, in part, by modulating ITGB1 expression. To test this, we first confirmed that knockdown of PML resulted in increased ITGB1 mRNA and protein accumulation in HUVECs (Fig. 5, B and C). In addition, TNF␣, IFN␣, and IFN␥ treatment, which increase PML level (Fig. 2), suppressed ITGB1 mRNA and protein levels in both HUVECs and HMVECs (Fig. 5, D and E). We also examined and found that knockdown of PML does not significantly alter the expression of VEGFA, VEGFB, and VEGFC (data not shown).
We also found that knockdown of STAT1 or PML by two different siRNAs relieved the TNF␣and IFN␣-mediated reduction of ITGB1 protein levels in HUVECs (Fig. 6, A and B). Furthermore, to verify that ITGB1 is important for endothelial cell migration, we used a neutralizing antibody to block ITGB1 from interacting with extracellular matrix in the in vitro wound-healing assays. We found that addition of the ITGB1 neutralizing antibody significantly impaired HUVEC migration and further inhibited the migration of HUVEC pre-treated with TNF␣ or IFN␣ (Fig. 6C). To test whether ITGB1 plays a role in STAT1-and PML-mediated inhibition of EC migration, we knocked down STAT1 or PML by two different siRNAs followed by the addition of ITGB1 neutralizing antibody in the in vitro wound-healing assays. We found that knockdown of STAT1 or PML promoted cell migration, while ITGB1 neutralizing antibody significantly impaired cell migration in STAT1 or PML knockdown HUVECs (Fig. 6D).
Because in vitro wound-healing assays measure EC migration in a context of EC monolayer integrity, to better assess the role of PML on EC migration, we carried out transwell migration assays. We found that concurrent treatment of TNF␣ and IFN␣ suppressed HUVEC transwell migration toward higher concentration of fetal bovine serum (FBS) (Fig. 7A). However, knockdown of PML promoted the HUVEC transwell migration in the presence of TNF␣ and IFN␣. Consistent with FBS-directed migration, we found that TNF␣ and IFN␣ inhibited HUVEC transwell migration to an independent attractant, extracellular matrix (ECM), and that knockdown of PML promoted HUVEC transwell migration and relieved TNF␣-and IFN␣-mediated inhibition (Fig. 7B). These results demonstrated that the ability of TNF␣ and IFN␣ to inhibit transwell migration partly depends on PML.

DISCUSSION
Angiogenesis is one of the hallmarks in tumorigenesis and inflammatory diseases (15)(16)(17)(18). Inflammation-associated cytokines, such as TNF␣ and IFN␣ invoke various cellular responses after binding to cell surface receptors, however their roles in angiogenesis are not fully understood (13, 38 -40). PML is a tumor suppressor and highly expressed in ECs but its role in ECs remains largely unknown. In this study, we show that PML has a novel inhibitory role in regulating EC network formation and migration in response to TNF␣ and IFN␣. Our data demonstrate that PML is an important mediator for TNF␣-and IFN␣-mediated inhibition of EC network formation and migration. We are first to show that besides IFN␣, TNF␣ also up-regulates PML mRNA, in part, through transcriptional factor STAT1 in ECs. Through microarray analysis in ECs, we identified ITGB1 as a potential PML target gene involved in cell migration and angiogenesis. In addition, our data indicate that TNF␣ and IFN␣ negatively regulate the expression of

The Role of PML in Endothelial Cells
ITGB1 in a PML-dependent manner. The ITGB1 antibody significantly blocked PML knockdown-mediated increase in EC migration, suggesting that ITGB1 is partially responsible for PML-regulated EC migration. Transwell chamber migration assays also confirmed that PML suppression of EC migration and its importance for TNF␣and IFN␣-mediated

The Role of PML in Endothelial Cells
23362 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 287 • NUMBER 28 • JULY 6, 2012 reduction of EC migration. Together, our study provides concrete evidence implicating an angiostatic role of PML in EC physiology. STAT1 is activated by IFN␣ and IFN␥ in several non-endothelial cells (41)(42)(43) and by IFN␥ in HUVEC (44). PML is an interferon-stimulated gene (34); however, whether STAT1 plays a role in IFN␣-induced PML mRNA accumulation in ECs and the mechanisms underlying this regulation remained largely unknown. Functionally, we show for the first time that knockdown of STAT1 significantly reduced basal as well as TNF␣and IFN␣-induced PML levels. These data indicate that STAT1 positively regulates PML gene transcription in response to cytokine treatment. Consistent with the hypothesis, we demonstrate that TNF␣ and IFN␣ treatment increased STAT1 binding to the PML promoter at two novel STAT binding sites. These data suggest that STAT transcription factors account for the transcriptional induction of PML expression following TNF␣ or IFN␣ treatment. To our knowledge, we are first to show that STAT1 can be induced by TNF␣ in ECs, though to a lesser extent than IFN␣. A possible mechanism is that TNF␣ treatment stimulates ECs to release interleukins (ILs) through NF-B pathway (12,14). ILs, such as IL15, have been shown to mildly activate STAT1 signaling (45). However, full understanding of this process requires future studies.
Cytokine signaling plays pivotal roles to modulate EC activities through mechanisms that are yet to be fully elucidated. PML is highly expressed in ECs (8); however its precise role in cytokine-modulated angiogenesis is not understood. Our data FIGURE 4. STAT1 participates in the TNF␣-and IFN␣-induced PML expression and inhibition of EC network formation in HUVECs. A, effects of STAT1 knockdown on PML mRNA levels analyzed by qRT-PCR. B, effects of TNF␣ and IFN␣ on endogenous STAT1 and PML protein levels in HUVECs assayed by Western blotting. C, effects of STAT1 knockdown by two siRNAs on PML protein levels following TNF␣ and IFN␣ treatment assayed by Western blotting. B and C, Western blotting bands were quantified by Image J as described in "Experimental Procedures." ND, not detectable. D, effects of STAT1 knockdown on PML NBs examined by immunofluorescence microscopy (DAPI, blue; STAT1, green; PML, red). The microscopic parameters were set identically for all images taken. E, association of STAT1 with the PML promoter following TNF␣ and IFN␣ treatment assayed by chromatin immunoprecipitation followed by quantification with qPCR. HUVECs treated with TNF␣ (20 ng/ml) or IFN␣ (10 3 units/ml) for 16 h followed by ChIP assays as described in "Experimental Procedures." Putative STAT binding sites were predicted with Genomatix. TNF␣ and IFN␣ increase acetylation of histone H3 flanking STAT1 binding sites at the PML promoter. F, effects of knockdown of STAT1 on TNF␣-and IFN␣-mediated inhibition of EC network formation in HUVECs. Quantitation of branch points after in vitro EC network formation assays was as described in Fig. 1, mean Ϯ S.D. (n ϭ 8). Statistics: unpaired two-tail t-tests (**, p Ͻ 0.001; ***, p Ͻ 0.0001; ns, not significant). show that TNF␣ and IFN␣ induced PML mRNA and protein accumulation in ECs. Furthermore, by a knockdown approach, we demonstrate that PML is an important effector that mediates the inhibitory effects by TNF␣ and IFN␣ on network formation in both HUVECs and HMVECs. Interestingly, we noticed that knockdown of PML does not fully abolish the inhibition of EC network formation by cytokines. This suggests that PML-independent pathways are also involved, although this could be due to incomplete knockdown of PML in cytokinetreated cells (Fig. 3). The TNF family member, TRAIL and IFN␣ have been reported to induce apoptosis in cancer cells (46,47). However, we did not observe significant differences in apoptosis between cells treated with vehicle, TNF␣ or IFN␣ in human primary endothelial cells, as judged by trypan blue and crystal violet cell viability assays (data not shown). Therefore, we do not think apoptosis accounts for the reduced tube formation observed in TNF␣or IFN␣-treated HUVECs. Instead, we conclude that the ability of PML to inhibit EC network formation is likely achieved through alterations in target gene expression that ultimately lead to remodeling of the extracellular matrix, migration, apoptosis, and proliferation. Consistent with this notion, PML Ϫ/Ϫ mice exhibited increased neovascularization following ischemic hypoxia, partly through down-regulation of HIF1␣, a master angiogenic regulator (40). While HIF1␣ regulates VEGF transcription, the PML Ϫ/Ϫ mouse of neovascular-ization model also reported a modest increase in plasma VEGF concentration (40), raising the possibility that PML regulates VEGF gene transcription. In addition, we previously demonstrated that PML is critical for basal and TNF␣-induced accumulation of MMP-10, a pivotal regulator in remodeling of the extracellular matrix (9). To further investigate the mechanisms underlying regulation of angiogenesis by PML in human endothelial cells, we performed a microarray analysis and identified an array of significantly changed genes involved in cell migration and angiogenesis. Surprisingly, our microarray analysis results suggest that knockdown of PML by two indepenent siRNAs in HUVECs does not significantly alter VEGF family gene transcription (VEGFA, VEGFB, and VEGFC). One possibility is that the reported modest increase of circulating VEGF in PML Ϫ/Ϫ mouse model (40) results from certain non-endothelium mechanism. Alternatively, we still cannot rule out the potential species difference between human and mouse. Interestingly, some other known angiostatic genes were down-regulated while other angiogenic factors were up-regulated as shown by our microarray analysis results in HUVECs. This expression profile conveys a picture of the molecular basis by which knockdown of PML promotes angiogenesis in ECs and identifies potential novel PML target genes in endothelial cells.
An early major event downstream of TNF␣ signaling is the activation of NF-B transcription factor (13). NF-B activation on TNF␣-and IFN␣-mediated reduction of ITGB1 expression assayed by Western blotting. C, effects of ITGB1 blocking antibody on HUVEC wound-healing migration following TNF␣ or IFN␣ treatment. The width of the initial wound was set as 100%, mean Ϯ S.D. (n ϭ 16). closed means wound width percentage is 0%. D, effects of ITGB1 blocking antibody on HUVEC wound-healing migration following knockdown of STAT1 or PML. siCTRL (control siRNA), two STAT1 siRNAs (siSTAT1-1 and siSTAT1-2), and two PML siRNAs (siPML-1 and siPML-2). An aliquot of the cells was used to determine STAT1 and PML knockdown efficiency as assayed by Western blotting (top panel), and the wound width percentages were quantified (bottom panel). Unpaired two-tail t test (**, p Ͻ 0.01; ***, p Ͻ 0.001; #, p Ͻ 0.00001).
has been shown to be angiogenic (48). Concurrent treatment with TNF␣ and IKK inhibitor VII, a previously reported (49 -51) selective IKK inhibitor which inhibits NF-B activation by blocking IB degradation, efficiently inhibited NF-B nuclear translocation (supplemental Fig. S3A) and target gene induction (supplemental Fig. S3B), but showed no obvious effects on TNF␣or IFN␣-mediated induction of PML protein accumulation (supplemental Fig. S3C). However, the observation that PML inhibits NF-B transcription activity (52) raises the possibility that part of the TNF␣ effect may depend upon the antagonism between PML and NF-B signaling. As a tumor suppressor (53,54), PML expression is subject to regulation by a host of extracellular stimuli (10). Interestingly, PML protein levels in many cancerous tissues are expressed at a lower level than that in normal tissues (54). We suspect that in some cancerous tissues, TNF␣ may have lost its ability to induce PML expression and thus becomes pro-angiogenic.
Previous studies have suggested that STAT1 negatively regulates angiogenesis (55). Knockdown of STAT1 in HUVECs blocks IFN␥-mediated inhibition of EC network formation. Furthermore, STAT1 inhibits angiogenesis by suppressing the expression of proangiogenic genes, including bFGF, MMP-2, MMP-9 in response to IFN␤ treatment (56). In addition, STAT1 is also involved in IFN␥-mediated inhibition of mono-cyte migration (57) and integrin-mediated cell adhesion (58). Our data expand the molecular understanding of the role of STAT1 in EC angiogenesis by demonstrating that STAT1 is important for TNF␣and IFN␣-mediated inhibition of angiogenesis in part by activating PML expression. Furthermore, we demonstrate that in response to TNF␣ and IFN␣, STAT1 regulates HUVEC migration, in part, through an ITGB1-dependent manner.
Endothelial cell migration is an important process during angiogenesis and transcriptional regulation of certain key genes could be one of the regulatory mechanisms involved in this multiple-faceted process. Among the significantly changed genes we identified by microarray analyses, we focused on ITGB1. We hypothesized that PML regulates EC migration at least partially by regulating ITGB1 expression. Indeed, we not only confirmed ITGB1 is suppressed by PML and inhibited by TNF␣ and IFN␣ in a STAT1-and PML-dependent manner, we also demonstrated that ITGB1 is functionally important in STAT1-and PML-regulated HUVEC migration. These data suggest that both PML and STAT1 are important regulators of ITGB1 expression in ECs. Furthermore, antibody experiments show that blocking ITGB1 delays the wound healing even in the presence of TNF␣ and IFN␣, though to a lesser extent compared with vehicle treatment (Fig. 6C). These data suggest that the ability of TNF␣ and IFN␣ inhibit EC migration requires both ITGB1-dependent and independent pathways. Interestingly, we also noticed that in ␣-ITGB1 antibody treated ECs, knockdown of PML resulted in relatively faster migration than the control siRNA-transfected cells and STAT1 knockdown cells (Fig. 6D). This suggests that PML inhibits EC migration through both ITGB1-dependent and -independent pathways. Overall, these data support a model in which ITGB1 is an important effector for EC migration downstream of cytokine signaling, STAT1 and PML.