MicroRNA-mediated integration of haemodynamics and Vegf signalling during angiogenesis.

Within the circulatory system, blood flow regulates vascular remodelling, stimulates blood stem cell formation, and has a role in the pathology of vascular disease. During vertebrate embryogenesis, vascular patterning is initially guided by conserved genetic pathways that act before circulation. Subsequently, endothelial cells must incorporate the mechanosensory stimulus of blood flow with these early signals to shape the embryonic vascular system. However, few details are known about how these signals are integrated during development. To investigate this process, we focused on the aortic arch (AA) blood vessels, which are known to remodel in response to blood flow. By using two-photon imaging of live zebrafish embryos, we observe that flow is essential for angiogenesis during AA development. We further find that angiogenic sprouting of AA vessels requires a flow-induced genetic pathway in which the mechano-sensitive zinc finger transcription factor klf2a induces expression of an endothelial-specific microRNA, mir-126, to activate Vegf signalling. Taken together, our work describes a novel genetic mechanism in which a microRNA facilitates integration of a physiological stimulus with growth factor signalling in endothelial cells to guide angiogenesis.

Within a blood vessel, flow exerts tangential and perpendicular forces upon endothelial cells, leading to cytoskeletal rearrangements and changes in gene expression4. While initial embryonic vascular patterning is largely independent of these hemodynamic forces, the onset of circulation drives subsequent remodeling of the circulatory system4. For example, flow plays an important role in the unilateral regression of the sixth AA during mouse development1. In zebrafish, the fifth and sixth AA arise after flow begins and form a persistent connection to the lateral dorsal aortae (LDA) that provides circulation to the trunk8, 9. These vessels continue to undergo angiogenesis throughout larval stages to comprise the gill vasculature9. To investigate how flow affects angiogenesis, we observed development of AA5 and 6 in zebrafish embryos by 2-photon time-lapse imaging (Supplementary Fig. 1a-c). To co-visualize endothelial cells and flow, we performed microangiography on Tg(kdrl:egfp) embryos, which display fluorescent green endothelial cells, using unconjugated Quantum dots (QDots). At 46 hpf, we observed AA perfusion, but no connection between the fifth and sixth AA and the LDA (data not shown and Supplementary Fig. 1d, 46h). Several hours later, the AA 5/6 connecting vessel (referred to as AA5x according to reference 8) sprouted from the left and right AAs (Supplementary Movie 1). At this point, the sprouts were sufficiently lumenized to allow perfusion with Qdots (Supplementary Fig. 1d, 53.75 magnified; Supplementary Movies 1 and 2). However, blood cells entering from the ventral aorta (VA) became trapped in AA5 and 6 (Supplementary Movie 3). AA5x sprouts then fused with the LDA to form a patent circulatory connection (Supplementary Fig. 1d, 59.75h, Supplementary Movie 1). Subsequently, the AA5x fully lumenized and blood flow through AA5 and 6 commenced (Supplementary Movie 4). These observations indicated that the AA5x develops via concomitant angiogenesis and lumenization in the presence of flow.

To determine if flow was required for this process, we performed unilateral laser microsurgery on Tg(kdrl:egfp) embryos to sever the connection between the VA and AA5 and 6 prior to AA5x sprouting (Supplementary Figs. 1e, 2a). Following microsurgery at 46 hpf, we observed normal AA perfusion on the unoperated side by microangiography at 72 hpf (Fig. 1a). By contrast, on the operated side (right) AA5 and 6 failed to bear flow (Fig. 1b), although cranial blood vessels and the AAs appeared morphologically normal (Fig. 1b; Supplementary Figure 2b). A dorsal view of the same embryo revealed that the AA5x formed on the left side of the embryo, but not on the right side where flow was blocked (Fig. 1c, Supplementary Table 1). To support these results, we treated Tg(kdrl:egfp) embryos beginning at 46 hpf with the myosin ATPase inhibitor 2,3-butanedione 2-monoxime (BDM) or the anesthetic Tricaine methanesulfonate to arrest the heart and block circulation10. In both treatments, embryos failed to form the AA5x (Fig. 1d, e; Supplementary Table 1), although vascular morphogenesis in other anatomical locations appeared normal (Supplementary Figure 3). 2-photon time-lapse microscopy of embryos without flow suggested that a failure to initiate sprouting, rather than vessel regression, was responsible for loss of AA5x (Supplementary Movies 5 and 6). Time lapse analysis using Tg(fli1a:negfp) embryos, in which endothelial cell nuclei are labeled with Egfp, revealed decreased migratory activity of cells within the aortic arches in the absence of flow when compared to wild type (Supplementary Movies 7 and 8). Interestingly, embryos injected with a gata1 Morpholino displayed normal AA5x development (Fig 1f, Supplementary Table 1), suggesting that shear stress from blood cells was dispensable for AA5x angiogenesis. Together, these results indicate that the AA5x forms via angiogenesis and that this process is dependent on flow.

Figure 1

AA5x angiogenesis requires flow and Vegf signaling

a-k, Tg(kdrl:egfp) embryos at 72 hpf (a-c), 60 hpf (d-f) or 65 hpf (g-k). a-c, g-k Embryos subjected to microangiography. Endothelial cells are green, flow is red. a, b, Lateral views, anterior to left (a), or right (b), dorsal is up. c-k, dorsal view, anterior is up. a, b, aortic arches (AA, numbered, indicated by arrows) after severing right AA5 and 6 from ventral aorta; opa – opercular artery. c, dorsal view of embryo in a, b. d, e, Stills from Supplementary Movies 5 and 6. Embryos treated beginning at 46 hpf with BDM (d), Tricaine (e), or injected with gata1 MO (f). g, kdrl mutant embryo at 65 hpf. h-k, Embryos treated with 2.5 μM SU5416 (h) or 0.1% DMSO (i) beginning at 46 hpf or injected with 3 ng of scrambled MO (j) or 3 ng of Vegfa MO (k). c-k, Arrows: lateral dorsal aortae (LDA), arrowheads: AA5x.

Vascular endothelial growth factor (Vegf) signaling has been implicated in flow-mediated AA remodeling in mouse embryos1. Accordingly, we observed AA expression of the zebrafish Vegf receptor-2 ortholog, kdrl, including expression in the developing AA5x at 48 hpf (Supplementary Fig. 4a). We also observed vegfa expression in the developing glomerulus (Supplementary Fig. 4b, c), which is located near the branch point of the dorsal aorta and towards which the AA5x sprouts (Supplementary Fig. 4d), and in cells surrounding the AA blood vessels (Supplementary Fig. 4e). Consistent with a role for Vegf signaling during AA5x angiogenesis, embryos bearing a kinase-dead mutation in Kdrl (referred to as kdrl; ref 11) failed to form a patent AA5x (Fig. 1g; Supplementary Table 1). Furthermore, treatment with the Vegf receptor inhibitor SU5416 from 46 to 65 hpf resulted in a block in AA5x formation, while DMSO had no effect (Fig. 1h, i; Supplementary Table 1). Similarly, partial reduction of Vegfa using a low Morpholino dose (3 ng; see reference 12) blocked AA5x development (Fig. 1j, k). Overall vascular morphology and circulatory function, including initial perfusion of the aortic arches, were normal following these manipulations (Supplementary Fig. 3). These observations demonstrate that AA5x formation requires Vegf signaling. In other developmental settings, Notch signaling coordinates Vegf-stimulated angiogenesis13, 14. However, we did not detect expression of Notch signaling molecules or Notch activation in the AAs (Supplementary Fig. 5a-c) and AA5x was not affected by loss of the Notch ligand dll4 (Supplementary Fig 5d). These results suggest that a Notch-independent mechanism is responsible for Vegf-stimulated AA5x angiogenesis.

A possible candidate gene responsible for integrating flow and Vegf signaling during AA5x formation was the zinc finger transcription factor, klf2, which is induced by flow in endothelial cells6, 7. We observed that zebrafish klf2a was expressed in the AA in a pattern similar to the endothelial marker, vascular-endothelial cadherin (cdh5; Fig. 2a) and was expressed in the developing AA5x (Supplementary Fig. 4f). Furthermore, AA expression of klf2a, but not cdh5, was reduced in cardiac troponin T2 (tnnt2)-deficient embryos, which lack circulation (Fig. 2a; Supplementary Table 2; Supplementary Fig. 6a, ref 15) and in embryos treated with Tricaine (Supplementary Fig. 6b; Supplementary Table 2). To determine if klf2a was required for AA5x angiogenesis, we utilized Morpholinos targeting either the klf2a exon 3 splice acceptor site (Supplementary Fig. 7a, b) or the klf2a start codon. Embryos injected with either Morpholino displayed normal morphology and grossly normal circulatory patterns, including perfusion of the aortic arches following angiography (Supplementary Fig. 7c, d and data not shown), consistent with recent work demonstrating relatively normal flow patterns and heart rate in klf2a-deficient zebrafish embryos at 48 hpf 16. However, the normal transient AA circulatory block persisted in klf2a-deficient embryos (compare Supplementary Movies 3, 4, and 9), suggesting a defect in AA5x formation. Indeed, while embryos injected with control Morpholino appeared normal, klf2a-deficient siblings failed to develop the AA5x (Fig. 2b; Supplementary Fig. 7e, f; Supplementary Table 3). Thus, despite the presence of flow, loss of klf2a mimics the AA5x defect observed in embryos lacking flow or Vegf signaling.

Figure 2

AA5x angiogenesis requires klf2a

a, Embryos subjected to whole mount in situ hybridization at 65 hpf using indicated riboprobes (klf2a, cdh5). Lateral view, anterior to the left, dorsal is up. Embryos were injected with 2 ng of scrambled control (left) or tnnt2 (right) MO. Aortic arch region is denoted by black boxes; aortic arches 4-6 are indicated by numbered arrows in cases where staining is present. b, Microangiogram of 72 hpf Tg(kdrl:egfp) embryo injected with 11 ng-embryo scrambled control MO (top) or klf2a ATG MO (bottom); arrowheads indicate position of normal AA5x formation, arrows denote lateral dorsal aortae. Dorsal views, anterior is up.

In Xenopus laevis embryos, klf2 is important for Vegf receptor-2 expression 17. However, kdrl expression appeared normal in klf2a-deficient zebrafish embryos (Supplementary Fig 6a, c). Similarly, neither kdrl nor vegfa were altered in embryos lacking circulation (Supplementary Fig. 6a, c) and we did not observe consistent reduction in other known klf2 responsive genes5-7 in the absence of flow or klf2a (Supplementary Fig. 6a). These results raised the possibility that a post-transcriptional mechanism linked flow, klf2a, and Vegf signaling. A candidate for this role was the endothelial-restricted microRNA, miR-12618, which can enhance Vegf signaling19, 20. While miR-126 expression was apparent in the embryonic vasculature prior to circulation (Supplementary Fig. 8a), at later stages its expression appeared much higher in the AAs (Fig 3a, Supplementary Fig. 8a). Strikingly, we found that AA miR-126 expression was dependent on both flow and klf2a expression. While control embryos expressed high levels of miR-126 within the AAs, tnnt2- or klf2a-deficient embryos did not (Fig. 3a-c; Supplementary Figs. 6a and 8b, Supplementary Table 2). By contrast, expression of cdh5 and let-7a was unchanged in the absence of flow or klf2a (Fig. 3d-f, Supplementary Fig. 8b), ruling out a general defect in endothelial gene expression or microRNA processing, respectively. Tricaine treatment to block flow similarly reduced miR-126 AA expression (Supplementary Fig. 6b; Supplementary Table 2). Embryos injected with a Morpholino to prevent miR-126 processing (Supplemental Fig. 8c) displayed blocked AA circulation (Supplementary Movie 10) and hemorrhage in this region by 60 hpf (Supplementary Fig. 8d). Similar to loss of klf2a, the AA5x did not form in miR-126-deficient embryos (Fig. 3g; Supplementary Table 3; Supplementary Movie 11). We also observed ectopic branching of segmental vessels and abnormal patterning of cranial blood vessels in miR-126-deficient embryos (Supplementary Fig. 8e). These results demonstrate that AA expression of miR-126 requires flow and klf2a and that miR-126 itself is required for AA5x angiogenesis.

Figure 3

miR-126 acts downstream of klf2a during AA5x development

a-f, Ventral view, anterior is up. Bracket: AA3-6. Expression of miR-126 (a-c) or cdh5 (d-f). Embryos injected with 11 ng control MO (a, d), 2 ng tnnt2 MO (b, e), 11 ng klf2a ATG MO (c, f). g-j, Tg(kdrl:egfp) embryos at 65 hpf, dorsal view, anterior is up. Endothelial cells are green, blood flow is red. Embryos injected with 20 ng of control or miR-126 MO (g), 2 ng klf2a ATG MO (h), 7 ng miR-126 MO (i), or co-injected with 2 ng klf2a ATG MO and 7 ng miR-126 MO (j). k-m, Tg(kdrl:egfp) embryos co-injected with 11 ng klf2a ATG MO and pTol-fli1ep:miR-126/mcherry (k, l) or 11 ng klf2a ATG MO and pTol-fli1ep:mcherry (m). k-m, Yellow indicates egfp and mcherry co-expression; dorsal views, anterior is up. g-m, Arrows: lateral dorsal aortae (LDA), arrowheads: AA5x.

Our results suggested that klf2a acted upstream of miR-126 to induce flow-stimulated angiogenesis. Consistent with this possibility, exogenous klf2a in embryos lacking blood flow restored AA miR-126 expression (Supplementary Fig. 9a-c). To further test their genetic interaction, we co-injected klf2a and miR-126 Morpholinos at suboptimal doses that individually caused no, or mild low penetrant aortic arch defects (Fig. 3h, i; Supplementary Fig. 9d,e; Supplementary Table 3). Co-injection of both Morpholinos in this case caused a drastic increase in the penetrance of AA5x defects, suggesting that miR-126 and klf2a act in a common pathway (Fig. 3j, Supplementary Fig. 9e, Supplementary Table 3). Interestingly, other vascular defects observed in miR-126-deficient embryos were not apparent in co-injected embryos (Supplementary Fig. 8e, data not shown), suggesting a specific genetic interaction between miR-126 and klf2a during AA5x development. To further confirm that miR-126 functioned downstream of klf2a, we drove mosaic endothelial expression of a miR-126/monomeric cherry (mcherry) transgene in klf2a-deficient embryos using the fli1ep promoter fragment (Supplementary Fig. 10a; ref 21). This construct drove flow-independent endothelial expression of mature miR-126 (Supplementary Fig. 10b, c and data not shown) and led to an increased proportion of klf2a-deficient embryos with AA5x formation as compared to injection of klf2a Morpholino alone (Supplementary Table 3). Rescued embryos displayed miR-126/mcherry transgene expression in AA5x endothelial cells, including cases of bi- and uni-lateral rescue (Fig. 3k, l), while the control fli1ep:mcherry transgene failed to rescue (Fig. 3m). These results indicate that miR-126 acts downstream of klf2a to drive flow-stimulated angiogenesis.

miR-126 promotes angiogenesis by repressing spred1 and pik3r2, which normally inhibit Vegf signaling19, 20. Our observations suggested that in the absence of flow and klf2a, reduced miR-126 expression allows upregulation of these molecules thereby preventing Vegf-induced AA5x angiogenesis. While miR-126 can repress the zebrafish spred1 3’UTR, it had no effect on pik3r2 in whole embryo miRNA sensor assays (Supplementary Fig. 11a). Using an endothelial autonomous miRNA sensor assay (Supplementary Fig. 11b), we further found that the spred1 3’ UTR prevented expression of a mcherry transcript in blood vessels, while egfp fused to a control 3’UTR was expressed (Fig. 4a, b; Supplementary Fig. 11c). By contrast, the mcherry-spred1-3’UTR transgene was robustly expressed in embryos lacking miR-126, blood flow, or klf2a (Fig. 4c-h, Supplementary Fig. 11c). These results support a genetic pathway in which spred1 repression is mediated by klf2a and miR-126 in response to flow. Accordingly, over-expression of mRNA encoding Spred1 blocked AA5x formation (Fig. 4i, j, Supplementary Table 3), while reducing Spred1 in miR-126-deficient embryos rescued AA5x development (Fig. 4k, l, Supplementary Table 3). Taken together, our findings support the existence of a genetic pathway in which flow induces klf2a and miR-126 (Fig. 4m). While our data suggest that the interaction between these genes occurs in AA endothelial cells, we cannot rule out the possibility of an indirect role for klf2a upstream of miR-126. Nevertheless, flow-stimulated miR-126 subsequently inhibits spred1 in endothelial cells to allow angiogenesis to proceed in response to Vegf (Fig 4m). In the absence of flow, klf2a and miR-126 are reduced allowing spred1 to repress Vegf-stimulated angiogenesis. Thus, miR-126 provides a crucial link between flow and Vegf signaling to promote angiogenesis. Importantly, flow, klf2a, and miR-126 were similarly required for angiogenesis in the zebrafish-xenograft model22 (Supplementary Fig 12), suggesting that this pathway may represent a general mechanism for flow-stimulated angiogenesis in the zebrafish.

Figure 4

Flow-mediated repression of Spred1 is required for AA5x angiogenesis

a-h, Cranial vessel expression of a miR-126 sensor at 65 hpf; lateral views, anterior to the left, dorsal is up. a, c, e, g, Expression of Egfp fused to control 3’UTR (green) and mCherry-spred1-3’UTR (red), coexpression is yellow. b, d, f, h, Expression of mCherry-spred1-3’UTR (red). Embryos co-injected with miR-126 sensor construct and 20 ng control MO (a, b), 20 ng miR-126 MO (c, d), 2 ng tnnt2 MO (e, f) , or 11 ng klf2a ATG MO (g, h). i-l, Tg(kdrl:egfp) embryos at 65 hpf, dorsal view, anterior is up. Endothelial cells in green, circulation in red. Arrow - Lateral dorsal aortae; AA5x - arrowheads. Embryos left uninjected (i), injected with 100 pg spred1 mRNA (j), 20 ng control MO (k), or 1 ng spred1 MO and 20 ng miR-126 MO (l). m, Pathway responsible for flow-stimulated angiogenesis.

The stereotyped pattern of the vertebrate circulatory system is initially established by conserved genetic pathways that act before circulation to drive endothelial differentiation and provide guidance cues. How haemodynamic forces subsequently modulate these pathways in vivo is largely unknown. Our current work provides new insights into how an endothelial cell's response to flow can be integrated with early developmental signals to drive angiogenesis in the presence of flow.

Methods summary

Zebrafish and their embryos were handled according to standard protocols23 and in accordance with University of Massachusetts Medical School IACUC guidelines. For laser-assisted microsurgery, embryos at 46 hpf were anesthetized and immobilized in 0.5% of low-melt agarose (Biorad). The connection between AA5 and AA6 and the ventral aorta was ablated using a Micropoint laser (Photonic Instrument, Inc) mounted on a Zeiss AX10 Imager M1. SU5416 (Calbiochem) was prepared and used as described previously11. Control embryos were treated with 0.1% dimethyl sulfoxide (DMSO). To arrest heartbeat, embryos were treated with 15 mM of 2,3-butanedione 2-monoxime (BDM; Sigma-Aldrich) or with buffered Tricaine methanesulfonate (Sigma-Aldrich) at 0.66 mg/ml in egg water for the indicated times. Two-photon time-lapse imaging, confocal microscopy and microangiography was performed as previously13, 24, with additional modifications as noted in Supplementary Methods. Antisense riboprobes against dll4, vegfa, kdrl, fli1a, and cdh5 were generated and used for whole mount in situ hybridization as described elsewhere25. A klf2a fragment was PCR amplified and cloned by Gateway recombination. The resulting clone was linearized with BglII and a DIG-labeled riboprobe was synthesized using T7 polymerase. Digoxigenin (DIG)-labeled locked nucleic acid (LNA) probes (Exiqon, Copenhagen) were used to detect mature miR-126 and let-7 using in situ hybridization or Northern analysis as described elsewhere18. Morpholinos, mRNA and Tol2-based plasmids were prepared and injected as previously11,21. In cases of co-injection with Morpholinos, Tol2-plasmids and transposase, a DNA/transposase mRNA mixture was initially injected, followed by Morpholino. Plasmid construction details are provided in the full methods section. Morpholinos against vegfa, tnnt2 and gata1 have been described elsewhere15, 26, 25; all other Morpholino and oligonucleotide sequences are provided in the full methods section.

Morpholino	sequences
klf2a ATG MO	GGACCTGTCCAGTTCATCCTTCCAC
klf2a splice blocking MO	CTCGCCTATGAAAGAAGAGAGGATT
miR-126 MO	TGCATTATTACTCACGGTACGAGTTTGAGTC
spred1 MO	AAACCTGTGGAAGGAGAAGGAAACC
control MO	Oligo-31N
tnnt2 MO	CATGTTTGCTCTGATCTGACACGCA
Primers
Klf2a attb1	gggg aca agt ttg tacaaaaaagcaggctacaggtcc ATGCATCTCAGC
Klf2a attb2	GGGGACCACTTTGTACAAGAAAGCTGGGTcattttccagagtccgttcC
Spred 3′UTR attb2	GGGG ACA GCT TTC TTG TAC AAA GTG G
	CGCTTGTGCCACCGCTGC
Spred 3′UTR attb3	GGGG AC AAC TTTGTATAATAA AGT TG
	GCAGAGCGTTTGGCGGTGT
PIk3r2 3UTR attb2	GGGG ACA GCT TTC TTG TAC AAA GTG G
	CACAACGACTCGCTGAATGT
PIK3r2 3UTR attb3	GGGG AC AAC TTT GTA TAA TAA AGT TG
	TCCTCAAGCTGGGATCATGT
Spred ORF attb1	ggggacaagtttgtacaaaaaagcaggctacatgagcgaa gaa cca aac
Spred ORF attb2	ggggaccactttgtacaagaaagctgggttcatcctgcagccttgt
miR126 chm8 F	TGCCATGCCTTGACGAGAAG
miR126 chm8 R	GTGATATTTAATGTAACATGCC
qpcr Bact F	AGCTTGAAACTCGCCAAGTG
qpcr Bact R	CAGCTTTATAGCCGGCACTG
qpcr vegf F	GCCAAAGGCAGAAGTCAAAG
qpcr vegf R	TGCAGGAGCATTTACAGGTG
qpcr Kdrl F	TCTACTGGGCTTTTCCCTCTC
qpcr Kdrl R	AGGGTTACTATGGTGACGTTGC
qpcr Zelastin F	GCTCGTCTCCATACAAAGCA
qpcr Zelastin R	CAGAACTCCTCCTGGTAGCC
qpcr Z Tie2 F	GCCGTCAAGAGGATGAAAGA
qpcr Z Tie2 R	GCTGCTGTGAGGAGAGTGTG
qpcr Znos1 F	TCAAATACGCCACCAACAAA
qpcr Znos1 R	GAGAAGAAGGGGCAAAACATC
qpcr ZItgb5 F	TGGGAAGGATGGACAAAGAG
qpcr ZItgb5 R	CGGGTGAATGAGGAGACACT
qpcr ZEdg1 F	CCACCGTATTCAGCGTTATT
qpcr ZEdg1 R	GTCAAGGAGGAGCAGGATGA
qpcr ZEdn 1 F	ATGAGGCGAAACCACAGAAG
qpcr ZEdn 1 R	AGAAACCACTTGAGCGAGG
F-qpcr klf2a	CCAACGCCAACCAAAGAGTA
R-qpcr klf2a	CTCGCCTGTGTGTGTTCTGT
Mir 126 qpcr F for miRscript	CGT ACC GTG AGT AAT AAT GC
R-RT-PCR klf2a	ACGTGGTACCCGCACGGCGAACTCACACTTG
F-RT-PCR klf2a	ACGTGGTACCacgcacggcgactcacacttg
miRNA Duplex
miR-126 sense	ucguaccgugaguaauaaugc
miR-126 antisense	gcauuauuacucacgguacga
miR-126 scramble antisense	gcaccacaacucaagguucga