Grb-2–associated binder 1 (Gab1) regulates postnatal ischemic and VEGF-induced through the protein kinase A–endothelial NOS pathway

Yao Lua,1, Yan Xionga,1, Yingqing Huoa, Jingyan Hanb, Xiao Yangc, Rongli Zhangd, De-Sheng Zhue, Stefan Klein-Heßlingf, Jun Lie, Xiaoyu Zhanga, Xiaofan Hana, Yanli Lia, Bin Sheng, Yulong Heg, Masabumi Shibuyah, Gen-Sheng Fengi, and Jincai Luoa,2

aLaboratory of Vascular Biology, Institute of Molecular ; and bDepartment of Integration of Chinese and Western Medicine, School of Basic Medical Sciences, Tasly Microcirculation Research Center, Peking University, Beijing 100871, China; cGenetic Laboratory of Development and Disease, Institute of Biotechnology, Beijing 100071, China; dNon-Human Primate Center, Institute of Molecular Medicine; and eAnimal Center, Peking University, Beijing 100871, China; fDepartment of Molecular Pathology, Institute of Pathology, University of Würzburg, D-97080 Würzburg, Germany; gLaboratory of Vascular and Cancer Biology, Model Animal Research Institute, Nanjing University, Nanjing 210061, China; hDepartment of Molecular Oncology, Tokyo Medical and Dental University, Tokyo 113-8510, Japan; and iDepartment of Pathology, School of Medicine, University of California at San Diego, La Jolla, CA 92093

Edited* by Napoleone Ferrara, Genentech, South San Francisco, CA, and approved January 5, 2011 (received for review June 30, 2010) The intracellular signaling mechanisms underlying postnatal an- phosphorylation and association with PI3K and the tyrosine phos- giogenesis are incompletely understood. Herein we show that phatase Shp2 (12). Recent in vitro studies showed that Gab1 is es- Grb-2–associated binder 1 (Gab1) plays a critical role in ischemic sential for VEGF signaling in promoting cell migration, survival, and VEGF-induced angiogenesis. -specific Gab1 KO and tube formation via interacting with Shp2 and p85 (13, 14). (EGKO) mice displayed impaired angiogenesis in the ischemic hin- However, so far, no direct in vivo evidence demonstrates a role of dlimb despite normal induction of VEGF expression. Matrigel plugs Gab1 in angiogenesis. with VEGF implanted in EGKO mice induced fewer capillaries than Homozygous disruption of the Gab1 gene results in embryonic those in control mice. The vessels and endothelial cells (ECs) de- lethality with multiple defects in the placenta and , along with rived from EGKO mice were defective in vascular sprouting and abnormal liver growth (15, 16). To study the functions of Gab1 in tube formation induced by VEGF. Biochemical analyses revealed angiogenesis in vivo, we created endothelium-specific Gab1-KO MEDICAL SCIENCES a substantial reduction of endothelial NOS (eNOS) activation in (EGKO) mice. EGKO mice were viable but displayed severe Gab1-deficient vessels and ECs following VEGF stimulation. Inter- defects in postnatal - and VEGF-induced angiogenesis. estingly, the phosphorylation of Akt, an enzyme known to pro- Here, through characterizing EGKO mice and their isolated mote VEGF-induced eNOS activation, was increased in Gab1- vessels and ECs, we show that Gab1 mediates VEGF-induced deficient vessels and ECs whereas protein kinase A (PKA) activity eNOS activation in endothelial tube formation via a Shp2-PKA– was significantly decreased. Introduction of an active form of PKA dependent signaling pathway. rescued VEGF-induced eNOS activation and tube formation in EGKO ECs. Reexpression of WT or mutant Gab1 molecules in EGKO Results ECs revealed requirement of Gab1/Shp2 association for the activa- Gab1 Mediates Flow Recovery and Collateralization Following tion of PKA and eNOS. Taken together, these results identify Gab1 Limb Ischemia. In our previous studies (17), the Cre-loxP system as a critical upstream signaling component in VEGF-induced eNOS was successfully used to analyze tissue-specific functions of Gab1. activation and tube formation, which is dependent on PKA. Of To study the role of Gab1 in ECs, we generated EGKO mice flox flox note, this pathway is conserved in primary human ECs for VEGF- (Gab1 / :Tie2-Cre/+) by crossing mice carrying a floxed Gab1 induced eNOS activation and tube formation, suggesting consider- allele (17) with a transgenic mouse line expressing Cre under the able potential in treatment of human ischemic diseases. control of a Tie2 endothelium-specific promoter (18). PCR flox analysis showed a Cre-mediated recombination of the Gab1 NO | Gab1 protein complex | collateralization | Tie2-Cre | PKA substrate allele specifically in ECs in various organs (Fig. 1A). Immunoblot analysis demonstrated that Gab1 protein expression was de- schemia-induced neovascularization is critical for blood flow creased by approximately 75% in ECs isolated from EGKO mice flox + Irecovery and tissue injury repair in ischemic tissues (1, 2). An- compared with those from control mice (in this study, Gab1 / : giogenesis is a complex process that includes endothelial pro- Cre/+ mice were used as controls in most experiments), whereas liferation, migration, and tube formation, involving several growth the expression of Gab2 or Gab3 remains unchanged (Fig. 1B). factors and related signaling networks. Among them, VEGF sig- Homozygous EGKO mice were born normally and were fertile naling is a crucial step (1–3). The endothelial NOS (eNOS) is (Table S1) without apparent gross abnormality. In addition, the critical for VEGF-triggered postnatal angiogenesis (4, 5). Several vasculature in organs examined is comparable between EGKO protein kinases, such as Akt, AMP-activated protein kinase and control mice (Fig. S1), suggesting that Gab1 is dispensable (AMPK), and protein kinase A (PKA), are known to activate for embryonic vascular development. To evaluate the role of eNOS (6). Among them, Akt has emerged as a central regulator for eNOS activation by VEGF (7). Interestingly, in endothelial cells (ECs) from Akt1-KO mice, the eNOS activation is still in- Author contributions: J. Luo designed research; Y. Lu, Y.X., Y. Huo, R.Z., X.Z., X.H., and B.S. performed research; J.H., X.Y., D.-S.Z., S.K.-H., J. Li, Y. Li, M.S., and G.-S.F. contributed new duced to a certain extent by VEGF, strongly suggesting the in- reagents/analytic tools; Y. Lu, Y.X., R.Z., B.S., Y. He, and J. Luo analyzed data; and J. Luo volvement of other enzymes, such as PKA, in this process (8, 9). wrote the paper. So far, however, little is known about the role of PKA in VEGF- The authors declare no conflict of interest. induced eNOS activation and thereby angiogenesis, although *This Direct Submission article had a prearranged editor. VEGF has been shown to activate cAMP/PKA signaling (10, 11). 1Y. Lu and Y.X. contributed equally to this work. – Grb-2 associated binder 1 (Gab1), a scaffolding adaptor, belongs 2To whom correspondence should be addressed. E-mail: [email protected]. to a family of signaling proteins consisting of Gab1, Gab2, and Gab3 This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. (12). Upon stimulation by growth factors, Gab1 undergoes tyrosine 1073/pnas.1009395108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1009395108 PNAS Early Edition | 1of6 Downloaded by guest on October 1, 2021 a poor recovery of flow in the ischemic limb. We then used an angiographic assay to test whether the underperfused status in the hindlimbs of EGKO mice was a result of reduced collater- alization. Consistent with the blood flow findings, at 4 wk after ligation, the ischemic hindlimbs of control mice showed a mod- erate establishment of collateral vessels whereas those of EGKO mice displayed impaired development of collateral vessels around the ligation site (Fig. 1E). These data suggest that Gab1 is required for collateral vessel establishment in response to acute hindlimb ischemia.

Gab1 Mediates Ischemic and VEGF-Induced Angiogenic Responses. To explore the mechanisms underlying collateral defects in the is- chemic hindlimbs of EGKO mice, we performed histochemical analyses by H&E and platelet–EC adhesion molecule-1 (i.e., CD31) staining to measure capillary density. The capillary density in the nonischemic anterior tibial skeletal muscles was comparable between EGKO and control mice (Fig. 2 A and B). However, 6 d after ligation, an obvious formation of new capillary tubes was observed in the ischemic anterior tibial skeletal muscles of control mice, but not in those of EGKO mice, and was accompanied by evident muscle atrophy. Quantification of CD31 staining revealed that the capillary density was approximately sixfold lower in EGKO than in control mice (Fig. 2B). In addition, lack of local invasion of leukocytes, a hallmark of ischemic injury repair (20), Fig. 1. EGKO mice displayed defective reperfusion and collateralization in few centrally situated nuclei, and weak COX and NADH activities response to limb ischemia. Genotypic analysis of EGKO mice were conducted (markers of mitochondrial respiratory chain function) were ob- using PCR (A) and immunoblot (B). (A) Genomic DNA from EGKO and control served in the ischemic muscles of EGKO (Fig. 2A), but not control mice were subjected to PCR with the use of primers for genotyping. The mice, showing a defect in collateralization and regeneration of fl flox/flox larger fragment ( ox, 630 bp) indicates Gab1 allele, and the smaller muscle in EGKO mice. These results indicate that Gab1 is im- fragment (Δ, 150 bp) indicates WT allele. Left: Representative PCR results with DNA from organs and purified ECs of EGKO mice. Right: Representative portant for the ischemic angiogenic response. PCR results with DNA from livers, purified hepatocytes (Hep), and liver si- Because ischemic angiogenesis induced by femoral artery li- nusoid ECs (LSEC) of EGKO and control mice. (B) ECs isolated from EGKO and gation is VEGF-dependent (3, 4), we assessed VEGF expression control mice were lysed and the expression of Gab1∼3 were analyzed using in ischemic muscles. VEGF expression was strikingly induced in specific antibodies with GAPDH as a loading control. (C–E) Phenotypic the ischemic anterior tibial skeletal muscles after femoral ligation, analysis of EGKO and control mice at 4 wk after femoral artery ligation of at comparable level between EGKO and control mice (Fig. 2C). left hindlimbs. (C) Left: The ischemic hindlimb of EGKO mice displayed ne- This finding indicates a defect in Gab1-mediated VEGF re- crosis (arrow), whereas that of control mice did not. Right: Clinical scores as sponsiveness in EGKO mice. To confirm this, we assessed VEGF- an index of severity of limb ischemia, based on published standard (26): induced angiogenesis in vivo by using a s.c. Matrigel implant as- Control, n = 14; EGKO, n = 10. **P < 0.01 compared with control. (D) Left: Serial laser Doppler analysis of blood in hindlimbs of EGKO and say. Histochemical analysis showed that Matrigel plugs with control mice. Note that ischemic hindlimbs of EGKO displayed a poor blood VEGF in the control mice induced more CD31-positive cellu- perfusion (arrow). Right: Quantitative analysis of blood flow using per- larity and channels containing red blood cells than those in centage of the ischemic limb relative to the control limb. Control, n = 10; EGKO mice (Fig. 2D). The difference of functional microvas- EGKO, n =7.**P < 0.01 compared with control. (E) Angiographic analysis of culature formation in Matrigel plugs between EGKO and control ischemic (Ligation) and nonischemic (Non-Ligation) hindlimbs in EGKO and mice was visualized by blood perfusion of FITC-conjugated dex- control mice. Asterisks indicate sites of femoral artery ligation. Arrows in- tran (Fig. 2D). These results demonstrate that Gab1-mediated dicate newly established arteries and recovered collateral circulation in is- endothelial signaling is essential to the VEGF-induced angio- chemic hindlimb of control but not EGKO mice. genic response. Subsequently, we isolated aortic rings from EGKO and control mice and cultured in Matrigel to compare Gab1 in postnatal angiogenesis, we first analyzed retinal angio- their vascular responses to VEGF ex vivo. In the presence of genesis in EGKO mice. By postnatal day 5 (P5), compared with VEGF, the aortic rings of EGKO mice produced significantly control mice, retinal capillary growth in EGKO was significantly fewer and shorter vascular sprouts relative to control (Fig. 2E). delayed, which caught up at approximately P15 (Fig. S2). Be- cause retinal angiogenesis is initiated and regulated by hypoxia Gab1 Mediates the Activations of PKA and eNOS, Which Are Required (19), we then examined ischemia-induced angiogenesis in hin- for VEGF-Induced Endothelial Tube Formation. To study the nature dlimbs of EGKO mice created by femoral artery ligation. Four of the alterations in the VEGF-induced angiogenic response in endothelial Gab1-deficient mice, we first characterized ECs weeks after ligation, all control mice completely recovered the isolated from EGKO and control mice. Remarkably, the capacity use of the limb without evident tissue necrosis, whereas 60% to of EGKO ECs for tube formation in response to VEGF was 70% of the EGKO mice had exacerbated clinical symptoms in severely impaired (Fig. 3A), whereas VEGF-induced pro- the ischemic limb, such as reduced spontaneous mobility, distal liferation and survival of ECs were not affected by Gab1 deletion C fl necrosis, or limb loss (Fig. 1 ). Serial examination of blood ow (Fig. S3). Besides, the migratory ability of EGKO ECs was also with laser Doppler imaging in control mice demonstrated that slightly decreased compared with control cells (Fig. 3B). Because the perfusion ratio of ischemic (ligated; left side, Fig. 1D)to the expression levels of the VEGF receptors, VEGFR1/Flt1 and nonischemic (nonligated; right side, Fig. 1D) hindlimbs pro- VEGFR2/Flk1, in ECs were comparable in EGKO and control gressively increased after ligation, indicating a recovery of blood mice (Fig. S4), the previous results indicate a defect in Gab1- flow. By contrast, the increase of perfusion ratio in EGKO mice mediated intracellular signaling in VEGF-induced tube forma- was insignificant even 4 wk after ligation (Fig. 1D), showing tion in EGKO ECs. As eNOS is a critical effector of VEGF

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1009395108 Lu et al. Downloaded by guest on October 1, 2021 MEDICAL SCIENCES

Fig. 2. Endothelial Gab1 deficiency leads to defective ischemia/VEGF-in- duced angiogenic responses. (A) Histological staining analyses of H&E, COX, and NADH for ischemic and nonischemic anterior tibial skeletal muscles from Fig. 3. Gab1-deficient ECs are defective in VEGF signaling and tube for- EGKO and control mice. Solid arrows in the ischemic muscle section from mation. ECs isolated from EGKO and control mice were serum-starved and control mice show inflammatory cell infiltration, and empty arrows indicate stimulated with VEGF (50 ng/mL) for indicated time periods, followed by the centralized nuclei in regenerating muscle cells. (B) Immunohistochemical analyses of in vitro endothelial Matrigel (A) and wound healing (B) assays, analysis of ischemic and nonischemic anterior tibial skeletal muscles from immunoblot (C and D), and NO release measurement (E). (A) Representative EGKO and control mice, using anti-mouse CD31 antibody. Arrows indicate results of Matrigel tube formation assay. Arrows indicate endothelial tube CD31-positive (brown) capillaries. Microvascular density was quantified as branches. (B) Results of endothelial wound healing assay (*P < 0.05 vs. CD31-positive area relative to the entire area. The results represent the control). (C) Immunoblots of phosphorylations of eNOS, Akt, Erk1/2, and mean (± SEM) of six sections from three animals. **P < 0.01 compared with AMPK in EGKO and control ECs. The relative band intensities, which repre- control. (C) Immunoblot analysis of VEGF expression in ischemic and non- sent the average of three independent experiments (the same is true for all ischemic anterior tibial skeletal muscles. GAPDH was used as loading control. of the gel quantification data in the following figures), of phosphorylated (D) Representative results of Matrigel implant assay. Matrigel containing eNOS (P-eNOS) and phosphorylated Akt (P-Akt) are indicated under each VEGF (200 ng/mL) were injected into the abdominal s.c. tissues (300 μL per band. The differences in the levels of P-eNOS and P-Akt of EGKO and control mouse). After 14 d, Matrigel plugs were harvested from EGKO and control groups were significant (P < 0.02). (D) Immunoblots of phosphorylations of mice 20 min after injection of FITC-dextran into tail veins. Upper: Matrigel eNOS and PKA substrate, and of PKA Cα and Cβ subunits in EGKO and plug photographs were taken under natural light. (Insets: Confocal FITC control ECs, pretreated with or without myristoylated PKI 14–22 amide (PKI; images of microvasculature in implanted Matrigel plugs.) Lower: Sections of 10 μM), a specific PKA inhibitor, before VEGF stimulation. Arrows indicate Matrigel stained with CD31. (E) Representative results of ring assay of tho- phosphorylation bands of PKA substrates, which are PKI-sensitive, and sub- racic aortae isolated from EGKO and control mice. Thoracic aortic rings were stantial decrease in EGKO ECs compared with control cells. The difference in embedded in Matrigel containing VEGF (20 ng/mL), incubated for 6 d, and P-eNOS levels of EGKO and control groups was significant (P < 0.01). (E)NO photographed every 3 d. (Scale bars: A and B,50μm; D, 100 μm.) release measurement of EGKO ECs pretreated with or without PKI (10 μM) before VEGF stimulation (**P < 0.01 vs. control.) (F) Representative results of Matrigel tube formation assay of EGKO and control ECs, pretreated with or signaling in the induction of tube formation (5), we assessed the without N-nitro-L-arginine methyl ester (L-NAME; 5 mM), an inhibitor of activation of eNOS by measuring the levels of phosphorylation eNOS, and PKI (10 μM) before VEGF stimulation (50 ng/mL). (G) Immunoblots on Ser1176 (Ser1177 in the human sequence) and NO produc- of phosphorylation of eNOS, Akt, Erk1/2, AMPK, and PKA substrate in EGKO tion in the ECs from EGKO and control mice. Indeed, VEGF- and control aortic rings after VEGF stimulation (100 ng/mL). The differences in the levels of P-eNOS and P-Akt of EGKO and control groups were signif- induced eNOS phosphorylation and NO production in EGKO < fi < < fi icant (P 0.02). Asterisks indicate statistical signi cance (*P 0.05, **P cells were signi cantly lower than in control cells despite their 0.01) for EGKO versus control. comparable basal levels (Fig. 3 C–E). The studies reported thus far suggest that Akt, which is known to function downstream of Gab1-mediated signaling (12), is a central kinase for activating the activation of other kinases known to activate eNOS in other eNOS in VEGF signaling (7). Unexpectedly, the phosphoryla- signaling pathways, AMPK and PKA (6, 8). Indeed, following tion levels of Akt in EGKO ECs were higher than in control cells VEGF stimulation, PKA activation (as measured by a pan-spe- stimulated with VEGF (Fig. 3C). This result led us to examine cific phosphor-PKA substrate antibody, confirmed by using

Lu et al. PNAS Early Edition | 3of6 Downloaded by guest on October 1, 2021 a PKA activation kit) was substantially decreased in EGKO ECs and AMPK activation was also slightly increased compared with control cells (Fig. 3 C and D and Fig. S5A). Consistently, pre- treatment of ECs with a PKA-specific inhibitor suppressed VEGF-induced eNOS phosphorylation and NO production (Fig. 3 D and E). Furthermore, pretreatment of ECs with the inhib- itors of NOS or PKA suppressed VEGF-induced tube formation, validating the roles of eNOS and PKA activation in promoting VEGF-induced tube formation (Fig. 3F). Finally, we compared the phosphorylation levels of eNOS and PKA in aortic rings isolated from EGKO and control mice. Phosphorylation of eNOS and PKA was also significantly decreased in the aortic rings of EGKO mice compared with control mice (Fig. 3G and Fig. S5B), strongly suggesting a pathophysiological association of PKA and eNOS activation in VEGF signaling in vivo. To further confirm the role of Gab1 in mediating VEGF signaling and tube formation, we reconstituted EGKO ECs with Gab1-expressing lentiviral vector. Gab1 reconstitution res- cued the capacity of EGKO ECs for VEGF-induced tube for- mation, and improved the activation of eNOS and PKA (Fig. 4). Similar phenotypic rescues were also observed in EGKO ECs expressing active forms of eNOS and PKA (Fig. 4), strengthen- ing the hypothesis of downstream roles of PKA and eNOS in the Gab1-medated VEGF signaling cascade in the induction of Fig. 4. Rescue of EGKO ECs with Gab1, eNOS, and PKA restores VEGF sig- tube formation. naling and tube formation. The ECs from the control and EGKO mice, or EGKO ECs overexpressing vector, Gab1, constitutively active eNOS (eNOS Gab1 Scaffolds the Formation of a Multiple-Protein Complex, Which Is S1176D) or constitutively active PKA (caPKA) were serum-starved and stim- Important for VEGF-Induced eNOS Activation and Tube Formation. To ulated with VEGF (50 ng/mL) for indicated time periods, followed by im- define the region responsible for Gab1-mediated VEGF signaling munoblot analysis (A and B), NO release measurement (C), and Matrigel cascade in the regulation of eNOS activation, we made constructs tube formation assay (D). (A) Immunoblots of phosphorylations of eNOS with mutations in these sites: Gab1-3YF (Δp85), in which all and PKA substrate, and of Gab1 and PKA Cα, in control cells, parental, three tyrosines for binding the p85 subunit of PI3K were mutated; and vector-, Gab1- or caPKA-rescued EGKO cells after VEGF stimulation. The Gab1-2YF (ΔShp2), which lost the two tyrosines responsible for differences in P-eNOS levels of Gab1- or caPKA-rescued EGKO cells were Δ significant compared with vector rescued EGKO cells (P < 0.01). (B) Immu- Shp2 binding; and PH domain deletion ( PH), in which the noblot of eNOS in parental EGKO cells or vector- or eNOS S1176D-introduced pleckstrin homology domain was deleted. We assessed the effects EGKO cells. (C) Results of NO release measurement of parental EGKO cells, or of expression of these Gab1 mutants on VEGF-induced activa- EGKO cells introduced with Gab1, eNOS S1176D, constitutively active PKA, or tions of eNOS and PKA in EGKO ECs. Compared with Gab1 empty vector, pretreated with or without PKI (10 μM) before VEGF stimu- WT, the ΔShp2 mutant was significantly impaired in restoring the lation. (E) Matrigel tube formation analysis of parental EGKO cells, or EGKO phosphorylation of eNOS and PKA substrates (Fig. 5A), sug- cells introduced with Gab1, eNOS S1176D, constitutively active PKA, or gesting an essential role of the interaction of Gab1 with Shp2 in empty vector, pretreated with or without PKI (10 μM) before VEGF stimu- fi < < mediating the activations of eNOS and PKA by VEGF. To in- lation. Asterisks indicate statistical signi cance (*P 0.05, **P 0.01) for vestigate whether Gab1 physically interacts with Shp2, PKA and EGKO versus control. eNOS, we conducted a coimmunoprecipitation assay with anti- Gab1 or anti-PKA antibodies. Indeed, Gab1 can form a complex fi B subunits signi cantly ablated VEGF-induced eNOS activation with Shp2, PKA, and eNOS in response to VEGF (Fig. 5 ). but slightly enhanced Akt activation (Fig. 6B). Further, down- However, the complex formation is disrupted when Gab1 lost its regulation of Shp2 and PKA significantly attenuated VEGF-in- Shp2 binding sites, suggesting a critical role of Shp2 in the com- duced NO production and tube formation of HUVECs (Fig. 6 C plex formation (Fig. 5C). Consistently, overexpression of ΔShp2 and E). These results clearly demonstrate that Gab1 also mediates mutant inhibited endothelial tube formation induced by VEGF the Shp2–PKA–eNOS pathway in human ECs, and is important (Fig. 5D). In summary, the formation of Gab1–PKA–eNOS for VEGF-induced tube formation. Conversely, because Akt complex mediated via Shp2 is critical for VEGF-induced activa- fi tions of PKA and eNOS as well as endothelial tube formation. knockdown also signi cantly inhibited VEGF-induced eNOS phosphorylation and NO production (Fig. 6 B and C), the role of Gab1-Mediated PKA-Dependent eNOS Activation and Tube Formation Akt cannot be excluded from Gab1-mediated eNOS activation. in Primary Human ECs in Response to VEGF. To further test whether Discussion Gab1 mediates the PKA–eNOS pathway in VEGF-induced tube formation by primary human vascular ECs (HUVECs), we se- A thorough understanding of intracellular signaling mechanisms quentially examined the effects of shRNA-mediated knockdown underlying VEGF-induced postnatal angiogenesis would be in- of Gab1, Shp2, and PKA (the shRNAs of both Cα and Cβ subunits strumental in designing new treatment for human ischemic dis- were used to create double knockdown) on VEGF-induced eNOS eases. It is well accepted that eNOS is an essential downstream phosphorylation and tube formation, with Akt shRNA used as effector of VEGF signaling in promoting postnatal angiogenesis a control. Consistent with the findings in the vessels and ECs (4, 5), whereas the upstream molecules and pathways that acti- isolated from EGKO mice, the down-regulation of Gab1 by vate eNOS have not been fully defined. By using a tissue-specific shRNA significantly reduced eNOS activity in HUVECs, ac- KO approach, we provide genetic evidence that endothelial companied by decreased activation of PKA, but not Akt and Gab1 is critical for postnatal angiogenesis in response to limb AMPK (Fig. 6A and Fig. S5C). Similarly, down-regulation of Shp2 ischemia and VEGF. In addition, this study also reveals Gab1 also inhibited VEGF-induced activations of PKA and eNOS (Fig. as an important upstream signaling modulator that mediates 6B and Fig. S5C). In addition, down-regulation of PKA catalytic VEGF-induced eNOS activation and tube formation.

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1009395108 Lu et al. Downloaded by guest on October 1, 2021 MEDICAL SCIENCES

Fig. 6. Knockdown of Gab1, shp2, or PKA in HUVECs led to defective VEGF Fig. 5. Shp2 binding site in Gab1 is critical for VEGF-induced eNOS and PKA signaling and tube formation. Fives days after infection, shRNA-expressing substrate phosphorylation. The ECs from the control and EGKO mice, or EGKO HUVECs were serum-starved and stimulated with VEGF (50 ng/mL) for indicated ECs overexpressing WT Gab1 or mutants and HUVECs overexpressing vector, time periods, followed by immunoblot analysis (A and B), NO release measure- Gab1 WT or mutants were serum-starved and stimulated with VEGF (50 ng/mL) ment (C), or tube formation assay (D and E). (A) Immunoblots of phosphor- for indicated time periods, followed by immunoblot analysis (A–C) and tube ylations of eNOS, Akt, Erk1/2, AMPK and PKA substrate, and Gab1 in HUVECs formation assay (D). (A) Immunoblot analysis of phosphorylations of eNOS and expressing scrambled or Gab1 shRNAs. The differences in P-eNOS and P-Akt PKA substrate, and of Gab1 in control cells, parental and Gab1 WT, or indicated levels of Gab1 and scrambled shRNA-expressing groups were significant (P < mutant-rescued EGKO cells. The differencesinP-eNOSlevelsofGab1mutant- 0.02). (B) Immunoblots of phosphorylations of eNOS, Akt and PKA substrate,and rescued EGKO cells were significant compared with Gab1 WT-rescued EGKO Akt, Shp2, PKA Cα,andPKACβ in HUVECs expressing scrambled, Shp2, PKA (α cells (P < 0.01). (B) Immunoblot analysis of anti-Gab1 or anti-PKA Cα immuno- and β), or Akt shRNAs. The differences in P-eNOS and P-Akt levels of Shp2, PKA (α precipitates from the lysates of HUVECs overexpressing Gab1. P85 was used as and β), or Akt shRNA-expressing groups were significant compared with loading control. (C) Immunoblot analysis of anti-Gab1 immunoprecipitates from scrambled control (P < 0.01). (C) Results of NO release measurement of HUVECs the lysates of HUVECs overexpressing vector, Gab1 WT, or indicated mutants. expressing scrambled, Gab1, shp2, PKA, or Akt shRNAs after VEGF stimulation. (D) Tube formation analysis of HUVECs overexpressing vector, Gab1 WT, or (D and E) Tube formation analysis of HUVECs expressing scrambled or gene- indicated mutants. Asterisks indicate statistical significance (*P < 0.05, **P < specific shRNAs of Gab1, shp2, PKA, or Akt pretreated with or without L-NAME 0.01) for EGKO versus control. (5 mM) or PKI (10 μM) (D) before VEGF stimulation. Asterisks indicate statistical significance (*P < 0.05, **P < 0.01) for EGKO versus control. Acute limb ischemia induced by femoral artery ligation in mice has been widely used as a model to study postnatal angiogenesis G induced by endogenous VEGF, whose expression is strikingly in- vessels isolated from EGKO mice (Fig. 3 ). Furthermore, fi duced by ischemia/hypoxia in vivo and is essential for ischemia- shRNA-mediated Gab1 knockdown signi cantly reduced eNOS induced angiogenesis (4, 5, 21). We thus used this model to activation, NO production, and tube formation of primary human address the role of Gab1 in postnatal angiogenesis induced by ECs stimulated by VEGF (Fig. 6). Taken together, these studies ischemia and VEGF. Ischemia-induced angiogenesis was se- establish a critical role of eNOS in Gab1-mediated VEGF sig- verely impaired in EGKO mice despite the fact that VEGF ex- naling leading to tube formation. eNOS is a crucial regulator of pression was normally induced (Fig. 2). To directly determine vascular homeostasis and (6). Interestingly, two the role of Gab1, we used a Matrigel implant assay and found recent studies suggested a role of Gab1 in shear stress-induced that VEGF-induced angiogenesis was significantly reduced in eNOS activation and ex vivo vasodilation (22). EGKO mice (Fig. 2). These data revealed a crucial role of Gab1 It is noteworthy that, so far, Akt has been regarded as a central in ischemia and VEGF-induced angiogenesis. enzyme of the activation of eNOS by VEGF, whereas we are Subsequent studies identified eNOS as a primary downstream aware of no reports on the role of PKA in VEGF-induced eNOS effector of the Gab1-mediated signaling pathway in promoting activation. This study presents several lines of evidence sup- endothelial tube formation in response to VEGF. VEGF-induced porting the idea that PKA plays an important role in Gab1- eNOS activation was severely impaired in Gab1-deficient ECs mediated eNOS activation by VEGF. First, VEGF-induced (Fig. 3), which were rescued by Gab1 reconstitution or intro- phosphorylation of eNOS (Ser1176) and NO production in ECs duction of a constitutively active form of eNOS (Fig. 4). In ad- from EGKO mice were substantially reduced despite the phos- dition, decreased eNOS activation by VEGF was detected in the phorylation of Akt being slightly increased, whereas PKA acti-

Lu et al. PNAS Early Edition | 5of6 Downloaded by guest on October 1, 2021 vation was significantly decreased (Fig. 3). Second, inhibition of ments of this Gab1-mediated PKA-dependent pathway, such as PKA activity by a specific inhibitor or PKA shRNAs ablated the binding sites on Gab1 for Shp2, Shp2, and PKA, may po- VEGF-induced eNOS activations as well as tube formation tentially be efficient therapeutic targets for pharmacological in- (Figs. 3 and 6). Third, reconstitution of Gab1 rescued the acti- tervention to treat human ischemic diseases. vation of PKA and eNOS of EGKO ECs (Fig. 4); conversely, Gab1 shRNAs suppressed VEGF-induced eNOS activation and Materials and Methods tube formation in primary human ECs (Fig. 6). Therefore, we Generation of EGKO Mice. To generate mice lacking Gab1 in endothelia, we fl fl identified a Gab1-mediated PKA-dependent pathway leading to crossed Gab1 ox/ ox mice (17) with Tie2-Cre transgenic mice (18) to obtain fl fl eNOS activation and tube formation. It is unexpected that Gab1 EGKO mice (Gab1 ox/ ox:Tie2-Cre/+). Genotyping for the Gab1 locus and Tie2- deficiency in ECs slightly increases Akt activity. Previous studies Cre transgene were performed by PCR analysis of tail genomic DNA as de- have suggested that cAMP/PKA pathway may inhibit Akt acti- scribed previously (17, 18). Age-matched male litters (12–18 wk) were used for vation, although its mechanism is still unclear (23). It is possible experiments. Animal procedures were carried out according to the rules of that Gab1 deficiency leads to the increase of Akt activity via American Association for the Accreditation of Laboratory Animal Care In- relieving the inhibition of PKA on Akt activation. It is also ternational and approved by the animal care committee of Peking University. possible that Gab1 deficiency blocks a p-IRS-1Ser612–mediated negative regulatory route and hence increases p-Akt, as sug- Mouse Model of Hindlimb Ischemia and Evaluation of Blood Flow and Col- gested in liver-specific Gab1 KO mice (17). VEGF is known to lateralization. A mouse ischemic hindlimb model was established and evaluated activate IRS-1 (24). Our preliminary data suggest the association as previously described (25, 26) and is detailed in SI Materials and Methods. of decreased phospho-IRS-1Ser612 with Gab1 deficiency (Fig. S6). Because Akt knockdown also reduced the level of PKA substrate Matrigel Aortic Ring and Endothelial Tube Formation Assays. Mouse aortic ring phosphorylation (Fig. 6), there might be a reciprocal regulation and endothelial tube formation assays were conducted as described (25) and between these two kinases, although the detailed mechanism are detailed in SI Materials and Methods. remains unclear and the exact mechanism awaits further study. fi Isolation of Mouse ECs. Isolation, culture, and characterization of ECs from Our ndings from the experiments that used Shp2-binding mouse heart and liver are described in previous studies (21). The subsequent mutants or Shp2 shRNA clearly demonstrated an essential role of purification was carried out using magnetic beads (MACS MicroBeads, Mil- Gab1–Shp2 association in modulating PKA/eNOS activation by – – tenyi Biotec) with biotin-conjugated anti-mouse platelet EC adhesion mole- VEGF. However, it remains a mystery how the Gab1 Shp2 as- cule-1 (i.e., CD31) antibody (BD Pharmingen), according to the manufacturer’s sociation mediates downstream signaling of PKA/eNOS activa- instructions. Isolated mouse ECs were used between passages two and four. tion in VEGF signaling. Interestingly, shear stress-induced eNOS activation also requires Gab1–Shp2 association via a PKA- Statistics. Results are expressed as mean ± SEM or SD on the basis of triplicate dependent pathway (22). Further investigation is required to experiments. Statistical analysis was made using Student t test (two-tailed). determine whether Shp2 regulates PKA activity, considering that A P value lower than 0.05 was considered statistically significant. Shp2 interacts with a PKA catalytic subunit (Fig. 5), or Shp2 di- rectly targets eNOS as a substrate with regard to the regulation of ACKNOWLEDGMENTS. We thank Drs. Peace Cheng, Xian Wang, Iain Bruce, eNOS activation by tyrosine phosphorylation. Xiuqin Zhang, and Lin Pan for helpful discussions and critical comments and In summary, by using EGKO mice combined with functional Dr. Jeng-Shin Lee for providing the pHRST lentiviral system. We thank Yue fi Feng, Yuli Liu, and Ning Hou for their excellent technical assistance. This analyses using isolated vessels and ECs, we identi ed an im- study was supported by National Science Fund Grants 30671030 and portant PKA-dependent pathway for VEGF-induced eNOS ac- 90607004 and Major State Basic Program of China 2007CB512100 (to J. Luo); tivation in parallel to Akt pathway, which may be critical for Key Project for Drug Discovery and Development in China 2009ZX09501-027 (to X.Y.); German Research Foundation Transregio TR52 (to S.K.-H.); Special angiogenesis induced by ischemia and VEGF. A very recent Project Research on Cancer–Bioscience 17014020 from the Ministry of Edu- study suggested that PKA activity is required for postnatal an- cation, Culture, Sports, Science and Technology of Japan (to M.S.); and Na- giogenesis and ischemic angiogenesis (25, 26). Thus, the ele- tional Institutes of Health Research Grant R01HL096125 (to G.-S.F.).

1. Ferrara N, Alitalo K (1999) Clinical applications of angiogenic growth factors and their 15. Itoh M, et al. (2000) Role of Gab1 in heart, placenta, and skin development and inhibitors. Nat Med 5:1359–1364. growth factor- and cytokine-induced extracellular signal-regulated kinase mitogen- – 2. Carmeliet P (2000) Mechanisms of angiogenesis and . Nat Med 6:389 395. activated protein kinase activation. Mol Cell Biol 20:3695–3704. 3. Shibuya M, Claesson-Welsh L (2006) Signal transduction by VEGF receptors in 16. Sachs M, et al. (2000) Essential role of Gab1 for signaling by the c-Met receptor in regulation of angiogenesis and lymphangiogenesis. Exp Cell Res 312:549–560. vivo. J Cell Biol 150:1375–1384. fi – 4. Couf nhal T, et al. (1998) Mouse model of angiogenesis. Am J Pathol 152:1667 1679. 17. Bard-Chapeau EA, et al. (2005) Deletion of Gab1 in the liver leads to enhanced 5. Murohara T, et al. (1998) Nitric oxide synthase modulates angiogenesis in response to glucose tolerance and improved hepatic insulin action. Nat Med 11:567–571. tissue ischemia. J Clin Invest 101:2567–2578. 18. Lan Y, et al. (2007) Essential role of endothelial Smad4 in vascular remodeling and 6. Dudzinski DM, Michel T (2007) Life history of eNOS: Partners and pathways. integrity. Mol Cell Biol 27:7683–7692. Cardiovasc Res 75:247–260. 7. Lin MI, Sessa WC (2006) Vascular endothelial growth factor signaling to endothelial 19. Stalmans I, et al. (2002) Arteriolar and venular patterning in retinas of mice selectively – nitric oxide synthase: More than a FLeeTing moment. Circ Res 99:666–668. expressing VEGF isoforms. J Clin Invest 109:327 336. 8. Somanath PR, Razorenova OV, Chen J, Byzova TV (2006) Akt1 in endothelial cell and 20. Hoefer IE, van Royen N, Buschmann IR, Piek JJ, Schaper W (2001) Time course of angiogenesis. Cell Cycle 5:512–518. arteriogenesis followingfemoral artery occlusion in the rabbit. Cardiovasc Res 49:609–617. 9. Chen J, et al. (2005) Akt1 regulates pathological angiogenesis, vascular maturation 21. Takeshita K, et al. (2007) Critical role of endothelial Notch1 signaling in postnatal and permeability in vivo. Nat Med 11:1188–1196. angiogenesis. Circ Res 100:70–78. 10. Mirzapoiazova T, Kolosova I, Usatyuk PV, Natarajan V, Verin AD (2006) Diverse effects 22. Dixit M, et al. (2005) Gab1, SHP2, and protein kinase A are crucial for the activation of of vascular endothelial growth factor on human pulmonary endothelial barrier and the endothelial NO synthase by fluid shear stress. Circ Res 97:1236–1244. migration. Am J Physiol Lung Cell Mol Physiol 291:L718–L724. 23. Hong K, Lou L, Gupta S, Ribeiro-Neto F, Altschuler DL (2008) A novel Epac-Rap-PP2A 11. Xiong Y, et al. (2009) VEGF Receptor-2 Y1175 signaling controls VEGF-induced vWF signaling module controls cAMP-dependent Akt regulation. J Biol Chem 283:23129–23138. release from endothelial cells via PLCgamma1- and PKA-dependent pathways. J Biol 24. Senthil D, Ghosh Choudhury G, Bhandari BK, Kasinath BS (2002) The type 2 vascular Chem 284:23217–23224. endothelial growth factor receptor recruits insulin receptor substrate-1 in its 12. Gu H, Neel BG (2003) The “Gab” in signal transduction. Trends Cell Biol 13:122–130. signalling pathway. Biochem J 368:49–56. 13. Dance M, et al. (2006) The adaptor protein Gab1 couples the stimulation of vascular 25. Venkatesh PK, et al. (2010) Dipyridamole enhances ischaemia-induced arteriogenesis endothelial growth factor receptor-2 to the activation of phosphoinositide 3-kinase. J Biol Chem 281:23285–23295. through an endocrine nitrite/nitric oxide-dependent pathway. Cardiovasc Res 85: – 14. Laramée M, et al. (2007) The scaffolding adapter Gab1 mediates vascular endothelial 661 670. growth factor signaling and is required for endothelial cell migration and capillary 26. Ackah E, et al. (2005) Akt1/protein kinase Balpha is critical for ischemic and VEGF- formation. J Biol Chem 282:7758–7769. mediated angiogenesis. J Clin Invest 115:2119–2127.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1009395108 Lu et al. Downloaded by guest on October 1, 2021