Snail Is a Direct Target of Hypoxia-inducible Factor 1α (HIF1α) in Hypoxia-induced Endothelial to Mesenchymal Transition of Human Coronary Endothelial Cells*

Background: Imbalanced TGFβ/BMP-mediated signaling has been identified as a principal stimulus of EndMT. Results: The EndMT master regulator SNAIL is a direct target of HIF1α. Conclusion: Hypoxia-induced EndMT is mediated by HIF1α through direct targeting of SNAIL. Significance: This study provides conceptual clues of how endothelial cells undergoing EndMT relate to tip cells associated with sprouting angiogenesis in response to hypoxia. Endothelial to mesenchymal transition (EndMT) was originally described in heart development where the endocardial endothelial cells that line the atrioventricular canal undergo an EndMT to form the endocardial mesenchymal cushion that later gives rise to the septum and mitral and tricuspid valves. In the postnatal heart specifically, endothelial cells that originate from the endocardium maintain increased susceptibility to undergo EndMT as remnants from their embryonic origin. Such EndMT involving adult coronary endothelial cells contributes to microvascular rarefaction and subsequent chronification of hypoxia in the injured heart, ultimately leading to cardiac fibrosis. Although in most endothelial beds hypoxia induces tip cell formation and sprouting angiogenesis, here we demonstrate that hypoxia is a stimulus for human coronary endothelial cells to undergo phenotypic changes reminiscent of EndMT via a mechanism involving hypoxia-inducible factor 1α-induced activation of the EndMT master regulatory transcription factor SNAIL. Our study adds further evidence for the unique susceptibility of endocardium-derived endothelial cells to undergo EndMT and provides novel insights into how hypoxia contributes to progression of cardiac fibrosis. Additional studies may be required to discriminate between distinct sprouting angiogenesis and EndMT responses of different endothelial cells populations.


Endothelial to mesenchymal transition (EndMT) was originally described in heart development where the endocardial endothelial cells that line the atrioventricular canal undergo an
EndMT to form the endocardial mesenchymal cushion that later gives rise to the septum and mitral and tricuspid valves. In the postnatal heart specifically, endothelial cells that originate from the endocardium maintain increased susceptibility to undergo EndMT as remnants from their embryonic origin. Such EndMT involving adult coronary endothelial cells contributes to microvascular rarefaction and subsequent chronification of hypoxia in the injured heart, ultimately leading to cardiac fibrosis. Although in most endothelial beds hypoxia induces tip cell formation and sprouting angiogenesis, here we demonstrate that hypoxia is a stimulus for human coronary endothelial cells to undergo phenotypic changes reminiscent of EndMT via a mechanism involving hypoxia-inducible factor 1␣-induced activation of the EndMT master regulatory transcription factor SNAIL. Our study adds further evidence for the unique susceptibility of endocardium-derived endothelial cells to undergo EndMT and provides novel insights into how hypoxia contributes to progression of cardiac fibrosis. Additional studies may be required to discriminate between distinct sprouting angiogenesis and EndMT responses of different endothelial cells populations.
Endothelial to mesenchymal transition (EndMT) 4 refers to a cellular process through which endothelial cells delaminate from their organized layer and migrate away, possibly invading the surrounding connective tissue (1)(2)(3). The acquired mesenchymal phenotype is associated with increased expression of mesenchymal marker proteins such as ␣-smooth muscle actin (␣-SMA) and collagen 1A1 (Col1A1) and decreased expression of typical endothelial markers such as CD31 (Pecam-1) and vascular endothelial (VE)-cadherin. EndMT was originally described in heart development where the endocardial endothelial cells that line the atrioventricular canal undergo EndMT to form the endocardial mesenchymal cushion that later gives rise to the septum and mitral and tricuspid valves (4). Postnatal EndMT contributes to pathologies including cardiac fibrosis (5)(6)(7)(8)(9)(10). Although it is not clear whether all endothelial cells can undergo EndMT, recent studies have suggested that among cardiac endothelial cells it is specifically endocardial endothelial cells and endothelial cells of endocardial origin (such as coronary endothelial cells) that can undergo EndMT possibly as remnants of their embryonic plasticity (11).
The molecular pathways that control EndMT involving embryonic and postnatal endothelial cells are still incompletely understood. Although an imbalance of TGF␤/bone morphogenetic protein-mediated signaling has been identified as a principal stimulus of EndMT, additional pathways such as Notch signaling can also induce EndMT (12)(13)(14)(15). Several studies have implicated unequivocally that all inducers of EndMT culminate in an induction of at least one of the transcription factors Snail, Slug, and Twist (16). Because all three transcription factors also control various epithelial-mesenchymal transitions during embryogenesis and cancer progression, they have been highlighted as master regulators of cell plasticity, directly targeting numerous genes involved in cellular transitions (17)(18)(19).
Chronic hypoxia is a hallmark of cardiac fibrosis resulting from both rarefaction of microvessels (and subsequent oxygen supply) and increased oxygen consumption by activated inflammatory cells and fibroblasts (20). Chronic hypoxia itself contributes to aberrant ventricular remodeling and cardiac fibrosis (21). In this regard, a circulus vitiosus of hypoxia-induced EndMT contributing to microvascular rarefaction and subsequent chronification of hypoxia appears to be an attractive concept. Here we explored the possibility that EndMT associated with cardiac fibrosis could be a direct consequence of hypoxia independent of growth factors.
The effect of hypoxia on any given cell is mediated by hypoxia-inducible factor (HIF) transcription factors that respond to changes in available oxygen in the cellular environment with HIF1␣ being the most abundant isoform. Intracellular levels of HIF1␣ under hypoxic conditions is caused by increased expression mostly through impaired proteolytic degradation: under hypoxic conditions, HIF1␣ is stabilized, dimerizes with HIF1␤, and translocates into the nucleus where it can bind to hypoxia response elements within promoter regions of target genes and induce transcriptional activity. Once normoxia is restored, prolyl hydroxylase hydroxylates HIF1␣, causing its association with the von Hippel Lindau tumor suppressor protein, ultimately causing ubiquitin-dependent degradation of HIF1␣. Vhl-null mice in which HIFs accumulate independently of hypoxia due to impaired degradation develop spontaneous cardiac fibrosis (22), prompting us to explore the involvement of HIF1␣ in hypoxia-induced EndMT.
Here we aimed to elucidate a potential role for hypoxia in the induction of EndMT involving coronary endothelial cells. Our studies demonstrate that hypoxia can induce EndMT through HIF1 signaling independently of TGF␤ and that Snail is a direct target of HIF1␣.

Experimental Procedures
Animal Experiments-C57Bl6 or Tie1Cre;R26RstopYFP mice strains were maintained at the breeding facility of the University of Göttingen. The investigation conforms to the Guide for the Care and Use of Laboratory Animals published by the National Institutes of Health (NIH Publication Number 85-23, revised 1996). The experimental protocols were approved by the Institutional Review Board of the University of Göttingen and the responsible government authority of Lower Saxony (Germany).
Ascending Aortic Constriction-Ascending aortic constriction was performed in 12-week-old mice under anesthesia with a mixture of ketamine (100 mg/kg) and xylazine (10 mg/kg) given at 0.1 ml intraperitoneally as described previously (10). Adequacy of anesthesia was monitored by observing muscle twitch and by tail pinch as well as documenting heart rate and body temperature. As postoperative analgesia, Buprenex (0.1 mg/kg) was given intraperitoneally every 6 -8 h for the first 48 h. All mice used weighed 25-30 g. For sham control operated mice, the chest was surgically opened under the same conditions as banded mice except that no aortic banding was performed. 4 weeks after operation, mice were sacrificed with a lethal intraperitoneal dose of pentobarbital sodium (100 mg/kg) followed by cervical dislocation.
Cell Culture and Immunofluorescence Staining-Human coronary artery endothelial cells (Genlantis) were grown in human coronary artery endothelial cell (HCAEC) culture medium (Genlantis) and passaged according to the company's instructions. At the fifth passage, 1 ϫ 10 6 cells were seeded into 10-cm plates (Griner), after 18 -24 h of attachment supplied with fresh medium, and incubated for an additional 72-96 h at normoxic or hypoxic conditions. For hypoxic conditions, cells were cultured in a gas mixture containing 1% O 2 in a hypoxia chamber (STEMCELL Technologies) following the protocol as described elsewhere (23,24) or treated in medium containing 400 M CoCl 2 (Sigma). To block the synthesis of HIF1␣, 100 nM digoxin (Sigma) was prior added into the medium before placed into a hypoxia chamber. For immunofluorescence staining, cells were seeded onto four-chamber culture slides (BD Falcon) after treatment with hypoxia and then fixed with ice-cold methanol/acetone (1:1) for 10 min at Ϫ20°C followed by permeabilization with 0.1% Triton X-100 in PBS and then blocking with 1% BSA in PBS for 30 min at room temperature. The cells were subjected to immunofluorescence staining with primary antibody CD31 (1:100; Dako) or ␣-SMA (1:100; Abcam) for 2 h at room temperature. The cells were washed with cold PBS and incubated with Alexa Fluor 568-labeled anti-rabbit (1:300) and Alexa Fluor 488-labeled anti-mouse (1:300) secondary antibodies (Life Technologies) at room temperature for 1 h. The cells were examined by fluorescence microscopy using a Zeiss Axiovert 200 microscope and AxioVision 3.0 software. The acquired imaged were processed using Photoshop CS3 software.
Histology and Immunohistochemistry-Ascending aortic constriction banded mouse hearts were formalin-fixed, paraffin-embedded, sectioned at 3 m, then deparaffinized in xylene, and rehydrated in ethanol containing distilled water. Masson's trichrome staining was performed as described previously (10). Immunohistochemistry were performed using carbonic anhydrase IX antibody (sc-25599, Santa Cruz Biotechnology) peroxidase-labeled with the Vectastain Universal Elite ABC kit (Vector Laboratories). 3-Amino-9-ethylcarbazole substrate chromogen (Dako) was used for peroxidase visualization according to the manufacturer's protocol. Mayer's hematoxylin solution (Sigma) was used for visualizing the cell nucleus.
RNA Extraction and Real Time PCR-Total RNA was extracted from cells or tissues using a PureLink RNA kit (Life Technologies) following the manufacturer's protocol. 1 g of total RNA was digested with DNase I (Sigma) and used for cDNA synthesis using the SuperScriptII system (Life Technologies). Diluted cDNA (1:10) was used as a template in Fast SYBR Master Mix (Life Technologies) and run on a StepOne Plus real time PCR system (Life Technologies) with real time PCR primers (sequences are listed in Table 1). Measurements were standardized to the GAPDH reaction using ⌬⌬Ct methods.
Protein Extraction and Western Blotting-Proteins were extracted from cells and tissues using Nonidet P-40 lysis buffer (Life Technologies) containing protease inhibitor mixture (Roche Applied Science). Protein samples were resolved by 4 -12% SDS-PAGE and transferred onto nitrocellulose membrane (Amersham Biosciences). After blocking with 5% dry milk in TBST (TBS, pH 7.2, 0.1% Tween 20), the membrane was incubated with primary antibodies (details and dilution factors are listed in Table 2) at 4°C overnight. On the 2nd day, after washing three times with 2% dry milk in TBST, the membrane was incubated with secondary HRP-conjugated antibodies (Cell Signaling Technology), and signals were detected using a chemiluminescence kit (Santa Cruz Biotechnology).
Chromatin Immunoprecipitation (ChIP)-ChIP assays were performed with a OneDay ChIP kit (Diagenode) and followed the protocols as described previously (25). Briefly, cells were first cross-linked and then lysed with a shearing kit (Diagenode) followed by sonication (Misonix). The sheared chromatin was immunoprecipitated with 5 g of DDK antibody (OriGene) or with IgG as a negative control (Diagenode) using the Diagenode OneDay ChIP protocol. Quantitative analysis of the immunoprecipitated DNA was performed using the StepOne Real-Time System (Applied Biosystems) using two pairs of primers (sequences are listed in Table 3) flanking the human SNAIL promoter region and one pair of primers flanking exon 2 as a negative control. The ChIP-quantitative PCR data were analyzed using the ⌬Ct method in which the immunoprecipitated sample Ct value was normalized with the input DNA Ct value, and the percentage of precipitation was calculated using the following formula: % input ϭ 2 Ϫ(Ct immunoprecipitated Ϫ Ct input) ϫ dilution factor ϫ 100%.
Transfection-For transfection experiments, cells were plated onto 10-cm plates (Griner), cultured overnight, and transfected with Lipofectamine 2000 (Life Technologies) according to the manufacturer's instructions. Briefly, the plasmid DNA (2 g each) and Lipofectamine 2000 were mixed in a ratio of 1:2 in a total volume of 500 l of Opti-MEM (Life Technologies) and allowed to form the complex by incubating for 20 min at room temperature. Then the DNA-Lipofectamine complex was added to the cells in endothelial basic medium (Genlantis). After 3 h of incubation, the medium was replaced by growth medium, and then the cells were subjected to hypoxic conditions. For the gene silencing experiment, pLKO.1 vector was used for generating pLKO.1-shHIF1␣ and pLKO.1-shSNAIL constructs (shRNA oligo sequences are listed in Table 4). For gene overexpression, pCMV6-HIF1␣ (RC202461) and pCMV-HIFAN (RC202843) were purchased from OriGene.
Cloning of SNAIL Promoter and Luciferase Reporter Assay-For generating a SNAIL promoter construct, a 1.6-kb fragment (Ϫ1638 to ϩ12 bp) located 5Ј upstream of the SNAIL coding sequence (26) was amplified using human HCAEC genomic DNA. The fragment was sequenced before being cloned in the KpnI/HindIII sites of pGL4.10 vector. Luciferase reporter assays were performed as described previously (27). Briefly, 1 day before transfection, cells were seeded (1 ϫ 10 5 cells/well) to a 6-well plate. The next day, the cells were transiently    transfected in duplicates with the indicated luciferase vectors, and 2 g of the reporter plasmid pGL4.10_SNAIL or empty vector pGL4.10 (Promega) was co-transfected with 0.2 g of pGL4.73 (Promega), a Renilla control vector for normalization of transfection efficiencies. The cells were lysed and assayed using the Dual-Glo Luciferase Assay System (Promega). Statistical Analysis-All quantitative PCR data for RNA expression analysis (two or more biological replicates) were calculated using the ⌬⌬Ct method. One-way ANOVA (multiple group analysis) and Student's t test (GraphPad Prism 5.1) were used to obtain calculations of statistical significance.
Cultivation of HCAECs under hypoxic conditions induced intracellular accumulation of Hif1␣ (Fig. 1, J and K), correlating with increased expression of mesenchymal marker ␣-SMA (Fig. 1J) and decreased expression of endothelial VE-cadherin (Fig. 1J). Moreover, EndMT transcription factor SNAIL colocalized with HIF1␣ under hypoxia (Fig. 1K). Based on this correlative evidence, we next aimed to further explore a causal link of Hif1␣ accumulation and EndMT.
HIF1␣ is the major transcription factor specifically activated during hypoxia. Because cobalt aberrantly increases Hif1␣ under normoxia by inhibiting the prolyl hydroxylase domaincontaining enzymes and subsequently inhibiting Hif1␣ degradation (28,29), we next exposed HCAECs to CoCl 2 to specifically induce Hif1␣ independently of hypoxia. Supplementation of cell culture media with CoCl 2 caused intracellular accumu-lation of Hif1␣ in a dose-dependent manner that correlated with acquisition of spindle-shaped morphology (Fig. 2, A and  B), acquisition of ␣-SMA, and decreased CD31 expression (Fig.  2, C-E). As under hypoxic conditions, loss of CD31 upon CoCl 2 treatment was associated with increased expression of FSP1, ␣-SMA, DDR2, and collagen 1A1 and loss of the other endothelial markers VE-cadherin and VWF (Fig. 2F). Correspondingly, EndMT key regulators SNAIL, SLUG, and TWIST as well as collagen 1A1 upon CoCl 2 treatment mimicked phenotypic changes observed upon cultivation under hypoxic conditions (compare Fig. 2, G--K, with Fig. 1). To further substantiate the link among hypoxia, Hif1␣, and EndMT, we next explored the impact of digoxin, an inhibitor of Hif1␣ protein translation (30) (Fig. 3A) and overexpression of a dominant-negative Hif1␣ mutant on the EndMT response to hypoxia. Both digoxin and dominant-negative Hif1␣ mutant overexpression effectively blunted expression of SNAIL, SLUG, and TWIST (Fig. 3, B-D) under hypoxic conditions. Furthermore, intracellular accumulation of Hif1␣ under normoxic conditions through superphysiologic transgenic overexpression (Fig. 3E) reduced CD31 expression and induced expression of EndMT master regulators (Fig. 3, F-H). Conversely, shRNA knockdown of HIF1␣ (using two different constructs; Fig. 3, I and J) reduced expression of EndMT transcription factors under hypoxic conditions (Fig. 3, K-M). In summary, our data suggested that hypoxia induces phenotypic changes typical of EndMT in HCAECs and that this effect is mediated by Hif1␣.
Because the EndMT master regulator SNAIL has been identified to be a direct transcriptional target of HIF1␣ (31), we next performed luciferase assays utilizing a SNAIL-pGL4.10 firefly reporter containing HIF1␣ target motifs within the SNAIL promoter to further elucidate this causal link with Hif1␣ transcription factor upon hypoxia and EndMT (Fig. 4A). Culture under hypoxic conditions, accumulation of endogenous Hif1␣ through addition of CoCl 2 to culture media, and overexpression of Hif1␣ at superphysiological levels all induced SNAIL promoter activity (Fig. 4, A-D). A ChIP assay further demonstrated that Hif1␣ directly binds to the promoter region of SNAIL containing Hif1␣ binding motifs (Fig. 4, E and F).
We therefore next aimed to test whether hypoxia-induced EndMT is SNAIL-dependent. Under hypoxic condition, shRNA knockdown of SNAIL (Fig. 5, A and B) led to reduced expression of the other two EndMT master genes, SLUG and TWIST, as well as   of collagen 1A1 as compared with scrambled control (Fig. 5, C-E). However, expression levels upon SNAIL knockdown did not reach the same low levels as under normoxic conditions (Fig. 5, C-E), suggesting an additional, Snail-independent pathway.
We next aimed to explore the possible contribution of hypoxia in the context of cardiac fibrosis. For this purpose, we utilized the mouse model of ascending aortic constriction that consistently causes cardiac fibrosis and in which EndMT was previously demonstrated to occur. Fibrotic tissue was detected by Masson's trichrome stain (Fig. 6A) associated with intrinsic hypoxia visualized by carbonic anhydrase IX immunohistochemistry (Fig. 6B). As expected, constriction of the ascending aorta caused both cardiac fibrosis throughout the myocardium and hypoxia especially in interstitial cells (Fig. 6, A and B). Cardiac fibrosis and intrinsic hypoxia were associated with increased expression of SNAIL, SLUG, and TWIST as well as of collagen 1A1 (Fig. 6, C-G), substantial accumulation of Hif1␣ (Fig. 6G), increased ␣-SMA, and decreased expression of CD31 (Fig. 6G), mirroring the link of hypoxia and EndMT that was observed in cell culture experiments. Finally, we utilized a Tie1Cre;R26RstopYFP endothelial lineage tracing mouse model with ascending aortic constriction-induced cardiac fibrosis and detected colocalization of SNAIL, HIF1␣, and YFP (Fig. 6H). Although these data support the occurrence of EndMT in HIF1␣-positive cells of endothelial origin in vivo, additional studies will be needed to prove HIF1␣ dependence in hypoxia-induced EndMT.

Discussion
Here we demonstrate that hypoxia induces phenotypic changes typical of EndMT in cultured human coronary endothelial cells, that this effect is mediated by HIF1␣, and that the FIGURE 4. HIF1␣ induces SNAIL expression through direct binding to its promoter. A, schematic illustration of pGL4.10_SNAIL luciferase reporter construct that was exposed to different conditions (normoxia, hypoxia, CoCl 2 treatment, and HIF1␣ overexpression (OE)). B-D, HCAECs were transfected with a pGL4.10_SNAIL luciferase reporter construct and a control Renilla expression vector. A luciferase assay shows that SNAIL promoter activity significantly increases under hypoxic conditions (B), CoCl 2 treatment (C), and HIF1␣ overexpression (D). The bar graphs display relative SNAIL promoter activities normalized to Renilla luciferase. E, simplified schematic showing the human SNAIL promoter along with exons (yellow boxes), translational start site (black arrow), HIF1␣ binding motifs (black diamonds), locations of SNAIL ChIP primers (amplicon1 and amplicon2), and negative control chip primer (neg) location. F, the binding properties of HIF1␣ to the SNAIL promoter region were analyzed by ChIP assay following qRT-PCR. IgG purified from the same species as HIF1␣ antibody was used as a negative control for ChIP (expression is presented as mean value; error bars represent S.D.; n ϭ 3; **, p Ͻ 0.01; ***, p Ͻ 0.001; n.s., no significance).
EndMT master regulator SNAIL is a direct target of HIF1␣. Because previous studies have reported that TGF␤ is a prototypical inducer of EndMT and that hypoxia is a stimulus for TGF␤ expression, it is important to note that our data demonstrate that EndMT involving HCAECs is also mediated by HIF1␣ independently of TGF␤, suggesting that these are distinct, potentially additive pathways.
Our finding that hypoxia-induced EndMT is mediated by HIF1␣ is in line with existing literature in the cancer epithelialmesenchymal transition field reporting that hypoxia contributes to cancer progression and metastasis through induction of epithelial-mesenchymal transition in an HIF1␣-dependent manner (32). Furthermore, our study is compatible with previous reports demonstrating that hypoxia-induced epithelialmesenchymal transition is mediated through epithelial-mesenchymal transition master transcription factors Snail, Slug, and Twist (33)(34)(35). In this regard, our study demonstrates that Snail is a direct target of HIF1␣ by revealing that HIF1␣ directly binds to its HIF1-binding sites within the SNAIL promoter. Future studies are needed to address which HIF1␣-binding sites are functionally relevant. Furthermore, our data are supported by a previous study demonstrating that cardiac ischemia-reperfusion injury is associated with induction of Snail within endothelial cells (36).
We are aware that hypoxia is a principal inducer of angiogenesis. In this regard, sprouting endothelial cells, so-called tip cells, share similarities with endothelial cells undergoing EndMT: both tip cells and EndMT cells display spindle-shaped morphology associated with increased expression of mesenchymal cytoskeletal constituents such as vimentin (37). A principal difference, however, is the decreased expression of endothelial markers such as CD31 and VE-cadherin that is typical of EndMT but is not observed in tip cells. Another difference is an increased expression of extracellular matrix proteins such as collagen types I and III that is only observed in endothelial cells undergoing EndMT but not in tip cells, raising the question of the molecular mechanisms discriminating between EndMT and sprouting angiogenesis (38).
In this regard, our studies demonstrate that primary HCAECs undergo EndMT in response to hypoxia via a mechanism involving HIF1-mediated induction of SNAIL. In contrast, hypoxia or SNAIL overexpression did not increase collagen expression in human umbilical vein endothelial cells (36). Similarly, targeted overexpression of SLUG (SNAI2) in human umbilical vein endothelial cells induced acquisition of spindle-shaped morphology, whereas VE-cadherin expression was unaltered (39). These differing observations are in line with previous reports that suggested that only distinct subpopulations of endothelial cells have the capacity to undergo EndMT. In the context of the heart, a previous study suggested that specifically endothelial cells that originate from the endocardium (that undergo EndMT during cardiac development to form the mesenchymal cushion) maintain increased susceptibility to undergo EndMT as rem- nants from their embryonic origin. HCAECs used in this study are derived at least in part from endocardial endothelial cells, whereas aortic endothelial or human umbilical vein endothelial cells are of other origin. The molecular mechanisms that determine distinct cellular responses to Snail/Slug/Twist activation in individual endothelial cell populations remain elusive for now and deserve further exploration.