The NuRD Chromatin-Remodeling Enzyme CHD4 Promotes Embryonic Vascular Integrity by Transcriptionally Regulating Extracellular Matrix Proteolysis

The extracellular matrix (ECM) supports vascular integrity during embryonic development. Proteolytic degradation of ECM components is required for angiogenesis, but excessive ECM proteolysis causes blood vessel fragility and hemorrhage. Little is understood about how ECM proteolysis is transcriptionally regulated during embryonic vascular development. We now show that the NuRD ATP-dependent chromatin-remodeling complex promotes vascular integrity by preventing excessive ECM proteolysis in vivo. Mice lacking endothelial CHD4—a catalytic subunit of NuRD complexes—died at midgestation from vascular rupture. ECM components surrounding rupture-prone vessels in Chd4 mutants were significantly downregulated prior to embryonic lethality. Using qPCR arrays, we found two critical mediators of ECM stability misregulated in mutant endothelial cells: the urokinase-type plasminogen activator receptor (uPAR or Plaur) was upregulated, and thrombospondin-1 (Thbs1) was downregulated. Chromatin immunoprecipitation assays showed that CHD4-containing NuRD complexes directly bound the promoters of these genes in endothelial cells. uPAR and THBS1 respectively promote and inhibit activation of the potent ECM protease plasmin, and we detected increased plasmin activity around rupture-prone vessels in Chd4 mutants. We rescued ECM components and vascular rupture in Chd4 mutants by genetically reducing urokinase (uPA or Plau), which cooperates with uPAR to activate plasmin. Our findings provide a novel mechanism by which a chromatin-remodeling enzyme regulates ECM stability to maintain vascular integrity during embryonic development.


Introduction
The extracellular matrix (ECM) plays a critical role in maintaining blood vessel integrity by serving as a scaffold to which endothelial cells and vascular support cells can adhere [1,2]. Signals between ECM components and endothelial cells likewise promote vessel stabilization by influencing endothelial cell proliferation, morphology, and survival. Aberrant degradation of ECM components is associated with life-threatening aortic and cerebral aneurysms, which are characterized by vascular fragility or rupture [3,4].
ECM composition and stability also impact vascular integrity during embryonic development. Mice with mutations in ECM components or in regulators of ECM deposition die during midgestation with vascular rupture and hemorrhage [5,6]. The timing of these lethal phenotypes is likely influenced by the combination of weakened vascular ECM and increasing intravascular pressure that results from embryonic heart rate acceleration during midgestation [7]. Mice with mutations that cause excessive vascular ECM degradation also succumb to vascular rupture at midgestation [8,9]. ECM proteases are produced by activated endothelial cells during vascular development and are necessary for sprouting angiogenesis [10]. However, ECM proteases must be tightly regulated to prevent excessive proteolysis of the vascular matrix and blood vessel fragility. Currently, the transcriptional regulation and coordination of ECM proteolysis during vascular development is poorly understood.
Epigenetic factors are important transcriptional regulators of developmental processes, and increasing evidence supports their roles in cardiovascular development. Enzymes that add or subtract covalent modifications from histone tails affect transcription of target genes that impact heart and blood vessel development [11,12]. Likewise, ATP-dependent chromatin-remodeling complexes, which transiently displace nucleosomes at gene regulatory regions, mediate transcriptional regulation of genes involved in cardiovascular development [13,14]. Mammalian Nucleosome-Remodeling and Histone Deacetylase (NuRD) chromatin-remodeling complexes contain both histone-modifying and chromatinremodeling enzymes [15][16][17]. These multi-subunit complexes incorporate histone deacetylases 1 and 2 (HDAC1 and HDAC2), which remove acetyl groups from histone tails. NuRD complexes also contain ATPase catalytic subunits belonging to the SNF2 superfamily: the Chromodomain-Helicase-DNA-binding proteins CHD3 and CHD4 (also called Mi-2a and Mi-2b, respectively). CHD3 and CHD4 contain protein-and DNA-binding domains in addition to an ATPase domain that provides energy for nucleosome remodeling [18,19]. NuRD complexes were traditionally thought to act in a transcriptionally repressive manner, since histone deacetylation is associated with gene silencing. However, evidence from developing blood and nerve cell lineages indicates that NuRD complexes can facilitate target gene activation in addition to silencing [20][21][22]. Although our lab previously showed NuRD negatively regulates vascular Wnt signaling in the developing extraembryonic yolk sac [23], a role for NuRD in the embryonic vasculature has not yet been described.
We now report that embryos depleted of the NuRD ATPase CHD4 in vascular endothelial cells died by embryonic day 11.5 (E11.5) with blood vessel rupture and massive hemorrhage. Our studies reveal coordinated chromatin-based mechanisms by which CHD4-containing NuRD complexes transcriptionally regulate ECM proteolysis to maintain vascular integrity during embryonic development.

Chd4 fl/fl ;Tie2-Cre + Embryos Die at Midgestation
We previously deleted Chd4 from embryonic endothelial cells with a Tie2-Cre transgenic mouse line in order to assess the role of CHD4-containing NuRD complexes in vascular development. We found Wnt signaling was upregulated in Chd4 fl/fl ;Tie2-Cre + yolk sac vessels at E10.5, but embryonic vascular phenotypes were not apparent at this developmental time point [23]. In order to determine whether Chd4 fl/fl ;Tie2-Cre + embryos could survive development, we mated Chd4 fl/+ and Chd4 fl/+ ;Tie2-Cre + mice together and expected to obtain 12.5% Chd4 fl/fl ;Tie2-Cre + offspring. However, no Chd4 fl/fl ;Tie2-Cre + animals were detected at weaning ( Figure 1A). These results indicated that expression of Chd4 on developing endothelial cells was important for embryonic survival.
Since many mutants with vascular defects die during midgestation [24], we focused on this time period for preliminary analysis of our Chd4 fl/fl ;Tie2-Cre + embryos. CHD4 is broadly expressed at E10.5, and we detected it by immunostaining in endothelial cells of both large and small vessels ( Figure S1). We performed dissections on E10.5-12.5 embryos and consistently found Chd4 fl/fl ; Tie2-Cre + embryos with massive hemorrhage in the trunk region at E11.5 ( Figure 1E). Chd4 fl/fl ;Tie2-Cre + embryos and yolk sacs appeared pale in comparison to littermate controls due to pooling of embryonic blood ( Figure 1B-1E). Prior to hemorrhage, mutant embryos were visibly normal and displayed minimal developmental delay compared to E10.5 littermate controls. Thus, we determined that Chd4 fl/fl ;Tie2-Cre + embryos died at E11.0-11.75 from sudden and massive hemorrhage, since embryos appeared normal at E10.5 and were found dead by E12.5 ( Figure 1F). Chd4 fl/fl ;Tie2-Cre + Dorsal Aortae and Cardinal Veins Are Prone to Rupture The localized hemorrhage observed at E11.5 suggested vascular rupture in the trunk region of Chd4 fl/fl ;Tie2-Cre + embryos. Histological analysis revealed rupture of the dorsal aortae and cardinal veins of Chd4 fl/fl ;Tie2-Cre + embryos at E11.25 ( Figure 1J and 1L), corresponding closely with the time of death. However, there was no indication of impending vascular rupture or evidence of blood leakage from Chd4 fl/fl ;Tie2-Cre + dorsal aortae or cardinal veins at E10.5 ( Figure 1H). Vascular patterning within Chd4 fl/fl ; Tie2-Cre + embryos was largely normal at E10.5 ( Figure S2A-S2F). Likewise, vascular patterning was comparable in control and Chd4 fl/fl ;Tie2-Cre + yolk sacs at E10.5 ( Figure S2G-S2K), as we previously reported [23]. Chd4 fl/fl ;Tie2-Cre + hearts showed slight evidence of hypotrabeculation compared to littermate controls at E10.5 ( Figure S3D). This hypotrabeculation was accompanied by a subtle decrease in cardiac ECM, as assessed by Alcian blue staining ( Figure S3B). Since cardiac ECM is critical for supporting trabeculation [25], we suspect that Chd4 fl/fl ;Tie2-Cre + hypotrabeculation is secondary to reduced cardiac ECM. Nevertheless, hypotrabeculation and diminished cardiac ECM are not associated with vascular rupture in other mutants [25][26][27][28][29][30]. Therefore the sudden vascular rupture and lethal hemorrhage in Chd4 fl/fl ; Tie2-Cre + embryos did not likely result from grossly observable cardiovascular anomalies. Chd4 fl/fl ;Tie2-Cre + Endothelial Cells Have Long Cytoplasmic Extensions but Intact Intercellular Junctions Prior to Vascular Rupture In order to evaluate the morphology of Chd4 fl/fl ;Tie2-Cre + endothelial cells prior to vascular rupture, we processed E10.5 control and mutant littermate embryos for transmission electron microscopy (TEM). We found that endothelial cells lining Chd4 fl/fl ; Tie2-Cre + vessels had long, thin extensions between cells that were not seen in control endothelial cells ( Figure 1M and 1N). In addition, we observed smooth muscle cells attached tightly to endothelial cells outside control dorsal aortae ( Figure 1M, arrowheads), but these attachments were disrupted outside Chd4 fl/fl ;Tie2-Cre + dorsal aortae ( Figure 1N). These phenotypes indicated that Chd4 fl/fl ;Tie2-Cre + vessels were fragile and lacked closely apposed supporting smooth muscle cells prior to rupture.
We also evaluated Chd4 fl/fl ;Tie2-Cre + endothelial cell proliferation and viability prior to vascular rupture. At E10.5, no differences were seen in numbers of proliferating endothelial cells within dorsal aortae of control and Chd4 fl/fl ;Tie2-Cre + embryos, as assessed by immunostaining with antibodies against the proliferation markers phosphorylated histone H3 (PPH3) or Ki67 (Figure

Author Summary
Blood vessels are surrounded by an extracellular matrix (ECM), which provides structural support and helps vessels withstand the biomechanical pressure of flowing blood. As blood vessels grow during embryonic development, the ECM is locally degraded to allow new vessels to sprout from existing ones. This localized ECM degradation must be tightly regulated, however, because excessive matrix breakdown leaves vessels susceptible to rupture and lethal hemorrhage. We now present a novel mechanism by which production of proteins that promote or repress ECM degradation is coordinated. We show that a chromatinremodeling enzyme called CHD4 is critical for maintaining blood vessel integrity during embryonic development. CHD4 achieves this task by limiting production of proteins that degrade ECM and promoting production of proteins that protect ECM around developing blood vessels. A better understanding of how ECM degradation is regulated during development could have an impact on our ability to combat adult pathologies, such as aneurysms, which are associated with excessive ECM degradation and vascular fragility.
S4A-S4D). Likewise no differences were seen in numbers of apoptotic endothelial cells in control and Chd4 fl/fl ;Tie2-Cre + embryos following TUNEL staining or immunostaining with an antibody against active Caspase 3 ( Figure S4E-S4H). We concluded that the long endothelial cell extensions and loosely connected smooth muscle cells lining Chd4 fl/fl ;Tie2-Cre + vessels at E10.5 were not influenced by aberrant endothelial cell apoptosis or deficient proliferation.
Since mutant embryos with defective endothelial cell junctions can display thin, extended endothelial cells prior to hemorrhagic events [31], we analyzed endothelial junctional components and morphology in control and Chd4 fl/fl ;Tie2-Cre + embryos prior to vascular rupture. We confirmed that intercellular junctions appeared morphologically similar by TEM in control and Chd4 fl/fl ;Tie2-Cre + endothelial cells at E10.5 ( Figure 1O and 1P). Likewise, E10.5 control and Chd4 fl/fl ;Tie2-Cre + dorsal aortae immunostained with antibodies against the adherens junction protein vascular endothelial cadherin (VE-cadherin; Figure S5A and S5B) or the tight junction protein zonula occludens-1 (ZO-1; Figure S5C and S5D) revealed comparable levels and patterns of expression. Therefore, our data suggested that defective cell junctions did not contribute to the thin endothelial cell extensions we saw at E10.5 or to vascular rupture at E11.5.
Extracellular Matrix Components Are Diminished around Chd4 fl/fl ;Tie2-Cre + Rupture-Prone Vessels In addition to endothelial cell junctions, the extracellular matrix plays an important role in the maintenance of vascular integrity [2]. One element of the ECM, the basement membrane, is a sheetlike structure that lines and supports endothelial cells. We immunostained sections of control and Chd4 fl/fl ;Tie2-Cre + embryos with an antibody against type IV collagen-a major component of the basement membrane-to assess its expression around ruptureprone dorsal aortae. Type IV collagen was diminished around Chd4 fl/fl ;Tie2-Cre + versus control dorsal aortae at E10.5 ( Figure 2A and 2B). Likewise, fibronectin, another ECM component that is abundantly expressed around dorsal aortae at midgestation [32], was downregulated around Chd4 fl/fl ;Tie2-Cre + versus control dorsal aortae at E10.5 ( Figure 2C and 2D). Therefore ECM components were diminished around Chd4 fl/fl ;Tie2-Cre + dorsal aortae prior to the vascular rupture we consistently saw in mutant embryos at E11.5. Interestingly, type IV collagen and fibronectin were not significantly diminished around small vessels we examined in mutant limb buds (data not shown), and these microvessels were not prone to rupture in Chd4 fl/fl ;Tie2-Cre + mutants.
Because smooth muscle cells (SMCs) provide mechanical support for developing arteries, we also assessed E10.5 Chd4 fl/fl ; Tie2-Cre + dorsal aortae for expression of the SMC differentiation marker alpha-smooth muscle actin (a-SMA). Cells surrounding Chd4 fl/fl ;Tie2-Cre + dorsal aortae expressed less a-SMA than those surrounding control dorsal aortae ( Figure 2E and 2F). Since a-SMA expression is influenced by ECM components such as type IV collagen [33], we suspected the reduced type IV collagen around rupture-prone Chd4 fl/fl ;Tie2-Cre + dorsal aortae contributed to diminished a-SMA levels on SMCs around those vessels. Alternatively, the reduced a-SMA expression and loose attachments between endothelial cells and SMCs observed by TEM ( Figure 1N) could indicate that mutant endothelial cells failed to recruit SMCs effectively. To address this possibility, we examined expression of several genes implicated in SMC recruitment toward endothelium in E10.5 control versus Chd4 fl/fl ;Tie2-Cre + endothelial cells (Tgfb1, Tgfbr1, Eng, Angpt1, Angpt2, and S1pr1; Table S1) or in an endothelial cell line depleted of CHD4 (Hbegf and Tgfbr2; Figure S6). We did not see significant expression changes in these genes, although we did see downregulation of Tie2 (Tek; Table S2), which plays important roles in communication between endothelial cells and SMCs [34]. Additionally, we saw significant upregulation of Pdgfb in CHD4-depleted endothelial cells ( Figure  S6), which reduces a-SMA expression on vascular SMCs [35,36]. Therefore, a-SMA expression on cells surrounding rupture-prone Chd4 fl/fl ;Tie2-Cre + vessels could be diminished due to multiple causes. Overall, we hypothesized that Chd4 fl/fl ;Tie2-Cre + vessels were susceptible to rupture because their walls were weakened from a paucity of ECM components and differentiated SMCs.

CHD4 Directly Inhibits Expression of uPAR
Since CHD4 typically exerts its cellular actions through transcriptional regulation of target genes, we sought to determine which endothelial CHD4 target genes could impact ECM composition. We isolated endothelial cells from E10.5 littermate control and Chd4 fl/fl ;Tie2-Cre + and embryos and identified significant gene expression changes using two commercially available quantitative real-time PCR (qPCR) arrays containing a total of 157 genes. We chose these arrays because they contained many genes with important roles in ECM composition or proteolysis. Notably, we did not see significant downregulation of genes encoding midgestational ECM components, such as collagens, laminins, and fibronectin (Table S1). Nor did we see upregulation of matrix metalloproteinases that could contribute to ECM proteolysis. However, the arrays revealed the urokinase-type plasminogen activator (uPA, encoded by the gene Plau) was significantly upregulated in Chd4 fl/fl ;Tie2-Cre + endothelial cells (Table S2). This result was notable since we had previously identified the uPA receptor (uPAR, encoded by the gene Plaur) as a CHD4 target gene in yolk sac vessels but had not determined the functional relevance of its upregulation in Chd4 fl/fl ;Tie2-Cre + endothelial cells [23]. uPAR regulates ECM proteolysis by binding and localizing uPA to cell surfaces where it activates the serine protease plasmin [37]. Therefore, we hypothesized that upregulation of uPA/uPAR and subsequent plasmin activity might contribute to ECM fragility, vascular rupture, and lethality in Chd4 fl/fl ;Tie2-Cre + embryos.
In order to validate the qPCR array results and determine whether Plau expression was altered in our mutant embryos prior to vascular rupture, we isolated endothelial cells from E10.5 control and Chd4 fl/fl ;Tie2-Cre + embryos and performed qPCR using primers specific for Plau and Plaur. Indeed, we found both Plau and Plaur were significantly upregulated in Chd4 fl/fl ;Tie2-Cre + endothelial cells ( Figure 3A). We next performed chromatin immunoprecipitation (ChIP) assays to determine whether Plaulike Plaur-is a direct CHD4 target gene. In order to obtain the number of cells required for ChIP, we used the C166 yolk sacderived murine endothelial cell line [38]. We chose this cell line for ChIP because knockdown of CHD4 in C166 cells resulted in upregulation of Plaur mRNA ( Figure 3C), as we saw in primary Chd4 fl/fl ;Tie2-Cre + endothelial cells. Plau expression trended upward as well in CHD4 knockdown C166 cells, but it was not significantly upregulated like Plaur ( Figure 3C). We found no evidence that CHD4 bound the Plau promoter in wildtype C166 cells by ChIP but confirmed that CHD4 bound the Plaur promoter in a region approximately 900 bp upstream of the transcription start site (TSS) ( Figure 3D). Therefore, our data indicated that CHD4 directly inhibited Plaur-but not Plau-expression in endothelial cells. Since Plau gene expression is influenced by uPAR signaling [39,40], we suspected that the upregulated Plau expression we saw in Chd4 fl/fl ;Tie2-Cre + endothelial cells was secondary to upregulated Plaur. Importantly, we also detected binding of HDAC1 to the same region of the Plaur promoter at which we detected CHD4 binding by ChIP ( Figure 3D). Since HDAC1 can participate in NuRD complexes [15], these data indicated that CHD4/HDAC1-containing NuRD complexes repressed Plaur transcription in endothelial cells.
To further assess the impact of CHD4 on Plaur transcription, we transfected a CHD4 expression plasmid into C166 cells and assessed Plaur transcription by qPCR. We detected significant downregulation of Plaur with the overexpression of CHD4 ( Figure 3E), which complemented our genetic and knockdown data and suggested that CHD4 inhibited Plaur transcription. Finally, we generated a luciferase reporter containing a fragment of the Plaur promoter that included the region in which we detected positive CHD4 and HDAC1 binding by ChIP ( Figure 3F), and examined reporter activity in C166 endothelial cells. We saw a significant increase in reporter activity upon knockdown of endogenous CHD4, and overexpression of CHD4 significantly decreased reporter activity ( Figure 3G). Altogether, these data indicated that CHD4 directly and negatively regulated Plaur transcription in endothelial cells.
To assess how CHD4 mediated negative regulation of Plaur transcription, we analyzed the acetylation status of H3K9 and H3K27, two covalent histone marks associated with transcriptional activation that can be influenced by NuRD deacetylase activity [41,42]. Following CHD4 knockdown we saw no significant changes in H3K9Ac or H3K27Ac marks in the region of the Plaur promoter that bound CHD4 and HDAC1 ( Figure S7). Therefore CHD4/HDAC1-containing NuRD complexes do not appear to repress Plaur transcription through deacetylation of H3K9 or H3K27 in the promoter region we analyzed. Further analysis of the Plaur promoter in CHD4 knockdown cells will be required to discern the precise mechanism by which NuRD negatively regulates Plaur transcription in endothelium.

CHD4 Directly Promotes Expression of Thrombospondin-1
In addition to Plau, our commercial qPCR arrays also revealed that Thrombospondin-1 (Thbs1) was misregulated in Chd4 fl/fl ;Tie2-Cre + endothelial cells (Table S2). However, unlike Plau, which was upregulated in the absence of CHD4, Thbs1 was significantly downregulated. This result was noteworthy because THBS1 inhibits the ECM protease plasmin as well as matrix metalloproteinases involved in ECM degradation [43,44]. Therefore, downregulation of Thbs1 could contribute to the excessive degradation of ECM components seen in our Chd4 fl/fl ;Tie2-Cre + embryos.
To validate our array data, we performed direct qPCR with Thbs1-specific primers and found a significant decrease in the transcript level of Thbs1 in endothelial cells isolated from Chd4 fl/fl ; Tie2-Cre + embryos ( Figure 3A). Thbs1 transcripts were similarly decreased in CHD4 knockdown C166 endothelial cells ( Figure 3C). Immunostaining confirmed that THBS1 protein expression was diminished around the dorsal aortae of E9.5 Chd4 fl/fl ;Tie2-Cre + embryos ( Figure S8D and S8F). Furthermore, the matrix metalloproteinase MMP9-which is normally inhibited by THBS1 [44]-was upregulated around E10.5 Chd4 fl/fl ;Tie2-Cre + dorsal aortae ( Figures S8H and S8J). ChIP assays revealed that CHD4 and HDAC1 associated with the Thbs1 promoter approximately 800 bp upstream of the transcription start site ( Figure 3D). Furthermore, overexpression of CHD4 in C166 endothelial cells resulted in significantly increased Thbs1 transcription, as measured by qPCR ( Figure 3E). Altogether our data indicated that CHD4 directly promoted Thbs1 expression in endothelial cells.
To examine the mechanism by which CHD4 promoted Thbs1 expression in endothelial cells, we assessed histone acetylation by ChIP in the region of the Thbs1 promoter where CHD4 and HDAC1 bound. As with the Plaur promoter, we saw no changes in H3K9 or H3K27 acetylation upon CHD4 knockdown ( Figure S7). Therefore, the mechanism by which CHD4-containing NuRD complexes promote Thbs1 expression in endothelial cells remains unknown.
Plasmin Activity Is High around Rupture-Prone Chd4 fl/fl ; Tie2-Cre + Vessels Since uPAR and THBS1 respectively promote and inhibit activation of the ECM protease plasmin [43,45], we hypothesized that upregulated uPAR/uPA and downregulated THBS1 activities led to excessive plasmin activation around rupture-prone Chd4 fl/fl ; Tie2-Cre + vessels. To test this hypothesis, we conducted in situ zymography assays on sections of E10.5 control and Chd4 fl/fl ;Tie2-Cre + embryos ( Figure 4A-4D). Plasmin activity within tissue sections was detected with the quenched fluorescent plasmin substrate casein, which fluoresces upon cleavage. After addition of the plasmin zymogen plasminogen, which normally circulates in plasma but must be added exogenously to tissue sections, plasmin activity was significantly elevated around Chd4 fl/fl ;Tie2-Cre + versus control dorsal aortae ( Figure 4D versus Figure 4B; quantification in Figure 4E). These results indicated that proteolytic activators of plasmin were upregulated around rupture-prone Chd4 fl/fl ;Tie2-Cre + blood vessels. Elevated plasmin activity was also detected in E10.5 Chd4 fl/fl ;Tie2-Cre + hearts ( Figure S9), which may explain the slight and Thbs1 was performed on endothelial cells isolated from E10.5 littermate control and Chd4 fl/fl ;Tie2-Cre + embryos. Data were normalized to the relative expression of control samples. Error bars represent SD of results from three independent experiments. (B and C) C166 endothelial cells were transfected with nonspecific (NS) or CHD4-specific siRNA oligonucleotides for 24 h. (B) Western blot analysis was performed on cell lysates using antibodies that recognize CHD4 or GAPDH. A representative blot from 3 independent experiments is shown. (C) RNA was isolated, cDNA was synthesized, and qPCR was performed using Plau-, Plaur-, or Thbs1-specific primers. Data were normalized to the relative expression of NS siRNAtreated samples. Error bars represent SD of results from three to four independent experiments. (D) Chromatin immunoprecipitation (ChIP) assays were carried out in C166 endothelial cells using antibodies against normal mouse IgG (negative control), CHD4, or HDAC1. Immunoprecipitated DNA was analyzed by qPCR to examine CHD4 and HDAC1 enrichment at the Plau, Plaur, and Thbs1 promoters. A transcriptionally inactive region approximately 5 kb upstream of the Fzd5 transcription start site (Fzd5UP) was assessed as a negative control for CHD4 and HDAC1 binding. Data are represented as a percent of total input chromatin. Error bars represent SD of results from three independent experiments. Results for Plau and Fzd5UP ChIP experiments are magnified in the insets. For the Plaur and Thbs1 ChIPs, CHD4 and HDAC1 binding were statistically compared against IgG binding at the respective loci or against CHD4 and HDAC1 binding at the Fzd5UP locus; both sets of comparisons revealed significant enrichment. (E) qPCR with gene-specific primers for Chd4, Plaur, and Thbs1 was performed on C166 endothelial cells transfected with 0.02 ng of a CHD4 expression plasmid or the analogous empty vector backbone (control). Data were normalized to the relative expression of control samples. Error bars represent hypotrabeculation seen in mutant hearts ( Figure S3). We did not see elevated plasmin activity around microvessels in mutant limb buds (data not shown), which was consistent with the undiminished ECM components we observed around these vessels and our finding that they were not prone to rupture. Therefore, although CHD4 is expressed in endothelium lining both large and small vessels ( Figure S1), it appears to regulate plasmin production preferentially around large truncal vessels that are subject to higher hemodynamic forces at midgestation.
Plasmin can be activated by two independent triggers: uPA or the tissue-type plasminogen activator (tPA). We assessed expression of tPA (Plat) by qPCR in CHD4 knockdown C166 endothelial cells to determine if tPA-like uPA-was aberrantly upregulated in the absence of CHD4. We did not see significant chances in Plat transcription ( Figure S10). To further confirm that elevated uPA triggered the excessive plasmin activity seen around rupture-prone Chd4 fl/fl ;Tie2-Cre + dorsal aortae, we performed in situ zymography assays on sections of control and mutant embryos in the presence of the uPA-specific inhibitor amiloride, which does not inhibit tPA activity [46]. Amiloride treatment caused a dramatic decrease in the amount of plasmin activity around Chd4 fl/fl ;Tie2-Cre + dorsal aortae ( Figure 4G versus 4D). These data indicated that elevated plasmin activity around rupture-prone Chd4 fl/fl ;Tie2-Cre + vessels resulted from upregulated uPA. We suspected that this elevated plasmin activity was responsible for the ECM degradation and vascular rupture seen in Chd4 fl/fl ;Tie2-Cre + embryos ( Figure 4H), since excessive plasmin production is associated with vessel wall destruction and aneurysm formation in adult mice [47].

Discussion
The extracellular matrix plays a key role in supporting vascular integrity, however little is known about how ECM production and degradation are regulated and coordinated during embryonic blood vessel development. The first evidence that epigenetic factors could impact embryonic vascular integrity through regulation of ECM composition came from genetic deletion of the histone deacetylase HDAC7. Conditional deletion of Hdac7 in endothelial cells results in lethal vascular rupture and hemorrhage by E11.5 associated with upregulated expression of the matrix metalloproteinase Mmp10 [8]. The ATP-dependent chromatinremodeling enzyme BRG1 similarly regulates endocardial expression of the secreted matrix metalloproteinase Adamts1, although vascular Brg1 mutants suffer from defective cardiac trabeculation rather than vascular rupture due to increased ADAMTS1 activity [25]. Our work now builds on these precedents by demonstrating that ATP-dependent chromatin-remodeling complexes also contribute to the regulation of embryonic vascular integrity by modulating expression of ECM regulators in endothelial cells. We show that NuRD coordinates differential expression of at least two genes to achieve a common outcome of ECM homeostasis. Importantly, genetic reduction of ECM proteolysis partially rescues vascular rupture in our NuRD-deficient mutants, providing convincing evidence that NuRD promotes ECM stability to maintain vascular integrity.
We hypothesize that excessive plasmin production contributes significantly to Chd4 fl/fl ;Tie2-Cre + vascular rupture. Importantly, we did not see evidence of significant plasmin generation around large or small vessels in control embryos by in situ zymography, suggesting that plasmin levels are normally low at midgestation. Plasmin mediates basement membrane degradation required for sprouting angiogenesis in maternal vessels during embryonic implantation and in tumor vessels during neovascularization [48,49]. However, it is questionable whether plasmin plays a SD of results from three independent experiments. (F) Schematic of the region of the murine Plaur promoter that was cloned into a luciferase (LUC) reporter plasmid for use in G. The Plaur promoter fragment encompasses the region to which CHD4 and HDAC1 were shown to bind by ChIP in D. (G) Luciferase assays were performed in C166 cells co-transfected with 250 ng of the reporter schematized in F and 10 ng of a constitutive Renilla luciferase control plasmid. Cells were also transfected with either 10 pmol of non-specific (NS) siRNA or CHD4 siRNA oligonucleotides to knock down endogenous CHD4 or with the CHD4 expression plasmid or its relevant control (empty vector) described in E. All transfections were performed for 24 h. Ratios of relative luciferase:renilla activity were normalized to results from the control samples. Error bars represent SD of results from four independent experiments (with triplicate samples) for the siRNA-transfected samples and from five independent experiments (with triplicate samples) for the CHD4/control plasmid-transfected samples. All statistical calculations for Figure 3 were performed using a two-tailed Student's t test (*, p,0.05). doi:10.1371/journal.pgen.1004031.g003 comparable role during embryonic angiogenesis. Knockouts for various plasmin-generating components such as plasminogen (Plg), uPA (Plau), and uPAR (Plaur) all survive development without notable angiogenesis defects [50][51][52][53]. Moreover, plasmin-mediated activation of the matrix metalloproteinases MMP1 and MMP10 causes vascular regression rather than sprouting in vitro [54][55][56]. These findings are supported by the report that excessive activation of MMP10 in vivo results in vascular rupture similar to that seen in our Chd4 fl/fl ;Tie2-Cre + embryos [8]. Therefore, plasmin may not promote vessel formation in the midgestation embryo but rather may trigger detrimental vessel breakdown by activation of pro-regressive proteinases. If true, minimization of plasmin production or activity-particularly in large vessels that are subject to high hemodynamic forces and increased fragility-would be desirable during embryonic development.  In adult tissues excessive plasmin generation and activity are curtailed by an anti-plasmin system that includes the plasminogen activator inhibitors 1 and 2 (PAI1/2) and a2-antiplasmin. However, expression levels and activities of these inhibitors in the midgestation embryo are unclear. Mice genetically depleted of PAI-1 (Serpine1), PAI-2 (Serpinb2), or a2-antiplasmin (Serpinf2)either alone or in combination-survive development without hemorrhagic events [57][58][59]. These findings suggest that the traditional anti-plasmin system is not required for suppressing excessive plasmin production and protecting vascular integrity during embryonic development. Instead, we propose that a chromatin-based mechanism for regulating transcription of plasminogen activators may have evolved to limit plasmin production prior to the developmental time point at which a robust anti-plasmin system is established.
Plasmin directly cleaves the ECM components laminin and fibronectin and activates multiple matrix metalloproteinases (MMPs) that degrade the ECM components collagen and elastin [60]. One of the MMPs activated by plasmin, MMP9, is directly inhibited by THBS1 [44]. Therefore, THBS1 may influence ECM proteolysis and vascular integrity during development by inhibiting both plasmin and MMP9. In addition, THBS1 is a chemoattractant for vascular smooth muscle cells [61], so diminished Thbs1 expression in Chd4 fl/fl ;Tie2-Cre + endothelial cells may contribute to the reduced a-SMA staining and the disrupted attachments we saw between smooth muscle cells and endothelial cells around mutant vessels. Given the multiple ways that THBS1 can impact ECM stability and vascular integrity, it is possible that the unexplained 25% embryonic lethality observed in Thbs1 2/2 mice [62] may be due to vascular rupture and lethal hemorrhage such as that seen in Chd4 fl/fl ;Tie2-Cre + embryos. We suspect the partial rescue of vascular rupture seen in Chd4 fl/fl ;Plau +/2 ;Tie2-Cre + embryos would be more penetrant if this cross were able to rescue additional functions of THBS1 beyond plasmin inhibition.
It is also likely that CHD4/NuRD has other primary or secondary genomic targets beyond Thbs1 and Plaur whose misregulation contribute to the vascular rupture phenotype seen in Chd4 fl/fl ;Tie2-Cre + embryos. For example, integrin b3 (Itgb3) was significantly downregulated in Chd4 fl/fl ;Tie2-Cre + endothelial cells (Table S2). Integrin b3 mediates interactions between the ECM and endothelial cells [63], and vascular integrity is compromised in Itgb3 2/2 mice, which experience postnatal hemorrhage [64]. We could not detect binding of CHD4 to the Itgb3 promoter by chromatin immunoprecipitation, so we suspect Itgb3 downregulation in Chd4 fl/fl ;Tie2-Cre + endothelial cells was a secondary consequence of Chd4 deletion. Nevertheless, decreased Itgb3 in Chd4 fl/fl ;Tie2-Cre + endothelial cells may partially contribute to vascular fragility at midgestation. Likewise, it is possible that misregulated Wnt signaling contributes to vascular fragility in Chd4 fl/fl ;Tie2-Cre + embryos. We had previously determined that the Wnt-responsive transcription factor Tcf7 and a subset of Wnt target genes-including Plaur-are upregulated in Chd4 fl/fl ;Tie2-Cre + yolk sac endothelial cells [23]. However, this upregulation of Wnt signaling does not appear to impact vascular patterning, as is seen in vascular gain-of-function mutants for the central Wnt signaling component b-catenin (b-catenin lox(ex3)/lox(ex3) ;Tie2-Cre + ) [65]. This phenotypic dissimilarity may be due to the fact that fewer Wnt target genes are misregulated in Chd4 fl/fl ;Tie2-Cre + vasculature than in b-catenin lox(ex3)/lox(ex3) ;Tie2-Cre + blood vessels. Although b-catenin lox(ex3)/lox(ex3) ;Tie2-Cre + embryos show no evidence of vascular rupture or hemorrhage prior to lethality at E11.5 [65], we cannot exclude the possibility that upregulated vascular Wnt signaling might contribute to blood vessel fragility in Chd4 fl/fl ; Tie2-Cre + embryos.
Finally, it will be important to determine if the increased ECM proteolysis and vascular rupture seen in Chd4 fl/fl ;Tie2-Cre + embryos are recapitulated in postnatal mice with induced vascular Chd4 excision. It is possible that plasminogen/plasmin inhibitors such as PAI-1 and a2-antiplasmin reduce the requirement for CHD4 in adult vasculature. However, if CHD4 limits plasmin production in adult endothelium as it does in the embryo, it may play an important preventative role in vascular pathologies such as abdominal aortic aneurysms, which are associated with excessive ECM degradation around vessel walls [66]. uPA/uPAR and plasmin have previously been recognized as promising therapeutic targets for combating aneurysm progression [67], and we now present a novel chromatin-based mechanism by which the activity of these proteins can be regulated during embryonic vascular development. Future studies will determine if NuRD influences proteolytically triggered aneurysms in adults.

Ethics Statement
The

Mice
Chd4-floxed mice (Chd4 fl/fl ), uPA +/2 mice, and Tie2-Cre transgenic mice have been described [20,52,68]. All mice were maintained on a mixed genetic background at the Oklahoma Medical Research Foundation animal facility. Tie2-Cre transgenic and Chd4-floxed embryos and mice were genotyped as described [23,69]. uPA +/2 mice and embryos were genotyped by PCR, using primer sequences obtained from Jackson Laboratory. The wildtype uPA allele was amplified with a forward primer (59-TCTGGAGGACCGCTTATCTG-39) and a reverse primer (59-CTCTTCTCCAATGTGGGATTG-39) that yielded a 153 bp amplicon. The targeted uPA allele was amplified by a neomycinspecific forward primer (59-CACGAGACTAGTGAGACGTG-39) and the same reverse primer used for detecting the wildtype allele, to yield a 337 bp amplicon.

Histological Staining
Hematoxylin and eosin (H&E) staining was performed on embryo cryosections as described previously [70]. Alcian blue staining was performed on paraffin sections of embryonic hearts. Sections were dewaxed, rehydrated, and stained with a 1% solution of Alcian blue 8 GX (Sigma) in acetic acid for 30 min. After Alcian blue staining, sections were counterstained with Nuclear Fast Red (Vector Laboratories) for 5 min.

Transmission Electron Microscopy
Tissue samples were fixed by immersion in a mixture of 2% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M sodium cacodylate for 1 h, followed by postfixation in 1% osmium tetroxide for 90 min and 1% tannic acid for 1 h. The samples were subsequently dehydrated in a graded ethanol series and embedded in epoxy resin (Electron Microscopy Sciences). Ultrathin sections (70 nm) were obtained using an RMC 7000 ultramicrotome (Boeckeler Instruments) equipped with a diamond knife. Sections were stained with uranyl acetate and lead citrate before being viewed with a Hitachi H-7600 electron microscope equipped with a 4 megapixel digital monochrome camera and AMT-EM image acquisition software (Advanced Microscopy Techniques).

Western Blot
Total protein was harvested from siRNA-transfected C166 cells, fractionated in a 9% SDS polyacrylamide gel, and transferred to a PVDF membrane for Western blot analysis with antibodies to CHD4 (Abcam, ab72418) and GAPDH (Sigma, G9545).

Luciferase Assays
A fragment of the murine Plaur promoter spanning the transcription start site (2952 bp to +303 bp) was amplified by PCR and cloned into a modified chromatinizable pCEP9 vector (gift of Jesse Raab). The luciferase gene from a pGL3 luciferase reporter vector was also cloned into the pCEP9 vector. The complete Plaur promoter reporter was transfected into C166 endothelial cells (250 ng) using Lipofectamine 2000. A renilla luciferase plasmid (pRL-TK) was co-transfected (10 ng) as a control. Reporter activity was assayed by transfection of 10 pmol NS siRNA or CHD4 siRNA oligonucleotides (Life Technologies, AM4637 and 4390816, respectively) or with 20 ng of the CHD4 expression plasmid (or analogous control plasmid) described above in ''cell culture and transfections.'' Twenty-four hours after transfection, cells were harvested and analyzed with the Dual-Luciferase Reporter Assay System (Promega) and a Glomax 20/20 Luminometer (Promega).

Fluorescence Microscopy and Image Acquisition
Gross embryonic images were obtained with a Nikon SMZ800 stereomicroscope and Nikon DS-Fi1 camera and monitor. Brightfield histological images were obtained with a Nikon Eclipse 80i microscope using a 106 (NA 0.3) and 406 (NA 0.75) objective and a Nikon DS-Fi1 camera. Fluorescent images were obtained with a Nikon Eclipse 80i microscope using 106 (NA 0.3), 206 (NA 0.5), and 406 (NA 0.75) objectives, an X-cite 120Q light source, and a Nikon DS-Qi1Mc camera. NIS-Elements AR3.0 (Nikon) software was used for all brightfield and fluorescent image acquisition and assembly. Relative fluorescence was determined using ImageJ software (National Institutes of Health).

Confocal Microscopy and Image Acquisition
Confocal images for VE-Cadherin and ZO-1 immunostained sections were obtained on an Olympus IX81 motorized inverted microscope with a Lumens 200 light source. Z stacks were acquired using an Olympus DSU spinning disc confocal system and analyzed using Slidebook 5.0 imaging software (Intelligent Imaging Innovations).
In Situ Zymography E10.5 embryos were cryosectioned without prior fixation. 20 mm sections of control and Chd4 fl/fl ;Tie2-Cre + embryos were affixed to the same slide and were overlaid with in situ zymography solution (ISZS) consisting of 1% low melting point agarose (Invitrogen), 0.1 mg/mL quenched BODIPY FL casein from the EnzChek Protease Assay Kit (Invitrogen), 1 U/mL human Glu-plasminogen (Sigma), and Hoechst (20 mg/mL). Overlaid sections were coverslipped and incubated for 1 h at 37uC. Casein cleavage was detected by fluorescence microscopy. Plasminogen-free ISZS was used as a negative control to verify that casein cleavage was plasmin-dependent. For amiloride treatment, 0.1 mM amiloride (Sigma, A7410) was mixed with the ISZS prior to overlaying embryo sections. qPCR To analyze transcript levels, total RNA isolated from primary vascular endothelial cells was purified using the RNeasy Mini Kit (QIAGEN) according to the manufacturer's instructions. cDNA was prepared using the MultiScribe Reverse Transcriptase kit (Applied Biosystems), and real-time quantitative PCR was performed using 26 SYBR green qPCR master mix (Life Technologies) and the CFX96 Real-Time System (Bio-Rad) with gene-specific primers. The following primers were used for qPCR: Chd4

qPCR Analysis
The relative fold change in transcription was determined using the comparative C T method and the housekeeping genes, Gapdh and b-actin, as internal controls. Data from at least 3 independent sets of littermate control and mutant embryos or from 3 independent siRNA knockdown experiments or overexpression experiments were combined and presented as the mean plus standard deviation (SD). Statistical differences were detected using a two-tailed unpaired Student's t test.

Angiogenesis and Extracellular Matrix qPCR Arrays
Total RNA from primary vascular endothelial cells was isolated using Trizol (Life Technologies) according to the manufacturer's instructions. The DNA-free kit (Life Technologies) was used to digest any contaminating DNA, and cDNA was prepared using the RT 2 First Strand kit (SABiosciences/QIAGEN). Three independent experiments were performed using Mouse Angiogenesis RT 2 Profiler PCR Array and Mouse Extracellular Matrix and Adhesion Molecules RT 2 Profiler PCR Array (SABiosciences/ QIAGEN) according to the manufacturer's instructions. Data analysis was performed using the web-based PCR array data analysis tool available through the SABiosciences/QIAGEN website.

Chromatin Immunoprecipitation (ChIP)
Sub-confluent C166 yolk sac endothelial cells were crosslinked with 1% formaldehyde for 10 min and processed for ChIP with the MAGnify Chromatin Immunoprecipitation System (Invitrogen) according to the manufacturer's instructions. A CHD4specific antibody (Abcam; ab70469) was used to immunoprecipitate protein-DNA complexes, and a normal mouse IgG antibody (Invitrogen; 100005291) was used as a negative control for CHD4 ChIP assays. An HDAC1-specific antibody (Abcam, ab46985) and normal mouse IgG antibody were used for HDAC1 ChIP assays.
For H3K9Ac and H3K27Ac ChIP assays, chromatin was harvested from C166 cells that were transfected for 24 h with nonspecific or CHD4-specific siRNAs. Sonicated chromatin was immunoprecipitated using an anti-histone H3K9Ac antibody (Active Motif; 39137) or an anti-histone H3K27Ac antibody (Abcam; ab4729), and isotype matched IgG antibodies (Invitrogen) were used as negative controls. All ChIP assays were performed with 10 mg of experimental or control antibodies. Real-time quantitative PCR was performed using RT 2 Fast SYBR green qPCR master mix (SABiosciences) and the CFX96 detection system (Bio-Rad) with specific primers. For the Thbs1 gene, the primers 59-AGCCAGATGGTTCAGCAAAT-39 and 59-CCCTCAAGGCAAAATTGAAA-39 (located 834 bp and 732 bp upstream of the TSS, respectively) were used for CHD4, HDAC1, H3K9Ac and H3K27Ac ChIP assays. For the Plaur gene, the primers 59-ACTGAGCCGCTCTGAGTGAT-39 and 59-CCAGGGGAAAAACAAGTTGA-39 (located 922 bp and 723 bp upstream of the TSS, respectively) were used for the ChIP assays described above. For the Plau gene, the primers 59-GACATGTGGGAGCCTTTGTT-39 and 59-CTCGCTGAT-CTCAAGCCTCT-39 (located 507 bp and 349 bp upstream of the TSS, respectively) were used for CHD4 and HDAC1 ChIP assays. Primers designed for a transcriptionally inactive region far upstream of the Fzd5 TSS were used to amplify a negative control region for CHD4 binding and activity. These Fzd5UP primers, 59-GGTGACTTAGGGCAAAACCA-39 and 59-AGGCCACCA-TACCAGGTTCT-39, were located 4856 bp and 4648 bp upstream of the TSS, respectively.

ChIP Analysis
A defined fraction of input chromatin that was subjected to sonication was set aside prior to immunoprecipitation but was otherwise treated identically to immunoprecipitated samples through reverse crosslinking and DNA purification steps. DNA from input chromatin was diluted and analyzed by qPCR along with DNA from immunoprecipitated samples. Input C T values were adjusted for dilution and used to calculate % input values for immunoprecipitated samples. A two-tailed Student's t test was used for statistical comparisons between samples immunoprecipitated with negative control (i.e. IgG) or experimental antibodies at a particular locus. Comparisons between immunoprecipitated samples at an experimental locus (i.e. Plaur or Thbs1) versus a negative control locus (i.e. Fzd5UP) were also analyzed.

Whole-Mount Embryo Staining
Whole-mount immunostaining for PECAM-1 was performed using a rat anti-mouse PECAM-1 antibody (BD Pharmingen, 557355) as described [72]. Figure S1 CHD4 is expressed in endothelial cells lining large and small vessels at E10.5. Immunostaining for the endothelial cell marker PECAM-1 (red) and for CHD4 (green) was performed on sections of E10.5 wildtype embryos. Hoechst (blue) was used to counterstain nuclei.  Figure S6 Genes involved in smooth muscle cell recruitment toward endothelium are expressed at normal or elevated levels in CHD4 knockdown endothelial cells. C166 endothelial cells were transfected with nonspecific (NS) or CHD4-specific siRNA oligonucleotides for 24 h. RNA was isolated, cDNA was synthesized, and qPCR was performed using gene-specific primers for heparin-binding EGF-like growth factor (Hbegf), platelet derived growth factor B (Pdgfb), or transforming growth factor b receptor II (Tgfbr2). Data were normalized to the relative expression of NS siRNA-treated samples. Error bars represent 6 SD of results from three independent experiments. Statistical analysis was performed using a two-tailed Student's t test (*, p,0.05). (TIF) Figure S7 No changes in H3K9Ac or H3K27Ac enrichment are seen at the Plaur and Thbs1 promoters in CHD4 knockdown endothelial cells. Chromatin immunoprecipitation (ChIP) assays were carried out in C166 endothelial cells transfected for 24 h with either non-specific (NS) siRNA or CHD4-specific siRNA oligonucleotides. Immunoprecipitation was performed using antibodies against H3K9Ac (A) or H3K27Ac (B). Immunoprecipitated DNA was analyzed by qPCR to examine enrichment of H3K9Ac or H3K27Ac at the Plaur and Thbs1 promoters. A transcriptionally inactive region approximately 5 kb upstream of the Fzd5 transcription start site (Fzd5UP) was assessed as a negative control for antibody binding. Data are represented as a percent of total input chromatin. Error bars represent SD of results from three independent experiments. For statistical analysis, NS siRNA-and CHD4 siRNA-treated samples were compared at each locus using a two-tailed Student's t test; no significant differences were detected. Likewise, no significant differences were detected between NS siRNA-and CHD4 siRNA-treated samples when H3K9Ac and H3K27Ac marks were normalized against total H3 pulldown at the Plaur and Thbs1 promoters (data not shown).  Figure S9 Plasmin activity is elevated in E10.5 Chd4 fl/fl ;Tie2-Cre + hearts. In situ zymography was performed on sections of E10.5 littermate control (A) and Chd4 fl/fl ;Tie2-Cre + (B) embryonic hearts for detection of plasmin activity, as described in Figure 5. Casein cleavage (green fluorescence) was substantially higher in the Chd4 fl/fl ; Tie2-Cre + hearts versus the control hearts. Hoechst (blue) was used as a nuclear counterstain. V = ventricle. Scale bars: 100 mm. (TIF) Figure S10 Expression of tissue plasminogen activator (tPA; Plat) is normal in CHD4 knockdown endothelial cells. C166 endothelial cells were transfected with nonspecific (NS) or CHD4-specific siRNA oligonucleotides for 24 h. RNA was isolated, cDNA was synthesized, and qPCR was performed using Plat-specific primers. Data were normalized to the relative expression of NS siRNAtreated samples. Error bars represent 6 SD of results from three independent experiments. (TIF) Figure S11 Rupture-prone dorsal aortae are significantly rescued from hemorrhage in Chd4 fl/fl ;Plau +/2 ;Tie2-Cre + embryos. (A-D) Hematoxylin and eosin (H&E) staining of E12.5 littermate control and Chd4 fl/fl ;Plau +/2 ;Tie2-Cre + embryos revealed an intact dorsal aorta (DA; insets in panels A and B). However, blood was seen aberrantly pooling in the central canal (CC) of the spinal cord in Chd4 fl/fl ;Plau +/2 ;Tie2-Cre + embryos (D). Scale bars: 500 mm (A-B); 100 mm (C-D).

(TIF)
Table S1 Genes with unchanged expression levels in E10.5 Chd4 fl/fl ;Tie2-Cre + endothelial cells. Endothelial cells from E10.5 littermate control and Chd4 fl/fl ;Tie2-Cre + embryos were isolated, RNA was purified, cDNA was synthesized and qPCR was performed using two commercial qPCR arrays containing a total of 157 genes important for extracellular matrix composition and angiogenesis (SABiosciences/QIAGEN). The 148 genes that were identified as having insignificant changes in expression levels (p.0.1) after three different experiments are listed. Data analysis was performed using the web-based PCR Array Data Analysis tool recommended for use with these arrays: (http://www. sabiosciences.com/pcrarraydataanalysis.php). See Table S2 for the list of genes with significant expression changes from the arrays. (DOC)

Table S2
Genes misregulated in E10.5 Chd4 fl/fl ;Tie2-Cre + embryonic endothelial cells. Endothelial cells from E10.5 littermate control and Chd4 fl/fl ;Tie2-Cre + embryos were isolated, RNA was purified, cDNA was synthesized and qPCR was performed using two commercial qPCR arrays containing a total of 157 genes important for extracellular matrix composition and angiogenesis (SABiosciences/QIAGEN). The nine genes that were identified as having significantly differential expression levels (p,0.1) after three different experiments are listed along with their average fold change. Data analysis was performed using the web-based PCR Array Data Analysis tool recommended for use with these arrays: (http://www.sabiosciences.com/pcrarraydataanalysis.php). See Table S1 for the list of genes with insignificant expression changes from the arrays. (DOC)