The transcriptional regulator CCCTC-binding factor limits oxidative stress in endothelial cells

Abstract

The CCCTC-binding factor (CTCF) is a versatile transcriptional regulator required for embryogenesis, but its function in vascular development or in diseases with a vascular component is poorly understood. Here, we found that endothelial Ctcf is essential for mouse vascular development and limits accumulation of reactive oxygen species (ROS). Conditional knockout of Ctcf in endothelial progenitors and their descendants affected embryonic growth, and caused lethality at embryonic day 10.5 because of defective yolk sac and placental vascular development. Analysis of global gene expression revealed Frataxin (Fxn), the gene mutated in Friedreich's ataxia (FRDA), as the most strongly down-regulated gene in Ctcf-deficient placental endothelial cells. Moreover, in vitro reporter assays showed that Ctcf activates the Fxn promoter in endothelial cells. ROS are known to accumulate in the endothelium of FRDA patients. Importantly, Ctcf deficiency induced ROS-mediated DNA damage in endothelial cells in vitro, and in placental endothelium in vivo. Taken together, our findings indicate that Ctcf promotes vascular development and limits oxidative stress in endothelial cells. These results reveal a function for Ctcf in vascular development, and suggest a potential mechanism for endothelial dysfunction in FRDA.

Introduction

The CCCTC-binding factor (CTCF)³ is a highly conserved versatile transcriptional regulator that interacts with DNA and multiple protein partners (1–3). CTCF is mainly known for its function as a genomic insulator, and as a mediator of long-range genomic interactions (1, 4–7); however, it can also promote gene expression as a member of transcriptional activation complexes (1, 3). Constitutive Ctcf depletion in mice results in death during early embryogenesis (8). We are only beginning to understand how Ctcf controls specific mammalian development processes (7, 9–14). For instance, Ctcf interacts with myogenic master regulators to control myogenic cell differentiation and muscle development (2, 3). In addition, Ctcf acts as a mediator of long-range genomic interactions to control limb and heart development (6, 9, 14). Overexpression experiments revealed that CTCF limits retinal angiogenesis by preventing enhancer-mediated activation of the gene encoding vascular endothelial growth factor (VEGF) (15). Whether CTCF controls development of the vascular system has not been investigated.

Development of the mouse vasculature begins during gastrulation at embryonic day (E) 6.5 with migration of a subpopulation of mesodermal precursors from the primitive streak toward the embryo proper and extraembryonic tissues, i.e. the yolk sac and placenta (16–19). Development of the vascular network begins with formation of new vessels by vasculogenesis, followed by branching of preexisting vessels by angiogenesis in the yolk sac (16). Blood starts circulating after formation of the primitive vascular plexus in the yolk sac and the embryo proper at E8.25, promoting vascular plexus remodeling into a complex network (16). Vasculogenesis after chorioallantoic fusion at E8.0 initiates placental vascular development, and branching angiogenesis forms a complex placental vascular network known as the labyrinth, which mediates nutrient and gas exchange between the mother and the developing embryo (16, 17). Endothelial transcriptional programs coordinate vascular development (18–21). Transcriptional misregulation in endothelial cells in embryonic and extra embryonic vasculature can cause cardiac and vascular defects leading to disease (22, 23).

Friedreich's ataxia (FRDA) is the most common hereditary neurodegenerative disease (24). Vascular defects and endothelial dysfunction may contribute to FRDA. Impaired vascularization might contribute to muscle fatigability (25). In addition, cardiomyopathy in FRDA is associated with microvascular disease (26). Furthermore, FRDA patients exhibit decreased flow-mediated dilation in the brachial artery, suggesting that endothelial dysfunction may contribute to FRDA (27). FRDA is caused by abnormal expansion of GAA trinucleotide repeats at intron 1 of the Frataxin (FXN) gene that results in decreased protein levels (28), and increased levels of reactive oxygen species (ROS) and oxidative stress (29, 30). FXN is a mitochondrial protein involved in the assembly of iron and sulfur clusters (31) expressed mainly in tissues with high metabolic rates, such as the heart and brown fat (32). GAA trinucleotide repeat expansion triggers silencing of FXN gene expression via epigenetic mechanisms (33). For instance, in FRDA patients' cells and mouse models, histones located near the expanded GAA repeats are occupied with the repressive mark histone H3 lysine 9 trimethylation (H3K9me3), and have reduced levels of acetylated core histones, which mark transcriptionally active genes (34–36). These modifications might interfere with the activity of transcriptional regulators controlling FXN expression. Studies on fibroblasts and cerebellum from FRDA patients showed that CTCF binding is required to maintain transcriptionally active chromatin, as its depletion from the 5′ untranslated region (5′-UTR) of FXN results in heterochromatin formation. Whether CTCF controls FXN gene expression in endothelial cells, and regulates vascular development is unknown.

Results

Ctcf is expressed in developing and adult mouse vascular endothelium

Ctcf is broadly expressed (3). However, its expression in embryonic or adult vascular endothelial cells has not been investigated. To visualize Ctcf protein in developing vascular endothelium, we performed immunofluorescence for Ctcf and platelet endothelial cell adhesion molecule 1 (Pecam-1), an endothelial marker, on sagittal sections of mice at E9.5, E11.5, postnatal day 2, and 6-week-old adults. Ctcf was detected in nuclei ubiquitously, and was present in vascular endothelial cell nuclei in the third branchial arch, outflow tract, aorta, and pulmonary artery (Fig. 1A). Thus, Ctcf is expressed in vascular endothelial cells throughout embryogenesis and in adulthood.

Figure 1.

Ctcf in endothelial cells is essential for embryogenesis. A, immunostaining of Ctcf (red), and Pecam-1 (green) on sagittal sections on the third branchial arch, outflow tract (OFT), aorta, and pulmonary artery from E9.5, E11.5, postnatal day 2 (P2), and 6-week-old adult mice. Arrows point to examples of endothelial cells expressing Ctcf. Lu, blood vessel lumen. Scale bar = 50 μm. B, whole control (Ctcf^fl/fl) and Ctcf mutant (Ctcf^fl/fl;Tie2-cre) embryos at E9.5 and E10.5. C, head of control and Ctcf mutant embryos at E9.5 and E10.5 stained for Pecam-1 by immunohistochemistry. White asterisks indicate major brain vessels. Scale bars = 1 mm. D, number of major brain vessels in control and Ctcf mutant embryos at E9.5 and E10.5. NS, nonsignificant. E, diameter of brain vessels in control and Ctcf mutant embryos at E9.5 and E10.5. Error bars represent the mean ± S.D. *, p < 0.05.

Ctcf in endothelial progenitors and their derivatives is essential for embryogenesis

Ctcf controls important developmental processes (7, 9–13), but its function in vascular development is unknown. To uncover the function of Ctcf in vascular development, we conditionally inactivated Ctcf in mouse endothelial progenitors and their derivatives by cre-mediated homologous recombination of a floxed allele. Exons 3 to 12 of Ctcf are flanked by LoxP sites in the Ctcf floxed allele (37), which was crossed with Tie2-cre transgenics (38). Efficiency of Ctcf depletion in endothelial cells was evaluated by immunofluorescence for Ctcf and Pecam-1. Ctcf mutants had over 90% fewer cell nuclei that were double positive for Pecam-1 and Ctcf compared with controls (Fig. S1, A and B).

Embryos with Ctcf-deficient endothelial progenitors and derivatives died by E11.5 (Fig. S1C). Ctcf mutant embryos at E9.5 had no gross morphological defects. E10.5 embryos were smaller (Fig. 1B), suggesting deficient growth. Mutants developed an overall normal heart, with normal ventricular wall thickness, but appeared to have reduced trabeculae at E10.5 (Fig. S1, D and E). To determine whether deficiency of Ctcf affects embryonic vasculature patterning we stained Pecam-1 in whole control and Ctcf mutant embryos. Ctcf mutants had an overall normal vasculature pattern (Fig. S1F). Accordingly, the number of major branches of the cerebral vasculature was comparable between control and mutant embryos at E9.5 and E10.5 (Fig. 1, C and D). In contrast, quantification of cerebral vessel diameter revealed narrower vessels in Ctcf mutants than controls at E9.5 and E10.5 (Fig. 1E). Thus, endothelial Ctcf is required for embryogenesis and might regulate vascular development.

Ctcf is required for yolk sac vascular remodeling

The developing vascular network extends into the yolk sac and placenta (16). Defects in the yolk sac vasculature can compromise embryonic development (20, 39, 40). Defective yolk sac vasculature might affect embryogenesis in Ctcf mutants. To test this possibility, we analyzed the vascular network in the yolk sac of Ctcf mutant and control embryos at E8.5, E9.5, and E10.5. Ctcf mutant yolk sacs at E8.5 and E9.5 appeared to be properly irrigated, however, E10.5 embryos had a pale yolk sac (Fig. 2A), suggesting vascular defects. Immunofluorescence of Pecam-1 in whole yolk sacs revealed the vascular network. We analyzed the yolk sac vascular network in control and Ctcf mutants using AngioTool (41). Control and Ctcf mutant yolk sacs at E8.5 had comparable vessel area, vessel length and lacunarity, a measure of the average gap between blood vessels and reflective of vessel disorganization (41) (Fig. 2, B and C). In contrast, E9.5 mutant yolk sacs had a decreased vessel area and vessel length, and increased lacunarity. Similarly, E10.5 Ctcf mutant yolk sacs had decreased vessel area and increased lacunarity (Fig. 2, B and C). These changes were not because of deficient endothelial cell proliferation or increased apoptosis. The number of cells double positive for Pecam-1 and phosphorylated histone H3, were comparable between yolk sacs of control and Ctcf mutant embryos at E9.5. Cells double positive for Pecam-1 and activated caspase 3 were absent in control and mutant yolk sacs (Fig. S2).

Figure 2.

Vascular remodeling defects in yolk sac of Ctcf mutants. A, whole mount images of control (Ctcf^fl/fl) and Ctcf mutant (Ctcf^fl/fl;Tie2-cre) yolk sacs attached to placentae and embryos at E8.5, E9.5, and 10.5. Scale bar = 1 mm. B, immunofluorescence for Pecam-1 (green) on yolk sacs. Lower panels are skeletons generated with AngioTool from fluorescent micrographs. Scale bar = 100 μm. C, quantification of vessel area, total vessel length, and lacunarity from skeletons using AngioTool. Error bars represent the mean ± S.D. *, p < 0.05.

Blood circulation causes shear stress and induces pressure on blood vessels, stimulating yolk sac vascular remodeling from E8.5 to E9.5 (16, 42). Lack of circulating blood in cultured embryos blocks remodeling of the yolk sac vasculature and causes a dramatic down-regulation of the mechanosensor, endothelial nitric-oxide synthase (eNOS) (encoded by Nos3) (16, 42). Deficient yolk sac vascular remodeling in Ctcf mutants might be caused by decreased blood flow; therefore, we analyzed the expression of Nos3 by quantitative PCR (qPCR) and eNOS protein abundance by Western blotting. Nos3 mRNA, and eNOS protein levels were comparable between control and Ctcf mutant yolk sacs at E9.5 (Fig. S3), which agrees with a normally irrigated yolk sac at this stage (Fig. 2A). This suggests that vascular remodeling defects in yolk sac up to E9.5 in Ctcf mutants are not secondary to decreased blood flow because of heart defects or reduced circulating blood.

Ctcf is required for placental vascular development

Defective labyrinth development can affect embryonic growth (17, 43). Whole placentae still attached to Ctcf mutant embryos through the umbilical cord appeared normal at E9.5. However, the placentae and umbilical cords in Ctcf mutants appeared improperly irrigated at E10.5 (Fig. 3A), suggesting defects in the labyrinth. Accordingly, E9.5 Ctcf mutant placentae had both embryonic and maternal blood vessels. In contrast, only maternal blood vessels were observed in histological sections of E10.5 Ctcf mutant placentae (Fig. 3B). To visualize the labyrinth, we incorporated the cre-dependent GFP reporter Rosa^mT/mG (44) into Ctcf floxed mice carrying the Tie2-cre transgene (38), resulting in GFP-labeled endothelial cells. Quantification on placenta sections stained for GFP revealed a comparable labyrinth area in Ctcf mutants at E9.5, and a significant decrease at E10.5 (Fig. 3, C and D). Decreased labyrinth expansion was not caused by decreased endothelial cell proliferation or increased cell death, as the numbers of endothelial cells positive for phosphorylated histone H3, or activated caspase 3, were comparable between control and Ctcf mutant placentae at E9.5 and E10.5. Caspase 3–positive cells were absent in E10.5 placentae (Fig. S4). Nos3 mRNA and eNOS protein levels were also comparable between control and Ctcf mutant placentae at E10.5 (Fig. S3), suggesting that defective labyrinth expansion is not secondary to heart defects or decreased blood flow. Thus, endothelial Ctcf is required for extraembryonic vascular development.

Figure 3.

Decreased labyrinth expansion in Ctcf mutants. A, whole control (Ctcf^fl/fl) and Ctcf mutant (Ctcf^fl/fl;Tie2-cre) placentae attached to E9.5 and E10.5 embryos. Scale bar = 100 μm. CP, chorionic plate; UM, umbilical cord; EM, embryo. B, histological sections of labyrinth. EB, embryonic blood; MB, maternal blood. Scale bar = 50 μm. C, GFP (green) immunostaining on sections of placentae at E9.5 and E10.5 in control (Ctcf^fl/+;Tie2-cre;Rosa^mT/mG/+) and mutant (Ctcf^fl/fl;Tie2-cre;Rosa^mT/mG/+) embryos. Scale bar = 200 μm. D, quantification of GFP fluorescence area in E9.5 and E10.5 placentae. Error bars represent the mean ± S.D. NS, nonsignificant; AU, absorbance unit. *, p < 0.05.

Genome-wide expression profile of E9.5 endothelial cells from WT and Ctcf mutant placentae

To uncover genes and pathways regulating vascular development downstream of Ctcf we performed high-throughput RNA-Seq on endothelial cells sorted from placentae of control (Ctcf^fl/+;Tie2-cre;Rosa^mT/mG/+) and Ctcf mutant (Ctcf^fl/fl;Tie2-cre;Rosa^mT/mG/+) embryos at E9.5. Sorted GFP-positive cells expressed significantly higher levels of GFP and the endothelial markers kinase insert domain receptor (Kdr), and TEK receptor tyrosine kinase (Tek, also known as Tie2), than GFP-negative cells (Fig. S5), indicating that the sorted cell population is enriched for endothelial cells. Consistent with loss of Ctcf protein (Fig. S1, A and B), Ctcf mRNA was drastically reduced in endothelial cells sorted from mutant embryos, as shown by qPCR (Fig. S5B). RNA-Seq analysis revealed 232 genes that were up-regulated, and 155 genes that were down-regulated over 1.5-fold in Ctcf mutant endothelial cells (Fig. 4A). Analysis of misregulated genes using DAVID revealed that up-regulated genes are enriched for processes important for vascular development including focal adhesion, extracellular matrix–receptor interaction, and adherens and tight junction. Down-regulated genes are enriched for processes related to cell cycle and genomic stability, and GSH metabolism (Fig. 4B). To validate the RNA-Seq results, we performed qPCR on selected genes that were highly misregulated in Ctcf mutants, including genes with known functions in vascular development. qPCR was performed on endothelial cells sorted from placentae and yolk sacs from control and Ctcf mutant embryos. This analysis confirmed down-regulation of frataxin (Fxn), Gstz1, C3ar1, E2f2, E2f8, and Flt4, and up-regulation of several developmental regulators including Msx1, Tcf7, Celsr1, Ralgds, Pitx1, Tead1, and Notch1 (Fig. 4, C and D and Fig. S6). Changes in gene expression levels in placenta, but not yolk sac, were largely consistent with the RNA-Seq (Fig. 4C), suggesting that placenta and yolk sac endothelial cells have unique gene expression programs, or that Ctcf controls specific transcriptional pathways in endothelial cells in different organs. RNA-Seq revealed that Frataxin (Fxn) was the most down-regulated gene in Ctcf mutant placental endothelial cells. Western blotting revealed a slight but statistically significant decrease in Fxn protein in Ctcf mutant placentae (Fig. 4, E and F). This analysis was carried out on labyrinth tissue; therefore, it likely underestimates protein decrease in endothelial cells. qPCR showed dramatic Fxn down-regulation consistently in both placental and yolk sac endothelial cells (Fig. 4, C and D), suggesting that Fxn might be a Ctcf target important for vascular endothelial cell development.

Figure 4.

Gene expression profile in Ctcf mutant placental endothelial cells. A, MA plot for differential expression in isolated placental labyrinth endothelial cells of control (Ctcf^fl/+;Tie2-cre;Rosa^mT/mG/+) and Ctcf mutant (Ctcf^fl/fl;Tie2-cre;Rosa^mT/mG/+) embryos at E9.5. Black dots represent genes with no significant change in expression. Red dots represent genes that underwent significant (p < 0.05) change in expression. Blue dots indicate some of the genes whose misregulation was confirmed by qPCR. X axis indicates the normalized mean expression level across all samples, and the y axis indicates the log 2-fold change. B, gene ontology categories enriched in genes up-regulated (red) and down-regulated (green) in endothelial cells sorted from control and Ctcf mutant placentae at E9.5. C, heat map of log 2 expression relative to Pgk1, as determined by qPCR on endothelial cells sorted from control and Ctcf mutant placentae and yolk sacs, of genes found by RNA-Seq to be misregulated in Ctcf mutant placental endothelial cells. D, qPCR on endothelial cells sorted from control and Ctcf mutant placentae (Fxn, Gstz, C3ar1) and yolk sacs (Fxn). Expression levels are relative to Pgk1. E, Western blotting of Fxn on labyrinths from control and Ctcf mutant placentae. Beta actin (Actb) was used as loading control. F, quantification of band intensities relative to Actb. Error bars represent the mean ± S.D. of three biological replicates. *, p < 0.05.

CTCF activates the FXN gene promoter

It has been proposed that CTCF activates FXN expression by maintaining an open chromatin configuration (33, 45–47). Others and we have shown that Ctcf regulates developmental processes by activating gene expression as a transcription factor (1, 3). To determine whether CTCF activates FXN expression as a transcription factor we assessed the capacity of CTCF to activate the FXN promoter in an episomal luciferase reporter. Comparison of the 5′ region immediately upstream the mouse Fxn gene against a database of validated CTCF-binding sites (48, 49) identified a motif spanning nucleotides −4 to −23 relative to the transcription start site (46). This binding site is conserved in the human FXN promoter and has high identity with other previously validated CTCF-binding sites (Fig. 5, A and B). We cloned a 386-bp DNA fragment corresponding to the 5′ regulatory region of the human FXN gene that includes the identified CTCF-binding motif. CTCF significantly activated the FXN promoter in transient cotransfections in bovine aortic endothelial cells (BAECs) (Fig. 5C). Thus, Ctcf is a transcriptional activator of FXN in endothelial cells.

Figure 5.

Ctcf activates the Fxn promoter, limits ROS accumulation, and promotes angiogenesis in vitro. A, scheme showing location of a Ctcf-binding site in the mouse (Mm) and human (Hs) FXN promoter relative to the transcription start site (arrow). B, alignment of the Ctcf-binding site in the mouse (Fxn Ctcf site Mm) and human (FXN CTCF site Hs) FXN promoter, with validated Ctcf-binding sites REN_20, MIT_LM2, MIT_LM7, and MIT_LM23. C, Luciferase activity normalized to Renilla activity of a control reporter (pGL3) or a FXN reporter cotransfected into BAECs with an empty (−) vector or a CTCF overexpression vector. Error bars represent the mean ± S.D. of a representative experiment of four performed with three technical replicates each. *, p < 0.05. D, Western blotting for CTCF in lysates of cells transfected with a negative control, two nonoverlapping anti-CTCF or anti-FXN siRNAs. GAPDH was used as loading control. E, densitometric quantification of Western blots from D. Data shown are CTCF normalized to GAPDH ± S.E., of three biological replicates. *, p < 0.05 compared with negative control (crtl). F, expression of FXN mRNA in HUVEC transfected with negative control, anti-CTCF, or anti-FXN siRNAs. Data shown are mean relative expression ± S.D. of three biological replicates. *, p < 0.05. G, immunofluorescence for 8-OHG and F-actin on HUVECs transfected with control (ctrl), anti-CTCF, or FXN siRNAs or treated with 200 μm H₂O₂. Nuclei were counterstained with DAPI. Scale bar = 10 μm. H, quantification of 8-OHG–positive nuclei in HUVECs transfected with control (ctrl), anti-CTCF, or FXN siRNAs or treated with 200 μm H₂O₂. Error bars represent the mean ± S.D. of three biological replicates. *, p < 0.05. I, fluorescent images of Matrigel tube formation assay from HUVEC transfected with negative control, anti-CTCF, or anti-FXN siRNA. J, total tube length calculated from images in I. Data shown are average tube length of three biological replicates ± S.E. *, p < 0.05 compared with negative control.

Ctcf prevents oxidative stress in endothelial cells

Frataxin deficiency in yeast (50) and in cells from patients with FRDA causes increased oxidative stress (29, 30), and enhanced oxidative stress is known to negatively affect vascular development (51). Our RNA-Seq analysis revealed that genes down-regulated in Ctcf mutant endothelial cells participate in GSH metabolism, including Fxn and GSH S-transferase zeta 1 (Gstz1) (Fig. 4, B and D), which modulate ROS generation (52, 53). This suggests a potential function of Ctcf as an oxidative stress regulator. To determine whether Ctcf deficiency causes oxidative stress in endothelium we analyzed ROS-mediated DNA damage in human umbilical vein endothelial cells (HUVECs) with reduced CTCF levels. CTCF was efficiently knocked down in HUVECs using two nonoverlapping siRNAs (Fig. 5, D and E). CTCF depletion led to decreased levels of FXN mRNA (Fig. 5F). CTCF-depleted cells were stained using an antibody against 8-hydroxyguanosine (8-OHG), a modified base that occurs in DNA as a result of oxidative stress (54). As positive controls, HUVECs with decreased levels of FXN or treated with H₂O₂ had increased levels of 8-OHG. More CTCF-depleted HUVECs had nuclei that were positive for 8-OHG compared with cells transfected with a control siRNA (Fig. 5, G and H and Fig. S7). Importantly, enhanced oxidative stress was associated with defects in angiogenesis, as CTCF or FXN knockdown (Fig. 5F) cells had decreased tube length in a Matrigel tube formation assay (Fig. 5, I and J).

We assessed whether enhanced oxidative stress was also present in Ctcf mutant embryos. In sections of placentae at E10.5, significantly more 8-OHG foci were found in endothelial cell nuclei in Ctcf mutant embryos, than in controls (Fig. 6, A and B). Increased ROS promotes lipid peroxidation in a humanized mouse model of FRDA (55). Western blotting revealed increased levels of 4-hydroxynonenal (4HNA), a common byproduct of lipid peroxidation (56), in labyrinth tissue from Ctcf mutants compared with controls (Fig. 6, C and D). Furthermore, endothelial cells in Ctcf mutant placentae had higher levels of 4HNA than controls (Fig. 6, E and F). Thus, Ctcf protects endothelial cells from oxidative stress.

Figure 6.

Ctcf limits ROS in endothelial cells in vivo. A, immunofluorescence for 8-OHG and Pecam-1 on sections of control (Ctcf^fl/+;Tie2-cre;Rosa^mT/mG/+) and Ctcf mutant (Ctcf^fl/fl;Tie2-cre;Rosa^mT/mG/+) placentae at E9.5. Boxes locate close-ups in lower panels. Arrows point to 8-OHG–positive endothelial cell nuclei in Ctcf mutant placentae. Scale bar = 10 μm. B, count of 8-OHG foci per nucleus of endothelial cells in control and Ctcf mutant placentae. Error bars represent the mean ± S.D. of cells quantified in three embryos. *, p < 0.05. C, Western blotting of 4HNA on control and Ctcf mutant labyrinths at E9.5. Gapdh was used a loading control. D, quantification of 4HNA relative to Gapdh from blot in C. Error bars represent the mean ± S.D. of three labyrinths. *, p < 0.05. E, immunofluorescence for 4HNA on E9.5 control and mutant placentae. Endothelial cells are revealed by staining for GFP. Nuclei were counterstained with DAPI. Scale bar = 10 μm. F, fluorescence for 4HNA in control and mutant E9.5 placentae normalized against fluorescence in the cytoplasmic region of Pecam-1 negative cells. G, Western blotting for cytochrome c (Cyt c), on labyrinth tissue from control and Ctcf mutant placentae at E9.5. B-actin was used as loading control. H, protein quantification relative to B-actin from blots in G. Error bars represent the mean ± S.D. of three biological replicates. *, p < 0.05. I, immunofluorescence for IscU1/2 and Pecam-1 on sections of control and mutant placentae at E9.5. Arrows point to nuclei. Boxes are close-ups in lower panels. J, fluorescence of IscU1/2 in control and mutant E9.5 placentae normalized against fluorescence in nuclei in Pecam-1 negative cells. Error bars represent the mean ± S.D. of three and four labyrinths. *, p < 0.05.

Mitochondrial dysfunction leads to ROS accumulation and lipid peroxidation in a humanized mouse model of FRDA (55). Cytochrome c, an essential component of the mitochondrial electron transport chain indispensable for energy production (57), is decreased in FXN knockdown and mutant cells (58, 59). Accordingly, Western blot analysis showed that cytochrome c is decreased in labyrinth tissue from Ctcf mutant placentae (Fig. 6, G and H). The iron-sulfur cluster assembly enzyme (IscU), which regulates mitochondrial iron homeostasis (60), is also decreased in Fxn mutant mouse tissues (61). Immunofluorescence on sections of labyrinth from control embryos revealed cytoplasmic and nuclear staining for IscU1 and 2 (IscU1/2) (Fig. 6I). This is consistent with cytoplasmic and nuclear localization of IscU1 in mammalian cells; however, the function of iron-sulfur cluster assembly in nucleus is not clear (62). Immunofluorescence revealed decreased levels of IscU1/2 in endothelial nuclei in placentae of Ctcf mutant embryos (Fig. 6, I and J). Thus, similar to FXN-deficient cells, Ctcf deficiency leads to a decrease in cytochrome c and IscU proteins.

Discussion

The contribution of Ctcf to vascular growth during development and postnatally remains poorly understood. Previously, it was shown that Ctcf can bind the promoter of Vegf to prevent surrounding enhancers from activating its expression (15, 63). Accordingly, depletion of Ctcf by shRNA injection in the subretinal space causes excess intraretinal vascularization (15). In contrast, we found that Ctcf inactivation in endothelial cells negatively affects embryonic vascular development, and that Vegf expression was not altered in Ctcf mutant endothelial cells in our RNA-Seq and qPCR analysis (Fig. S6). This suggests alternative or context-specific functions for Ctcf in developing vascular endothelium. We found that Ctcf limits oxidative stress in endothelial cells. ROS modulate key signaling pathways controlling vascular development during embryogenesis and regenerative processes (64). Moderate oxidative stress and ROS levels can favor, whereas excessive oxidative stress can be detrimental to, vascular development (51, 65). Our results suggest that Ctcf is an important modulator of oxidative stress that prevents excessive ROS accumulation in endothelial cells to promote vascular development.

Ctcf is known to affect the proliferation and survival of particular cell types. For example, Ctcf promotes T cell proliferation in the thymus (37). In contrast, Ctcf deficiency in the developing limb or heart does not affect mesenchyme cell (9) or cardiomyocyte (14) proliferation, however, it induces mesenchyme cell apoptosis (9). We found that Ctcf deficiency does not affect proliferation nor does it induce apoptosis in embryonic endothelial cells, suggesting that Ctcf regulates cell growth and maintenance cell specifically. Proliferating endothelial cells produce higher ROS levels than quiescent cells (66). ROS induces activation of signaling pathways that promote endothelial cell proliferation and survival (67). It is possible that increased ROS production might have prevented imbalanced proliferation and apoptosis in Ctcf-depleted endothelial cells at least before E10.5. Alternatively, other mechanisms of cell death, including ferroptosis, might have been affected in Ctcf mutants. Ferroptosis is an iron-regulated route of cell death (68, 69) dependent on fatty acid synthesis and cysteine transport (70). These processes are linked to 4HNA (70), which was increased in Ctcf mutant placentae (Fig. 6, C–F). To the best of our knowledge, our work demonstrates for the first time that Ctcf regulates ROS accumulation. Future experiments will be required to directly test the extent to which Frataxin is responsible for the oxidative phenotype in Ctcf knockdown endothelial cells. If Ctcf regulates ROS levels in other cell types, it will be of interest to determine whether sensitivity to different ROS levels underlies cell-specific functions of Ctcf.

Decreased ability of cells to relieve oxidative stress has been implicated in cancer, diabetes, and aging and in neurodegenerative (71) and cardiovascular pathogenesis (51, 72). Neurons and cerebellar granule cells from a mouse model of FRDA generate ROS, resulting in decreased GSH (55). In addition, models of frataxin deficiency in yeast, fly, mouse, and cells in culture (73) support a function for frataxin in preventing ROS-induced toxicity in FRDA pathology (74, 75). Accordingly, reducing ROS prevents early mortality in frataxin-deficient Drosophila (76), and improves electrical contraction, coupling, and decay velocity of calcium kinetics in cardiomyocytes derived from stem cells from FRDA patients (77). Endothelial dysfunction has been associated with FRDA (27). However, how oxidative stress in endothelial cells contributes to FRDA has not been investigated. Our finding of decreased frataxin and increased ROS in Ctcf-deficient endothelial cells opens the possibility to investigate the endothelial component of FRDA.

Oxidative stress can potentially alter global gene expression patterns (78). We found that Ctcf-depleted developing endothelial cells are under oxidative stress and misregulate hundreds of genes. It will be of interest to determine whether Ctcf depletion alters interaction of distal regulatory elements resulting in dysregulated oxidative stress responses in endothelial cells during development. Our results indicate that Ctcf is an important regulator of oxidative stress in developing endothelial cells and open the possibility to investigate the endothelial component of diseases associated with oxidative stress.

Experimental procedures

Mice

All animal procedures were approved by the Animal Care Committee at the Hospital for Sick Children and followed the guidelines of The Centre for Phenogenomics. The following strains were used: Ctcf^fl/fl (37), Tie2-cre (38), and ROSA26^mT/mG (44). Presence of vaginal plugs indicated E0.5. Embryos not carrying the ROSA26^mT/mG transgene were obtained by crossing Ctcf^fl/+; Tie2-cre males with Ctcf^fl/fl females. Embryos carrying the ROSA26^mT/mG transgene were obtained by crossing Ctcf^fl/+;Tie2-cre males with Ctcf^fl/fl;ROSA26^mT/mG/mT/mG females.

Genotyping

Tail clips, ear notches, and yolk sacs were digested in 300 μl of 50 mm NaOH at 95 °C for 10–30 min, and 100 μl of 0.5 μm Tris-HCl were added to neutralize the reaction (79). 1 μl of the digestion was used for PCR. Amplification conditions used to identify floxed Ctcf alleles were the following: 94 °C 3 min, 94 °C 30 s, 63 °C 30 s, 72 °C for 1 min, steps 2–4 for 35 cycles, and 72 °C for 5 min. Amplification conditions used for Cre were the following: 95 °C 2 min, 95 °C 40 s, 55 °C 50 s, 72 °C 1 min, steps 2–4 for 35 cycles, and 72 °C for 10 min. Primers are in Table S1.

Fixation and histology

Freshly dissected embryos in cold PBS were fixed in 4% paraformaldehyde (PFA) overnight at 4 °C. Tissues were washed twice in PBS for 30 min at 4 °C and stored in 70% ethanol overnight at 4 °C. Tissues were washed in the following ethanol and xylene series at room temperature: 85% EtOH for 30 min twice, 95% EtOH for 30 min twice, 100% EtOH for 45 min four times, 100% xylene for 5 min, and 100% xylene for 10 min three times. Tissues were incubated in 50% xylene:wax at 60 °C for 30 min and stored at room temperature overnight. The next day, tissues were incubated at 60 °C for 30 min, washed with 100% wax for 1 h at 60 °C twice, and after refreshing the wax, incubated for 2 h at 60 °C. Tissues were then embedded in wax blocks that were mounted on histology cassettes. 4- to 8-μm-thick sections were generated, mounted on glass slides, and stained with hematoxylin and eosin as follows. Slides were washed in xylene for 10 min twice, 100% EtOH for 2 min twice, 90% EtOH for 2 min, 70% EtOH for 2 min, 50% for 2 min, 30% for 2 min, and quickly washed with tap water three times. Slides were stained with 100% hematoxylin for 10 min, quickly washed three times with tap water, placed in 0.5% acid alcohol for 5 s, quickly washed three times with tap water, placed in 1% lithium carbonate for 5 s, washed with tap water three times, 30% EtOH, 50% EtOH, and 70% EtOH for 1min. Slides were then stained in 3% eosin for 10 min, washed in 90% EtOH for 1 min, 100% EtOH twice for 1 min, and placed in xylene twice for 5min. Slides were mounted with Permount (Fisher).

Immunofluorescence and whole mount immunostaining

Dissected tissues were fixed in 4% PFA overnight at 4 °C, washed in PBS three times for 10 min at room temperature, and kept in 30% sucrose/PBS at 4 °C overnight or until tissues sank down. Tissues were embedded in OCT compound and sectioned. 4-μm frozen sections were mounted on glass slides, fixed in 4% PFA for 5 min, and washed in PBS three times, 5 min each. Slides were blocked with 5% goat serum, 0.1% Triton X-100 in PBS for 15 min and incubated with primary antibodies overnight at 4 °C in a humidified chamber. Slides were washed in PBS three times for 10 min each, incubated with secondary antibodies diluted in blocking buffer for 1 h at room temperature. Slides were washed in PBS three times 5 min each and PBS with 0.05% Tween 20 for 5 min. Slides were mounted in Vectashield Mounting Medium with DAPI (Vector Laboratories). Antibodies and dilutions were as follows: Ctcf (1/300) (80) (Santa Cruz Biotechnology, G-8), phosphorylated histone H3 (1:100) (Santa Cruz, SC-8656-R), cleaved caspase 3 (1:100) (Cell Signaling Technologies, 9661), CT3 (1:200) (Development Studies Hybridoma Bank), CD31/Pecam-1 (1:100) (BD Pharmingen, 553370), IscU1/2 (1:1000) (Santa Cruz Biotechnology, sc-373694), and GFP (1:1000) (GeneScript, A01694).

Embryos for whole mount immunostaining were fixed in 4% PFA overnight and washed three times in PBS for 5 min each. Embryos were permeabilized and blocked for 1 h at room temperature in PBT (1× PBS with 0.2% Triton X-100) with 0.1% BSA and 2% goat serum. Embryos were incubated in anti–Pecam-1 antibody diluted in blocking buffer (0.1% BSA and 2% normal goat serum) overnight at 4 °C. Samples were washed five times for 5 min with PBS, blocked in PBT with 2% normal goat serum for 1 h at room temperature, and incubated with secondary antibodies, diluted in 0.15 BSA and 2% goat serum, for 1 h in the dark at room temperature. Samples were washed five times for 5 min each with PBS and cleared in 1:1 glycerol:PBS for 3 h at 4 °C, and then in 80% glycerol:PBS for 1 h at 4 °C before imaging (81).

Cells were transfected at 50% confluence on 10 μg/ml fibronectin-coated Permanox 8-well chamber slides (Thermo Fisher) and allowed to grow to confluence. Cells were then fixed with 4% paraformaldehyde for 20 min, followed by permeabilization for 5 min with 0.2% Triton X-100, and blocking with 5% BSA/PBST for 1 h. Cells were then incubated with mouse anti–8-hydroxyguanosine (1:100) (Santa Cruz Biotechnology), followed by incubation with rabbit anti-mouse IgG 488 (1:200) (Invitrogen), and counterstained and mounted in Vectashield Mounting Medium with DAPI (Vector Laboratories). For actin staining, after incubation with secondary antibody, cells were washed and incubated with rhodamine-phalloidin (1:40) (Invitrogen) for 30 min at room temperature. Images were taken on an Olympus FV1000 Confocal microscope using a LumPlanFI40X/0.8NA objective. The total number of nuclei was divided by the number of nuclei that had an 8-OHG signal.

Gene expression analysis

GFP-positive cells were sorted from control Ctcf^fl/+;Tie2-cre;Rosa^mT/mG and mutant Ctcf^fl/fl;Tie2-cre;Rosa^mT/mG yolk sacs and placentae, and RNA was isolated using the Direct-zol^TM RNA Miniprep kit (Zymo Research). cDNA was synthesized using the SuperScript^® VILO cDNA synthesis kit (Thermo Fisher Scientific). cDNA was used in qPCR done using SsoAdvanced^TM Universal SYBR^® Green Supermix (Bio-Rad) on a CFX384 Touch^TM Real-Time PCR Detection System (Bio-Rad). Data were analyzed using CFX Manager Software (Bio-Rad) and normalized to Pgk1 expression levels. qPCR primer sequences are in Table S1.

RNA-seq

Endothelial cell RNA was isolated from GFP-positive cells (82, 83) sorted from individual dissected placental labyrinth (84) of control Ctcf^fl/+;Tie2-cre;Rosa^mT/mG and mutant Ctcf^fl/fl;Tie2-cre;Rosa^mT/mG embryos using Direct-zol^TM RNA Miniprep Plus Kit (Zymo Research) and treated with DNase according to the manufacturer's protocol. RNA integrity was verified using the Agilent Bioanalyzer (Agilent Technologies). Three biological replicates were used for each group. RNA-Seq libraries were prepared using the Ovation^® Single Cell RNA-Seq System (NuGEN Technologies) and sequenced in single-end sequence reads (50 bp in length) on the Illumina HiSeq 2500 platform. The first 8-bp sequences of the 5′ end of the sequencing reads were trimmed using Trimmomatic (85), as recommended by the Ovation^® Single Cell RNA-Seq System (NuGEN) protocol. Trimmed high-quality reads were then mapped to the mouse genome (mm10) using STAR v2.4.2a (86). Mapped read counts were obtained using HTseq (87). Differential expression analysis and MA plots was performed using DESeq2 (88). Enrichment of gene ontology categories in differentially expressed genes was determined using DAVID (89).

Cell culture and transfection

HUVEC and BAEC (ScienCell Research Laboratories) were cultured in Endothelial Cell Medium (ScienCell), 5% fetal bovine serum (FBS), 1% endothelial cell growth supplement (ScienCell), and 1% penicillin/streptomycin. Cells were grown on attachment factor (Gibco)–coated culture dishes and used from passage 3 to passage 6 in experiments. BAECs (Lonza) were cultured in DMEM high glucose (Gibco) with 10% FBS, and 1% penicillin/streptomycin. HUVEC were transfected with 40 nm siRNA using RNAi Max (Invitrogen) per manufacturer's instructions. After 48 h, knockdown was confirmed via qRT-PCR and/or Western blotting.

Tube formation assay

Matrigel (Corning) was polymerized in μ-Slide Angiogenesis chambers (ibidi) for 1 h at 37 °C. siRNA transfected HUVECs were incubated with 5 μm Cell Tracker Green (Thermo Fisher) for 30 min at 37 °C and seeded onto Matrigel for 8 h. Images were taken on a stereo microscope (Leica M165FC), and total tube length was calculated using angiogenesis analyzer (ImageJ). Two fields of view were analyzed per condition and averaged each experiment.

siRNA and plasmids

Nontargeting Silencer Select Negative Control No. 1 siRNA, CTCF siRNA No. 1 and siRNA No. 2 (assay IDs: s20967 and s3855), and Frataxin siRNA (assay ID: s5360) were from Ambion/Invitrogen. The luciferase reporter was constructed by cloning a 386-bp PCR product corresponding to the Frataxin promoter into pGL3 Basic. Primers are in Table S1. The CTCF overexpression plasmid PCI-7.1 was described previously.

Luciferase assays

Confluent BAECs were transfected with 0.5 μg of FXN luciferase construct, 0.5 μg of human CTCF overexpression construct, and 0.1 μg of pRenilla construct using Lipofectamine 2000 (2 μl) in Opti-MEM (Invitrogen). Transfection was performed in 12-well dishes and media were changed back to Endothelial Cell Medium after 5 h. After 24 h, dual luciferase (Renilla and Firefly) was measured using a GloMax 20/20 Luminometer (Promega) using the Dual-Luciferase Reporter Assay System (Promega).

Western blotting

siRNA transfected cells were lysed in 2× Laemelli buffer and boiled at 95 °C for 10 min and centrifuged. Samples were then loaded on precast SDS-PAGE gels (Bio-Rad) for Western blot analysis. Antibodies used were CTCF (1:500) (Santa Cruz Biotechnology, sc-271474), GAPDH (1:5000) (Santa Cruz Biotechnology, sc-47724), 4 hydroxynonenal (1:200) (Abcam, ab46545), Frataxin (1:100) (Abcam, 175402), cytochrome c (1:1000) (Santa Cruz Biotechnology, sc-13156), IscU1/2 (1:1000) (Santa Cruz Biotechnology, sc-373694), and HRP-conjugated goat anti-mouse IgG (1:3000) (Cell Signaling Technology, 7076). Blots were processed using MicroChemi 4.2 (DNR Bio-Imaging Systems).

Microscopy and imaging

Nikon SMZ1500 and Nikon Eclipse Ni microscopes were used. Images were analyzed and quantified using ImageJ Cell Counter and Angiogenic Analyzer tools.

Statistical analysis

Data are presented as the mean ± S.D. or S.E., as indicated. Data were compared by Student's t test. p <0.5 was considered significant. At least three biological replicates were compared in all analyses.

Author contributions

A. R. R. designed and performed experiments, analyzed data, and wrote the manuscript with P. D. O., A. A., L. C., P. V. D., and N. K. designed and performed experiments and analyzed data. J. E. F. designed experiments, analyzed data, and edited the manuscript. M. D. W. analyzed data. N. G. provided the Ctcf^fl/fl mouse line and P. D. O. conceived the study, designed and performed experiments, analyzed data, and wrote the manuscript with A. R. R with input from all authors.