Blood vessels develop through two consecutive processes, vasculogenesis and angiogenesis. Vasculogenesis occurs by means of differentiation of endothelial cells from hemangioblasts in the blood island of the extra embryonic mesoderm. Angiogenesis is a physiological process involving the growth of new blood vessels from the endothelial cells of existing blood vessels beginning at embryonic day 8.5 in the mouse (Risau and Flamme,1995; Risau,1997; Mojzis et al.,2008). Vasculogenesis and angiogenesis are regulated differently. Although the exact mechanisms are still not entirely defined, both vasculogenesis and angiogenesis are driven by growth factors like vascular endothelial growth factor (VEGF) and angiopoeitins, and additionally involve cell–cell and cell–matrix interactions (Tanjore et al.,2008).
Nuclear factor-κB (NF-κB) regulates expression of several genes involved in immune and inflammatory responses, cellular growth, survival, differentiation and development (Chen and Castranova,2007; Hayden and Ghosh,2008). There are five NF-κB family members in mammals: p50 (NF-κB1), p52 (NF-κB2), p65 (RelA), c-Rel, and RelB. These factors share an N-terminal Rel homology domain (RHD) responsible for both DNA binding and homo- and heterodimerization (Hayden and Ghosh,2008). In resting cells, NF-κB complexes are retained in the cytoplasm by interaction with IκB inhibitory proteins. Upon stimulation by inflammatory signals, such as TNF-α, IL-1, or LPS, IκB proteins become phosphorylated and ubiquitinated and are subsequently degraded by the proteasome. This results in the translocation of NF-κB proteins to the nucleus where they regulate expression of target genes (Hayden and Ghosh,2008). Phosphorylation of IκB proteins is mediated by IκB kinase (IKK). The IKK complex contains two highly homologous kinase subunits, IKKα/IKK1 and IKKβ/IKK2, and a regulatory subunit NEMO/IKKγ (Hacker and Karin,2006). IKKβ is involved in canonical NF-κB signaling in response to inflammatory signals, while the noncanonical pathway depends on the IKKα subunit that phosphorylates NF-κB2/p100 leading to its processing to the p52 protein (Bonizzi and Karin,2004).
The NF-κB pathway has been previously implicated in angiogenesis. VEGF gene expression can be regulated by NF-κB (Chilov et al.,1997; Ishdorj et al.,2008). Other genes that are important for angiogenesis, such as MMPs, uPA, VCAM-1, ICAM-1, COX-2, GRO-1, iNOS, jagged-1, HMGB1 receptor, and HIF-1α are also regulated by NF-κB (Koch et al.,1995; Taylor et al.,1998; Stetler-Stevenson,1999; Yasuda et al.,2002; Zhou et al.,2003; Rius et al.,2008; Sainson et al.,2008; van Beijnum et al.,2008). Ikkβ-deficient mice died between E12.5 and E14.5 because of severe liver defects (Li et al.,1999a, b; Tanaka et al.,1999). Thus, it is not possible to study the role of the NF-κB pathway in endothelial cells using this model.
To directly address the potential roles of the NF-κB pathway in angiogenesis, we took an in vivo genetic approach using the Cre/LoxP system to specifically disrupt Ikkβ in Tie2-Cre–positive endothelial cells. Conditional Ikkβ ablation in endothelial cells led to embryonic lethality between E13.5 and E15.5. The blood vessel network in the embryonic liver was specifically affected in the mice with endothelial cell deletion of Ikkβ. This led to apoptosis of hepatocytes and liver degeneration. Our results demonstrate that IKKβ is critical for fetal liver angiogenesis in vivo.
Ablation of Ikkβ in Tie2-Positive Endothelial Cells Leads to Embryonic Lethality Between E13.5 and E15.5
Conditional deletion of Ikkβ using Tie2-Cre (Kisanuki et al.,2001) was used to study the function of the canonical NF-κB pathway in endothelial cells, as previous work has demonstrated that targeting this kinase leads to ablation of the canonical NF-κB pathway (Li et al.,1999a, b; Tanaka et al.,1999). Initially mice homozygous for the conditional Ikkβ floxed allele (IkkβF/F/Tie2-Cre) were studied. The majority of IkkβF/F/Tie2-Cre embryos died between E13.5 and E17.5, but infrequently mice with this genotype were born and grew to adulthood (e.g., 25 expected of genotype IkkβF/F/Tie2-Cre, but only 2 survived; Supp. Table S1). Endothelial cells were isolated from the aorta of the atypical mice that survived to adulthood. Polymerase chain reaction (PCR) genotyping of the aortic endothelial cells demonstrated that the null Ikkβ allele was not present (Supp. Fig. S1A, which is available online). In addition, analysis of IKKβ by Western blot analysis revealed that IKKβ was expressed at near normal levels in endothelial cells from the mice that survived to adulthood (Supp. Fig. S1B).
Because these results indicated that the Tie2-Cre-mediated deletion of the Ikkβ conditional allele was incomplete, we adopted the strategy of studying mice with one conventional knockout allele and one conditional allele, Ikkβ−/F, so that only one allele needed to be deleted in endothelial cells through the action of Tie2-Cre. The strategy used to generate mice of genotype of Ikkβ−/F/Tie2-Cre is outlined in Figure 1A. Briefly, a conventional knockout allele was created by breeding IkkβF/+ mice with Sox2-Cre mice (Hayashi et al.,2002). Mice with the knockout allele were bred with Tie2-Cre mice and Ikkβ−/+/Tie2-Cre male mice were identified. The Ikkβ−/+/Tie2-Cre males were bred to IkkβF/F female mice to generate the experimental and control genotypes.
Mice of Ikkβ−/F/Tie2-Cre genotype, expected at a Mendelian frequency of 25%, were not observed beginning at E15.5 (Table 1). Further analysis indicated that Ikkβ−/F/Tie2-Cre mice were present at expected frequencies at E12.5. However, embryonic lethality was observed beginning at E13.5 (Table 1). All other possible genotypes from this genetic cross were observed at expected frequencies (not shown). These results indicated that embryonic lethality occurred in the period between E13.5 and E15.5.
Table 1. Lethality of Ikkβ-/F/Tie2-Cre Mice Happened Between E13.5 and E15.5 in FVB/N Backgrounda
Expected Ikkβ-/F/Tie2-Cre embryos
Expected Mendelian ratio is 25%. Viable: heartbeat can be detected at dissection.
To confirm that IKKβ was absent in endothelial cells, PECAM/CD31-positive cells were isolated from E12.5 embryos, before embryonic lethality occurred, by high speed fluorescence-activated cell sorting (FACS). The PECAM-1–positive cells constituted approximately 5% of the total cells isolated from each individual embryo (Supp. Fig. S2). Total RNA was extracted from the enriched PECAM-1–positive cells and subjected to qRT-PCR analysis. The level of Ikkβ mRNA in Ikkβ−/F/Tie2-Cre endothelial cells was reduced by 16-fold compared with controls, similar to levels observed in conventional knockout mice (Fig. 1B). VE-cadherin, another endothelial cell marker, was expressed at the similar level in Ikkβ−/F/Tie2-Cre, Ikkβ−/F and Ikkβ−/− embryos (Fig. 1B). Additionally, the expression level of six NF-κB target genes in the PECAM-1–positive cells was also studied by qRT-PCR (Fig. 1C). The mRNA levels of three target genes, IL-6, VEGF-C, and COX-2, were reduced twofold in Ikkβ−/F/Tie2-Cre endothelial cells compared with Ikkβ−/F control cells (Fig. 1C). The relatively small effect on expression of these NF-κB target genes may reflect that basal levels and not cytokine induced levels of expression were determined. The data are consistent with a deletion of the Ikkβ gene and down-regulation of the NF-κB pathway in endothelial cells isolated for Ikkβ−/F/Tie2-Cre embryos.
As a complementary approach to confirm Ikkβ deletion in endothelial cells, IKKβ expression was studied by immunohistochemistry in experimental and control mice (Fig. 1D). Significantly reduced IKKβ expression was detected in the majority of endothelial cells of Ikkβ−/F/Tie2-Cre embryos compared with controls (arrows in Fig. 1D; quantified in Fig. 1E). This commercial antibody did not work in our hands for indirect immunofluorescence using frozen sections (data not shown).The results support the conclusion that Ikkβ was deleted in Tie2-positive endothelial cells throughout the embryo.
Angiogenesis in Livers of Ikkβ−/F/Tie2-Cre Embryos Is Impaired
Ikkβ−/F/Tie2-Cre embryos at E13.5 had normal appearance at dissection (Fig. 2A). Histological analysis of sagittal sections of viable E13.5 Ikkβ−/F/Tie2-Cre embryos revealed specific abnormalities in the livers of these mutant embryos compared with controls (Fig. 2B). The architecture of the liver was disrupted, and there were regions of apparent cell death. This phenotype was very similar to Ikkβ−/− embryos, but cell death occurred regionally in the livers of Ikkβ−/F/Tie2-Cre embryos compared with through the entire livers in Ikkβ−/− embryos at the same stage. No obvious defects were detected in other organs or regions of the embryos (Fig. 2C).
To determine whether angiogenesis was affected in Ikkβ−/F/Tie2-Cre embryonic livers, PECAM-1 antibody was used to visualize endothelial cells by indirect immunofluorescence (Fig. 3A). The density of blood vessels was reduced by 2.6-fold in Ikkβ−/F/Tie2-Cre embryonic livers (Fig. 3B). This decrease in blood vessel density was seen throughout the entire embryonic liver. In contrast, angiogenesis through the remainder of the embryonic body was not affected in Ikkβ−/F/Tie2-Cre embryos, as determined by whole mount PECAM-1 immunohistochemistry staining (Fig. 3C). Blood vessels are composed of endothelial cells and supporting cells, such as vascular smooth muscle cells and pericytes (Cleaver and Melton,2003). To study the blood vessel integrity in Ikkβ−/F/Tie2-Cre embryos, PECAM-1 and smooth muscle α-actin double immunofluorescence staining was performed on frozen sections (Fig. 3D). The analysis indicated that there was no obvious blood vessel disorganization in the embryonic vasculature in Ikkβ−/F/Tie2-Cre embryos.
Hematopoiesis occurs in the liver beginning at E10 (Godin and Cumano,2005). Since liver degeneration was observed in Ikkβ−/F/Tie2-Cre embryos, and because Tie2-Cre was also active in hematopoietic stem cells (Kisanuki et al.,2001), the status of fetal liver hematopoiesis was examined at E12.5 by performing flow cytometry with different hematopoietic cell lineage markers. Populations of c-Kit+, Ter119+, F4/80+, Mac-1+, and Gr-1+ cells were intact in Ikkβ−/F/Tie2-Cre fetal livers (Fig. 3E). These results indicated that hematopoiesis in mutant livers was not impaired. Immunostaining with F4/80 indicated no significant differences in the Kupffer cells populations between mutant and control embryos (data not shown).
Decreased Proliferation and Increased Apoptosis in Livers of Ikkβ−/F/Tie2-Cre Embryos
Cell proliferation was measured in sagittal sections of Ikkβ−/F/Tie2-Cre embryos using a BrdU (5-bromo-2′-deoxyuridine) incorporation assay (Fig. 4). BrdU-positive cells were reduced by fivefold at E13.5 in Ikkβ−/F/Tie2-Cre embryonic livers compared with Ikkβ−/F controls (Fig. 4A, top panels, quantification in Fig. 4B), indicating that the number of cells in S phase was significantly reduced at E13.5 in Ikkβ−/F/Tie2-Cre embryonic livers. In contrast, no obvious difference between mutants and controls was detected at E12.5 or in other organs at E13.5 (Fig. S3) (Fig. 4A, bottom panels, Fig. 4C).
To further characterize the phenotype of Ikkβ−/F/Tie2-Cre embryonic livers, immunohistochemistry with cleaved Caspase-3 (Asp175) antibody was performed on both Ikkβ−/F/Tie2-Cre and Ikkβ−/F sections at E13.5 (Fig. 5A) and E12.5 (Supp. Fig. S4) to analyze potential changes in apoptosis. Cells that were undergoing apoptosis increased 10-fold in Ikkβ−/F/Tie2-Cre embryonic livers at E13.5 compared with controls. This level of apoptosis was approximately 3.5-fold lower than in the conventional Ikkβ knockout mice (Fig. 5B). Apoptosis in the liver at E13.5 was regional and occurred in an apparently random manner. This was in contrast to the changes in blood vessel density which occurred throughout the fetal liver (Fig. 3A). At E12.5, before embryonic lethality was first detected, Ikkβ−/F/Tie2-Cre embryonic livers showed only a twofold increase in apoptosis compared with controls (Supp. Fig. S4; Fig. 5B). There was no increased apoptosis detected in the heart, lung and kidney (Fig. 5C). Apoptosis was also verified by TUNEL staining in an independent set of embryos (data not shown).
Two-color immunofluorescence with the cleaved Caspase-3 antibody and either PECAM-1 antibody staining to detect endothelial cells or alpha-fetoprotein antibody staining to detect embryonic hepatocytes, was performed to identify the cell type undergoing apoptosis. The experiments revealed that apoptosis was significantly increased in alpha-fetoprotein–positive hepatocytes in mutant versus control embryos (Fig. 6A). However, there was no significant increase in apoptosis in cells that were PECAM-1–positive (Fig. 6B). These results indicated that hepatocytes were the major cell type undergoing apoptosis in Ikkβ−/F/Tie2-Cre embryos.
To rule out the possibility that Tie2-Cre may have nonspecifically deleted Ikkβ from hepatocytes leading to cell death, immunohistochemistry with IKKβ antibody was performed on transverse sections of Ikkβ−/F/Tie2-Cre embryonic livers. Robust IKKβ staining was detected in Ikkβ−/F/Tie2-Cre embryonic livers (Fig. 6C), consistent with the expression of IKKβ in hepatocytes. In a control experiment, tissue from Ikkβ−/− embryos was also studied, demonstrating the absence of IKKβ staining in these embryos (Fig. 6C). In conventional Ikkβ knockout mice, apoptosis in the liver occurs due to increased expression and sensitivity to TNF-α (Li et al.,1999a, b; Tanaka et al.,1999). However, TNF-α RNA levels were no different in Ikkβ−/F/Tie2-Cre compared with controls (Fig. 6D).
Conditional deletion of Ikkβ using Tie2-Cre resulted in embryonic lethality with obvious defects in fetal liver angiogenesis and liver degeneration. The restricted phenotype was not due to liver-specific ablation of Ikkβ as deletion of the gene was pan-endothelial. An inductive role for endothelial cells in liver development beginning at E8.5 has been previously established (Matsumoto et al.,2001). However, based on the later onset of the phenotype in the Ikkβ−/F/Tie2-Cre mice, a role for the IKKβ/NF-κB pathway in early endothelial inductive events appears unlikely. The results presented here indicate that the IKKβ/NF-κB pathway plays a novel role in liver homeostasis during embryonic development. The sinusoidal endothelium of the liver is structurally and functionally distinct from the general endothelium of the embryo (Couvelard et al.,1996; Braet and Wisse,2002) and these differences may explain the apparent lack of phenotype in other tissues and organs.
The phenotype of the Ikkβ−/F/Tie2-Cre embryos was unexpectedly similar to the phenotype observed in the conventional knockout, albeit not as severe (Li et al.,1999a, b; Tanaka et al.,1999). Hepatocyte apoptosis occurs later in the Ikkβ−/F/Tie2-Cre mice and hepatocyte apoptosis was initially more regional, limited to a few lobes at E13.5, compared with the Ikkβ deficient mice at the same or earlier stages. In contrast, hepatocyte-specific knockout of Ikkβ using the Alfp-Cre transgene did not result in significant embryonic lethality, despite the activity of this Cre-driver beginning around E10.5 (Kellendonk et al.,2000; Luedde et al.,2005). Taken together, this information suggests that ablation of Ikkβ in endothelial cells could contribute to the conventional knockout phenotype. One caveat concerning this possible interpretation is that a different targeted allele of Ikkβ, a construct that targeted exons 6–7, was used in the Alfp-Cre experiments (Luedde et al.,2005). It is possible this allele is hypomorphic, or that Alfp-Cre is not expressed early enough in embryonic development to duplicate the results with the conventional knockout allele. A second caveat is that in the results reported here, the one copy of Ikkβ deleted in hepatocytes may be necessary for the phenotype seen with Tie2-Cre ablation of Ikkβ in endothelial cells. However, hepatocyte apoptosis was also observed in IkkβF/F/Tie2-Cre studied in embryos at E15.5 (data not shown), indicating that the phenotype observed does not depend on deletion of one allele of Ikkβ in hepatocytes.
The decreased cell proliferation and increased cell death observed in hepatocytes in Ikkβ−/F/Tie2-Cre embryos is non–cell-autonomous, dependent apparently on the deletion of the Ikkβ gene in endothelial cells. It is well appreciated that communication between the distinct cell types present in liver is required to affect organ structure and functions. For example, the endothelium has been demonstrated to protect hepatocytes independent of its evident role in supplying nutrients and oxygen (LeCouter et al.,2003). Similarly, co-cultivation of hepatocytes with endothelial cells in vitro can potentiate expression of hepatocyte-specific genes, such as albumin and ApoA-I (Harimoto et al.,2002; Kurosawa et al.,2005). Additionally, crosstalk between hepatocytes and closely juxtaposed endothelial cells mediated by means of vascular endothelial growth factor is essential for sinusoidal endothelial growth and differentiation (Edwards et al.,2005). Our results define the IKKβ/NF-κB pathway as a key element in this intercellular communication during embryonic development. The exact molecular nature of the intercellular dialogue, and how its disruption results in the phenotype observed in these studies, remains to be determined. One obvious candidate, TNF-α does not seem to be involved, because hepatocytes still expressed IKKβ and TNF-α RNA Levels were not altered. Which factors or cytokines are responsible for this crosstalk deserves further investigation.
Floxed mice (IkkβF/F) were described previously (Li et al.,2003). Ikkβ conventional knockout mice were generated by crossing IkkβF/F mice with Sox2-Cre mice (Hayashi et al.,2002). Tie2-Cre mice were described previously (Kisanuki et al.,2001). All animals were maintained on a FVB/N background (>10 generations). Animals were housed in the animal facility at The Ohio State University Biomedical Research Tower under sterile conditions maintaining constant temperature and humidity and were fed a standard diet. Mice were genotyped by PCR analysis from prepared tail DNA. All protocols for the use of mice were approved by the OSU Institutional Animal Use and Care Committee.
Quantitative Real-time PCR
Total RNA was extracted from CD31 sorted endothelial cells by Trizol (Invitrogen, Carlsbad, CA) according to the manufacturer's recommendations. RNA was purified using the TURBO DNA-free kit (Ambion, Austin, TX). Total RNA was reverse transcribed by Superscript III reverse transcriptase (Invitrogen) with random hexamer primers. qPCR was performed as previously described (Wei et al.,2004). All PCR reactions were performed in duplicate. Primer sequences are available upon request.
Histology and Immunohistochemistry
Whole-mount embryo immunohistochemistry was performed as described (Suri et al.,1996). Tissue was fixed in 10% buffered formalin (Fisher, Middletown, VA). After routine processing, paraffin-embedded sections (5 μm thick) were stained with hematoxylin and eosin (H&E) for histological analysis. Immunohistochemistry was performed on paraffin-embedded sections (5 μm thick), and immunofluorescence studies were performed on cryopreserved sections (7 μm thick). Paraffin sections were dewaxed and rehydrated through a graded series of ethanol washes, and maintained in phosphate buffered saline (pH 7.4; PBS). Frozen sections were air dried for 30 min and then placed in PBS. Sections were treated with 3% H2O2 for 10 min to quench endogenous peroxidase activity and placed in Dako target retrieval solution at sub-boiling temperature for 60 min to unmask the antigens. Protein Block (Dako, Carpinteria, CA) was applied for 30 min to reduce nonspecific binding of the antibodies. Sections were treated with primary antibody, followed by biotin conjugated secondary antibody (Rockland, Gilbertsville, PA). Alexa Fluro dye 488 or 594 streptavidin (Molecular Probes, Eugene, OR) were used for immunofluorescence. Vectorstain ABC kit (Vector Laboratories, Burlingame, CA) was used for immunohistochemistry, followed by DAB to visualize immunocomplexes. Primary antibodies were: anti-PECAM antibody (MEC13.3, rat anti-mouse monoclonal, Pharmingen, San Diego, CA), anti-smooth muscle α-actin antibody (Spring Bioscience, Fremont, CA), anti-cleaved Caspase-3 antibody (Asp175) (9661, cell signaling, Danvers, MA), anti-BrdU antibody (M0744, Dako), anti-AFP antibody (MAB1368, R&D Sytems, Minneapolis, MN), and anti-IKKβ antibody (ab55404, abcam, Cambridge, MA).
Embryos and embryonic livers were isolated at E12.5. DNA prepared from yolk sacs was used to genotype the embryos. A single cell suspension of embryonic livers or whole embryos was prepared by established method (Motoike et al.,2000) and treated with red blood cell lysing buffer (Sigma, St. Louis, MO). 1 × 106 cells for flow or 1 × 107 cells for sorting were incubated with antibodies (0.5 mg/1 × 106 cells) in DMEM +2% fetal calf serum (FCS) on ice for 30 min in the dark. Cells were washed once with 1× PBS + 1% FCS. Cells fixed in 1% formalin were analyzed with the FACS Calibur (Beckton-Dickinson, San Jose, CA), or directly sorted without fixation on the FACS Aria (Beckton-Dickinson). Monoclonal antibodies against the following antigens were used: Gr-1- FITC, F4/80-PE, Mac-1-FITC, Ter119-PE, c-Kit- FITC, and CD31-PE (eBioscience, San Diego, CA and Caltag, Carlsbad, CA). Appropriate isotype controls were also analyzed.
BrdU Incorporation Assay
One hundred micrograms of BrdU (5-bromo-2′-deoxyuridine, Sigma) per gram of body weight in 1× PBS was injected ventrally into the abdomen of pregnant mice avoiding the embryos. Mice were killed 2 hr after injection. Embryos were fixed in 10% buffered formalin and embedded in paraffin. BrdU incorporation was measured by anti-BrdU antibody immunohistochemistry. Three individual embryos were analyzed in each genotype group at each time point.
We thank Masashi Yanagisawa for Tie2-Cre transgenic mice. We are grateful to the staff of mouse facilities for animal husbandry, and to the Flow Cytometry Shared Resource and Histology Services of Ohio State University for technical support. We thank Denis Guttridge for critical discussions and advice. M.C.O. was funded by the NIH.