Attribution of vascular phenotypes of the murine Egfl7 locus to the microRNA miR-126

MicroRNAs (miRNAs) are essential regulators of physiology and pathophysiology (Zhao and Srivastava,2007). The inadvertent dysregulation of intronic miRNAs has been predicted to be a general complication in the design and interpretation of mouse knockout studies (Osokine et al.,2008). miR-126 (Mirn126 - Mouse Genome Informatics) is an endothelial miRNA residing within intron 7 of Egfl7, resulting in pan-vascular developmental coexpression of miR-126 and Egfl7 and their abundant expression in cultured endothelium (Fitch et al.,2004; Kloosterman et al.,2006; Kuehbacher et al.,2007; Poliseno et al.,2006). Egfl7 is an endothelial secreted extracellular matrix protein, which, in zebrafish, regulates embryonic vascular tube assembly (De Maziere et al.,2008; Parker et al.,2004). In vitro, various functions have been ascribed to Egfl7, including the regulation of endothelial or vascular smooth muscle migration and adhesion (Campagnolo et al., 2005; Parker et al.,2004; Soncin et al.,2003). Two different mouse knockout alleles of Egfl7 have been described: a gene-trap insertion into intron 2, and an IRES lacZknock-in replacing exons 5-7, both upstream of miR-126 in intron 7. Both the Egfl7 gene-trap and lacZ knock-in are associated with edema, angiogenic deficits and ∼50% embryonic lethality(Schmidt et al., 2007). Here,we explored the functions of both Egfl7 and its embedded miRNA, miR-126, using floxed alleles to selectively disrupt each gene without reciprocal perturbation.

For targeting Egfl7, a loxP site (Pl452) and a neomycin selection cassette plus a loxP site (Pl451) were cloned into an AflII site 5′ of exon 5 and into an NheI site 3′ of exon 7,respectively. For targeting the 73-bp miR-126 precursor, Pl452 and Pl451 were cloned into an NheI site 194 bp 5′ of miR-126 and an NsiI site 22 bp 3′ of miR-126,respectively (flanking 289 bp total) (for details, see Figs S1 and S2 in the supplementary material). Delta (Δ) alleles were generated by crossing to CMV- or HPRT-Cre mice. Mutant mice were analyzed in a mixed 129sV/C57Bl/6 genetic background. All mice were treated according to the Stanford Institutional Animal Care and Use Committee and the Stanford Administrative Panel on Laboratory Animal Care.

In situ hybridization was performed as described(Obernosterer et al., 2007). Mouse miR-126 locked nucleic acid (LNA) probes were from Exiqon.

Rabbits were immunized against the bacterially expressed C-terminal 112 amino acids of murine Egfl7 fused to the C-terminus of maltose binding protein(MBP). Antiserum was affinity purified against the C-terminal 112 amino acids of Egfl7 fused to the C-terminus of glutathione-S-transferase (GST). PFA-fixed frozen uterus sections were stained with 0.1 μg of affinity-purified rabbit anti-Egfl7 antibody and imaged with a Zeiss Z1 Axioimager with Apotome.

miR-126 expression was analyzed using the Taqman MicroRNA Assay(Applied Biosystems) utilizing looped RT primers to detect processed miR-126, and expression was normalized to that of miR-16. Egfl7 expression was determined using the SYBR Green Quantitect PCR Kit(Qiagen) and normalized to that of Gapdh. Egfl7 primers:5′-TGCGACGGACACAGAGCCTGCA-3′ and 5′-CAAGTATCTCCCTGCCATCCCA-3′. Assays were performed in triplicate and results from at least three independent experiments are presented.

P5 eyes were dissected and fixed in 4% paraformaldehyde (PFA) in PBS overnight at 4°C. Retinas were isolated, blocked in PBS containing 1% BSA and 0.5% Triton X-100 overnight at 4°C, incubated overnight with 10 μg of FITC-conjugated isolectin B4 (Vector Labs) in 500 μl of the same solution, washed and then flat mounted.

Antibodies used were: rabbit anti-p85, rabbit anti-phospho-Akt (Akt1 -Mouse Genome Informatics) (Ser 473), rabbit anti-phospho-Erk (Mapk1 - Mouse Genome Informatics) (all from Cell Signaling), rabbit anti-Spred1, rabbit anti-p85β, rabbit anti-α-actin (all from Abcam) and rat anti-HA(Roche).

Anti-miR-126 hairpin inhibitors (Thermo Scientific Dharmacon) or negative control inhibitor were transfected into HUVEC at 100 nM using Dharmafect1. Cells were assayed for protein expression 48 hours after transfection.

HUVEC were serum starved overnight 24 hours after transfection of miRNA inhibitors, and scraped with a sterile P200 tip to generate a cell-free zone. Cells were stimulated with human VEGF₁₆₅ (R&D Systems) (10 ng/ml) for 24 hours. Migration was quantified by counting the number of cells per scratched area (n=6).

The corneal micropocket assay was performed as described(Kuo et al., 2001).

The 3′UTR of Pik3r2 and Spred1 were amplified and cloned downstream of a Renilla luciferase reporter gene. The miR-126 binding sites were mutated from 5′-ACGGTAC-3′ to 5′-GTAACGA-3′ and from 5′-GGTACG-3′ to 5′-AAGCAT-3′ in the 3′UTR of Pik3r2 and Spred1, respectively. The Lin41 (Trim71 - Mouse Genome Informatics) 3′UTR was used as a negative control. 293T cells in 24-well plates were transfected with 3.35 ng/well of firefly luciferase, 0.667 ng/well of Renilla 3′UTR construct, and either 0, 10 or 100 ng/well of miR-126 expression vector. Empty vector was added to provide a total of 337 ng of DNA per transfection. Forty-eight hours after transfection, the Renilla/firefly luciferase was measured using the Dual Reporter Luciferase Kit (Promega).

Akt/Erk phosphorylation assays were performed as described(Gerber et al., 1998).

P-values were determined using a two-tailed Student's t-test assuming unequal variances.

We explored the mouse Egfl7/miR-126 locus using selectively floxed Egfl7^Δ and miR-126^Δalleles to replace either a 289 bp segment of intron 7 containing miR-126 or exons 5-7 of Egfl7 with a single loxP site,without disruption of the reciprocal gene or miRNA(Fig. 1A; see Figs S1 and S2 in the supplementary material). miR-126^Δ/Δ, but not Egfl7^Δ/Δ, embryos exhibited loss of miR-126 expression as assessed by in situ hybridization or quantitative PCR (qPCR) using looped RT primers to detect processed miR-126 (Fig. 1B,C). Conversely, Egfl7^Δ/Δ, but not miR-126^Δ/Δ, mice exhibited loss of Egfl7 by qPCR and by immunofluorescence with an affinity-purified rabbit anti-Egfl7 antiserum (Fig. 1C,D). Furthermore, sequencing of the Egfl7 ORF amplified from miR-126^Δ/Δ cDNA revealed a lack of occult Egfl7 splicing alterations resulting from the miR-126 microdeletion (Fig. 1E). These studies indicated the successful generation of two monospecific Δ alleles for miR-126 and Egfl7,respectively.

Surprisingly, Egfl7^Δ/Δ mice were phenotypically normal and born at the expected Mendelian ratios despite previous reports from gene-trap and conventional knockout alleles(Schmidt et al., 2007)(Fig. 2A). By contrast, miR-126^Δ/Δ mice recapitulated numerous previously described Egfl7 mutant phenotypes(Schmidt et al., 2007)including ∼50% embryonic lethality(Fig. 2B), which appeared obligately associated with the development of prominent subcutaneous embryonic edema by E14.5 (Fig. 2C,D). At E15.5, multifocal, progressive hemorrhage of varying severity from ruptured blood vessels was observed in ∼20% of the miR-126^Δ/Δ embryos, most prominently in the jugular and subcutaneous regions(Fig. 2E), with resultant embryonic lethality becoming first apparent at E16.5. The embryonic edema was phenocopied by miR-126^flox//Δ;Tie2-Cre embryos,consistent with a cell-autonomous mechanism in the endothelium(Fig. 2F).

Surviving miR-126^Δ/Δ neonates,which were obtained at ∼50% of the expected frequency(Fig. 2B), exhibited delayed postnatal retinal angiogenesis (Fig. 3A-C). This was particularly notable in terms of compromised radial migration, a decreased area of retinal vascularization, and abnormally thickened endothelial sprouts (Fig. 3A-C), as previously described in Egfl7 gene-trap and knock-in mice (Schmidt et al.,2007). miR-126^Δ/Δ mice further displayed delayed developmental cranial angiogenesis(Fig. 3D), again reminiscent of previously described Egfl7 mutant phenotypes(Schmidt et al., 2007). Consistent with a more substantial role for miR-126 in the regulation of adult angiogenic processes, surviving miR-126^Δ/Δ mice demonstrated impaired angiogenesis in a VEGF-dependent corneal micropocket assay(Fig. 3E,F). None of the aforementioned phenotypes was observed in Egfl7^Δ/Δ or wild-type mice(Fig. 3A-F), with the deficits in retinal, head and corneal vasculature all supporting the in vivo regulation of angiogenesis by miR-126.

The mechanisms of miR-126 regulation of angiogenesis were further explored in cultured endothelial cells. Transfection of an RNA hairpin inhibitor induced a greater than 95% depletion of mature miR-126 in HUVEC (Fig. 4A). This was accompanied by significant decreases in migration in scratch assays, as well as impaired VEGF-dependent activation of the downstream kinase Akt(Fig. 4B,C). The basis for this impaired VEGF signaling in miR-126-deficient endothelium was examined at the level of miRNA target genes. miR-126 directly repressed expression of the Pik3r2-encoded p85β subunit of PI3 kinase(PI3K) in co-transfection assays, whereas p85β protein was increased in both primary miR-126^Δ/Δ endothelium and miR-126 knockdown HUVEC (Fig. 4D-G). Either the knockdown of miR-126 or the overexpression of the target p85β in HUVEC was sufficient to impair VEGF-mediated activation of the PI3K downstream target Akt, paralleling inhibition of insulin receptor tyrosine kinase signaling by p85 overexpression(Barbour et al., 2005; Brachmann et al., 2005; Ueki et al., 2002)(Fig. 4C,H). miR-126knockdown additionally impaired VEGF activation of Erk(Fig. 4C), further reiterating compromised signal transduction in angiogenesis by miR-126 knockdown in vitro. In this regard, the Erk pathway inhibitor Spred1(Taniguchi et al., 2007) was directly repressed by miR-126 co-transfection and was upregulated in miR-126 knockdown HUVEC (see Fig. S3 in the supplementary material).

Overall, the current studies describe essential in vivo regulation of angiogenesis by a miRNA as evidenced by delayed developmental vascularization in retina and brain, impaired adult VEGF-dependent corneal angiogenesis, and in vitro regulation of motility. Edema was a prominent feature of miR-126^Δ/Δ embryos and was tightly correlated with the lethality observed in ∼50% of embryos. This edema did not appear secondary to intrinsic cardiac defects (data not shown). The incompletely penetrant embryonic lethality and angiogenic delay of miR-126^Δ/Δ mice contrast with the more classical embryonic lethal angiogenic phenotypes(Gale and Yancopoulos, 1999)and appear consonant with the comparatively subtle action of miRNAs in fine-tuning global gene expression profiles(Kloosterman and Plasterk,2006; Zhao and Srivastava,2007).

Consistent with its vascular expression pattern, these miR-126phenotypes occur cell-autonomously in endothelium as judged from the compartment-specific deletion phenotypes of miR-126^flox//Δ;Tie2-Cre embryos. Mechanistically, this cell-autonomous action allows miR-126deficiency to derepress and overexpress the p85β regulatory subunit of PI3K and Spred1, which represent negative regulators of PI3K and MAP kinase signaling, respectively (see Fig. S4 in the supplementary material). Although p85β and Spred1 dysregulation clearly appears contributory to the miR-126 phenotype, the promiscuous action of miRNA suggests the likely action of numerous additional target genes.

During the preparation of this manuscript, miR-126 deletion phenotypes in mouse and knockdown in zebrafish were described with impaired angiogenesis and vascular integrity via dysregulation of Spred1 and p85β(Fish et al., 2008; Wang et al., 2008). These phenotypes are both reinforced by similar findings in the current report and are extended by our analysis of endothelial-specific deletion in miR-126^flox//Δ;Tie2-Cre embryos. Furthermore, an added significant feature of the current study is the unexpected lack of abnormalities in Egfl7^Δ/Δ mice and the widespread phenocopying by miR-126^Δ/Δ mice of vascular deficits of previously described Egfl7 alleles, consisting of a gene-trap in intron 2 and a lacZ insertion into exons 5-7, both upstream of intron 7 that contains miR-126(Schmidt et al., 2007). These data indicate that miR-126 might well regulate the collective migration of endothelium as has been proposed for Egfl7(Schmidt et al., 2007). These miR-126^Δ/Δ mice should facilitate additional exploration of miR-126 function in settings of adult angiogenesis, as well as of divergent miR-126 roles such as in metastasis suppression (Tavazoie et al.,2008). Conversely, the Egfl7^Δ/Δ mice will allow selective in vivo analysis of Egfl7 without the confounding influence of miR-126.Our data by no means exclude novel and essential Egfl7-specific functions, either alone or in conjunction with the paralog Egfl8, or as described in zebrafish knockdown, mouse overexpression and in vitro studies(Campagnolo et al., 2005; Lelievre et al., 2008; Soncin et al., 2003; Xu et al., 2008).

The inadvertent disruption of miRNA expression by conventional deletion and gene-trap knockout approaches in mice was recently predicted in a bioinformatics analysis by McManus and colleagues(Osokine et al., 2008). Our results comparing miR-126^Δ/Δ and Egfl7^Δ/Δ mice provide the most extensive documentation of this complication to date, which might be more widespread than anticipated; this possibility was not formally examined in the mouse miR-126 deletion, as a parallel Egfl7 knockout was not engineered (Wang et al.,2008). From these studies, evaluation of intronic miRNA should be a general consideration in the design and interpretation of mouse knockout studies (Osokine et al.,2008), and complications thereof might be avoided by utilizing minimally disruptive strategies such as floxed alleles covering small genomic regions. Finally, the regulation of angiogenesis by a mammalian miRNA suggests novel methods for the therapeutic modulation of vascularization, for instance during cancer or macular degeneration.

We are indebted to Gerald Crabtree, Andrew Fire, Cecile Chartier and Cullen Taniguchi for helpful discussions. Hprt-Cre mice and recombineering plasmids were kind gifts of Liqun Luo and Neal Copeland, respectively. This work was supported by grants from the NIH (1 R01 CA95654-01, 1 R01 NS052830-01 and 1 R01 HL074267-01) and the Brain Tumor Society to C.J.K.