RNA interference (RNAi) is a gene silencing pathway by which specific messenger RNAs (mRNAs) are either degraded or translationally suppressed 1 . It is mediated by microRNA (miRNA) or short interfering RNA (siRNA), both non coding RNAs of 20–22 nucleotides which are matured by the RNase Dicer and are involved in base pairing with target mRNAs. In mice, dicer is critical for early mouse development because its abrogation prevents the production of functional interfering RNAs resulting in embryonic lethality at E7.5 2 . A second study reported death at E13.5 which was associated with angiogenesis defects 3 but both studies were unable to decipher the role of Dicer in specific vascular cell types. Conditional ablation of dicer developed to investigate its function in limb buds 4 , in immune cells 5 , and heart development 6 have suggested important roles of RNA interference in various biologic processes such as cell survival, proliferation, differentiation, and maintenance of cell function.

In angiogenesis, the role of Dicer-regulated miRNAs was further suggested in mice expressing a hypomorphic Dicer1 allele, which resulted in female infertility caused by corpus luteum insufficiency and defective ovarian angiogenesis 7 . In addition, Dicer has been shown to have multiple roles in vascular biology. Tamoxifen-inducible and smooth muscle cell (SMC)-specific deletion of Dicer achieved by Cre-Lox recombination showed that miRNAs are necessary for vascular smooth muscle growth, differentiation, and function 8 9 . Dicer-deficient mice exhibited a dramatic reduction in blood pressure due to significant loss of vascular contractile function and SMC contractile differentiation as well as vascular remodeling. This phenotype pointed to miRNAs as important mediators for the modulation of the VSMC phenotype by targeting transcription factors and the cytoskeleton, which acts as molecular switches for VSMC differentiation 10 . In these cells, the Mir143/145 gene cluster plays a major role in regulating the contractile phenotype and controling responses to various types of injury 11 12 13 .

The reduction of endothelial miRNAs by inactivation of Dicer both in vitro 14 and in vivo using Cre-recombinase under the regulation of tie2 promoter/enhancer or tamoxifen inducible expressed Cre-recombinase (Cre-ER^T2) under the regulation of vascular endothelial cadherin promoter was shown to reduce postnatal angiogenic response to a variety of stimuli, including exogenous VEGF, tumors, limb ischemia, and wound healing 15 . In vitro studies demonstrated the presence of miRNAs in endothelial cells 16 17 and silencing of Dicer using short interfering (si)RNA in human endothelial cells resulted in impaired capillary-like structures and reduced cell growth 18 19 20 21 . The angiogenic properties of members of the mir 17–92 cluster have been extensively studied 15 22 23 . Also, miR-92a, miR-15a, miR-126 were identified to target mRNAs corresponding to several proangiogenic proteins, such as FGF2 and VEGF 22 24 25 26 27 28 . In addition, recent studies reported the role of miR-99b, miR-181a, and miR-181b in the differentiation of human embryonic stem cells to vascular endothelial cells 29 . In the vascular endothelium, recent findings have shown that miRNAs such as mir-210 orchestrate the response to hypoxia 30 31 and that down-regulation of Dicer under chronic hypoxia is an adaptive mechanism that serves to maintain the cellular hypoxic response through HIF-α and miRNA-dependent mechanisms 29 . Functional deficiency of Dicer in chronic hypoxia is relevant to both HIF-α isoforms and hypoxia-responsive/HIF target genes. The regulation of Prox1 by miR-181 further highlighted the contribution of RNA interference in the induction of lymphatic endothelium. Indeed, miR-181 is highly expressed in the blood vasculature, but significantly reduced in lymphatic endothelial cells, reciprocally to Prox1 expression 32 .

However, whether Dicer could regulate angiogenesis, especially during development when hypoxia is a major stimulus remains largely unclear. There is still insufficient evidence for the involvement of RNA interference during the early stages of vascular cell development, and particularly in the control of endothelial arterial-, venous-, and lymphatic- fate specification. Here, we show that conditional inactivation of Dicer in mice expressing Cre recombinase under the control of the tie2 promoter causes no major alterations in EC fates and differentiation but leads to unexpected functional and morphologic alterations in the separation of blood and lymphatic vasculature.

To bypass the early embryonic lethality of dicer-null mice 2 , we developed a new Cre-loxP-mediated conditional deletion of dicer in tie2-expressing cells in order to investigate its role in vascular development. To this end, we crossed dicer-floxed mice (dicer ^fl/fl ) 4 with tie2-Cre transgenic mice 4 34 . The resulting heterozygous double transgenic mice (dicer ^ΔEC/+ ) were viable. Intercrosses of dicer ^ΔEC/+ male with dicer ^fl/fl females yielded no dicer ^ΔEC/ΔEC pups out of 293 viable offspring at birth (see Table 1). These data suggested that mice bearing dicer gene deficiency in tie2-expressing cells do not survive embryogenesis. To determine when the dicer ^ΔEC/ΔEC mice died, embryos were examined from E10.5 to birth. Embryos were removed and embryonic DNA was analyzed for homo- or heterogeneity of the floxed allele. This genotype was then correlated with the viability of the embryo. Mendelian ratios were observed from E10.5 to E13.5 (see Table 1). Mutant embryos from E10.5 to E11.5 were macroscopically indistinguishable from the control littermates. At later stages, E12.5 onwards, macroscopic examination revealed the presence of hemorrhages and edema in mutant embryos that increased in size and number with age (Figure 1A). Genotyping PECAM⁺ endothelial cells showed efficient dicer inactivation in E13.5 dicer ^ΔEC/ΔEC embryos compared to E13.5 dicer ^ΔEC/+ embryos here used as controls (Figure 1B).

Recombination was also monitored in tie2-expressing cells using the ROSA26 (R26) reporter line 35 . We then crossed the dicer ^ΔEC/+ males with homozygous dicer ^fl/fl :R26/R26 females to generate dicer ^ΔEC/+ :R26/+ (heterozygous, here as a control) and dicer ^ΔEC/ΔEC :R26/+ (mutant) triple transgenic embryos. As indicated by whole-mount X-Gal staining, the recombination was efficient in blood endothelial cells thereby allowing us to compare the pattern of the vascular network in mutant and control embryos using LacZ staining (Figure 2). Between E10.5 and E12.5, dicer ^ΔEC/ΔEC embryos did not display obvious blood vascular defects: avascular regions were not observed in control embryos. X-Gal-stained blood vessels formed properly and vascular density was comparable in both control and mutant embryos (Figure 2). Whole-mount staining using an anti-PECAM antibody confirmed these observations as reported in Figure 3A showing that vascular patterning of blood vessels was comparable to controls in E11.5 dicer ^ΔEC/ΔEC embryos. In order to study the development of the blood vessel network in greater details, branchpoints of the cranial vascular network (internal carotid artery) were quantified on E11.5 embryos. The number of branchpoints in the internal carotid artery was not statistically different in dicer ^ΔEC/ΔEC embryos compared to WT embryos (Figure 3B).

Between E12.5 and E14.5, extensive edema gradually progressed on the back of the embryos and was sometimes filled with blood cells in dicer ^ΔEC/ΔEC embryos which were all dead at E14.5 (Figure 1 and Table 1). This also phenocopies the effects observed upon genetic deletion of Prospero homeobox 1 (prox-1) 36 , Src homology domain-containing leukocyte protein-76 (slp-76) 37 or C-type lectin-like receptor 2 (clec-2) 38 . All show impaired lymphatic vessel development and die in utero with severe edema and hemorrhages. To establish whether dicer ^ΔEC/ΔEC embryos also present defects in lymphatic vessels development, we examined transverse sections of mutant embryos. At E13.5, we never observed any disruption of the main blood vessels i.e. the thoracic aorta or the cardinal vein in dicer ^ΔEC/ΔEC embryos (Additional file 1: Figure S1 and Figure 4). The lymph sacs, the first lymphatic structure that emerges from the cardinal vein 39 during development, also appeared normal (Additional file 2: Figure S2). At E13.5 however, in contrast to control embryos, these lymph sacs were filled with blood cells in dicer ^ΔEC/ΔEC embryos (Figure 4A). The lymphatic identity of the blood-filled structures was confirmed by the expression of lymphatic markers VEGFR-3 (Figure 4A) and PROX-1 (Additional file 3: Figure S3). Moreover, LYVE-1 whole-mount immunostaining evidenced a complete overlap between blood-filled structures and the lymphatic vasculature in E14.5 mutant embryos (Figure 4B), confirming the blood-filled lymphatics phenotype.

Altogether, these data indicate that dicer inactivation in tie2 expressing cells leads to embryonic lethality at E14.5, and to a failure in the separation of lymphatic vessels during embryonic angiogenesis.

Here, using Cre/loxP-mediated inactivation of dicer in tie2-expressing cells, we demonstrate for the first time that embryonic venous-lymphatic separation is submitted to epigenetic control by RNA interference. Previous studies using a similar approach of conditional dicer deficiency using tie2-Cre and ve-cadherin-CRE-ERT2 have reported reduced postnatal angiogenesis but no developmental defects 15 . The likely explanation for this discrepancy probably relies on the use of a different dicer-floxed mouse leading to the presence of residual Dicer protein levels in tie2-Cre:dicer ^fl/fl endothelial cells, reflecting an incomplete excision of the dicer allele 15 . Thus, these mice were hypomorphic for dicer in ECs and tie2-Cre:dicer ^fl/fl newborn litters were overtly normal and indistinguishable from their littermate controls. In contrast, in the present study, efficient dicer inactivation was evidenced in PECAM⁺ endothelial cells which showed complete excision of dicer in dicer ^ΔEC/ΔEC embryos. The present study thus shows that dicer gene deletion in Tie2 expressing cells leads to embryonic lethality at E14.5. Mutant embryos, which display hemorrhages and edema, showed blood-filled lymphatics without evident angiogenesis defects at early stages.

We here used the well-documented tie2-Cre transgenic mice that express Cre in a pan-endothelial fashion for vascular endothelial targeting 34 . With the Rosa26 reporter line, we showed recombination in lymphatic vessels (Additional file 4: Figure S4). Using the same tie2-Cre ROSA26 strain, Srinivasan et al. demonstrates that at E11.5, Prox1⁺ endothelial cells in the anterior cardinal vein and those budding from it were lacZ ⁺. Similarly, all E13.5 and E14.5 Prox1⁺ endothelial cells in the lymph sacs were lacZ ⁺ 40 . Nevertheless, it should be noted that it has also been reported that tie2-Cre transgenic mice express Cre in blood island progenitors 41 42 . Recent studies have highlighted the role of hematopoietic cells during the process of separation between the venous and the lymphatic vasculature. It has been shown that podoplanin, a transmembrane protein expressed on lymphatic endothelial cells, engages the platelet receptor CLEC-2 leading to Syk-Slp-76-dependent platelet activation 43 . Deletion of these genes leads to aberrant vascular connection between blood and lymphatic vessels. Similar lymphovenous connections were also observed in mice deficient for the homeodomain transcription factor Meis1 (myeloid ecotropic viral integration site 1) which completely lack megakaryocyte/platelets and for the transcription factor Runx1 which lack hematopoietic stem cells 40 44 . It should also be noted that runx1 mutant embryos, which lack platelets, present hemorrhages in the brain 45 , which could also be observed in some dicer ^ΔEC/ΔEC embryos. Because platelets also act to maintain vascular integrity and as the brain and lungs are more susceptible to haemorrhage in a mouse model of acute severe thrombocytopenia induced by platelet depletion 46 , these hemorrhages most likely occur secondary to the lack of platelets. These data showed that platelets are required during embryonic lymphangiogenesis for the separation of the nascent lymphatic vasculature from blood vessels 47 48 . However, recent studies by Yang et al. 49 and Hägerling et al. 50 have disproved a direct involment of platelets in the emergence of the first jugular lymph sacs. Podoplanin expression only starts after lymphatic endothelial cells leave the cardinal vein suggesting that platelets have a role restricted to the region where lymphatics and blood vessels coalesce, in the lymphovenous valves. Nevertheless, the presence of blood cells in lymphatic vessels may also indicate an incomplete separation of blood and lymph vessel, but could also result from de novo connections of previously separated blood and lymph vessels. Recently, Hess et al. proved that platelets interact with lymphatic endothelium valves specifically at the thoracic duct-subclavian vein junction 51 . Blood-filled lymphatics arise due to backfilling of the lymphatic vascular network from this site either due to a lymphovenous valve defect or due to a platelet aggregation defect. We therefore looked at the thoracic duct-subclavian vein junction and we determined that the lymphovenous valves appears normal (Additional file 5: Figure S5) suggesting a defect in platelet aggregation.

We therefore sought to decipher whether perturbing dicer expression in megakaryocytes could also reproduce a blood-filled lymphatic phenotype during development by generating pf4-cre:dicer ^fl/fl mice. Pf4-cre express Cre-recombinase in the megakaryocytic lineage as previously shown 52 and are a useful tool to study megakaryopoiesis, and platelet function. These mice were born at normal mendelian ratio and the separation of the lymphatic vasculature from the blood vessels was not disrupted during development (Additional file 6: Figure S6 and Table 2). Recombination was observed in liver megakaryocytes before venous-lymphatic separation, as soon as E11.5 (data not shown) and persisted at E16.5 (Additional file 4: Figure S4B). However, the pf4-Cre transgene is also partially expressed in other hematopoietic lineages and the recombination pattern during early embryogenesis is not clear 53 . A megakaryocyte specific promoter that could allow earlier deletion might be useful but does not exist.

Also, cells from the myeloid lineage play a critical role in this separation. Abnormal infiltration of a specific monocyte population in syk-deficient mice leads to lymphatic hyperplasia, vessel dilation and blood-lymphatic shunts 54 . Tie2 is expressed in the early yolk sac mesoderm suggesting that recombination may occur in hematopoietic cells 55 . The use of a more endothelial specific strains such as ve-cadherin-CRE-ERT2 56 or pdgfb-CRE-ERT2 57 would also be very useful for understanding the specific role of Dicer in the endothelium. However, the CRE activation is tamoxifen-dependent making these models more suitable for postnatal angiogenesis as recombination at a precise embryonic time point might be somewhat difficult to achieve in a very reproducible manner.

MicroRNAs are involved in many aspects of physiological and malignant hematopoiesis but surprisingly, no existing studies have focused on the role of dicer during hematopoietic development. However, dicer invalidation in adult has been described. Buza-Vidas et al. showed that dicer is required during erythroid lineage differentiation 58 . It was also suggested that Dicer is involved in the regulation of the hematopoietic stem cell niche as well as the regulation of hematopoietic stem cell number 59 60 . The blood filled phenotype that we observed could result from either a defect of hematopoiesis or a volume expansion of the blood stream indirectly affecting lymphatic development. We therefore believe that further experiments, outside of the scope of the present manuscript, will be needed to determine precisely whether hematopoiesis is modulated in dicer ^ΔEC/ΔEC embryos and to fully decipher the cellular and molecular mechanisms responsible for the blood-filled lymphatic phenotype in these mice.

Taken together, these results show a new role for RNA interference in epigenetic control of embryonic venous-lymphatic separation and provide a knowledge base for further investigations to validate functional roles for microRNAs.

CLEC-2: C-type lectin-like receptor 2; Cv: Cardinal vein; E: Embryonic day; Ica: Internal carotid artery; Ls: Lymph sac; miRNA: microRNA; mRNA: Messenger RNA; Pecam-1: Platelet endothelial cell adhesion molecule 1; siRNA: Short interfering RNA; Vegfr-3: Vascular endothelial growth factor receptor 3; WT: Wild type.

The authors declare that they have no competing interests.

SGe, SGa designed experiments. SGa, JP and ML performed experiments. SGe, SGa, IG and MT wrote the paper. All authors read and approved the final manuscript.