|Home | About | Journals | Submit | Contact Us | Français|
Angiogenesis requires coordination of distinct cell behaviors between tip and stalk cells. While this process is governed by regulatory interactions between the Vascular endothelial growth factor (Vegf) and Notch signaling pathways, little is known about the potential role of microRNAs. Through deep sequencing and functional screening in zebrafish, we find that miR-221 is essential for angiogenesis. miR-221 knockdown phenocopied defects associated with loss of the tip cell-expressed Flt4 receptor. Furthermore, miR-221 was required for tip cell proliferation and migration, as well as tip cell potential in mosaic blood vessels. miR-221 knockdown also prevented “hyper-angiogenesis” defects associated with Notch deficiency and miR-221 expression was inhibited by Notch signaling. Finally, miR-221 promoted tip cell behavior through repression of two targets: cyclin dependent kinase inhibitor 1b (cdkn1b) and phosphoinositide-3-kinase regulatory subunit 1 (pik3r1). These results identify miR-221 as an important regulatory node through which tip cell migration and proliferation are controlled during angiogenesis.
During embryogenesis, vascular development proceeds through two distinct stages (Poole and Coffin, 1989). De novo formation of blood vessels, or vasculogenesis, begins with the emergence of angioblast progenitor cells from the lateral mesoderm and migration to target tissues where they coalesce, assemble into cords, and lumenize to form the major blood vessels. Subsequently, smaller caliber vessels sprout from pre-existing ones by angiogenesis( Poole and Coffin, 1989), which requires careful orchestration of distinct cell behaviors between adjacent endothelial cells (Gerhardt et al., 2003). Initial sprouting is driven by pro-angiogenic factors in the surrounding extracellular environment that induce selected cells to emerge from an established blood vessel. These leading tip cells exhibit extensive filopodia and pathfinding behavior, while trailing stalk cells do not (Gerhardt et al., 2003). In some vascular beds (e.g. mouse retinal vasculature), the vascular plexus grows via proliferation of stalk cells as tip cells migrate, while in others (e.g. zebrafish segmental vessels), tip cells exhibit both migration and proliferation (Gerhardt et al., 2003; Siekmann and Lawson, 2007). Despite these differences, the distinction between tip and stalk cells is an essential conserved process that allows blood vessels to grow while maintaining their connection to the patent vascular system (Thurston et al., 2007).
The signals that coordinate tip and stalk cell behaviors have been the subject of intense investigation. Initial tip cell emergence is driven by Vascular endothelial growth factor A (Vegfa), which is acutely required for tip cell filopodia activity and migration (Gerhardt et al., 2003). Accordingly, tip cells express receptors for both Vegfa (Vegfr-2) and Vegfc (Vegfr-3/Flt4), the latter of which is highly dynamic and becomes restricted to the tip cell during sprouting( Covassin et al., 2006b; Gerhardt et al., 2003; Siekmann and Lawson, 2007; Tammela et al., 2008). Vegfa induces tip cell expression of the Notch ligand, dll4, and subsequent Notch activation in the stalk cell suppresses tip cell behaviors, in part, by repressing flt4 expression (Hellstrom et al., 2007; Lobov et al., 2007 Siekmann, 2007 #901; Tammela et al., 2008). Consistent with this model, mouse or zebrafish embryos lacking dll4 display excessive blood vessel branching and endothelial proliferation (Hellstrom et al., 2007; Leslie et al., 2007; Siekmann and Lawson, 2007; Suchting et al., 2007), which can be normalized by reducing Vegfr-3/flt4 signaling (Hogan et al., 2009; Siekmann and Lawson, 2007; Tammela et al., 2008). More recent observations have demonstrated competition between endothelial cells for the leading tip cell position during sprouting (Jakobsson et al., 2010). Endothelial cells with lower Vegfr-2 or increased Notch signaling are often excluded from the tip cell position (Jakobsson et al., 2010), consistent with similar results in mosaic zebrafish blood vessels (Siekmann and Lawson, 2007). Thus, tip and stalk cell behaviors are coordinated through negative interactions between the Vegf and Notch pathways. However, little is known about the possible role of post-transcriptional control of this dynamic process.
MicroRNAs are expressed as autonomous transcripts, or are found in introns where they are expressed within pre-mRNA (Ghildiyal and Zamore, 2009). Autonomous microRNA precursors are cleaved in the nucleus by Drosha (Lee et al., 2003), leaving short hairpin RNAs that are transported to the cytoplasm and cleaved by the ribonuclease Dicer ( Bernstein et al., 2001; Hutvagner et al., 2001). The mature strand from the resulting double stranded ~22 nucleotide microRNA is subsequently incorporated into a ribonucleoprotein silencing complex (RISC; Czech and Hannon, 2011; Gregory et al., 2005) and used as a complementary guide sequence that binds to sites in the 3′ UTR of a target mRNA. Upon RISC binding, a target transcript is deadenylated leading to its degradation and translational repression (Giraldez et al., 2006; Guo et al., 2010; Wu et al., 2006). Thus, in most cases, microRNAs provide a mechanism for post-transcriptional repression of gene expression. Since microRNAs have been implicated in multiple aspects of vascular growth control (Suarez and Sessa, 2009), we reasoned that they might play a role to coordinate tip and stalk cell behaviors. By applying deep sequencing and functional screening of microRNAs in zebrafish embryos, we identified mir-221 as an essential regulator of tip cell proliferation and migration.
To identify candidate microRNAs required for angiogenesis, we sequenced small RNAs from zebrafish endothelial cells at 24 hours post fertilization (hpf), at which time there is extensive vascular growth (Isogai et al., 2003; Siekmann and Lawson, 2007; Figure S1A). Approximately 20 microRNAs displayed 2 fold or greater enrichment in GFP+ cells isolated from Tg(kdrl:egfp)la116 embryos compared to GFP- cells and more than 50 were highly expressed (>1000 sequence tags) in GFP+ cells, many of which are also expressed in human endothelial cells (Table S1; Kuehbacher et al., 2007). We also noted hematopoietic-expressed microRNAs (e.g. miR-223, miR-451) consistent with kdrl:egfp expression in hematopoietic cells (Bertrand et al., 2010).
We validated candidate microRNA expression by whole mount in situ hybridization using locked nucleic acid (LNA) probes. Candidates were selected based on enrichment (e.g. miR-20b) or high expression (e.g. miR-221) in kdrl:gfp-positive cells. Several candidate microRNAs displayed expression similar to the endothelial-expressed transcription factor, fli1a, at 48 hpf including miR-107, miR-20b, miR-221, and miR-222 (Figure 1A, Table S1, and data not shown). To assess the function of these microRNAs, we injected Morpholinos (MOs) to block their maturation (Figure S1B and data not shown) and observed vascular and overall morphology in Tg(kdrl:egfp)la116 embryos. In most cases, we noted normal development and ISV formation at 30 hpf, except for miR-221-knockdown, which caused partial ISV formation (Figure 1B). We also observed distinct cardiovascular defects associated with loss of other microRNAs. For example, miR-107-deficient embryos displayed vascular stability defects (Figure S1C), while embryos lacking miR-20b exhibited an apparent block in erythroid differentiation (Figure S1D). Based on our interest in identifying microRNAs involved in angiogenesis, we investigated the function of miR-221 in greater detail.
Consistent with our deep sequencing data, miR-221 is detected at high levels in endothelial cells from Tg(kdrl:egfp)la116 embryos at 24 hpf (Figure S1E). While a high proportion of miR-221-deficient embryos displayed partial ISV formation at 30 hpf (Figure 1B, S1F), general embryonic development was not delayed and overall morphology appeared normal (Figure 1B). Loss of miR-221 did not perturb development or differentiation of structures adjacent to ISVs, such as somites, neural tube, and notochord (Figure S1G). Closer inspection revealed defects in formation of the primordial hindbrain channel, a transient vein in the zebrafish hindbrain (Isogai et al., 2003), in embryos lacking miR-221, while the adjacent lateral dorsal aortae formed normally (Figure 1C). We also noted lack of the thoracic duct, a primary lymphatic vessel, at 5 days post fertilization while overall morphology was normal at this stage (Figure 1D, E). A second MO targeting miR-221 caused the same vascular defects (Figure S1H, I and data not shown). Although there is a high degree of homology between miR-221 and miR-222, both MOs were specific for miR-221 (Figure S1J) and simultaneous knockdown of miR-221 and miR-222 did not enhance or cause additional vascular defects (data not shown). We did not note any changes in expression of either vegfa, or its receptor kdrl (Figure S2A), which are essential for ISV sprouting, (Covassin et al., 2006b) in miR-221 deficient embryos. Likewise, ISV and vein-specific expression of flt4 and artery-specific expression of its ligand vegfc, as well as ephrinb2a, notch3, and dll4 were not affected by loss of miR-221 (Figure S2A, B). Circulation and heart function were also normal in miR-221 deficient embryos (data not shown). Together, these results demonstrate that miR-221 is highly expressed in endothelial cells and is required for angiogenesis.
In the zebrafish embryo, kdrl and flt4 cooperate to drive ISV growth, while flt4 alone is required for vein and lymphatic development and kdrl alone is required for artery differentiation and morphogenesis (Covassin et al., 2006b; Habeck et al., 2002; Hogan et al., 2009). Strikingly, the phenotypes associated with loss of miR-221 were identical to those in embryos lacking flt4 (Covassin et al., 2006b; Hogan et al., 2009). By contrast, miR-221-deficient embryos did not display hallmarks of Vegfa deficiency, such as arteriovenous circulatory shunts or loss of artery differentiation (data not shown and Figure S2A). These observations suggested that miR-221 preferentially affects the Vegfc/Flt4 pathway. To determine if this was the case, we first assessed the genetic interaction between miR-221 and kdrl or flt4 by quantifying ISV length in embryos lacking combinations of these genes. Embryos injected separately with either 1 ng of flt4 or 5 ng of miR-221 MO displayed ISVs of normal length at 30 hpf similar to embryos injected with control MO (Figure 2A–C, G). However, co-injection of these sub-phenotypic MO doses resulted in significantly shorter ISVs (Figure 2D, G), similar to a more complete knockdown of either gene alone at higher doses (Figure 2G). Furthermore, we did not observe any enhancement of the ISV defect in embryos co-injected with these higher MO doses (Figure 2G), suggesting that flt4 and miR-221 act in the same pathway. By contrast, reduction of miR-221 in mutant embryos bearing a null kdrlum19 allele decreased ISV length compared to control kdrlum19 mutant embryos (Figure 2E–G), which normally display variable partial shortening in ISV length (Covassin et al., 2009). The enhanced ISV defect in embryos lacking miR-221 and kdrl is similar to the effect of simultaneously reducing flt4 and kdrl (Covassin et al., 2006b). These observations suggest that miR-221 acts parallel to vegfa and kdrl and is likely acting in a common pathway with vegfc and flt4. Despite this genetic interaction, we did not detect changes in miR-221 levels in the absence of flt4 (Figure S2D) and, as noted above, flt4 levels were normal in miR-221 deficient embryos (Figure S2A, B).
We next determined if miR-221 function was required for Vegfc/Flt4 signaling by assessing the response of miR-221-deficient embryos to ectopically expressed Vegfc. We used the heat shock 70 (hsp70) promoter to inducibly express Vegfc in frame with monomeric red fluorescent protein (mCherry) separated by the viral 2A peptide sequence in zebrafish embryos. Mosaic Vegfc over-expression following heat shock at 15 somite state caused ectopic branching of ISVs along the horizontal myoseptum in control embryos at 33 hpf (Figure 2H, L), but not in flt4-deficient embryos (Figure 2I, L) indicating that this effect is dependent on the Flt4 receptor. Likewise, miR-221 deficient embryos displayed significantly fewer ectopic vessels in response to Vegfc (Figure 2J, L). The loss of ectopic sprouting was not generally attributable to partial ISV formation caused by flt4- or miR-221-deficiency, as kdrlum19 mutant embryos, which also exhibit partial ISV sprouts (see above), form ectopic vessels in response to Vegfc (Figure 2K, L). Together these results suggest that miR-221 is required for Vegfc/Flt4 signaling during ISV sprouting.
In sprouting blood vessels, flt4 expression is highly dynamic and becomes restricted to endothelial tip cells (Covassin et al., 2006b; Siekmann and Lawson, 2007; Tammela et al., 2008). Since miR-221 was required for ISV growth and Vegfc/Flt4-induced angiogenesis, we reasoned that it might play a role in governing endothelial tip cell behaviors. To determine if this was the case, we performed two-photon time-lapse microscopy on embryos lacking or over-expressing miR-221. In control embryos, ISVs generally formed by migration of endothelial tip cells from the dorsal aorta followed by a second trailing cell (Figure 3A, 19:30 and 20:54; Supplementary Movie 1), as shown previously (Siekmann and Lawson, 2007). Once reaching the horizontal myoseptum, tip cells in most ISVs divided (Figure 3A, 20:54 to 24:36) and a single daughter cell subsequently migrated dorsally to form the DLAV (Figure 3A, 27:01). In the absence of miR-221, the initial rate of movement for tip and trailing cells from the dorsal aorta into the ISV was relatively normal (Figure 3B, 19:30 to 20:50; Supplementary Movie 2; Figure S3A). However, subsequent migration of tip cells from the horizontal myoseptum to the DLAV was significantly slower than in control embryos (for example, compare Figure 3B, 26:51 to Figure 3A, 24:36; Supplementary Movie 2, Figure S3A). In addition, tip cells often failed to divide in miR-221 deficient embryos (Figure 3B, Supplementary Movie 2). Together, these defects resulted in delayed formation of the ISV and DLAV, which were ultimately comprised of fewer endothelial cells in miR-221 deficient embryos (see below). By contrast, injection of miR-221 duplex induced precocious division of tip cells prior to reaching the horizontal myoseptum (Figure 3C, 19:30 to 20:20; Supplementary Movie 3). In addition, excessive numbers of cells from the dorsal aorta continued to migrate into the ISV (Figure 3C, 20:20 to 23:07), further increasing cell numbers in the ISVs (Figure 3C, 28:37 and see below), although ISV growth rate was not faster than in control embryos (data not shown). Consistent with our time-lapse observations, fewer tip cells in miR-221-deficient ISVs incorporated BrdU than in control embryos when pulsed beginning at 20 hpf, while exogenous miR-221 expression increased BrdU incorporation in cells that contributed to the DLAV (Figure S3B, C). Interestingly, the number of BrdU-positive cells in the dorsal aorta and posterior cardinal vein did not change significantly in either case (Figure S3D), suggesting a preferential effect of miR-221 on ISV tip cell proliferation. These observations indicate that miR-221 is required for both migration and proliferation of endothelial tip cells during ISV angiogenesis.
We have previously assessed the potential of an endothelial cell to be a tip cell (referred to hereafter as “tip cell potential”) by determining its ability to contribute to the DLAV, which, based on time lapse analysis, is initially formed from a daughter of the initial ISV tip cell. Using this assay, we have successfully demonstrated that endothelial cells lacking Notch activity, which exhibit excessive proliferation and migration, preferentially localize to the DLAV (Siekmann and Lawson, 2007), while recent studies demonstrate a similar behavior for endothelial cells partially deficient in plexin signaling (Zygmunt et al., 2011). To assess the tip cell potential of miR-221-deficient or over-expressing cells, and to confirm the endothelial autonomy of these effects, we transplanted donor cells from Tg(fli1a:egfp)y1 embryos injected with miR-221 MO or miR-221 duplex, respectively, into wild type Tg(kdrl:ras-cherry)s916 host embryos. Control donor cells contributed well to all host blood vessels in the zebrafish trunk (Figure 4A, D, E; control MO into wt, N=20; control duplex into wt, N=16), while miR-221-deficient endothelial cells contributed less frequently to the DLAV (Figure 4B, D; N=17), consistent with their reduced proliferation and migratory rates observed in time-lapse analysis (see above). On the contrary, donor cells expressing exogenous miR-221 showed an enhanced ability to contribute to the DLAV (Figure 4C, E; N=21). Donor cells with either gain or loss of miR-221 expression otherwise contributed to most other blood vessel positions, ruling out a general defect in endothelial cell specification (Figure 4A–E). Furthermore, ISV formation was normal in embryos where miR-221-deficient donor cells contributed only to surrounding neural tube or somite tissue but not endothelial cells (Figure 4F; N=10). Thus, miR-221 increases tip cell potential during ISV angiogenesis and does so in an endothelial autonomous manner.
The defects caused by exogenous miR-221 were similar to those associated with loss of dll4 (Hellstrom et al., 2007; Leslie et al., 2007; Siekmann and Lawson, 2007), suggesting an antagonistic relationship between miR-221 and Notch signaling. Therefore, we investigated the genetic interaction between miR-221 and dll4 by comparing ISV length and cell numbers in embryos following knockdown of one or both of these genes. Similar to previous observations, dll4 deficiency resulted in significantly more ISV endothelial cells at 27 hpf than in control embryos (Fig. 5A–B, J), similar to injection of miR-221 duplex (Fig. 5C, J). By contrast, mir-221 knockdown significantly decreased cell numbers, as well as ISV length (Fig. 5D, J, K), consistent with time-lapse analysis (see Figure 3). Likewise, simultaneous reduction of mir-221 and dll4 decreased ISV cell number and length compared to dll4 knockdown alone (Figure 5E, J, K). At 35 hpf, dll4-deficient embryos displayed excessive cell numbers compared to control embryos (Figure 5F, G, J), while ISVs in mir-221-deficient embryos, which have recovered to form a DLAV and exhibit normal ISV length (Figure 5K, H), displayed fewer ISV cells (Figure 5H, J). Furthermore, dll4-deficient embryos lacking miR-221 displayed normalized numbers of cells similar to wild type ISVs (Figure 5I, J). Taken with our finding that exogenous miR-221 drives excessive migration and proliferation, these observations suggest that miR-221 is required for excessive angiogenesis associated with loss of Notch.
Notch signaling blocks angiogenesis, in part, by repressing flt4 expression (Siekmann and Lawson, 2007). Since miR-221 acted in the flt4 signaling pathway, we determined whether it might also be repressed by Notch. In embryos injected with a MO targeting the Rbpsuh DNA binding protein, which is required for Notch signaling (Bailey and Posakony, 1995; Fortini and Artavanis-Tsakonas, 1994), we found that miR-221 is upregulated (Figure 5L). By contrast, injection of mRNA encoding an activated form of Notch repressed miR-221 levels when compared to control injected embryos (Figure 5L). To demonstrate that the effect of Notch signaling on miR-221 levels was occurring in endothelial cells, we utilized an endothelial autonomous microRNA sensor assay (Nicoli et al., 2010). In this case, we generated a transgenic line (Tg(fli1ep:egfp;mcherry-pik3r1)um28) using an endothelial cell-specific bicistronic vector driving mcherry fused to a 3′ UTR containing two miR-221 binding sites (Figure S4A and see below) and egfp fused to a control 3′UTR. Tg(fli1ep:egfp;mcherry-pik3r1-utr)um28 embryos exhibited robust green fluorescence in trunk blood vessels at 28 hpf, while mcherry levels appeared lower (Figure 5M). Consistent with miR-221-mediated repression of the mcherry transcript, we observed increased red fluorescence in Tg(fli1ep:egfp;mcherry-pik3r1)um28 embryos injected with miR-221 MO (Figure 5M, S4B). By contrast, Mcherry expression was significantly repressed following injection of rbpsuh MO (Figure 5M, S4B), consistent with our observation that miR-221 levels are increased in Notch deficient embryos (Figure 5L). Similar results were observed with a second 3′ UTR bearing miR-221 sites (see below and Figure S4B, C). Taken together, our results demonstrate that Notch negatively regulates miR-221 to block endothelial tip cell proliferation and migration.
In tumor cells miR-221 drives proliferation by repressing cyclin dependent kinase inhibitor 1b (cdkn1b; Galardi et al., 2007), which blocks cell cycle progression. Since our studies demonstrated a block in endothelial proliferation in miR-221 deficient embryos, cdkn1b was a viable candidate target during angiogenesis. In addition, we identified pik3r1, the p85-alpha regulatory subunit of the phosphoinositide-3-kinase (PI3K) complex, as a candidate target of miR-221. Given the importance of PI3K signaling in vascular development (Graupera et al., 2008), we reasoned that this transcript was also a possible functional target of miR-221 during ISV angiogenesis.
Both cdkn1b and pik3r1 3′ UTRs contain miR-221 binding sites (Figure S4A and data not shown) and were repressed by miR-221 in whole embryo reporter assays (Figure 6A–D) and endothelial cells in vivo (Figure 5M, S4B, C). We also detected low-level expression of both transcripts in the trunk blood vessels and isolated kdrl:egfp cells (Figure S5A, B). Consistent with the possibility that increased levels of cdkn1b and pik3r1 contributed to defects associated with loss of miR-221 (see above), overexpression of cdkn1b or pik3r1, or both together, reduced ISV length and cell numbers (Figure 6E, F), without an overt affect on general development (Figure S5C). Furthermore, mosaic analysis demonstrated that cells expressing both pik3r1 and cdkn1b contribute less frequently to the tip cell position than control (Figure 6G; control mRNA>wt, N=20, cdkn1b/pik3r1 mRNA>wt, N=25), similar to miR-221 deficient cells (Figure 4B, D). These results indicate that increased levels of cdkn1b and pik3r1 can negatively affect tip cell potential in an endothelial autonomous manner, further supporting that they are normally targeted by miR-221 during ISV sprouting.
Our results suggest that miR-221 controls ISV growth by inducing proliferation and PI3K activity through repression of cdkn1b and pik3r1, respectively. To further investigate the importance of proliferation and PI3K during ISV growth, we treated embryos at 20 hpf with either 5-hydroxyurea and aphidocolin (HUA/AP) to block cell cycle or LY294002 to inhibit PI3K. In both cases, we observed a significant decrease in ISV length and cell number (Fig. 7A–C), although embryos were normal and did not exhibit changes in vegfa or kdrl expression (Figure S5D and data not shown). The decrease in ISV cell number was more modest in LY294002-treated embryos than those treated with HUA/AP (Figure 7C). Furthermore, similar to miR-221 knockdown, HUA/AP, prevented BrdU incorporation into ISV tip cells while LY294002 did not (Figure 7A, D), suggesting that PI3K inhibition does not affect endothelial cell proliferation during sprouting. Our results demonstrate that pik3r1 over-expression and PI3K inhibition similarly blocked ISV growth, suggesting that increased pik3r1 in the absence of miR-221 blunts PI3K signaling output. To determine if this was the case, we generated transgenic zebrafish (referred to as Tg(fli1ep:phaktegfp-2A-mcherry)um63) bearing EGFP fused to the pleckstrin homology domain of human Akt1 (PH-AKT-EGFP), which localizes to the membrane in response to PI3K activation (Astoul et al., 1999), and a co-expressed red fluorescent protein. In Tg(fli1ep:phaktegfp-2A-mcherry)um63 embryos lacking miR-221, we observed decreased filopodial localization of PH-AKT-EFP compared to control (Figure S5E, F). Furthermore, partial reduction of pik3r1 to levels that do not affect ISV formation (see below) restores filopodial PH-AKT-EGFP localization in miR-221-deficient Tg(fli1ep:phaktegfp-2A-mcherry)um63 embryos (Figure S5E, F). These observations suggest that increased levels of pik3r1 in the absence of miR-221 cause reduced or mis-localized PI3K output in sprouting endothelial cells, which contributes to the observed defects in ISV growth.
Our results suggested that upregulation of pik3r1 and cdkn1b is the likely cause of ISV growth defects in miR-221 deficient embryos. If this were the case, then reducing their levels would rescue the phenotypes in miR-221 deficient embryos. To investigate this possibility, we injected embryos with MOs to inhibit splicing of cdkn1b or pik3r1 (Figure S6A–D). In all cases, we co-injected tp53 MO to eliminate off-target toxicity. Reduction of tp53 alone did not rescue defects associated with miR-221 deficiency (Figure 8A, E, J, L). Wild type embryos lacking cdkn1b exhibited excessive numbers of ISV endothelial cells (Figure 8A, B, I), similar to embryos injected with miR-221 duplex (see Figure 5C, J), but were otherwise normal (Figure S6D). Embryos injected with 10 ng of MO to reduce pik3r1 levels displayed variable shortening of ISV length and a slight, but non-significant decrease in ISV cell numbers and were otherwise overtly normal (Figure 8C,I,K; S6B–D). Injection of 5 ng pik3r1 MO did not cause a phenotype (Figure 8D, I, K). We attempted to rescue miR-221 deficiency by co-injecting 5 ng of either cdkn1b or pik3r1 MO with 10 ng of miR-221 MO. In both cases, we observed a partial restoration of ISV length, although not to control levels (Figure 8E–G, L). Interestingly, knockdown of cdkn1b, but not pik3r1, restored ISV endothelial cell number to control levels in the absence of miR-221 (Figure 8J), consistent with the differential effects of HUA/AP and LY294002 on endothelial proliferation and ISV growth noted above. Simultaneous knockdown of ckdn1b and pik3r1 in miR-221 deficient embryos fully restored both ISV length and cell number (Figure 8H, J, L), indicating that these two targets together mediated the function of miR-221 during ISV growth. Furthermore, endothelial cells with reduced levels of cdkn1b and pik3r1 localized to the DLAV at a greater frequency than control cells in mosaic embryos (Figure 8M; control MO data are same as those shown in Figure 4; cdkn1b/pik3r1 MO>wt, N=24), similar to cells expressing exogenous levels of miR-221 (see Figure 4E). Thus, miR-221 induces tip cell proliferation through down-regulation of cdkn1b and promotes optimal PI3K output by reducing pik3r1. Together, these effects contribute to endothelial tip cell migration and proliferation during angiogenesis.
The ability of an endothelial cell to dynamically respond to pro-angiogenic cues is essential to coordinate distinct cellular behaviors with its neighbors. Without this coordination, productive angiogenesis is hindered. Our present work provides evidence that post-transcriptional regulation by microRNAs plays an important role in this process. In particular, we identify miR-221 as an essential regulator of angiogenesis in embryonic zebrafish and provide evidence that supports a role for miR-221 in endothelial tip cell proliferation and migration.
miR-221 appears to act in sprouting endothelial cells through repression of two distinct target transcripts: pik3r1 and cdkn1b. Our results suggest that these target genes control distinct tip cell behaviors. While miR-221-mediated down-regulation of cdkn1b was required for proliferation, PI3K signaling was dispensable for this behavior in tip cells. Although repression of pik3r1 by microRNAs in other contexts can cause p53-dependent apoptosis (Park et al., 2009), endothelial tip cell survival appeared normal following loss of miR-221, pik3r1, or PI3K signaling and tp53 knockdown did not rescue miR-221 deficiency. Instead, we believe that pik3r1 is important for tip cell migration, consistent with the role for the PI3K catalytic p110α subunit in endothelial cells (Graupera et al., 2008). Our results further suggest that miR-221 appropriately tunes PI3K signaling output by controlling the levels of a PI3K regulatory subunit. Pik3r1 usually exists in a 1:1 ratio with a catalytic subunit of the PI3K signaling complex (Geering et al., 2007), such as p110α, and both proteins are otherwise unstable( Yu et al., 1998). Pik3r1 inhibits the catalytic subunits, but is also required for PI3K activity following membrane localization of the Pik3r1/p110 complex and activation by a receptor tyrosine kinase (Vanhaesebroeck et al., 2010). A central question is how increased Pik3r1 may alter PI3K signaling to block sprouting. In some contexts, Pik3r1 acts independently of PI3K catalytic subunits (Garcia et al., 2006), suggesting that increased Pik3r1 may interfere with other signaling pathways required for angiogenesis. However, in this case excess Pik3r1 would not affect PI3K output, yet we observe decreased PI3K activity in the filopodia of miR-221 deficient ISVs. A more likely possibility is that increased Pik3r1 alters the appropriate balance in regulatory and catalytic subunits, which can be composed of several different isoforms that are present in endothelial cells (Graupera et al., 2008). In turn, a shift in stoichiometry may squelch receptor tyrosine kinase signaling output and PI3K activity. It is also possible that these changes lead to inappropriate subcellular localization of PI3K complexes. In this regard, it is of note that PI3K signaling is active in endothelial filopodia in developing ISVs. Increased Pik3r1 may alter the subcellular localization of PI3K complexes, thereby reducing filopodia PI3K activity, possibly without a change in total cellular PI3K output. In any event, microRNA regulation of PI3K signaling appears to be an emerging theme in the control of this crucial signaling regulator as miR-126 similarly regulates Vegfr-2 signaling output through modulation of Pik3r2 (Fish et al., 2008; Wang et al., 2008).
The importance of miR-221 for endothelial tip cell migration and proliferation is consistent with its requirement for signaling through flt4. Interestingly, neither vegfc, nor flt4 expression is regulated by miR-221, while miR-221 expression is normal in flt4-deficient embryos. We believe that the central point of interaction between miR-221 and flt4 is at the level of pik3r1 regulation. Pik3r1 possesses SH2 domains to facilitate interaction with activated receptor tyrosine kinases and is known to interact with Flt4 following binding to Vegfc (Wang et al., 2004). Thus, miR-221 may directly affect Flt4 signaling output through its regulation of pik3r1 levels. While miR-221 expression is independent of flt4 itself, we find that it is repressed by Notch activation, similar to flt4. Together, our findings suggest a model in which miR-221 promotes migration and proliferation by potentiating flt4 signaling through regulation of pik3r1, while repressing cdkn1b. By contrast, Notch activation in stalk cells represses both flt4 and miR-221. Subsequently, cdkn1b levels increase to limit proliferation, while an increase in pik3r1 serves to dampen, or qualitatively shift PI3K output.
In contrast to our findings, exogenous miR-221 is anti-angiogenic in human venous or lymphatic endothelial cells (Chen et al., 2010; Poliseno et al., 2006; Wu et al., 2011). These effects were mediated through numerous distinct targets depending on the study and included the ETS1 transcription factor (Zhu et al., 2011), the stem cell factor receptor C-KIT (Poliseno et al., 2006), and the transcriptional repressor ZEB2 (Chen et al., 2010). miR-221 can also increase lymphocyte adhesion through repression of ETS1, leading to a decrease in angiotensin II expression (Zhu et al., 2011). Interestingly, this latter work did not note a migration defect in response to miR-221 over-expression, as cited in other studies. A possible explanation for these discrepancies is that miR-221 levels vary significantly in response to both serum and Vegfa (Chen et al., 2010; Suarez et al., 2008), suggesting that growth conditions influence microRNA function in cultured cells. These observations raise the possibility that miR-221 governs context-specific changes in endothelial behavior depending on cell type or developmental stage. Thus, while miR-221 is an important pro-angiogenic signal during embryonic development, it may play different roles in the mature circulatory system.
While miR-221 is only modestly enriched in endothelial cells and is present at high levels in non-endothelial cell types, miR-221 deficiency caused remarkably specific developmental vascular defects. There may several reasons for this observation. First, MOs only provide partial knockdown and the observed phenotypes may be manifest only in cell types (e.g. endothelial tip cells) where high gene dosage is required. Second, miR-221-deficient embryos may display subtle defects in other tissues and developmental processes. Indeed, the observed vascular defects were only obvious when using a transgenic background to visualize blood vessel morphology. Thus, more careful molecular and cellular analysis applied to other organ systems may reveal further defects associated with miR-221 reduction. Finally, microRNAs play an important regulatory role by tuning gene expression to appropriate levels. As such, single microRNA deficiency often has very subtle effects on animal development. In this regard our findings are consistent with those in other models. Interestingly, we also noted specific cardiovascular defects associated with knockdown of other microRNAs (e.g. miR-107 and miR-20b), suggesting a widespread role for small RNAs in vascular development. Further screening of microRNA function in zebrafish will likely reveal additional roles for small RNAs in vascular formation, function and homeostasis.
Zebrafish used in this study are described in Supplementary Experimental Procedures.
Endothelial cells were isolated from Tg(kdrl:egfp)la116 embryos as previously(Covassin et al., 2006a). Small RNA (18–24 nt) purification, adapter ligation, cDNA synthesis and library amplification were performed as described (Gu et al., 2009). Deep sequencing was performed at the Center for AIDS Research at UMass Medical School. Reads were mapped to known zebrafish microRNAs using the latest version of mirDeep (Friedlander et al., 2008).
For Northern analysis, 5 to 10 μg of total RNA was resolved on a 15% acrylamide/7 M urea gel and transferred by electrophoresis to positively charged nylon membrane (Millipore) using a Trans-Blot SD apparatus (Biorad). Hybridization and detection using digoxigenin (DIG)-labeled locked nucleic acid (LNA) probes (Exiqon) were performed as described (Kloosterman et al., 2006). Whole mount in situ hybridization using LNA or antisense RNA probes was performed as previously (Kloosterman et al., 2006; Nicoli et al., 2010). miRNA-quantitative-PCR was performed using the miScript PCR system (Qiagen). Probe and primer sequences can be found in the Supplementary Experimental Procedures.
MO sequences and primers designed for RT-PCR validation are described in Supplementary Experimental Procedures. Knockdown of the zebrafish cdkn1b and pik3r1 genes was accomplished using MOs targeting splice junctions within these genes which were validated by RT-PCR and qRT-PCR on RNA isolated from injected embryos. MOs targeting rbpsuh, flt4 and dll4 are described elsewhere (Covassin et al., 2006b; Siekmann and Lawson, 2007). MOs were synthesized by Gene Tools, LLC and dissolved in DEPC water.
Whole embryo and endothelial autonomous microRNA sensor assays were carried out as previously(Nicoli et al.). For details see Supplementary Experimental Procedures.
10 to 20 embryos injected with 3′utr sensor and control mRNAs at 24 hpf were dechorionated and triturated in calcium free ringer solution (Covassin et al., 2006a) and lysed as described (Rand et al., 2004). 40 μg of total protein was separated on a 12 % SDS-PAGE gel. Western blotting was performed according to standard protocols. GFP and Cherry were sequentially detected using rabbit anti-GFP (1:1000, Invitrogen) and rabbit anti-DsRed (1:1000, Clontech), respectively, followed by chemiluminscent immunodetection using anti-rabbit HRP conjugate (1:20,000, Invitrogen).
To block cell division, we treated embryos simultaneously with 150 μM aphidicolin (Sigma) and 20 mM hydroxyurea (Sigma) as described (Lyons et al., 2005). To block PI3K signaling we used 25 μM LY294002 (Sigma). Drug treatments were performed at 20 hpf on dechorionated embryos in agarose-coated dishes. BrdU labeling was performed by injection of 50 mM BrdU in 0.2 M KCl directly into the yolk of Tg(kdrl:egfp)la116 embryos. Embryos were injected at 20 hpf, transferred to embryo media, and incubated at 28.5°C until 28 hpf. For whole mount immunostaining, embryos were fixed in 4% paraformaldehyde for 2 hours at room temperature and transferred to methanol overnight at −20°C. Embryos were rehydrated to PBST (PBS + 0.5% Triton X-100) and incubated in 2N HCl for 1 hour at room temperature, washed in PBST and placed in blocking solution (PBST + 1% DMSO + 1% BSA + 0.2% goat serum) for 30 minutes at room temperature. To detect BrdU, embryos were immunostained with Alexa-594 anti-BrdU antibody (1:200, Invitrogen) and Alexa-488 anti-GFP antibody (1:300, Invitrogen).
Whole mount live and fixed embryos were analyzed using a MZFLIII microscope equipped with epifluorescence. Digital images were captured using a Zeiss mRC digital camera and AxioVision software. Confocal stacks in green (ex. 488 nm laser) and red (ex. 651 nm laser) channels were acquired sequentially using a Leica SP2 confocal microscope. Two-photon imaging was performed using a LSM7 MP Laser scanning microscope (Zeiss) equipped with a Chameleon Ti:Sapphire pulsed laser (Coherent, Inc.) and images acquired using ZEN 2009 software. For detection of BrdU and Egfp in Tg(kdrl:egfp)la116 embryos, we sequentially scanned embryos at 1040 nm (70% power) and 920 nm (20% power). Measurements of ISVs were made straight from the edge of the aorta to the leading edge of the sprout. 2-dimensional projections were generated using Imaris (Bitplane). Except where otherwise indicated, all pairwise comparisons were analyzed for significance using a Student’s 2-tailed t-test. Error bars in all graphs represent standard deviation (SD).
Transplantations were performed as described (Covassin et al., 2009). Donors were injected with 10 ng of control MO or miR-221 MO to assess loss-of-function effects or 500 μM of miR-221 duplex or control duplex to assess gain-of-function effects. The contribution of donor cells was assigned as a percentage of total number of host embryos that display GFP endothelial cells in the indicated vessels. To assess non-autonomous effects, donor cells were injected with miniRuby (Invitrogen) and transplanted into Tg(fli1a:egfp)y1 embryos. To determine cell autonomous function of pik3r1 and cdkn1b, donor embryos were co-injected with 5 ng of cdkn1b and 5 ng of pik3r1 MOs or 200 pg of cdkn1b and 300 pg of pik3r1 mRNA respectively. Control transplants were performed using donors injected with 500 pg mcherry mRNA. The proportions of embryos exhibiting contribution to each vessel type following the indicated experimental manipulations were compared using Fisher’s exact test.
We thank Jacques Villefranc and Arndt Siekmann for critical review of the manuscript. We are grateful to Daryl Conte and Craig Mello for providing a protocol for small RNA library construction. We thank Tom Smith for excellent technical assistance and John Polli for fish care. This work was supported by R01HL093467 awarded to N. D. L.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.