FOXO1-Mediated Activation of Akt Plays a Critical Role in Vascular HomeostasisNovelty and Significance
Rationale: Forkhead box-O transcription factors (FOXOs) transduce a wide range of extracellular signals, resulting in changes in cell survival, cell cycle progression, and several cell type-specific responses. FOXO1 is expressed in many cell types, including endothelial cells (ECs). Previous studies have shown that Foxo1 knockout in mice results in embryonic lethality at E11 because of impaired vascular development. In contrast, somatic deletion of Foxo1 is associated with hyperproliferation of ECs. Thus, the precise role of FOXO1 in the endothelium remains enigmatic.
Objective: To determine the effect of endothelial-specific knockout and overexpression of FOXO1 on vascular homeostasis.
Methods and Results: We show that EC-specific disruption of Foxo1 in mice phenocopies the full knockout. Although endothelial expression of FOXO1 rescued otherwise Foxo1-null animals, overexpression of constitutively active FOXO1 resulted in increased EC size, occlusion of capillaries, elevated peripheral resistance, heart failure, and death. Knockdown of FOXO1 in ECs resulted in marked inhibition of basal and vascular endothelial growth factor–induced Akt-mammalian target of rapamycin complex 1 (mTORC1) signaling.
Conclusions: Our findings suggest that in mice, endothelial expression of FOXO1 is both necessary and sufficient for embryonic development. Moreover, FOXO1-mediated feedback activation of Akt maintains growth factor responsive Akt/mTORC1 activity within a homeostatic range.
The FOXO family of transcription factors belongs to the winged helix or forkhead box class of transcription factors.1 Invertebrates possess 1 FOXO gene, termed daf-16 in the worm and dFOXO in the fly. Mice and humans possess 4 FOXO members: FOXO1, FOXO3, FOXO4, and FOXO6. FOXO1, FOXO3, and FOXO4 are highly related homologs with overlapping patterns of expression and transcriptional activities. The FOXO protein family is regulated primarily by post-translational modifications, including phosphorylation, acetylation, monoubiquitination, and polyubiquitination.2,3 These various modifications control subcellular localization and protein levels, as well as efficacy of DNA binding and transcriptional activity. Most notably, FOXO1, FOXO3, and FOXO4 have 3 conserved amino acids that are targets for phosphorylation by Akt or serum/glucocorticoid-regulated kinase (SGK). Phosphorylation at these sites leads to nuclear exclusion of the transcription factor. FOXOs have been shown to play a role in many physiological processes, including the control of cell proliferation and survival, cell cycle progression, DNA repair, oxidative stress resistance, energy metabolism, and cell differentiation.4–6 Unrestrained FOXO activity can result in cellular senescence, autophagy, and atrophy and can promote a catabolic state.7,8
Both Foxo3- and Foxo4-null mice are viable.9,10 In contrast, mice that are null for Foxo1 are embryonic lethal at E11 because of impaired vasculogenesis.10,11 Thus, FOXO1 has a specific role in vascular development, which cannot be compensated for by other FOXO family members. FOXO1 is expressed in multiple cell types and tissues during development, including endothelial cells, smooth muscle cells, neural crest cells, cardiomycoytes, adipose tissue, somites, branchial arches, and trigeminal ganglia.10–13
In contrast to the embryonic knockout, widespread somatic deletion of Foxo1 in adult tissues predisposes to the development of vascular bed–specific hemangiomas, an effect that is accentuated by combined deletion of Foxo3 and Foxo4.14 One previous study showed that mice with endothelial cell–specific knockout of Foxo1 are born at term in normal Mendelian ratios and show no gross or metabolic abnormalities.15 However, another study reported that endothelial-specific deletion of Foxo1 is embryonic lethal.16 Conditional deletion of all 3 FOXO factors in the endothelium is compatible with survival and protects against atherosclerosis in low-density lipoprotein receptor knockout mice.17 Mice with lineage-specific knockouts of Foxo1 in other tissues, including bone,18 liver,19 heart,7 T cells,20 and teeth enamel,21 are viable but have wide-ranging phenotypes, implying an important role for this transcription factor in homeostasis.
The effect of FOXO activity on gene expression and cell function is highly cell-type-specific. For example, combined deletion of Foxo1, Foxo3, and Foxo4 in postnatal mice resulted in altered expression of 608 genes in liver endothelial cells and 610 genes in thymocytes, only 11 of which overlapped.14 Moreover, the expression profile differed between FOXO1/O3/O4–deficient liver and lung endothelial cells, as did the proliferative capacity of the cells in response to vascular endothelial growth factor (VEGF).14
In the current study, we have generated several genetic mouse models to address the role of endothelial FOXO1 in homeostasis more definitively. Using a combination of global knockout, endothelial cell–specific knockout, endothelial rescue, and endothelial overexpression mouse models, we provide evidence that endothelial FOXO1 is both necessary and sufficient for viability, and that a balanced level of FOXO1 activity is required for survival. Moreover, we show that a primary role of FOXO1 in the endothelium is to feed back and activate Akt-mTORC1, thus sensitizing cells to the effect of VEGF.
A detailed description of the Methods used in this study is provided in the Online Data Supplement.
Generation of Gene-Targeted and Transgenic Mice
The generation of Foxo1−/−, Foxo1EC−/−, tetracycline (Tet)-triple mutant (TM)-FOXO1, Tet-LacZ, and Foxo1-rescue (Foxo1-res) mice is detailed in the Online Data Supplement. VE-cadherin (VEC)-tetracycline transactivator (tTA) mice were a generous gift from Dr Laura Benjamin. Tie2-Cre (B6.Cg-Tg(Tek-cre)1Ywa/J) and VEC-Cre (B6;129-Tg(Cdh5-cre)1Spe/J) were obtained from Jackson Laboratory. All animal studies were approved by the Animal Care and Use Committee at the Beth Israel Deaconess Medical Center.
Endothelial and nonendothelial cells used in this study are described in the Online Data Supplement. All experiments with primary cell lines were performed between passage 3 and 6. Mouse endothelial cells were harvested and grown as previously described.22 Cells were either taken off or maintained on 10 mg/L of tetracycline. For VEGF treatment, cells were serum-starved in 0.5% FBS for 16 hours and treated with human VEGF-165 (50 ng/mL; Peprotech) for the times indicated.
Human umbilical vein endothelial cells (HUVEC) were infected with adenovirus (Ad)-cytomegalus virus (CMV)-β-galactosidase (Ad-bgal), Ad-wild-type (WT)-FOXO1, Ad-TM-FOXO1, or Ad-CMV-Akt1 (Vector Biolabs) as previously described.23
Transfection of Endothelial Cells With siRNA and shRNA
HUVEC were plated at a density 5×105 cells per 6-cm plate and transfected with 1 of 6 siRNAs against FOXO1 (FOXO1-1 through FOXO1-6) or 3 siRNAs against FOXO3 (FOXO3-1, FOXO3-2, and FOXO3-3) in Opti-MEM (Invitrogen) with lipofectin.23 siRNA sequences are shown in the Online Data Supplement (Online Table I). FOXO1-2, FOXO1-3, FOXO3-2, and FOXO3-3 are identical to those used by Potente et al.24 For shRNA experiments, HUVEC were incubated with lentiviral shRNA against FOXO1 or control shRNA at 20 MOI for 6 hours in the presence of 8 μg/mL polybrene. Forty-eight hours after infection, cells were selected with puromycin (10 μg/mL) for 1 week before they were expanded for experiments.
Quantitative Real-Time Polymerase Chain Reaction Assays
Quantitative polymerase chain reaction (qPCR) assays were performed as previously described.25 Primer sequences are available on request.
Cardiac Morphometric and Pressure–Volume Loop Studies
Heart tissues were harvested and fixed overnight at 4°C in 4% paraformaldehyde and sterile PBS. Fixed hearts were paraffin-embedded and sectioned on the sagittal plane at 6-μm thickness. Paraffin-embedded hearts were examined using a Wild Heerbrugg M400 Photomakroskop (Wild Heerbrugg Ltd). Ventricular volumes were quantified with ImageJ software (National Institutes of Health). Pressure–volume loop analysis was performed as previously reported.26
Western blots were performed as previously described.27 Blots were analyzed using antibodies detailed in the Online Data Supplement. The bands were visualized with a chemiluminescence detection kit (ECl; Amersham Biosciences) and were quantified with ImageJ software (National Institutes of Health).
Blood was collected via cardiac puncture. Serum creatinine levels in VEC-tTA; Tet-FOXO1 mice were measured using the i-STAT Chem8+ cartridge and analyzed using i-STAT analyzer (Abbot Point of Care, Inc). Urine samples were collected directly from the urinary bladder. Urine albumin levels were assayed using an ELISA kit from Exocell. Rescue mice blood samples were assayed by the Division of the Comparative Medicine, Massachusetts Institute of Technology.
Transmission Electron Microscopy
Transmission electron microscopy was performed as previously described.28
LacZ staining was performed as previously described.22
Aortic Ring Assay
Detailed procedures for this assay are provided in the Online Data Supplement.
Analysis of Cell Cycle, Thymidine Uptake, Modified Boyden Chamber, and Scratch Wound Assays
Detailed procedures for these assays are provided in the Online Data Supplement.
Data were expressed as mean±SD of 3 or more independent experiments. The mean statistical difference was determined using a t test with P<0.05 as statistically significant. Data analysis and generation of all graphs were performed in PRISM5 software (GraphPad).
Endothelial-Specific Deletion of Foxo1 Phenocopies the Full Foxo1 Knockout
We wished to determine the extent to which endothelium is responsible for mediating the vascular phenotype in Foxo1−/− mice. To that end, we generated both Foxo1-null mice (Foxo1−/−) and mice in which Foxo1 was conditionally deleted in the endothelium (Foxo1EC−/−; Figure 1A and 1B). FOXO1 protein was not detected in Western blot analyses of E10.5 Foxo1−/− embryos (Figure 6E). Consistent with what has been previously reported, targeted disruption of the Foxo1 gene in mice resulted in embryonic lethality around E11 (Online Table II). Foxo1+/+ and Foxo1−/− embryos were indistinguishable until E9.5 at which time mutant embryos were smaller, lacked a second branchial arch, and often exhibited marked pericardial swelling (Figure 2A–2C). In whole-mount E9.5 to E10.5 embryos stained for CD31, the dorsal aorta appeared thin, underdeveloped, and irregularly formed in Foxo1−/− embryos (Online Figure IA). This was confirmed in Hematoxylin and eosin–stained tissue sections (Online Figure IIA). Intersomitic vessels were similarly poorly developed in Foxo1−/− embryos (Figure 2D; Online Figure IA). The heart was smaller when compared with that of WT controls (Online Figure IIB). The head vasculature of Foxo1−/− embryos was arrested in the primary plexus stage (Online Figure IB and IC). Yolk sacs were pale and had poorly formed vasculature (Figure 2A and 2E).
Foxo1EC−/− mice were generated by crossing Foxo1-LoxP–targeted mice (Foxo1fl/fl) with the Tie2-Cre line. Similar to the full knockout, FOXO1EC−/− embryos died at ≈E11.0 (Online Table II). Cre-mediated excision of Foxo1 in the endothelium of mice resulted in a virtually identical phenotype to that observed with the complete knockout, with the exception that the second branchial arch was preserved (Figure 2F–2J; Online Figures ID–IK and IIC–IIE). Electron microscopy of aortas and intersomitic vessels from E9.5 Foxo1−/− and Foxo1EC−/− mouse aortas did not demonstrate any ultrastructural abnormalities (not shown). Although the Tie2-Cre line is the most commonly used driver strain in endothelial-specific knockout studies, it has been shown to yield Cre-mediated recombination in hematopoietic cells.29 To confirm our data, we bred transgenic mice that express Cre recombinase from VE-cadherin (VEC) regulatory sequences (VEC-Cre) with Foxo1fl/fl mice. The resulting offspring (Foxo1EC-VE−/−) demonstrated an identical phenotype to that of the Foxo1EC−/− mice (Figure IL–IM shows whole mounts at E10). Together, these findings indicate that endothelial expression of FOXO1 is necessary for normal development, and that its absence accounts for the majority of the phenotypic manifestations of the complete Foxo1 knockout.
Endothelial Expression of FOXO1 Partially Rescues the Full Foxo1 Knockout
Our next goal was to establish whether endothelial expression of FOXO1 could rescue the Foxo1−/− phenotype. To that end, we generated transgenic mice with a DNA cassette containing the Tie2 promoter/enhancer coupled to murine Foxo1 cDNA (Figure 1C). The Tie2 promoter/enhancer that we used has been shown to direct integration-independent endothelial-specific expression throughout the vasculature of transgenic mice.30 Two independent Tie2-FOXO1 lines were crossed with Foxo1+/-mice. FOXO1+/- mice hemizygous for the transgene were then crossed with FOXO1+/- mice. In both cases, offspring were obtained that expressed the Foxo1 transgene on a Foxo1-null background. Although the number of rescue mice (Foxo1-res) born was less than the expected Mendelian ratio (Online Table II), those that did survive were grossly indistinguishable from WT littermates (Figure 3A). There was no difference in body weight or weight of individual organs (with the exception of fat) between the rescue mice and WT littermates (Figure 3B). qPCR analysis of various mouse organs demonstrated near-normal mRNA levels of Foxo1 and endothelial-restricted FOXO1 target genes, including endothelial nitric oxide synthase (eNOS; Nos3), angiopoietin 2 (Ang2), and endothelial-specific molecule 1 (Esm1) (Figure 3C). Similarly, qPCR assays of primary mouse lung endothelial cells revealed detectable mRNA levels of Foxo1 and selected target genes (Figure 3D). Tunel staining of various tissues did not reveal any difference in the number of apoptotic cells between rescue mice and WT littermates (Figure 3E shows liver, heart, and lung). Moreover, vascular density was similar in rescue mice and WT controls, as assayed by CD31 staining (Figure 3F shows heart, kidney, and liver). It was previously reported that liver-specific knockout of Foxo1 results in fasting hypoglycemia.19,31 Consistent with these data, blood glucose levels were significantly reduced in overnight-fasted Foxo1-res mice when compared with WT littermates (Online Table III). Also consistent with the phenotype of liver-specific knockout mice, fasting Foxo1-res mice demonstrated reduced mRNA expression of G6pc (a rate-limiting enzyme for gluconeogenesis) and insulin receptor substrate 2 (Irs2) in the liver (Online Figure IIIA). However, hepatic expression of phosphoenolpyruvate carboxykinase 1 (Pck1), another gluconeogenic enzyme whose expression was reduced in liver-specific knockout mice, was unaffected in the Foxo1-res mice (Online Figure IIIA). In a previous study, the conditional knockout of Foxo1 in ameloblasts of the teeth was shown to result in a chalky, white tooth phenotype consistent with enamel hypomaturation and attrition.21 Foxo1-res mice also demonstrated abnormally white, chalky incisors when compared with WT littermates (Online Figure IIIB). Taken together, the data suggest that endothelial FOXO1 can rescue embryonic lethality in otherwise Foxo1-null mice.
siRNA-Mediated Knockdown of FOXO1 in Endothelial Cells Results in G1 Cell Cycle Arrest and Inhibition of VEGF-Mediated Migration and Proliferation
We next studied the mechanism by which deficiency endothelial FOXO1 interferes with vascular development. To that end, we transfected HUVEC with siRNA against FOXO1 (si-FOXO1-1) or control siRNA (si-CTR). As shown in Figure 4A, si-FOXO1 transfection resulted in an 80% decrease in FOXO1mRNA expression. In fluorescence-activated cell sorting (FACS) analyses, FOXO1 knockdown was associated with an increased number of cells in G1 and a corresponding decrease in the numbers in S and G2/M, indicating an arrest in G1 (Figure 4B). In a modified Boyden chamber assay, si-FOXO1 transfection resulted in a significant reduction in VEGF-induced cell migration (54% reduction; Figure 4C). In a scratch wound assay, si-FOXO1–transfected cells also showed a significant decrease in percentage of area covered under both under basal and VEGF-induced conditions (63% and 75% reduction, respectively; Figure 4D). Finally, in proliferation assays, basal and VEGF-mediated thymidine incorporation was reduced 7- and 16-fold, respectively, in si-FOXO1–treated HUVEC (Figure 4E). Similar results were observed in cell cycle, migration, and proliferation assays using 2 other siRNA against FOXO1 (FOXO1-2 and FOXO1-3; Online Figure IV). Thus, FOXO1 plays a role in cell cycle progression and in VEGF-mediated endothelial cell migration and proliferation.
siRNA-Mediated Knockdown of FOXO1 in Endothelial Cells Results in Altered Expression of Many Genes Involved in Vascular Health
Next, we wished to determine whether FOXO1 deficiency results in altered expression of established FOXO1 target genes and other genes that are known to be important for vascular development. To that end, we transfected primary human endothelial cells with si-CTR or si-FOXO1 and assayed the cells for mRNA expression using qPCR. As expected, the expression of ESM1, ANG2, IRS2, BMP2, SOD2, and CITED2 (established FOXO1 targets in endothelial cells) was downregulated (Figure 4F; Online Figures V and VI). Similar results were observed in HUVEC grown to different degrees of confluence (Online Figure VII). Importantly, FOXO1 knockdown also affected the expression of genes implicated in angiogenesis and vessel maturation, including VE-cadherin (CDH5), ephrin B2 (EFNB2), and ALK1 (Figure 4F; Online Figures V and VI). Of particular note were changes in the expression of genes involved in the Notch signaling pathway. For example, FOXO1-deficient endothelial cells demonstrated elevated expression of DLL4 and the downstream Notch target genes, NRARP (NOTCH-regulated ankyrin repeat protein), HES1 and HES2. Neuropilin 1 (NRP1), which is repressed by NOTCH1, was downregulated. Expression of FOXC1, which has been previously shown to induce expression of DLL4, was increased. These data are consistent with overactive Notch signaling (Figure 4F; Online Figures V and VI). si-FOXO1 increased VEGF mRNA expression, but absolute levels remained low (<2 copies mRNA per cell). Finally, consistent with cell cycle arrest at the G1/S boundary, siRNA against FOXO1 resulted in increased expression of p27kip1, p21WAFI/CIP1; decreased expression of MYC and CDK4; decreased expression of cyclins specific to the S and G2 phase of cell cycle, including CyclinE1 (CCNE1), CyclinB1 (CCNB1), CyclinB2 (CCNB1); and decreased expression of proliferating cell nuclear antigen (PCNA; Figure 4G; Online Figures V and VI). In contrast to a previous study showing that FOXO1 inhibits eNOS expression,24 si-FOXO1 did not result in increased eNOS mRNA levels. However, as discussed in the Online Data Supplement (Online Figure VIII), eNOS expression was increased in HUVEC with prolonged lentivirus shRNA-mediated knockdown of FOXO1, as well as in E9.5 Foxo1EC−/− mice. Together, these data suggest that FOXO1 deficiency in endothelial cells leads to altered expression of many genes implicated in vascular development and cell cycle control.
Endothelial Expression of Constitutively Active FOXO1 Leads to Endothelial Overgrowth, Vascular Occlusion, Cardiac Failure, and Death
Previous studies have shown that somatic deletion of Foxo1 (particularly when associated with the loss of an Foxo3/04) leads to the development of hemangiomas.14 Together with a plethora of published data implicating a role for FOXO1 as an antiproliferative and proapoptotic gene, these findings suggested that increased activity of FOXO1 in the intact endothelium might lead to widespread endothelial death. To test this hypothesis, we used homologous recombination in the hypoxanthine guanine phosphoribosyl transferase (Hprt) locus to generate transgenic mice that carry a tetracycline-responsive promoter (Tet) coupled to a gene that encodes constitutively active human FOXO1 in which the 3 Akt phosphorylation sites have been mutated to alanines (TM-FOXO1). These mice (Tet-TM-FOXO1) were crossed with endothelial-specific tTA mice (VEC-tTA) to generate inducible (Tet-Off) binary transgenic mice that express TM-FOXO1 in the endothelium on withdrawal of tetracycline from the drinking water (Figure 1D). Because commercially available FOXO1 antibodies perform poorly in immunohistochemistry, we decided to monitor efficacy of our binary system by generating a separate line of mice in which the Hprt locus was targeted with the tetracycline promoter coupled to LacZ (Tet-LacZ). Because both Tet-TM-FOXO1 and Tet-LacZ are inserted as single copies into an identical locus, LacZ expression should serve as a surrogate marker for TM-FOXO1 expression. Analysis of VEC-tTA; Tet-LacZ mice revealed minimal leakage of expression in mice on tetracycline and endothelial-restricted inducible expression off tetracycline (Online Figure IX). VEC-tTA; Tet-TM-FOXO1 mice that were maintained on tetracycline demonstrated low-level expression of human FOXO1 in various organs (<5 copies per 1×106 18S copies; Online Figure XA). Seven days after withdrawal of tetracycline from the drinking water, FOXO1 mRNA levels were significantly induced. TM-FOXO1 expression also resulted in increased expression of established endothelial cell–restricted FOXO1 target genes in several organs, including Esm1, Ang2, p21Cip1 (but not p27kip1), cyclin G2 (Ccng2), and Bcl6b (Online Figure XA shows Esm1 and Ang2). Despite the putative role of FOXO1 as a repressor of eNOS, inducible expression of TM-FOXO1 in the endothelium did not alter eNOS mRNA or protein levels in any tissue examined (Online Figure XA shows qPCR data). Nor did TM-FOXO1 expression affect the expression of the cell adhesion molecules, vascular cell adhesion molecule 1 (Vcam1) and intercellular molecule 1 (Icam1) (data not shown). Endothelial cells were isolated from the lungs and hearts of VEC-tTA; Tet-TM-FOXO1 mice and grown in the absence or presence of tetracycline. On removal of tetracycline from the culture medium, there was significant induction of FOXO1 and FOXO1 target genes, including Esm1, Ang2, p21Cip1, cyclin G2 (Ccng2), Gadd45a, and Sod2 (Figure 5A shows Esm1 and Ang2). Induction of TM-FOXO1 had no effect on eNOS mRNA levels in heart ECs and actually increased eNOS expression in lung ECs (Figure 5A). There was no change in Vcam1 or Icam1 expression (data not shown).
Inducible expression of TM-FOXO1 in the endothelium resulted in lethality after 7 days. The dry weight of the lung, but not other organs, was increased (not shown). Morphometric analyses of the heart revealed left ventricular enlargement (Figure 5B and 5C). Pressure–volume loop experiments demonstrated increased peripheral vascular resistance, lower mean arterial pressure, and decreased cardiac output (Figure 5D). There was no evidence of endothelial cell hyperproliferation or increased apoptosis, as measured by Tunel staining and bromodeoxyuridine (BrdU) incorporation/CD31 staining, respectively (Online Figure XB and XC). A previous study implicated a role of endothelial FOXO proteins in monocyte recruitment.17 However, CD45 staining of various mouse tissues did not reveal any difference in the number of CD45+ cells between Tet-On and Tet-Off mice (Online Figure XIA shows heart, kidney, and liver). Moreover, Hematoxylin and eosin stains did not demonstrate any evidence of inflammation (Online Figure XIB).
Electron microscopy revealed enlarged endothelial cells with large nuclei, abundant rough endoplasmic reticulum, and occasional multilamellar basement membrane in capillaries, resulting in a narrowed capillary lumen with trapped red blood cells (Figure 5E; Online Figure XII). There was focal loss in the fenestrae of the renal glomerular endothelium. The latter finding was associated with albuminuria and increased blood creatinine (Online Figure XD). Electron microscopy of organs from an unrelated Hprt-targeted mouse line VEC-tTA; Tet-placental growth factor (PlGF) mice, in which PlGF is inducibly expressed in the endothelium on withdrawal of tetracycline, did not reveal a similar phenotype, arguing against nonspecific effects of transgenic expression in the Hprt locus (data not shown).
To determine whether TM-FOXO1 expression in endothelial cells impaired sprouting angiogenesis, we performed aortic ring assays using samples from VEC-tTA; Tet-TM-FOXO1 mice. As demonstrated in Figure 5F, the inducible expression of TM-FOXO1 had no effect on sprouting distance or area. Together with the BrdU incorporation/CD31 staining of adult tissues, these findings argue against a significant antiproliferative role of FOXO1 in the endothelium.
Overexpression of Constitutively Active FOXO1 in Endothelial Cells Results in Increased Cell Size and Activation of Akt-mTORC1
The electron microscopy results suggested that the endothelial cells of mice overexpressing constitutively active FOXO1 in the endothelium were enlarged (thus compromising the lumen of small vessels). In some, but not all experiments, endothelial cells isolated from the heart and lung of VEC-tTA; Tet-TM-FOXO1 mice demonstrated increased cell volume. This lack of reproducibility may reflect a loss of larger cells during the isolation process. However, infection of HUVEC with adenovirus expressing TM-FOXO1 (Ad-TM-FOXO1) resulted in a significant increase in cell volume and size (Figure 6A). Cell volume is controlled primarily by the Akt-mTORC1 signaling pathway.32 Previous studies in nonendothelial cells have shown that FOXO1 may feed back to activate Akt, while inhibiting mTORC1, thus uncoupling Akt and mTORC1 signaling.33–35 In Western blot analyses of HUVEC, phospho (p)-Akt was elevated in cells infected with adenovirus expressing WT FOXO1 (in which the phosphorylation sites are intact) when compared with Ad-bgal–infected controls and was further increased in cells expressing the phosphorylation-resistant TM-FOXO1 (Figure 6B). The phosphorylation status of S6 and S6K is commonly used to evaluate mTORC1 activity.36 In Western blots, Ad-WT-FOXO1– and Ad-TM-FOXO1–infected HUVEC demonstrated progressively increased levels of p-S6 and p-S6K (Figure 6B). Interestingly, total levels of S6 were increased in TM-FOXO1–expressing cells, whereas total levels of S6K were decreased. To determine whether the effect of TM-FOXO1 (ie, activation of Akt and mTROC1) was specific to HUVEC, we repeated these experiments using other types of endothelial cells, as well as nonendothelial cells. As shown in Online Figure XIII, TM-FOXO1 induced both p-Akt and p-S6K in human coronary artery endothelial cells, human dermal microvascular endothelial cells, and human coronary artery vascular smooth muscle cells. In contrast, TM-FOXO1 inhibited p-S6K in HEK cells. Taken together, these results suggest that FOXO1 feeds back to activate Akt. However, in contrast to what has been reported in other cell types (and what we observed in HEK cells), Akt is free to activate the mTORC1 pathway. Thus, overexpression of constitutively active, nuclear FOXO1 may result in Akt-mTORC1–mediated cell growth.
Akt-mTORC1 Pathway Is Attenuated in FOXO1-Deficient Endothelial Cells
On the basis of the results of the TM-FOXO1–overexpressing cells, we hypothesized that FOXO1 deficiency may be associated with blunted Akt signaling in endothelial cells. Indeed, in Western blot analyses of HUVEC, siRNA against FOXO1 resulted in a significant reduction in basal and VEGF-inducible levels of p-Akt, p-S6, and p-S6K (Figure 6C; Online Figure XIVA–XIVC). By contrast, siRNA-mediated knockdown of FOXO3 had no effect on Akt levels (Online Figure XIVD). Similar results with si-FOXO1 were observed with human coronary artery endothelial cells and human dermal microvascular endothelial cells (Online Figure XIVE and XIVF). FOXO1 knockdown in HUVEC did not affect VEGF-mediated phosphorylation of VEGF receptor 2 (Online Figure XIVG), suggesting that FOXO1 exerts its effect on Akt signaling distal to the VEGF receptor. To determine whether overexpression of Akt could rescue the phenotype of FOXO1-deficient cells, we infected si-CTR– and si-FOXO1–tranfected cells with constitutively active Akt. In these experiments, siRNA against FOXO1 resulted in a marked diminution in thymidine uptake in the absence but not in the presence of CA-Akt (Figure 6D). We harvested E10.5 Foxo1−/− embryos for protein and performed Western blot analyses for Akt and mTORC1 activation. As shown in Figure 6E and Online Figure XVA, homozygous knockout embryos demonstrated reduced p-Akt and p-S6, despite an increased total level of S6. Western blots of Foxo1EC−/− embryos demonstrated a trend toward reduced p-Akt (some blots, such as the one shown in Figure 6F demonstrated marked reduction) and significantly reduced levels of p-S6 (Figure 6F; Online Figure XVB). Together, these data suggest that FOXO1 deficiency is associated with a loss of feedback activation of Akt-mTORC1.
Previous studies have implicated several FOXO1-responsive genes in mediating feedback activation of Akt, including Sestrin 3 (Sesn3),33 Rictor,33 and the pseudokinase, TRB3.35 In qPCR assays, siRNA-mediated knockdown of FOXO1 resulted in downregulation of SESN3 (Figure 6G). Expression of TM-FOXO1 in endothelial cells resulted in increased expression of RICTOR and reduced expression of TRB3 but no change in SESN3 mRNA levels (Figure 6G). Finally, there was an increase in both Rictor and Sesn3 mRNA levels in endothelial cells isolated from the hearts of VEC-tTA; Tet-TM-FOXO1 mice (Figure 6G).
Previous studies support an important role for FOXO1 in vascular homeostasis. Most importantly, 2 independent groups have shown that Foxo1−/− embryos die at E11 as a consequence of incomplete vascular development.10,11 We have confirmed these findings in the present study. The vascular phenotype of Foxo1−/− mice raises the important question as to whether endothelial FOXO1 is primarily responsible for the embryonic lethality of the knockout mouse. We and others have previously shown that FOXO1 is expressed in endothelial cells and is functionally relevant. Here, we show that endothelial-specific deletion of Foxo1 essentially phenocopies the full knockout, the one obvious difference being in the formation of the branchial arches. By contrast, embryonic development is not compromised in mice with lineage-specific deletions of Foxo1 in bone,18 liver,19 heart,7 T cells,20 and teeth enamel.21 Collectively, these findings suggest that embryonic lethality in the full knockout is attributed primarily to the loss of FOXO1 in the endothelium. Finally, we were able to show that endothelial expression of FOXO1 can rescue Foxo1−/− mice. Although the number of rescue mice born was less than the expected Mendelian ratio, the findings indicate that FOXO1 is also sufficient for embryonic survival. Taken together, our study supports a central role for endothelial-derived FOXO1 in vascular homeostasis.
The only detectable difference in phenotype between the full knockout and the endothelial-specific knockout of Foxo1 was the absence of the second branchial arch in the Foxo1−/− mice. It was previously shown that FOXO1 is expressed in neural crest cells migrating toward branchial arches at E8.5 and becomes localized to the first and second branchial arches between E8.5 and E9.5.11 Interestingly, other studies have demonstrated that neural crest cell invasion of branchial arch 2 involves an interaction between NRP1-expressing neural crest cells and VEGF-expressing ectoderm in the second arch.37,38 Given our findings that FOXO1 is required for VEGF signaling, it is tempting to speculate that loss of FOXO1 in neural crest cells in Foxo1−/− mice interferes with chemoattraction-mediated invasion of branchial arch 2.
FOXO transcription factors are widely considered to have antiproliferative and proapoptotic functions. According to the canonical pathway, growth factors, such as insulin or insulin-like growth factor, activate PI3K and Akt. Akt, in turn, phosphorylates FOXO proteins, leading to their nuclear exclusion and downregulation of FOXO-dependent death genes. Consistent with this model, Paik et al14 reported that the somatic deletion of 3 Foxo genes (Foxo1, Foxo3, and Foxo4) resulted in a proliferative endothelial phenotype in some (but interestingly not all) organs. Endothelial cells from the liver of Foxo1/o3/o4-null mice (but not the lung) demonstrated enhanced VEGF- and fibroblast growth factor–stimulated proliferation.14 In another study, conditional knockout of all 3 FOXO factors in the endothelium resulted in increased proliferation, reduced cellular senescence, and decreased apoptosis in aortic endothelial cells.17 Together, these data support the notion that FOXO proteins are antiproliferative and proapoptotic in the endothelium.
How, then, do we reconcile these published findings with the observation that whole-body or endothelial-specific deletion of Foxo1 leads to impaired vascular development and not overgrowth of endothelial cells? The observation that FOXO proteins exhibit vascular bed–specific properties in adult mice14 raises the possibility that FOXO1 functions differently in embryonic and postnatal endothelium. An alternative explanation is that FOXO1 alone is not antiproliferative, but rather exerts this function only when FOXO3 and FOXO4 are absent. In fact, our data suggest that endothelial FOXO1 has the opposite effect of promoting basal and VEGF-stimulated migration and proliferation, as well as cell cycle progression through G1. These results are at odds with those of Potente et al,24 who demonstrated that siRNA-mediated FOXO1 knockdown in HUVEC significantly increased endothelial migration, tube formation in the Matrigel assay and sprouting activity of endothelial spheroids. However, in our hands, the si-FOXOs used in the latter study had an inhibitory effect on endothelial cell migration, cell cycle, and proliferation (Online Figure IV). Moreover, lentivirus shRNA-mediated knockdown of FOXO1 also inhibited VEGF-mediated proliferation of endothelial cells (Online Figure VIIIE and VIIIF). At this time, we cannot explain the reason for the discrepant findings. However, as we discuss below, we think that our in vitro and in vivo data strongly support a model in which FOXO1 is necessary for Akt-mTORC1 signaling and growth/proliferation in endothelial cells.
Some,24,39 but not all,14,15,40 studies have shown that eNOS is negatively regulated by FOXO1. Lentiviral shRNA-mediated knockdown of FOXO1 in HUVEC increased the expression of eNOS, as did endothelial cell–specific knockout of Foxo1 at E10.5. Interestingly, overexpression of constitutively active FOXO1 also yielded variable results. For example, adenovirus-mediated high-level expression of constitutively active TM-FOXO1 in HUVEC (>4000 copies/cell) led to marked reduction in eNOS mRNA levels (Online Figure VIIIH), whereas mice overexpressing TM-FOXO1 in the endothelium (<70 copies/cell) demonstrated normal tissue levels of eNOS mRNA and protein. Together, these data suggest that the effect of FOXO1 levels on eNOS expression is highly context dependent and may vary according to absolute FOXO1 levels, to vascular bed type and to in vitro versus in vivo settings.
The finding that mice with an endothelial deletion of all 3 FOXO proteins are viable,17 whereas those with a single deletion of Foxo1 die at E11 indicates that FOXO3 and FOXO4 deficiency rescues the FOXO1 defect. It is possible that the hyperproliferative effect of the combined Foxo1/03/04 knockout compensates for the vascular phenotype in Foxo1−/− mice. Indeed, we showed that FOXO3 knockdown in endothelial cells partially reverses the inhibitory effect of FOXO1 deficiency on VEGF-mediated proliferation and rescues the expression of several genes that are critical for sprouting angiogenesis (Results in the Online Data Supplement; Online Figure VI). The data indicate that the difference between single knockout of Foxo1 and triple knockout of Foxo1/03/04 in the endothelium of mice is explained, at least in part, by the nonredundancy of FOXO factors.
The phenotype of mice expressing a constitutively active form of FOXO1 also argues against an antiproliferative function of FOXO1 in endothelial cells, at least in the postnatal period. These animals demonstrated normal CD31 counts and BrdU uptake in multiple vascular beds. Moreover, there was no evidence of increased apoptosis, as measured by Tunel assay. Tissues from the TM-FOXO1-expressing mice revealed normal eNOS mRNA and protein levels, arguing against a role of NO deficiency in mediating the increased peripheral resistance. Instead, electron microcopy of various organs suggested that endothelial cells were enlarged to such an extent that they were impinging on the blood vessel lumen, in some cases occluding small capillaries. This effect likely accounts for the increased vascular resistance and reduced cardiac output observed in pressure–volume loop studies. Consistent with the in vivo data, TM-FOXO1 expression in cultured endothelial cells resulted in increased cell size and activation of Akt-mTORC1, which is the principle signaling pathway for cell growth.
Recent studies have demonstrated that in nonendothelial cells, transcriptionally active FOXO1 activates Akt, creating a negative feedback loop. For example, in cardiomyocytes, forced expression of FOXO1 triggers Akt phosphorylation via a calcineurin/protein phosphatase 2A (PP2A)-dependent mechanism.34 Moreover, Ad-TM-FOXO1 delivery to the liver of mice resulted in a paradoxical increase in p-Akt.35 These findings, together with ours, indicate that FOXO1 engages in a feedback loop whereby in nutrient- or growth factor–depleted states, nuclear (unphosphorylated) FOXO1 activates Akt, thus preventing the complete extinction of Akt signaling and sensitizing the cell to subsequent growth factor signals. At the same time, the increased p-Akt will lead to phosphorylation and inactivation of FOXO1 (in other words, although the FOXO1-Akt arc is positive, the FOXO1-Akt-FOXO1 feedback loop is negative). Thus, it has been hypothesized that FOXO1 provides acute but not long-term reprieve from metabolic stress/starvation.17
Additional studies have provided insights into the mechanisms that underlie FOXO1-mediated activation of Akt. First, FOXO1 has been shown to induce Sestrin 3 expression, which in turn inhibits mTORC1.33 mTORC1 normally feeds back to inhibit Akt.41 Thus, the net effect of FOXO1-stimulated Sestrin 3 is to induce p-Akt levels. Second, FOXO1 may increase the expression of Rictor.33 Rictor forms part of the mTORC2 complex. mTORC2 activates Akt (both directly and by reducing the pool of mTORC1 by competing for mTOR). Therefore, the net effect of FOXO1-stimulated Rictor is to induce p-Akt levels. Finally, FOXO1 also been shown to suppress the expression of Trb3.35 Trb3, in turn, has been shown to inhibit Akt activity (without affecting mTORC1).42 Here, we have shown that in human endothelial cells, TM-FOXO1 increases RICTOR expression and dramatically inhibits TRB3 expression, whereas in mouse endothelial cells, TM-FOXO1 expression increases both Rictor and Sestrin 3 (Sesn3) mRNA expression.
FOXO1-mediated induction of Sestrin 3 and Rictor are associated not only with increased p-Akt but also with reduced mTORC1 activity. However, in endothelial cells (as well as vascular smooth muscle cells), FOXO1 promotes activation of both Akt and mTORC1, as evidenced by increased phosphorylation of S6 and S6K. Thus, a more likely explanation is the profound repression of Trb3, which activates Akt independently of any effects on mTORC1. Alternatively, mutual repression between mTORC1 and mTORC2 may be cell-type-specific,43 and absent in endothelial cells. In this case, FOXO1-mediated induction of Rictor may result in upregulation of both Akt and mTORC1. Regardless of the underlying mechanism, our data suggest that in contrast to certain other cell types, FOXO1 expression in endothelial cells does not uncouple Akt and mTORC1 activities. mTORC1 plays an important role in regulating cell growth by activating protein synthesis and suppressing autophagy.36,44 Thus, FOXO1-mediated activation of Akt and secondary activation of mTORC1 may account for the increase in endothelial cell size in TM-FOXO1–expressing endothelial cells.
Our finding that TM-FOXO1 activates Akt and mTORC1 raised the distinct possibility that FOXO1 deficiency inhibits p-Akt-mTORC1 signaling. To test this possibility, we examined FOXO1-deficient HUVEC, as well as Foxo1−/− and Foxo1EC−/− embryos, and found that indeed p-Akt and mTORC1 activity are significantly reduced. These data may explain why siRNA-mediated knockdown of FOXO1 attenuates VEGF signaling and proliferation. More importantly, they may provide an explanation for the lethal vascular phenotype of Foxo1−/− mice.
In summary, we have provided evidence that FOXO1 in endothelial cells feeds back to activate Akt-mTORC1 (Figure 7A). According to this model, there exists an optimal FOXO1 range in which endothelial cells are most responsive to growth factor signaling. When FOXO1 levels fall below that range, there is a loss of feedback, which results in reduced Akt-mTORC1 signaling (Figure 7B), G1 arrest, inhibition of proliferation, and reduced mTORC1-mediated metabolism, even in the presence of growth factors (Figure 7C shows proliferation). When FOXO1 levels exceed the normal range, accentuated feedback leads to increased p-Akt-mTORC1 signaling (Figure 7B), with a resulting increase in cell size. Excessively high levels of FOXO1 (especially if uncoupled from inhibition by p-AKT) may over-ride the proproliferative effect of p-Akt and induce G2 arrest and apoptosis (as occurs in Ad-TM-FOXO1–infected HUVEC; Figure 7C). A comparison with FOXO3, which does not activate Akt, shows that the presence or absence of FOXO-Akt feedback has a profound effect on the phenotypic response of endothelial cells to altered FOXO levels (Online Figure XVI).
Our findings raise interesting questions that have important mechanistic and therapeutic implications. What are the paracrine effects of Akt-mTORC1 activity in endothelial cells on other cell types and how do these alter organ function? Does FOXO1 activate Akt in every cell type? In those cells in which FOXO1 does feedback to activate Akt, is p-Akt associated with increased or reduced mTORC1 activity? Cells in which mTORC1 activity is concomitantly induced (eg, endothelial cells) are more likely to grow in response to overactivation of FOXO1, whereas those in which mTORC1 activity is inhibited are more likely to undergo autophagy and cell cycle arrest. Finally, does the FOXO1-Akt-mTORC1 feedback circuit behave differently in endothelial cells from different vascular beds? If so, therapeutic manipulation of FOXO1 activity in the endothelium may yield vascular bed–specific effects. These and related questions are ripe for further study.
Sources of Funding
This work was supported by National Institutes of Health: National Heart, Lung, and Blood Institute, grant HL077348 (W.C. Aird).
In April 2014, the average time from submission to first decision for all original research papers submitted to Circulation Research was 14.38 days.
The online-only Data Supplement is available with this article at http://circres.ahajournals.org/lookup/suppl/doi:10.1161/CIRCRESAHA.115.303227/-/DC1.
- Nonstandard Abbreviations and Acronyms
- angiopoietin 2
- endothelial nitric oxide synthase
- endothelial-specific molecule 1
- hypoxanthine guanine phosphoribosyl transferase
- human umbilical vein endothelial cells
- Akt-mammalian target of rapamycin complex 1
- quantitative polymerase chain reaction
- triple mutant
- tetracycline transactivator
- vascular endothelial growth factor
- Received December 13, 2013.
- Revision received May 27, 2014.
- Accepted May 28, 2014.
- © 2014 American Heart Association, Inc.
- Huang H,
- Tindall DJ
- Sengupta A,
- Molkentin JD,
- Paik JH,
- DePinho RA,
- Yutzey KE
- Castrillon DH,
- Miao L,
- Kollipara R,
- Horner JW,
- DePinho RA
- Hosaka T,
- Biggs WH III.,
- Tieu D,
- Boyer AD,
- Varki NM,
- Cavenee WK,
- Arden KC
- Furuyama T,
- Kitayama K,
- Shimoda Y,
- Ogawa M,
- Sone K,
- Yoshida-Araki K,
- Hisatsune H,
- Nishikawa S,
- Nakayama K,
- Ikeda K,
- Motoyama N,
- Mori N
- Evans-Anderson HJ,
- Alfieri CM,
- Yutzey KE
- Tanaka J,
- Qiang L,
- Banks AS,
- Welch CL,
- Matsumoto M,
- Kitamura T,
- Ido-Kitamura Y,
- DePinho RA,
- Accili D
- Yuan L,
- Janes L,
- Beeler D,
- Spokes KC,
- Smith J,
- Li D,
- Jaminet SC,
- Oettgen P,
- Aird WC
- Abid MR,
- Shih SC,
- Otu HH,
- Spokes KC,
- Okada Y,
- Curiel DT,
- Minami T,
- Aird WC
- Okada Y,
- Yano K,
- Jin E,
- Funahashi N,
- Kitayama M,
- Doi T,
- Spokes K,
- Beeler DL,
- Shih SC,
- Okada H,
- Danilov TA,
- Maynard E,
- Minami T,
- Oettgen P,
- Aird WC
- Yano K,
- Liaw PC,
- Mullington JM,
- Shih SC,
- Okada H,
- Bodyak N,
- Kang PM,
- Toltl L,
- Belikoff B,
- Buras J,
- Simms BT,
- Mizgerd JP,
- Carmeliet P,
- Karumanchi SA,
- Aird WC
- Abid MR,
- Nadeau RJ,
- Spokes KC,
- Minami T,
- Li D,
- Shih SC,
- Aird WC
- Yano K,
- Gale D,
- Massberg S,
- Cheruvu PK,
- Monahan-Earley R,
- Morgan ES,
- Haig D,
- von Andrian UH,
- Dvorak AM,
- Aird WC
- Schlaeger TM,
- Bartunkova S,
- Lawitts JA,
- Teichmann G,
- Risau W,
- Deutsch U,
- Sato TN
- Fingar DC,
- Salama S,
- Tsou C,
- Harlow E,
- Blenis J
- Ni YG,
- Wang N,
- Cao DJ,
- Sachan N,
- Morris DJ,
- Gerard RD,
- Kuro-O M,
- Rothermel BA,
- Hill JA
- Harrington LS,
- Findlay GM,
- Gray A,
- Tolkacheva T,
- Wigfield S,
- Rebholz H,
- Barnett J,
- Leslie NR,
- Cheng S,
- Shepherd PR,
- Gout I,
- Downes CP,
- Lamb RF
- Du K,
- Herzig S,
- Kulkarni RN,
- Montminy M
Novelty and Significance
What Is Known?
FOXO1 promotes cell death and inhibits proliferation in many cell types.
FOXO1-null mice die at E11 because of impaired vasculogenesis.
Endothelium-specific deletion of FOXO1 was shown to be embryonic lethal in one study but embryonic viable in another.
What New Information Does This Article Contribute?
Endothelial FOXO1 is necessary for embryonic development.
Overexpression of constitutively active FOXO1 in the endothelium results in increased size of endothelial cells (ECs) and occlusion of capillaries.
FOXO1 feeds back to activate Akt-mTORC1 in ECs.
FOXO1 affects multiple facets of cellular function in several organs, but its therapeutic potential is currently limited by a need to untangle its occasionally paradoxical, cell-type–specific effects. To define the role of FOXO1 in the endothelium, we generated several mouse models with altered FOXO1 activity in the endothelium. We found that endothelial-specific deletion of FOXO1 phenocopied the full knockout of FOXO1 (with the exception of branchial arch development), and endothelial expression of FOXO1 rescued FOXO1-null mice. Expression of a constitutively active form of FOXO1 in the endothelium of mice resulted in increased cell size, occlusion of capillaries, increased peripheral vascular resistance, and heart failure. Finally, we found that FOXO1 activates Akt and mTORC1 in ECs and that knockdown of FOXO1 in ECs results in marked inhibition of basal and vascular endothelial growth factor–induced Akt-mTORC1 signaling. In nonvascular cells, FOXO1 activated Akt but not mTORC1. These findings may explain why mice that are null for FOXO1 develop a lethal vascular phenotype. Moreover, the data support a model in which FOXO1-mediated feedback activation of Akt in ECs maintains growth factor responsive Akt/mTORC1 activity within a homeostatic range.